pic-a-star - · pdf filetrol the transceiver - and a basic utility is ... pic-/ pic-a-star...

© G3XJP PIC-A-STAR

PART 1General introductionTypical Tx/Rx block diagram

PART 2Software block diagramHow the software works; is packaged; and can beobtained

PART 3PCB Manufacturing processComponent list for Timer, Status, IF, DSP boards;and assembly thereof

PART 4Super VOXTimer board circuit diagram and PCB

PART 5Circuit diagrams for DSP mother; Processor andCODEC daughter boards

PART 6Component location for DSP mother board.DSP board test process

PART 7Overview of G3XJP Tx/RxDSP board construction techniqueDSP board PCB art

PART 8IF board description and circuit diagram

PART 9IF board component layout and PCB artDSP Assembly construction

PART 10DSP Assembly assembly and test

PART 11STAR UI backgroundPossible Tx/Rx layout and front panel

PART 12PicAdapter and Status board circuit diagrams.RS232 wiring.

PART 13Possible stereo amplifierPicAdapter and Status PCB component location andartwork

PART 14Some Pic 'N' Mix suggestions

PART 15DSP operational features and use

PART 16Magic Roundabout circuit diagram, options

PART 17Magic Roundabout PCB component layout and artOverall STAR Rx performance numbers

PART 18Band Pass Filters - discussion and circuit diagram

PART 19Band Pass Filters - PCB component layout andartwork

PART 20Conclusion

INDEX by Peter Rhodes, BSc, G3XJP PIC-A-STAR SOFTWARE TRANSMITTER AND RECEIVER

© G3XJP PIC-A-STAR Page 1

HIS IS A detailed constructionproject aimed at those of modestexperience who would like to en-

hance both their craft and technology skills.At the outset - like me - it may well be

that you don't have the skills or knowledgeto build this project. By the end, you willhave. That is, as I see it, the whole idea.

By design, this is a project without end.From my perspective, it is the basis foryears of happy building to come - and is myfirst investment in a new core transceiverplatform in some 25 years. A glance at thephoto tells you why I needed a new one.

From your perspective it is a source ofideas for improving an existing transceiver -not least, replacing the back-end with apowerful Digital Signal Processing (DSP)capability. There are also some craft tech-niques for handling small-size high-functioncomponents. So, there is something in thisfor everyone with an eye on the self-education requirement of their Licence.

SUMMARYTHE HEART of PIC-A-STAR is the DSPmodule. This provides both the back-endreceiver functionality - as well as SSB/CWgeneration on transmit. The bottom line isabsolutely superb audio quality on bothtransmit and receive. If you want to test theformer, come on the home-brew net frequency (seephoto) any day aroundlunch-time where you willfind at least one STAR inoperation most days. If youwant to test the latter thenyou will just have to makeone.

Being implemented bysoftware, it provides theopportunity to address bothabsolute performance as wellas the delights of operationalconvenience - at zero incre-mental cost. This is pre-cisely the basis for futuredevelopments, but the fun-damental functionality to-gether with some bells andeven the odd whistle hasbeen in daily use here forabout nine months. This isthe project on offer - but bythe time you get there it willhave moved on.

PIC-A-STAR is explicitly designed to beupgraded over the web, so there will noincremental DSP enhancement costs.

POSITIONING DSPYOU MIGHT reasonably expect the authorof a DSP project to have some seriousknowledge in the field. So would I! Actu-ally in many ways it is important to get thispublished before I acquire more thanenough to be merely dangerous.

If, like me, you are at least in your late50's it is unlikely that DSP theory featuredeven in a formal engineering education.And if equally like me you have neverworked in the engineering profession thenyou could reasonably start from the positionthat DSP is some kind of black magic whichyou could never understand in a life-time oftrying. You might well be correct in thisassumption because some of the theory isindeed very heavy.

But my personal discovery was that youdon't need to understand DSP at other thana superficial level to be able to build it athome and to use it.

From a position of not being able to spellDSP, it took me two weeks to get my firstDSP receiver working. The attraction isthat everything since then has been incre-mental and I have not been off-air for asingle day. Design mistakes - and therehave been many - have cost me my time butnever any money - which is about perfect

for a hobby. So this lends itself to a learn-as-you-go approach. In other words, unlikeconversational French, you don't have tolearn a lot before you can even get started.SKILLS AND FACILITIESA REQUIREMENT of all my projects is

that they can be built on the kitchen tablewith no access to professional facilities.Otherwise, it would not be amateur radio.

This one is no exception - though I havehad to acquire new skills and hone them tothe point of repeatability in order to buildsome of the hardware. This is all part of theadventure, part of the fun. And of coursethose skills have general application so theyopen up new avenues for the future.

A simple (and inexpensive) technique formaking precision PCBs will be covered -which includes mounting a 48-pin chip witha mere 0.5mm interval between pins. Andyou get to practice on a really easy one of128 pins by 0.8mm first. If the prospect ofthis puts you off, I really can't help. If itsparks a can-do spirit of adventure then weare in business.

INSPIRATIONTHREE THINGS made this project possi-ble. In the order in which I found them:-• "The Scientist and Engineer's Guide toDigital Signal Processing" by Steven W.Smith. This book is a little gem. If youflick through quickly you will see copiousexamples and illustrations. What you donot see is lots of equations and impenetrablenotation. I need just one quote:- " [thisbook] … is written for those who want to

use DSP as a tool, nota new career." Mykind of book!• The Analog De-vices web-site [1].This contains awealth of both theo-retical and practicalinformation - andspecifically the elec-tronic version of theabove book. Mostvaluable to me werelots of DSP codeexamples for theirADSP-218x proces-sors. The first incar-nation of STAR wasbuilt by six of us ontheir ADSP-2181EZLITE evaluationboard which had be-come somewhat astandard over the

years. Then over one fateful weekend whenthis project was 'finished', its price wentfrom $90 to $275 - which spurred the chal-lenge to home-brew a compatible and re-producible DSP board.• DSP-10, a 2m DSP transceiver project

Part 1 by Peter Rhodes, BSc, G3XJP PIC-A-STAR SOFTWARE TRANSMITTER AND RECEIVER

T

Early integration testing. Bottom left is my Third Method transceiver (borrowed front-end and PA), top right is Pic ′N Mix DDS still on its original breadboard (injection andcontrols) - and in the middle is the new DSP module under development. Note that Pic′N Mix provides all the transceiver controls, leading to a clean and compact front panel.


published by QST in Sept-Oct 1999 - andreviewed in RadCom, Feb 2000. Althoughfeatured for VHF/UHF applications, theDSP core is totally universal. This projectwas designed by Bob Larkin, W7PUA [2]and I am indebted to Bob not only for theinspiration for this project but for a signifi-cant amount of advice and help - includingsome code written specifically for PIC-A-STAR. Above all, Bob showed it can bedone and whenever I get into problems, hismaterial is the first place I look for under-standing.

INTEGRATING PIC-A-STARTHE DSP MODULE - designed to combinewith the Tx/Rx RF stages of your choice -operates at a final IF of 15kHz as shown inFig 1. This is a high enough frequency tomake it immune from image responses yetlow enough to be affordable. And it is not aDSP audio add-on, which coming after theproduct detector will always struggle.RF STAGESYour HF IF can be derived from any rea-sonable transceiver front-end. My ThirdMethod Transceiver [3] and G3TSO'sModular Transceiver [4] have both beentested as representative - and there are lotsof them out there. CDG2000 looks like apowerful approach and their front-end could

well be my next increment. The choice willsubstantially determine the overall receiverstrong-signal handling capability - but notthe basic effectiveness of downstream DSP.IF STAGESIn principle (and possibly in practice) youcould modify an existing IF board so that itsproduct detector produced a 15kHz outputinstead of straight audio but I don't recom-mend it in the long run. You would have tocompletely rebuild all the audio filtering topass 15kHz both on transmit and on receive.And unless the crystal filter were wider thanusual, you would be passing up the oppor-tunity to enjoy the pleasures of full-bandwidth SSB reception under good con-ditions. Some of the older IF amplifiers arenot noted for their noise figure, so, to cut along story short, this design includes an IFboard built for the job.

Details of this follow later, but it onlyneeds a modest roofing filter (at any HF IFof your choosing) since all the serious fil-tering is implemented in DSP.CONVERSION INJECTIONYou won't be surprised to see Pic ′N Mix[5] used as the injection source to mix fromRF to your chosen HF IF. This is not man-datory but my records show 281 of them outthere, so it is a non-trivial population.

COMMAND AND CONTROLYou need the ability to command the DSPfor all the functions normally associatedwith front-panel controls. You may besomewhat surprised to see Pic ′N Mix usedfor this purpose as well.

A small adapter board is used to fit amore versatile and powerful PIC - whichnot only controls all the original DDS capa-bility but extends the existing keypad, dis-play and tuning knob to control all thetransceiver features.

Although highly recommended, use ofPic ′N Mix is not mandatory. As an alterna-tive, you can use your PC to load and con-trol the transceiver - and a BASIC utility isprovided to achieve this.

PIC-A-STAR FEATURES IN BRIEF• SSB and CW detection and generation• a bank of high-performance Rx filters• impulse noise blanking• non-coherent noise reduction• auto-notch heterodyne(s) removal• variable AGC decay time• synthetic stereo reception• adjustable RF clipping on transmit• very fast VOX and QSK operation• the flexibility to change!

PIC

AD603

Right in Right out

15kHz IF

Existing (or typical) HF Tx/Rx RF stages

AFamp

PTT

PA

BPF x 9

LPF x 6

Xtal filter

eg 10.7MHz 2.8-15kHz wide

J310Mix

Left in Left out

Mic

SerialEEPROM

(DSP code)

DSP code at power-onthen User commands

Pic 'N' Mix DDS + small adapter board

Knob + shaftencoder

New DSP code releases(over the web) from yourPCvia RS232 link

Band select

AFamp

DSP Assembly = 15kHz IF board + DSP Processor board

LS

Simple20kHzLPF

AGC

Status LEDs

KeyPad (4x3)

RS232

Mix

Xtal osc10.7MHz+15kHz

Injection

Key

DSPProcessor

board(2-channel)

DAC

Bar-graph ormechanical

Timer board

T/R

Controls/times T/Rvoltage and change-

over switching

PIC Driver

Micamp

'S' meter

Stereoamp

6-digitdisplay

Postamp

AD9850DDS

Latches

Fig 1: A typical transceiver incorporating PIC-A-STAR at a final IF of 15kHz. See text for a discussion of the major hardware elements.


REFERENCES[1] www.analog.com[2] www.proaxis.com/~boblark/dsp10.htm[3] RadCom June-October, 1996[4] RadCom Oct-Nov, 1988[5] RadCom Jan-May, 1999


OFTWARE is this month's topic, namely anoutline of what it does, how it works, how itis packaged and how you obtain it.

The essentials of programming PICs hasalready been covered [6] so this project willconcentrate on the DSP dimension.

PIC-A-STAR DSPILLUSTRATED in Fig 2 is the functional-ity implemented in software. You will haveseen not dissimilar block diagrams imple-mented in analogue hardware - but not atthis price and not inside a 28mm2 chip!

Actually, there is an intrinsic overhead,namely that the analogue signals need con-verting to digital form before processing -and back to analogue after. This is the pur-pose of the CODEC (encoder/decoder)referenced on several inputs and outputs -and is implemented on a separate chip.

The greatest appeal of the software ap-proach, not least to the amateur, is the

flexibility to change the line-up at a touchof the keyboard, so to speak. This allowseasy experimentation (or overt tinkering, ifyou prefer) since at any time you can aban-don the change and go back to the previousversion. There are other subtleties.

For example, you will find five 15kHzoscillators scattered around the diagram. Infact their frequency changes depending onmode ie USB/LSB/CW. In DSP softwareterms the sinusoidal oscillator is simply asubroutine. To invoke it, all you need do istell it what frequency/phase you want onany given occasion - and it is done.

Another example is the 'delay' in the Rxfront-end image-cancelling I/Q mixer. Inone path there is a 90° phase shift, in theother a delay. The latter arises because ittakes real elapsed time to produce the phaseshift, so an equal amount of time has to be'wasted' in the other channel to maintainthat phase relationship.TIME IS OF THE ESSENCEThe basic understanding you need in orderto grasp how DSP works is to note that time

is the critical commodity. Every functionalbox in fig 2 takes time to execute. So doesevery individual instruction that goes tomake up that functionality.

This would be of little concern were itnot for our old friend Nyquist. He statedthat in order to faithfully process a signalyou must sample it at twice (at least) therate of the highest frequency present.

For example, the incoming Rx signal isaround 15kHz and so needs to be sampledat 30kHz or more. In fact, 48kHz is used toprovide a useful margin.

The consequence of this is that havinggrabbed one sample you have no more than20.83µs (by simple arithmetic) to do all theprocessing required before you have to getback to handle the next one. (Actually ifyou don't achieve it, the processor will in-terrupt whatever you are doing and dragyou back, so important is it.)

So just how much processing can beachieved in 20 millionths of a second? TheADSP-2181 processor in this design exe-cutes an instruction in 30 nanoseconds. Thesimplistic answer is therefore 666 in-


S

Mix

15kHzRx IF

LPF3kHz

15kHzosc (sin)

15kHzosc (cos)

Noisegate

Gaincontrol

Autonotch

Noisereducer

LPF3kHz

90°phaseshift

Delay

+/-

Impulsespike

detector

Synth-etic

stereo

AGCdetector

15kHz"stereo"

AF toCODEC

Mix

Filter bank:-'Wide'

'Medium''Narrow'(twice)

SSBSSBSSB

SSBSSBCW

Anti-VOX

Mic AFfrom

CODECVOX

T/Rcontrol

PTTKey

T/Rstatus

Commandinterface

To/fromCODEC

Key clickfilter/shaper

15kHzosc

Keyinggate

GaincontrolLPF

3kHz

Delay Gaincontrol

VOGADdetector

RFClip LPF

2.8kHz15kHzosc (cos)

15kHzosc (sin)

LPF3.7kHz

+/-

Delay

90°phaseshift

LPF18kHz

Gaincontrol

Mix

LPF3.7kHz

Mix

15kHzTx IF

SSB Power

USBLSB

Clip amount

Gaincontrol

CW Power

VOXpredelay

VOXgain

Gaincontrol

CWpitch

Filterselect

USBLSB

Pulsewidth

& level

Gaincontrol

AntiVOXgain

Beta Beta,Decay,Delay

AFgain

Stereoeffect

Micgain

Decay time

User commands fromPic 'N' Mix or PC

via RS232 link

Allcontrols

externalAGC to

DAC

KEY Control

Receive Transmit

User control

RF gain

Delay

QSKpredelay

Fig 2: Software block diagram of PIC-A-STAR DSP functionality. Not shown are simple on/off switches associated with VOX, noise blanker, auto-notch, noise reducer and the RF clipper. The filter bank also has an off (ie bypass) switch to give a net maximum bandwidth of some 3kHz.


structions-worth. But this is far from thewhole story. During one processor cycle itcan, for example, fetch two 16-bit numbers,multiply them to give a 32-bit product andadd the result to a 40-bit accumulator. ThisMAC (Multiply & Accumulate) instructionis the essence of filter implementation andis critical because you need to loop aroundit many times. Meanwhile, in the back-ground, the processor is also organisingdata samples in and out of the CODEC aswell as handling any serial communicationsport activity.

All these features (and more) characterisea processor capable of serious real-timeDSP.

Fig 3 shows a snatch of PIC-A-STARcode so you can visualise just how muchradio you get from each line of code.MULTI-RATE PROCESSINGThere is a more structural solution to theissue of buying some time - which equallyderives from Nyquist. Namely, once the Rxsignal has been mixed down to audio you

no longer need to process it at the 48kHzrate. Twice the audio frequency is fastenough.

PIC-A-STAR runs audio processing at8kHz - by grouping the audio functions into6 blocks and running one of them - but eachin turn - during 6 successive 20.83µs time-slots. At the end of each slot the data isagain processed at 48kHz because that isthe sample rate used by the CODEC foroutbound signals also.

This whole approach is known as "multi-rate processing"; getting the sample ratedown, "decimation"; and getting it back upagain, "interpolation". A similar process isused on the Tx side.

So you can see that all the way along theline Nyquist is satisfied - and so am I be-cause there is plenty of time for some exoticas well as the more mundane processing.

SOFTWARE PACKAGINGTHE DESIRE to provide choice and flexi-bility but above all upgradability (if that is

an English word) leads to some com-plication in describing the various modules.The context will become clearer once thehardware functionality has been covered.Suffice it to say at this stage that from anoperator's perspective the system is totallytransparent ie you just switch it on, waitabout 20 seconds (as if for the valves towarm up) and then use it. The softwarecomes in the following modules:-DSP BOOT UTILITYThis code resides in PROM on the DSPboard. At power-on time, besides runningsome basic hardware checks, it manages theon-board serial port to load the target DSPcode. This utility was written by BobLarkin, W7PUA for PIC-A-STAR based onthe original AD code.DSP TX/RX CODEThis runs on the DSP board and providesthe core functionality as in fig 2. It needs tobe loaded at power-on time, a processwhich takes some 20 seconds. Subsequentto loading it, you also need to be able to

command it.

DSP LOADERThis is a QBASICutility which runs onyour PC. It is writ-ten in very basicBASIC to enableyou to adapt it orport it if you wish.It has two distinctalternative func-tions:-• To load and

subsequently com-mand the DSP codedirectly to the DSPboard, via a COMport and a 9.6kBserial link.

• To load a new (or, of course, first)release of the DSP code to the PicAdapterboard (see next) in Pic ′N' Mix. Subse-quently, Pic ′N' Mix automatically loads thecode at power-on time - and provides thecommand user interface.

These two alternatives are not mutuallyexclusive. For early testing and use, theformer gets you going quickly. The latterfrees up your PC and in my view, gives amuch cleaner user interface - albeit with alittle practice. The choice is yours.

(There is a further option here. Youcould build a dedicated controller using anyprogrammable device with an RS232 capa-bility. The command syntax is simple andalso provided - and is in any event self-evident from the QBASIC code. With someloss of maintainability, you could also burnthe entire Tx/Rx code into the bootPROM/EPROM.)

PIC ′N' MIX PICADAPTERWritten in MicroChip Assembler, this code

runs on a 16F870 (which replaces the pres-ent 16x84) to provide all the original DDScontrol functionality of Pic ′N' Mix and inaddition, it now integrates the ability to:-• Download new release DSP code

from your PC (via the web)• Subsequently upload that same code

to the DSP board at power-on time• Command the DSP using the self-

same keypad, tuning knob and display asalready fitted to Pic ′N' Mix.TIMER BOARDAlso in MicroChip Assembler, this coderuns on a 16F627. It provides the sequenc-ing and timing of receive/transmit transi-tions - both ways - to make them as cleanand fast as possible. This board is designedto be general and will find uses on othertransceiver projects.BARGRAPH 'S' METERThis is both optional for PIC-A-STAR andequally of general application. Also run-ning on a 16F627 it controls a 12 LED bar-graph on the Status board. It was built at allbecause 10 LEDs as provided by most con-trol chips are not enough - and in any event,the PIC provides a lower cost solution.TO SUMMARISEThe programmed chips are a PROM, a16F870 and two 16F627s. These providethe base infrastructure. The target DSPcode - where most of the future enhance-ments will occur - is loaded from your PCusing the QBASIC utility. No further hard-ware (eg a programmer) is needed.

SOFTWARE DISTRIBUTIONALL THE SOFTWARE itemised abovewill be available for your personal use at nocharge. However, this does not mean itcomes entirely free. The 'price' is that youneed to send me an e-mail note requestingthe software - giving an explicit undertakingthat it is for your personal use and for thepurpose of self-education.

Not least, this allows me to maintain a listof 'customers' to advise when updates be-come available; as mentioned previously,this is, by intent, a project without end.

By software, I mean at this stage theloadable object code. Source code avail-ability mechanisms are still under consid-eration but at the least, it will not be formany months.

If however you want me to use my re-sources to program chips for you then I willsupply them ready-programmed at £8 perchip - plus return postage. It is worth point-ing out in this context that you could build aprogrammer yourself for about £10.

REFERENCES[6] "Pic-A-Switch" by Peter Rhodes,

G3XJP, RadCom, Sept - Dec 2001

Fetch Rx sample via CODEC and place in register mx0 …mx0=dm(Rx_in_buffer);

… and fetch current RF gain value and place in register my0. my0 = dm(RF_gain);

Multiply the two together to give a gain-controlled value …mr = mx0 * my0 (SU);

… and keep the gain-controlled signal in register my0. my0 = mr1;

Fetch the phase incremented value of LO and place in register ax0 …ax0 = dm(LO_ phase);

Pass the phase value to sin to get instantaneous sinusoid amplitude …call sin;

… and mix (ie multiply) it with the signal in register my0 mr=ar*my0(SS);

Fig 3: Some early lines of code for the receiver. Yes, the last line truly isa mixer (otherwise known as a product detector). The code continues byadding 90° to the phase value so that a call to sin returns the quad-rature LO - which is again mixed with the signal. After phase-shifting,the outputs of the two mixers are then literally added to give USB - orsubtracted for LSB.


ARDWARE infrastructure is thismonth's focus. Firstly, the processfor producing the PCBs. Specific

detail on mounting IC25 and IC27 and anyother board-specific notes follow later.Secondly, the project component list - soyou can contemplate costs and sources.Details of front-end components follow.

PCB MANUFACTUREIT HAS BECOME a tradition with myprojects that each has been produced by adifferent one-off PCB production technique- in an attempt to dispel any unwarrantedmystique and even some phobia. Andfrankly I want to advance my own craftskills with every project. On this occasion Iused iron-on laser film to speed up thedevelopment cycle - and as a matter ofnecessity for the DSP board - and found theresults to be excellent.

I am indebted to Ed, EI9GQand subsequently Harold,W4ZCB for sharing theirexperiences with thisapproach.PROCESS OVERVIEWIn outline, the process is tophotocopy the publishedartwork onto the film; thentransfer the toner from the filmto the board using heat andpressure from a clothes iron.This toner then acts as a superbresist while etching the copperin the normal way.

This is not entirely aprecision engineering processand there are some experientialskills. Cleanliness is every-thing! Although incrediblyfine lines and small spacingsreproduce well in myexperience, ironically there issometimes difficulty with largeblack areas. But equally, theseare the easiest to touch up withan indelible pen before etching - and attimes, you absolutely will need to.RESOURCESYou need access to a black-and-white laserphotocopier. Almost all modern machinesuse this technology. It may well be that thecopier in the corner shop will not be up tothe job and you should probably considerthe small cost of taking it to a professionalcopy shop as money well spent.

Some sheets of laser film (sold nowadaysby several suppliers for this purpose); ablock of flat scrap wood larger than thePCB; and a domestic clothes iron constitute

the tools.The latter should preferably not be a

steam-iron. If it is, ensure it is fully drainedsince water/steam and this process do notmix. Also the steam holes in the sole-plateare unhelpful. Avoid if possible the moremodern easy-iron technology which hasfine ridges on the sole-plate. Check that thesole-plate is flat. Some have a slightcurvature. If faced with these problems, itis best to use several (say, three)intervening layers of clean paper to providea more evenly distributed heat source.Experiment!

The PCB material (all double-sided) canstart out badly discoloured but must not bemechanically damaged ie no scratches. 1ozor more copper is better, but I used merely0.5oz on GRP for all my boards.THE PROCESS

(1) Firstly, test copy the artwork onto plainpaper in order to check for copier qualityand acceptable scaling error. Use themaximum contrast consistent with retaininga clean white background.(2) Copy the artwork onto the film. Whenviewed with the toner (matt) side down, youwant to end up looking at the tracking withthe correct orientation ie as if viewing thefinished board. For most published artworkthis requires the extra step of firstly copyingit to a transparency, flipping it over andthen copying that to the iron-on film. Inorder to avoid this extra step - with some

inevitable degradation - the PCB artwork inthis project will be printed pre-flipped so tospeak - and therefore should be copieddirectly to the film. The film itself is not'sided'.(3) Cut the PCB to size (or preferablysomewhat over-size for now). Remove allburrs and sharp edges. With cold water, weta soap-impregnated wire-wool pad and useit to polish the copper with increasinglylight strokes - until immaculate; do nottouch the surface thereafter. Polish bothsides and then wash off all traces of soapresidue with a clean paint-brush and coldwater - and dry with kitchen paper.(4) Place the PCB on the scrap wood andclean it with some kitchen paper(uncoloured) moistened in acetone,isopropyl alcohol or cellulose thinners.(5) Heat the iron to about 140°C (cottonsetting) and leave it for a few minutes to

attain an even temperature acrossthe sole-plate. At this temperatureit should just scorch plain 80gsmcopier paper.(6) Cut out the artwork to nolarger than the PCB and register ittoner side to the board.(7) With at least one sheet ofclean paper interposed, lower theiron vertically onto the middle ofthe board and let it rest there forsome 5 seconds. This willestablish the registration of theartwork to the copper.(8) If the board is bigger than theiron, continuously lift the iron offand relocate it every few seconds.Under no circumstances use an"ironing" motion. Simply raiseand lower it vertically - andfrequently until all the board hasseen the iron and some appliedpressure for about 20 seconds.For pressure, the weight of theiron plus about as much again isnear enough and is not critical.Too little pressure and the toner

will not transfer. Too much and the tonerwill migrate to widen the lines (ie smear)and reduce its depth. The former iscorrectable, the latter is absolutely not.(9) Inspect the result. You may see anyareas which have not transferred as stillretaining a somewhat glossy appearance.Repeat selectively as necessary. Payparticular attention to the edges.(10) Allow the board to cool naturally backto room temperature.(11) Carefully peel back the film from eachcorner and note that the toner hastransferred. If any critical areas have not


H

The completed CODEC board, illustrating the quality of PCBproduction available (0.2mm wide tracks at 0.5mm spacing) - usingdomestic kitchen resources. The CODEC chip is 7mm square.


taken, the artwork will still be registeredand you can selectively repeat.(12) Touch up any blemishes or areas ofvisibly thin toner with an indelible pen.(13) Now spray-mask the opposite side ofthe board and etch as normal.(14) Before removing the etch resist, centre-pop and/or drill the holes. They are easierto see at this stage. Clean off the etch resistwith cellulose thinners and gently re-polishthe board.(15) At this stage I lightly spray the boardwith SK10 which is both a protectivelacquer and a flux. Available from RapidElectronics, this makes for clean solderingand prevents contamination of the copper.

COMPONENT LISTRESISTORS, 1/8-1/4W, 5-10%R1 ................................................................. 47RR2, R48, R49, R76, R77 ............................ 100RR3 ............................................................... 220RR4 ............................................................... 470RR5, R6 ......................................................... 560RR7, R50, R64-R71 .......................................... 1kR10, R43, R52, R72 ..................................... 2k2R11 ............................................................... 2k7R12, R34, R37 .............................................. 22kR13-R15 ....................................................... 3k3R16-R19 ....................................................... 4k7R20 ............................................................... 5k6R21-R32, R45-R47, R53-R55,R63, R73, R78 .............................................. 10kR33, R51, R59-R61 ...................................... 47kR35, R36 ..................................................... 100kR38 ............................................................. 120kR56-R58 ....................................................... 10RR62 ............................................................. 100kR74 .......................... 1k x 4 SIL network (5 pin)R75 .......................... 1k x 8 SIL network (9 pin)RV1, RV2, RV6 ............. 100k horizontal presetRV3, RV4 ......................... 10k horizontal presetRV5 .................................. 47k horizontal preset

RESISTORS SMD 1206 SIZE, 5%R8, R9, R93 .................................................... 1kR39, R40 ....................................................... 22kR41, R42, R97 ............................................ 220kR44 ............................................................... 3k3R79-R89 ....................................................... 10kR90 ............................................................... 330R91, R92 ....................................................... 2k2R94-R96, R100-R103 .................................. 4k7R98, R99 ..................................................... 100kR104,R105 ................................................... 47k

CAPACITORS, 16V RATINGC1-C19 ........................ 1n solder-in feedthrough

(C11 and C12 may be replaced witha stereo jack socket)

C21 ......................................... 10p ceramic plateC20, C22, C23, C26, C69 ................. 220µ axialC24 ..................................................... 10µ radialC25, C75, C76, C84 ........................... 33µ radialC27 ...................................................... 47µ axialC28 ..................................................... 2µ2 radialC29 ..................................................... 4µ7 radialC30 ..................................................... 33µ radialC31-C34 ........................................ 220n ceramicC35, C36 ............................................ 1n ceramicC37-C44, C59, C70, C71 ................ 10n ceramicC45 .................................................. 22n ceramicC46-C58, C60-C65, C67, C68C74, C77-C83, C85-C89 ............... 100n ceramic

C72, C73 ................................ 15p ceramic plateC90, C91 ................................ 22p ceramic plateC92, C93 .............................. 270p NPO ceramicC94 .................................................. 47n ceramicC95, C96 ....................................... 220n ceramicC97 ....................................... 220p ceramic plateC120, C121 ............................ 18p ceramic plateC122, C131 ........................................ 4µ7 radialC123-C125, C128, C132 ................... 10µ radialC126, C127, C129, C130, C133 .......... 1µ radialC66, C98-C119 ................. 100n SMD 1206 sizeVC1-3 ................ 100p polyethylene trimmer (1)

INDUCTORSRFC1-RFC3,RFC6, RFC7 ........................... 1mH axial chokeRFC4, 5,RFC8-RFC10 ....................... 100µH axial chokeRFC12 .................................. 330µH axial chokeL1 ............. 4.85µH 38t 30swg on T50-2 (2), (3)T1 ........................... 6t:3t 30swg on FT37-43 (2)T2 ............................... 10t:3t 30swg on FT37-43T3 ............................. 20t:4t 24swg on T50-2 (3)FB .................... 1 turn through small ferrite bead

SEMICONDUCTORSIC1, IC13, IC15,IC17-IC19 .......................... 78L05 regulator, 5VIC2 ........................................... AD603AQ (DIL)IC3, IC4 ................................ FST3125M (SMD)IC5 ...................................................... SBL-1 (7)IC6, IC7, IC22 ............................... TL072 (DIL)IC8 ..................................... TLC7524CD (SMD)IC9 .................................. PIC 16F870-ISP (DIL)IC10 ...................................... 24LC256-IP (DIL)IC11, IC28 ...................... PIC 16F627-04P(DIL)IC12 ....................................... ULN2803A (DIL)IC14 .................................... 7810 regulator, 10VIC16 .................................................. 4094 (DIL)IC20 ....................................... 74LVX125 (SMDIC21 ............................................. ST232N (DIL)IC23 ........................................... 74HC14 (SMD)IC24 .............................. LE33CZ regulator, 3V3IC25 ................................................. AD1885JSTIC26 ........ 27C512 512KB PROM, programmedIC27 .................... ADSP-2181 KS130 or KS160TR1, TR2 .................................................... J310TR3-TR5, TR12 ..................................... 2N3904TR6 ......................................................... 2N3906TR7, TR8 ................................................ BC517TR9-TR11 ................ VP0300LS P-ch MOSFETD1-D6, D11, D15 ................................... 1N4148D7-D10 ............................................. BA244 (8)D12-D14, D34 ...................... 1N4007 or similarD16-D20, D35 ............................ 3mm LED, redD21 ..................... 3mm LED, red/green tricolourD22-D33 ....... 1.8mm LED, colours for S-meterZD1 ............................ 4V7 200mW Zener diode

MISCELLANEOUSFL1 ............................... Crystal roofing filter (4)X1 ........................... wire-ended BFO crystal (5)X2 ............................... 4MHz wire-ended crystalX3 .......................... 16.67MHz, probably customX4 ... 24.576MHz low profile, Rapid ElectronicsS-meter ......................................... 1mA, optionalS1 ....................... 8-pole on/off PCB DIL switchS2, S3 .................. skeleton push-to-make switchAll screened RF leads ............................. RG174Other screened leads ......... thin microphone leadSpacers and 3mm nuts/bolts ................ 4 sets (6)PL1 .. 2 strips of 9-way SIL header ie male/maleSK2-SK8 ................................... SIL socket stripPL2-PL8 ........................... mating SIL plug stripTurned pin socket IC6, 7, 10, 22 ................ 8 pinTurned pin socket IC16, 21 ...................... 16 pinTurned pin socket IC11, 12, 28 ................ 18 pinTurned pin socket used as spacer ............. 18 pin

Turned pin socket IC9 ................ 28 pin by 0.3inTurned pin socket IC26 .............. 28 pin by 0.6in

NOTES:-(1) These three trimmers may all bereplaced with fixed capacitors afteradjustment.(2) Assumes a 2k2Ω filter impedance.(3) Assumes X1 around 10.7MHz.(4) Filter width between 2.7kHz and15kHz. Around 3-6kHz is ideal.(5) Centre frequency of FL1 plus 15kHz.(6) Spacing specified to give clearancebetween FL1 and the DSP board.(7) The SBL-1 mixer may be replaced by astronger mixer and / or one using discretecomponents at your discretion. It isspecified here as a well established datumand works well.(8) These diodes can be replaced with1N4148 if you can't obtain RF switchingdiodes.SUPPLIERSMost of the components were procuredfrom Farnell or Rapid Electronics. Theexceptions are the toroids which I boughtfrom Mainline - and the crystal filter whosespecification is loose enough that it shouldbe possible to find a surplus one.

The two critical chips, namely IC25 andIC27 are manufactured by Analog Devices.They have an enlightened marketing policyfor promoting interest in their products suchthat for many of their devices, individualscan order up to two as free samples. Youdo this via their web-site - and the first timeround you have to register and answer a fewquestions. You do not have to be acommercial organisation; the occupation of"Amateur Radio" at your private addressand for the purpose of "training and self-education" are entirely acceptable.

Provided, of course, you do not attemptto abuse their generosity - which wouldreflect on the amateur community as awhole. Personally, I consider orderingchips whose MTBF exceeds your life-spanas spares; ordering just in case one day youmight find a use; or the simple inability toturn down a free offer - would all constituteabuse. My ethical test is that I never order afree sample if I can purchase smallquantities.

THE FINAL IRONYDON'T FORGET to reset the clothes iron toa more modest temperature before putting itaway. 140°C will melt many syntheticmaterials and add alarmingly to the projectcosts. Even this has been Beta tested!


HIS MONTH covers the concept ofSuper VOX. This apparentlyinnocent topic was picked as

exemplifying how a large number of littletweaks can be made in DSP at noincremental cost to significantly improveoperating pleasure and practice. And it istime to practice your ironing skills on aneasy one. And why Super? Well, read on.

I suppose that if there is one thing Idislike more than long monologue 'overs', itis long ‘doubles’. I have personally alwaysoperated VOX and indeed for many yearsdid not bother to fit any PTT capability atall. Perhaps it would help the psychology ifthat switch on the microphone were knownas ‘Lift To Listen’.

The system has two elements, namely atimed solid-state switch for the various DCT/R lines and any relays; and an intelligentVOX system, implemented within DSP.But note that the Timer board has beendesigned as a flexible and stand-alonegeneral solution to manage the T/Rswitching in any transceiver.

VOX PREREQUISITESTO BE EFFECTIVE, any VOX systemneeds both T/R transitions to be free fromclicks and thumps - both electrical andmechanical. The first step to achieving thisis to leave the maximum amount ofcircuitry powered up on both transmit andreceive. Certainly all DC switching shouldbe solid-state (hence the Timer Board) butthe RF changeover can be more of an issue,especially at higher power levels. I use thecircuit published in the 1988 ARRLHandbook, but hope to do better before theend of this series. Even if you use relaysthere are still significant benefits, thoughpersonally I just hate those acoustic rattlingnoises.

TIMER BOARDTHE T/R Timer board circuit is shown inFig 4 and manages both the R to T and theT to R transitions.

The benefit of this approach is that thetwo transition sequences and timing can be- as indeed they should be - different. Thiscannot be achieved with the typical windowcomparator approach.

This board has one significant input,namely T/R Status from the DSP Assembly.This is a +5v logic signal (or floating) whenon receive - and grounded to 0v to switch totransmit.

The PIC is a 16F627 which has thebenefit of not needing an external crystal iftiming accuracy requirements are modest.

And as a result the two crystal pins can beused for digital I/O purposes.

There are five timed outputs - which havebeen arbitrarily named for their mostobvious general use. The 'External Linear'and 'Local PA c/o' lines are grounded ontransmit, can sink 500mA - and are thussuitable for relay or solid-state switchcontrol. The other three lines are at +12vwhen active and are explicitly groundedotherwise. They can each source/sink500mA. They behave as follows:-RECEIVE TO TRANSMIT (R/T)The sequence for this transition follows,each step being followed by a timed delay:-

External linear to Tx ...........................T1Local PA c/o to Tx ..............................T212v Rx off ...........................................T312v Tx on ............................................T4Tx PA bias on

There then follows a re-triggerable hangtime, T5. This whole transition is notinterruptible, see below.

TRANSMIT TO RECEIVE (T/R)For this transition the sequence is:-

Tx PA bias offExternal linear to RxLocal PA c/o to Rx ..............................T612v Tx off............................................T712 Rx on

This transition is interruptible after step 1.That is, if you are part way through

dropping back to receive when a transmitdemand occurs, the T/R sequence will beaborted and the R/T sequence executedimmediately. This interrupt logic is basedon the view that it is always better to risklosing a moment of reception than to risk'hot' switching.

Note that the 12v Tx and 12v Rx linescan never be energised at the same time.

Following a T/R transition, the PIC goesto SLEEP; that is all dynamic activityceases including its internal clock. Thus itcan never act as a noise source to yourreceiver.

The process for adjusting the times for


T

R5147k

IC1378L05

10VDD

VSS

R522k2

16RA7

1

C20220

µ

+12V

3

IC1116F627

PIC

13

6

RA2

2RA3

14

5

IC12ULN2803ADarlingtondriver x 8

MCLR4

D111N4148

ProgrammerInterface

RB6

MCLR

RB7 RB713

RB612

R48

100R R49

100R

R501k

RB06T/R Status

RB17

RB28

RB39

RB410

RB511

R5310k

s d

g

R5610R

+12VRx

1

TR11VP0300LS

16

14

5

17

15

4

18

2

VSS

9

+5VIC147810

+10V

VDD

17RA0

8 Ext linear11

Local PA c/o127

S1

R5410k

s d

g

R5710R

+12VTx

TR10VP0300LS

R5510k

s d

g

R5810R

Tx PAbias

TR9VP0300LS

C7633µ

C77100n

C78100n

C79100n

C7533µ

C80100n

R59

47k

R60

47k

R61

47k

Ground

RA1 18

nc

RA615

T7

T6

T5

T4

T3

T2

T1

MOREor LESS

RA43

from DSP Assy

Fig 4: Timer board circuit diagram. This provides timed transitions between transmit and receive- in both directions. Sw1 allows you to set up the timing for your installation - and covers therange from slow relay-based RF T/R switching through to solid-state QSK.


your installation will be covered later.CONSTRUCTION NOTESFig 5 shows the PCB artwork ready forthe iron-on process described lastmonth. The 10v regulator chip IC14provides power to the STAR DSP boardand may be omitted (as in thephotograph) if you don't require thisunswitched rail. Mounting holes forIC14 and the board (optional) have notbeen specified.

Start by fitting IC13, C76, C78 andC80, soldering one lead to the topgroundplane. Then fit the socket forIC11 and solder pin 5 to thegroundplane. Then the socket for IC12with pin 9 grounded. The remainingconstruction sequence is not critical.Mount the otherwise symmetrical switchso that with the switches set away from thePIC, they are open circuit.

SUPER VOX IN DSPTHE CONCEPT USED here was inspiredby a conversation with Bill, W7AAZ.

Sadistically, I did not enable the PTT lineon the STAR Beta build for some sixmonths, so the whole VOX system - theonly way to get to transmit - has had a goodthrashing.CONVENTIONAL VOXWhen VOX detects the beginning of yourspeech it initiates the R to T transition -which is going to take at least 3ms tocomplete. Further, if you have a T/R relaythen the design must ensure the relay hassettled in the transmit position before lettingthe RF through, typically adding a further20-30ms. The result is that the leading edgeof your speech is clipped off. Not by muchin a good design, but often noticeable.

To disguise this effect, a VOX hang timeis incorporated which is set to drop back toreceive if you pause for breath, which at

least minimises the number of truncatedwords. The other workaround you oftenhear from VOX operators is that they do notanswer a direct question with "Yes". Theytend to say "um, yes", probablysubconsciously - in order to avoid it comingover as merely "esss".

How much better it would be if yourtransceiver started the R to T transition inanticipation, ie just before you started tospeak! Sounds fanciful? In effect, this iswhat Super VOX does. And by the way, itapplies equally to QSK CW operation.SUPER VOXThe idea is to trigger the R to T transitionimmediately on detection of your voice, butthen delay the ‘voice’ in DSP for the time ittakes for the transition to complete. Thusthe leading edge can never be clipped off.

Critically, this means in turn that youneed no hang-time, since there is now nodesire to minimise the number oftransitions. Of course your delayed voice isstill coming 'out of the antenna' for a fewmilliseconds after you stopped talking soyou need to stay on transmit for that time -

but absolutely no longer.The net effect is that at a normal

conversational speaking speed, you dropback onto receive not between breathsand sentences, but between every word -and often enough, between syllables -and if your T/R transitions are fastenough, you can listen through. Equallysomeone listening to your transmissionwould be totally unaware that you werespending a significant percentage ofyour over on receive - in short but veryfrequent bursts.

The overall effect is very close insensation to full duplex as in a normal(and therefore interruptible)conversation and if widely practicedwould do much to turn many a QSO intoa conversation rather than a series of

speeches.ANTI-VOXThis normally works by comparing themicrophone input with the speaker output -and if the same, concludes that it not youspeaking. STAR incorporates a furtherrefinement in that the microphone input iscompared to the output that did come fromthe speaker 4ms earlier. Why 4ms?Because this is the time it takes sound totravel 4ft in air, an assumed reasonabledistance between the speaker andmicrophone. The improvement isnoticeable and is worth having because thefew extra lines of code don't cost anything.AGC IMPLICATIONSNormally, the AGC voltage decays tonothing shortly after you go to transmit.The result is that the receiver comes backon full gain in VOX gaps - which is notvery comfortable in an 'S9' QSO.

The approach adopted by STAR is toretain the AGC level established by the last2 second period of continuous receive - andapply that level during the gaps. It isimportant to ignore AGC levels establishedduring the gaps for this purpose, so 2seconds was chosen as an arbitrary intervalwhich is clearly longer than a casual pause.And if at any time somebody other than youstarts speaking, the normal AGC attacktakes care of any adjustment in a fewmilliseconds.

So this is, if you like, extended hangAGC - where the 'hang' is extended overperiods of transmission.THE SOUND OF SUPER VOXOn the Members' Web site, you will find anumber of .WAV files so you can hear whata STAR sounds like in action. One of thesegives a good feel for the VOX operation.

PCB track and drilling templateNB:- This image is mirrored

PCB dims:- 2.25" x 2"(58 x 51mm)

Component location(groundplane side)

IC11 IC12

S1

C20

R48

IC14

R52

R50

D11

R49

R51

C76

R56R

60

R59

R54

C75

R57

R53

R55

R58

R61TR9

TR10 TR11

C80

IC13

C78

C79 C77

T/R

sta

tus

+12V

in

+12V

Tx+1

2VR

x

Ext

line

arLo

cal P

A c

/o

Tx P

A b

ias

+10V

out

Gro

und

GroundRB6RB7

MCLR

Optionalprogrammer

interface

Mor

e/Le

ssT7 T6 T5 T4 T3 T2 T1

groundboth sides

Fig 5: Timer board PCB - with the component side unetched. Countersink the ungrounded holeson the component side. The input/output connector (if any) is not specified, but is 0.1" pitch. Iused SIL plug/socket strip for this - and all other arbitrary connectors.

The Timer board, which will fit in a small corner in mosttransceivers and get rid of those DC switching relays.


HIS MONTH covers the DSP board circuitdiagram. The board is based on (and is notincompatible with) the Analog Devices2181 EZLITE board. That is, aspects ofthat board which are not used either bySTAR or by W7PUA's DSP-10 have beenomitted; the physical construction iscompletely different; and a currentproduction and superior CODEC chip hasbeen used. But conversely, the EZLITE

board can be (and has been) used in thisapplication and if you already have yourhands on one, e-mail me for further detail.Signal names as defined by Analog Devicesare used throughout.

MOTHER BOARDTHE MOTHER board with her twodaughters is shown in Fig 6. This form ofconstruction was adopted to spread the riskduring board manufacture and to allowupgrade of either the CODEC or Processorchips later.

Each board has its own regulator chips tospread the heat dissipation and to maintainmodularity.

The CODEC daughter converts analoguesignals to/from digital/analogue form forthe benefit of the Processor. The digitalsignals are passed back and forth using a12.288MHz industry standard AC '97 serialbus - which multiplexes data in, data outand commands. The Processor daughterdoes the DSP processing (no surprisesthere) - but has other control inputs/outputsas well.


T

6 IC23e74HC14

50

9

14

4

IC1978L05

C116100n

To IF board

CODECDAUGHTER

BOARD

SYNC

SDATA_OUT

BIT_CLK

SDATA_IN

+8V approx

PROCESSORDAUGHTER

BOARD

RFS0

DT0

SCLK0

DR0From IFboard

Rx 15kHz

Mic AF Left

Right

ToIFboard

Rx left AF

Rx right AFor Tx 15kHz

Left

Right

R91k

C3033µ

1

C66100n

IC8TLC7524CD

8-bit SMD DAC(6 used)

OUT 1 RFB

2 15OUT 2 REF3 14GND VDD

4 13DB75 12DB66 11DB5 DB07 10DB4 DB18 9DB3 DB2

CS

WRPF7

PF6

PF5

PF4

PF3

PF2

32

31

30

29

28

27RESETRESET

PF1 26

PF0 25

R4022k

R3922k

KEY

PTT

D2

1N4148

D1

1N4148

'Narrow' AGC

FL0 5

R44

3k3

T/R StatusTo Timer

board and Pic'N' Mix

1

IC21ST232N DIL

RS232 driver/receiver

C1+2 15V+ Gnd3 14C1-

VCC

4 13C2+5 12C2-6 11V- T1in7 10T2out T2in8 9R2in R2out

R1out

R1in

T1out

C119100n

C117100n C115

100n1

2

3

4

5

6

7

8

9DFR1/F1 & TFS1/IRQ1

15

5V

DT1/F016

D34

1N4007 C108100n

C69220µ

C109100n

C1314µ7

5V

7IC22aTL0725

6R96

4k7

R97

220k

C97

220p

R931k

R944k7

R954k7

C13210µ

5V

1IC22bTL0722

3 8

4

5V

C111

100n

To IF board'Wide' AGC

D4

1N4148 R8

1k

AGCSET

R41220k

C294µ7

D31N4148

TR52N3904

RV6100k

R42

220k

IC23b74HC14

31

C1331µ

R912k2

RESET SW2

R98100k

5V

911

C113100n

R922k2

INTERRUPT SW3

R99100k

5V

5D35

LED

R90

330

5V

13

R8910k C114

100n

C110

100n

32

31

30

29

28

27

26

25

33

PL/S

K5

4

5

6

1

2

3

PL/S

K6

4

2 13

11

9

12

10

5

3

6

PL/S

K7

8

PL8

7

1

14

IRQE

FL1

1

6

PL/S

K7

GND 1 1

GND 24 24

GND 50

SK8

IRQ2 2 40

IRQL0 3 38

IRQL1 4 39

FL2 8 35

34

GND 7

DSPcode

& usercommands

FromPic 'N' Mix

or PC

C112100n

notused

notused

+7V

2IC23a74HC14

8IC23d74HC14

10IC23c74HC14

12IC23f74HC14

7

DB9F socketnot fitted onPIC-A-STAR

C118100n

16 nc

16

Not

use

d on

STA

R

Fig 6: DSP mother / daughters relationship - and mother board circuit diagram.


Unlike the EZLITE board, the DSPBoard carries IC8 and IC22/Tr5 forgenerating AGC voltages for use on the IFBoard - accepts inputs from KEY and PTTlines - and generates the controlling systemT/R line as a function of mode and controlparameters eg VOX/QSK operation - thuscustomising it from the general to thisparticular transceiver application.

IC21 control RS232 communicationsfrom a host - either your PC or the PIC inPic 'N' Mix - and is used to upload theoperational DSP code. It also accepts usercommands to control the entire transceiver.IC23 buffers manual resets and interrupts -and drives an LED to show status.

PROCESSOR BOARDTHIS COMPRISES the processor chip -and some memory used only at power-on(or Reset) time to boot load the realoperational code. See Fig 7. For furtherdetail, see the ADSP-2181 data sheet.

Being mostly track, the board is veryquick and easy to build.

CODEC BOARDTHIS IS a standard (albeit minimal)implementation of the AD1885JST CODECchip. See Fig 8. For further detail, consultthe data sheet.

The CODEC uses a 3v3 digital rail, but5v on the analogue side; IC20 translates the5v logic signals from the Processor to this3v3 level. Outbound 3v3 lines to theProcessor are already within its logic 1/0definition range.

NEXT/SUBSEQUENT MONTHSCOVER THE component layout, PCBs and

construction detail of these boards.

WWWFor detail and applications of the ADSP-

2181 and AD1885, visit www.analog.com

IC27 ADSP-2181 DSP PROCESSOR

+7V

IC1778L05

RFS0

DT0

SCLK0

DR0

PF7

PF6

PF5

PF4

PF3

PF2

RESET

PF1

PF0

FL0

DR1/F1

DT1/F0

C98100n

C1224µ7

13

11

9

12

10

PL/SK7

PL/SK8

14

IRQE

FL1

1

6

nc

+5V

IRQ2

IRQL0

IRQL1

FL2

31

5

2

3

4

8

7

15

16

TFS1/IRQ1

32

31

30

29

28

27

26

25

2-23

24

49

50

1

FL0 33

FL1 34

FL2 35

nc 36-37

IRQL0 38

IRQL1 39

IRQ2 40

nc 41-48

IC2627C512512kbPROM

11 10A0 A012 9A1 A113 8A2 A214 7A3 A315 6A4 A416 5A5 A517 4A6 A618 3A7 A725 25A8 A826 24A9 A927 21A10 A1028 23A11 A1129 2A12 A1230 26A13 A13

28Vcc

20CE22OE

14XTAL

C12018p

34

41

42

43

44

45

46

48

49

50

58

51

55

52

8, 21, 23, 40, 54, 75,86, 88, 98, 112, 123

GND

7911 D8D08012 D9D18113 D10D28215 D11D38316 D12D48417 D13D58518 D14D68919 D15D79027 D16A14911 D17A15

CLKIN

C12118p

20

19

+5V

C103100n

Not

use

d on

PIC

-A-S

TAR

GND

9 24 39 69 87111

BMODE 35

MMAP 32

PWD33

BR66

IS124

C102100n

5 BMS3 RD

R7910k

+5V

+7V

R80

10k

1

128

127

126

122

121

120

119

+5V

R84

10k R78

10kR81

10k R82

10kR83

10k

C101100n

C100100n

C99100n

C107100n

+5V

16.67MHzX3

Fig 7: Processor daughter board circuit diagram. IC27 may be a KS-130 or KS-160 processor butin any event, the slower KS-130 device is assumed. To retain compatibility with EZLITE, theunused or unconnected lines on PL/SK8 are available on the mother board for non-STARapplications.

R88

10k

IC2074LVX125

Leveltranslator

IC25 AD1885JST CODEC

C89100n

11

C9022p

C9122p

X4 24.576MHz

SDATA_IN

BIT_CLK

SDATA_OUT

SYNC

RESET 2

4

6

3

5

8

7

9

1

SCLK0

DRO

RFS0

DT0

RESET

+7V

5

10

8

6

5

12

9

6

11

8

14 1, 2, 4,7, 10, 13

R87

10k

9C85100n

7DVSS2

DVDD2

DVSS1 4

XTL_OUT 3

XTL_IN 2

DVDD1 1

C86100n

C12310µ

LIN

E_IN

_R

LIN

E_IN

_L2324

JS0

JS1

R8510k

R8610k

4847464443424140

ncncAV

SS

3

AV

DD

3

ncHP_

OU

T_R

AV

SS

2

HP_

OU

T_L

AV

DD

2

nc

393837

C87100n

12nc

MIC

222

MIC

121

nc20

nc19

nc18

nc17

nc16

AU

X_R

15A

UX

_L14

nc13

AVDD125

26

27

28

29

30

31

32

33

34

35

AVSS1

VREF

nc

AFILT1

AFILT2

FILT_R

FILT_L

RX3D

CX3D

LINE_OUT_L36 LINE_OUT_R

C88100n

C106100n

C12810µ

C92270p C93

270p

C1291µ C130

1µ

C9447n

C105100n

C126

1µ C127

1µ

R10547k

R10447k

C96 220n

C95 220n

R103

4k7

R100

4k7

IC1878L05

IC24LE33CZ

C12410µ

C104100n

C12510µ

+7V +5V

+3V3

+3V3

+5V

R101

4k7

R102

4k7

1

2

3

Left in

Right in

Right out

Left out4

5

6

ANALOGUE GROUND PLANE DIGITAL GROUND PLANEFB

PL/S

K6

PL/S

K5

45

Pins brought out onto pads,but not used by PIC-A-STAR

Fig 8: CODEC daughter board circuit diagram. Note that the ground plane is split between analogue and digital to minimise noise.


HIS MONTH covers the componentlocation and external connectionsfor the DSP mother board - and its

two daughters. These are shown in Fig 9.Also, the procedure for a stand-alone test ofthe completed board is provided. Nextmonth concludes the 'DSP' phase of theconstruction with PCB artwork andconstruction notes.

COMPONENT SPECIFICATIONSMD COMPONENTS have been specifiedhere where space, cost, or performanceconsiderations require them - but nototherwise. 1206-size devices are used andthese are no more difficult to handle thanconventional leaded components.

Specifically, SMD electrolytic capacitorsare not used since these are expensive - andthe small space savings which areachievable are not needed.

All the small coupling and decouplingcapacitors are 100n and in general, they areSMD. However, on the CODEC board,C85-C89 are specified as wire-ended discceramic units because their leads are used tocouple power and ground between the twosides of the board.

SMD resistors are used throughout sincethese save a great deal of space. The singleexception is R78 - where I positivelyneeded a larger component to span anotherwise unbridgeable gap.

In any event, all components are mountedon the track surface, but in some cases,leads are also soldered underneath. Youneed not take this to extremes, but everyreasonable opportunity should be taken tointerconnect the top and bottom grounds.

EZLITE COMPATIBILITYIT IS ANTICIPATED that this DSP boardwill find application on other DSP projects.If you are contemplating this, contact theauthor of that project in the first place forcurrent status. This hardware is afunctionally compatible subset. It has thesame overall dimensions albeit withdifferent connector locations. The address,data and emulation expansion sockets havenot been provided on this board - and nor,realistically could they be. PIC-A-STARuses a different CODEC chip whichrequires a different DSP code module tohandle it. A source code shell for this isavailable on request.HARDWARE TESTAS THE DSP board is progressivelycompleted, it is highly desirable to test it instand-alone mode before moving on. Thisprocess also proves the interface to your PC

- which will be needed operationally later.PREREQUISITESThe first requirement is that you are runningQBASIC under Windows on your PC. Onolder machines it is a standard application;later it was provided on the archive disc andon Windows ME it is not provided at all -but does run. In any event, it is an absoluteprerequisite. The PC itself is totallyuncritical.

You need to make up a lead from yourPC serial port - but only two of the lines areused. These are pin 3, the signal - and pin5, the ground. These connect (temporarily)to the mother board at "DSP code and usercommands" as per fig 9. Ensure the groundlead is indeed grounded.SET UPOn your PC, establish a new directory. Thesoftware assumes C:\STAR but you can editthe software for any other location.

In that directory, place the file testxx.xjpwhere xx is the current version number ofthe test program - and XJPload.bas which isthe utility used to load all STAR DSPsoftware, not least this test program.

Open QBASIC and from there, openXJPload.bas. To run the test program, justfollow the on-screen instructions!PROCESSOR TESTThis requires the Mother board withProcessor daughter - but not necessarily theCODEC. The test process starts with D35flashing. Once you start to load the testprogram, the LED will be permanently lit.Once the program has loaded, the LED willbe off if the CODEC was successfullyinitialised, on if it was not (particularly if it

is not yet even fitted). In either event, ifyou press the Interrupt switch, Sw3, theLED will toggle on and off. This verifiesthat the code has loaded and that theprocessor is running and is in (or indeed,under) control. This also establishes yourcapability of loading any code over theserial link and unless and until you canachieve this, no further progress can bemade.CODEC TESTOnce the CODEC daughter has been fittedand the previous test successfully repeated,power down and connect a patch lead fromthe CODEC left and right outputs to astereo amplifier.

Power on again and re-load the testprogram. A damp finger placed on theCODEC left or right inputs should nowproduce a corresponding hum on therespective output. Should you prefersomething more exciting, you could connectup a microphone or any standard line-levelstereo input.

This is a test of a full loop-back on bothchannels. That is, the input is beingdigitised, sent to the processor where aminimal operation occurs in the digitaldomain before it comes back to the CODEC- where it is converted back to analogueform and thence to your ears.

Thus when this test works, you havecompletely proved the CODEC and the vastmajority of the processor functionality - andthe interface between them.

If, however, it should fail yet theprocessor successfully loaded the testprogram in the first place, the problemalmost certainly lies on the CODEC board


T

The DSP assembly. That is, the DSP mother board with CODEC and Processor daughters boards- mounted back-to-back with the IF board in its enclosure. The top, bottom and side screeningpanels are not fitted until after final test.


itself - or the link between it and theprocessor.

A 12.288MHz clock train is generated bythe CODEC on the Bit Clock line. Inresponse, the processor provides a 48kHzclock on the Sync line. If these are both

present and you can see data pulses on theData in/out lines, then the problem isprobably on the analogue side of theCODEC. But if you rigorously checked theboard in the first place, then there can't be aproblem, can there?

The DSP mother board, ready for daughter board fitting and test. NoteIC8 and its associated components are located under the Processordaughter board. In fact, neither they nor IC22 need be fitted for stand-alone testing.

The finished Processor daughter board. Note that the crystal X3 is fittedafter bending its leads - to reduce height. C120 and C121 are fittedunder the board. IC26, when fitted in its socket defines the overallheight of the complete DSP board.

IC17

R78

R79

C102 R81

R80

C10

0

R83R82

C10

1

C103

C98C99

R84

C122

C10

7

IC27

IC26

X3

IC19

C10

9C

108

D34IC22

C97

C132

R97

R93

R95

C11

2

C111D4

C29TR5 R42

R41

RV6 R8

S2

S3

C133

C113

R99

C114R90R98

R89

R92

IC23

D35 R

91 IC21

C11

6C

119

C117

C11

5

IC8 R9

C66

C30

R40R

39

D2D1

C104

IC24R88

R87 R85

R86

C85

C95C96

R10

1R

100

R10

2R

103

C88C87

IC18

C91 C90X4

C89

C105

C10

6

IC25

IC20

C124

C125

C127

C126

C123

C130

C129

C128

C92

C93

R104 R105C94

R44

C131

C69

C86

C120, C121mounted onX3 pads underboard

PL/SK5

PL/SK6

PL/SK7

PL8 on daughterSK8 below on mother

DSP code & usercommands fromPic 'N' Mix or PC

+8V approx in'Wide' AGC

'Narrow'AGC

T/Rstatus

KEY

PTT

Left in Right in Right outLeft out

FB

(a) Mother board (c) CODEC daughter board

(b) Processor daughter board

D3

Wire link,not fitted onPIC-A-STAR

Pin 2Pin 1

For componentspecification, see

RadComOctober, 2002

CODEC Daughter Board

ProcessorDaughter

Board

C110 R96 R94

C118

Fig 9: DSP boards component location diagram. SMD capacitors are shown as rectangles, disc ceramics with 'rounded' ends. The Processordaughter should be rotated clockwise through a right-angle to visualise the fit on the mother board A significant number of pins on PL8 and SK8are not used by PIC-A-STAR, but were included for compatibility with Analog Devices EZLITE board. These locations need not be populated forSTAR. For a photograph of the CODEC board, see Part 3.


AKING THE DSP board PCBs and assem-bling them is this month's specific topic.But before we get to that, since this series isapproaching the half-way point it is anopportune moment to take some time outfor a factual summary - as well as to share aspeculative prediction of where this "projectwithout end" may be going.

SUMMARYONE OF THE DELIGHTS of softwarecontrol is that once you have all the hard-ware elements, integrating them into aworking transceiver is very quick - giventhe lack of system cabling. See WWW. forsome web sites where you can see picturesof other STAR implementations.

The less good news is that you do stillneed hardware to run all this software on;but from now on the constructional pacepicks up and you will be lucky to keep up!

A few words on Beta testing since manyhave asked. All the engineering diagramsfor this project (eg circuit diagrams, PCBartwork etc) as published in RadCom arethe actual master drawings which severalpeople have used to verify the build. Inother words they were neither producedretrospectively - nor were they redrawn forpublication.

These drawings are therefore extremelyvaluable, especially if you are buildinghardware you may not completely under-stand (and therefore can't instinctively spotany errors).

The same principle applies to the soft-ware. You can either use mine, suppliedand tested; build your own; or migrate overtime. The self-education opportunity isobvious.HARDWAREThis month sees the completion of the DSPboard ie the mother and her two daughters.Subsequent months cover the IF board andits integration with the DSP board; theStatus board; and the PicAdapter board.That will complete the hardware covered bythe Components List in Part 3.

Following on from there is a 10-bandBand Pass Filter which has just completedBeta test as I write. This uses FST3126band switches which perform delightfully -and result in a compact layout which befitsthe scale of the rest of the transceiver. Youcan see it in the photographs.

A fundamental frequency injection H-mode mixer using home-made transformersis the next development - and a post-mix

amplifier. The next phase is a plug-inupgrade for the AD9850 in Pic 'N' Mix.Designs for a stereo audio amplifier and asolid-state T/R switch are also planned. Butlong before any of these developmentsappear, you should have a fully functionaltransceiver.SOFTWAREAll of the software described in Part 2 isnow up, running and available either ascode or as ready-programmed plug-and-gochips. It has been rigorously tested beforerelease. See Part 2 for the packaging detailand the release process.

Further development is now under way toallow tailoring of both transmit and receiveaudio - and the addition of some simpleutilities such as a 2-tonetest generator.

Instructions for buildingyour own DSP filters willalso be provided. Theseplug in as alternatives tomine as required.DSP CODEDEVELOPMENTIF YOU WANT TO de-velop your own code, youwill need the tools.

My code was developedusing Analog Devicesolder DOS-based develop-ment tools - which used tobe supplied with theirEZLITE board. These are

available from their FTP site - see WWW.Nowadays they supply their VisualDSP++environment which has the merit of a Ccompiler. As an evaluation package it alsohas a program memory limit though at thetime of writing, STAR would only useabout half this limit. This world can changevery quickly so visit the AD site for thelatest information.

DSP BOARD CONSTRUCTIONALTHOUGH targeted specifically at theSTAR DSP board, the technique formounting the chips is totally general.

There are no special tools required tomount these 'difficult' chips - except apositive attitude. I have heard much


M

G3XJP's STAR built in a PCB enclosure - shown with all compartment cover-plates removed.The overall dimensions of the case are 310mm deep by 240mm wide by 85mm high. This generoussize allows good in situ access to all the boards.

The view from underneath, traditionally somewhat less beautiful -so shown smaller.


moaning about how these chips spell theend of home-brew construction - but it turnsout the opposite is true. You can lay thesechips down with a minimum of histrionicsand I am indebted to Russ, AA7QU for theprocess - which is completely repeatable.TOOLSFirstly, the soldering iron. I used an AntexCS series iron (17w) with a 0.1mm tip, filedback from a mere point to a small chisel.Any bit about 1-2mm is fine. The otheringredients are:-• Laser film, Farnell 895-945• some common solder• desolder braid, 2.7mm or less• a flux pen, Farnell 891-186• jam (home-brew, of course) or tooth-

paste.The latter is for holding the CODEC chip

in place long enough to tack its legs downand needs to be home-brew so it is neitherrunny nor full of seeds. Seriously, anywater-soluble non-setting stick is fine.CONSTRUCTION SEQUENCEMake all three PCBs first as per Fig 10using the iron-on process previously de-scribed.

The daughter boards are double-sidedbut, by design, only just. Under all circum-stances, treat these as 2-pass single-sidedboards. Any attempt to etch both sides inone pass is simply taking unnecessary risks.Do the complex top-side first. If you wantto use the artwork for the second side, drillall the holes, register the artwork with pinsthrough those holes and then iron it on. Butmuch easier, just sketch the trivial track andground-plane in with an indelible pen,joining up the dots. When etching eitherside, merely spray mask the other.

When you have fully etched a board, ab-solutely check every track for continuity orshorts, either inter-track or to ground. Ifyou get an open-circuit track the likelihoodis that it will merely not work till you findthe problem. If you have shorted tracks,however, the likelihood is that you willcook a chip and never find the problem.

If you have not used SMD Rs and Cs be-fore, just tack one end down crudely whileholding it in position with a vertical screw-driver. Then solder the other end properly -and then revisit the first end.MOTHER BOARDBuild this first, less the daughter boardsockets. This board is completely unetchedon the reverse (ground plane) side. Theonly point to watch is the sockets for IC21and IC22. Cut all their pins back to theshoulder except the grounded ones, whichare soldered both sides. Check that all theobviously grounded areas on the board areindeed continuous and if not, add linksthrough to the ground plane side.

For the external connections, I simplycountersunk the holes on the ground side,soldered stub wires to the pads on the track

side - and then applied epoxy resin on theground side to fabricate instant feed-through insulators.PROCESSOR BOARDFit the inter-side links first. Then check theintegrity of the tracking.

IC26 socket comes next. Cut the pinsback which solder only to the top track;note that exceptionally, pins 14 and 28 aresoldered both sides.

Next the processor chip. Although it hasmore pins than the CODEC, it is somewhateasier to mount since the pin spacing isgreater and the chip is quite heavy so it isless inclined to skid around. The targettime to mount this 128-pin PQFP chip is nomore than 15 minutes - or you are doingsomething wrong!

Line the chip to the pads. Please checkthe orientation as you only have a 25%chance if you leave it to luck. The goodnews is that the correct quad-pack chiplocation on the board is totally unambigu-ous. Get someone else to hold it downwhile you roughly tack down a few legs inthe middle of each side. It sounds cruel, buttrust me, it feels no pain.

Running the iron and solder along eachside at the point where the pins meet thetrack, run in a fillet of solder paying (al-most) no attention to bridging the pins orthe tracks. The only requirement at thisstage is that every pin is indeed soldered toits track. 3 minutes elapsed.

Saturate some desolder braid with flux.Rest some fresh braid - over the top of thechip - on the bridged pins. Lightly applythe iron to the braid and when you see thesolder appear on the braid, withdraw. Thenrepeat as needed. Lay the braid on anybridged tracks - and repeat until all surplussolder has been removed. Do not draw thebraid across the tracks, only along them. 8minutes elapsed.

Using a continuity meter, preferably witha 'beep' - and fabricating some probes fromsewing needles, check that all bridges haveindeed been removed. Finally, wash off

any surplus flux under tepid water - and airdry. Job done, 7 seconds per pin.

Note that C120 and C121 mount on thepads of X3 on the underside of the board.Use SIL plug strip for both PL7 and PL8 -but use only the minimum populationneeded for the latter. Ensure the smallerdiameter end of the plugs mates with thesockets. For the sockets on the motherboard, cut back the pins - except thegrounded ones which solder both sides. Fitthe connectors dry to both the mother anddaughter to ensure alignment - and thensolder them to their respective boards.

The partial assembly may now be tested.Apply 8-10v power to the mother board andcheck the voltage rails before and then afterfitting IC21 and IC22. Then plug in theProcessor daughter and after power up, D35should flash at about 1Hz. Pressing theReset button S2 should cause a momentaryhesitation before the flash resumes. Nowrun the test program, details of which wereprovided last month..CODEC BOARDHaving established that the digital andanalogue ground planes are mutually iso-lated, fit a wire link via a ferrite bead (FB)to join them. Then mount the CODEC chipas described for the processor, but in thiscase, use a very small amount of jam to holdthe chip in register at first.

Then fit IC20 and C85-C89 and checkintegrity of ground and power. The othercomponents are easiest mounted workingoutward from the chip, leaving the electro-lytics till last.

Finally, after rigorous checking andprobing of every pin and every track (itmust be right first time), I mounted thedaughter to the mother using short lengthsof component lead. In the case of the leftand right inputs and outputs, their leads passright through the mother board.

With this approach, should you ever wantto remove the daughter subsequently, cuteach wire first and then desolder both ends.

CODEC chip before and after desoldering. Donot panic, it works!

Processor chip before and after removingexcess solder. The target time to mount this128-pin chip and clean up is 15 minutes.


The complete board may now be testedby again running the test program, details ofwhich were provided last month. Thepleasure and pride of success at this stage isindescribable!

(a) Mother board 90 x 139.7mm (b) Processor board, TOP 88.9 x 66.4mm

(c) Processor board, REAR

(d) CODEC board, REAR 39 x 43mm (e) CODEC board, TOP

Fig 10: DSP boards PCB artwork. NB all these images are mirrored for direct copying to laser film. Should you wish to use an indelible pen toapply the artwork to the back of the daughter boards, the image needs to be flipped (or simply viewed in a mirror). All holes are 0.7mm.

WWW.STAR beta build web sites:-http://www.barville.freeserve.co.uk/star_build.htmhttp://homepage.eircom.net/~ei9gq/pex.htmlhttp://www.w4zcb.com

Analog Devices FTP downloadftp://ftp.analog.com/pub/dsp/21xx/218x/ez-kit-lite/You need the two files titled: disk1.zip disk2.zip

HIS MONTH COVERS the IFboard circuitry; next month theconstruction detail. As you can see

from the photograph, this board is nothinglike as 'tight' as the DSP board but, bycontrast, there are many more components.

IF BOARD OVERVIEWThe block diagram was covered in Part 1.The board has a bi-directional IF port -which is then translated to / from 15kHzwhere the DSP takes over both on transmitand receive.

The IF frequency can be at any HFfrequency of your choosing, typically in therange 5-12Mhz. The determinant is theavailability of the crystal filter, FL1.

Pic 'N' Mix allows you to change the IFfrequency injection offset in a matter ofseconds, so there are no issues there.

CIRCUIT DESCRIPTIONREFERRING TO Fig ?, the 50Ω IF ismatched to Fl1 by L1 / VC3. This is astandard L-match and should be modified -applying the text-book L-match equations -for your filter's frequency and impedance.VC3 is adjusted for maximum output in thefirst place but thereafter for minimum pass-band ripple. The turns ratio of T1 alsoneeds to be established for your filterimpedance. The values given assume a10.7MHz filter with 2k2Ω impedance.

TR1 is the ubiquitous bi-directional J310IF amplifier. It offers modest and quietgain, a stable load and much convenience.

IC3 and IC4 provide fast (and silent) T /R signal switching - with high isolation andonly a few ohms on-resistance. Resistivedivider networks are used throughout tobias the signal paths to mid-rail.ON RECEIVEIC3 routes the signal to IC2, the AD603 IFamplifier. This is a quiet device with goodAGC characteristics. As used here, it has again range from 0-40dB - with a linear indB response to a linear DC control voltage(an increase in control voltage produces anincrease in gain). It is not the mostinexpensive device available, but if youhave been brought up on IF amplifierswhich emulate snakes, you will appreciatethe difference.

This 40dB AGC range is combined witha further 45dB in DSP to give some 85dB intotal - more than plenty by most standards.

The process for setting RV1 and RV2

follows later. The two AGC control signals('wide' and 'narrow') are generated on theDSP board and summed at the junction ofR38 and RFC3. The 'wide' AGC voltage isgenerated by detection over the full FL1bandwidth - and is there only for emergencygain back-off in the presence of a verystrong signal outside the DSP filterbandwidth. Normally the 'narrow' controlvoltage dominates - and it is also routed viaa buffer, IC7b, to drive the 'S' meter. Thislatter is shown as a 1mA movement - butlater I will be offering a bar-graphalternative - in which case RV3 sets thezero point and RV4 is not fitted at all.

The output from IC2 is routed via IC3 tothe SBL-1 mixer, IC5. You may ultimatelywish to fit a stronger device here -depending on the width of your roofingfilter and your operational needs. Themixer injection port is fed from a basiccrystal oscillator - and this could also be'beefed up' if required.

C38 and R1 terminate the sum (HF)mixer product - whereas RFC12, C39 andC51 pass the wanted 15kHz differencecomponent.

TR3 is a low-noise, modest-gainamplifier which feeds IC6b. This latter hasmodest gain at low frequencies with theresponse rolled off rapidly by heavynegative feedback provided by C59.ON TRANSMITIC7a provides modest shaping of themicrophone audio and significant gain toget the level up to that required by theCODEC on the DSP board.

You should alter the input arrangementsof IC7a to suit your microphone impedance- and C45 in particular for a good mid-range

audio response with your voice. Sometailoring options may later be added in DSPas well.

The output of IC7a is routedunconditionally to one input of the DSP -since it needs to continuously monitor themic input for VOX purposes.

The transmit signal next appears as a15kHz SSB or CW signal from DSP -which is routed via IC4 to the buffer IC6a.This in turn drives the complimentary pairTR4 and TR6 which are there to deliverpower into the low-impedance loadpresented by the SBL-1.

On transmit, the AD603 is out of circuitand IC3 routes the signal directly to theJ310, TR1 - and from there to the filter FL1- and thence out to your transmit IF strip.T/R SWITCHING CONTROLThe J310 is switched by the 12VRx and12VTx lines, the inactive one being taken tonear ground. All other T / R switching ismanaged by IC3 and IC4. Their switchingvoltages are derived from the 12VRx lineonly, with TR12 acting as a simple inverter.This approach is designed to prevent youfrom being on transmit and receive at thesame time - in the event of loss of either the12VTx or 12VRx supplies.

Part 8 by Peter Rhodes, BSc, G3XJP Pic-A-Star SOFTWARE TRANSMITTER AND RECEIVER

T

IC1 78L05

SBL-1

C37

10n

IC4 FST3125

IC3 FST3125

7

FL1 crystalroofing filter

7

4 85

8

4

C4110n

C3810n

RFC12

330µH

C39

10n

C51100n

R147R

TR32N3904

C33

220n

R205k6

C60100n

R3210k

R112k7

R143k3

C57

100n

TR42N3904

TR62N3906

D5

1N4148

D6

1N4148

R4470R

R153k3

Tx15kHz

Rx 15kHz toDSP board Right input

+10V

0V to Tx

2 3

C4310n

R2100R

R6560R

D9

BA244

TR1J310

D10BA244

R194k7C36

1n

R3422k

C35

1n

D7BA244

R184k7

D8

BA244

R5560R

C32220n

C2747µ

C22220µ

10t:3t6t:3t

C49100nC48

100n

+5V

R2410k C31

220n

R3010k

R3110k

C65

100n

20t:4tT3

C47100n

RFC10

100µH

TR2J310

R35100k

X1seetext

RV1100k

R2510k

R2610k

+5V(A)

14

+5V

0V to Rx

IC6aTL072

1

2

3

R2810k

R2910k

IC6bTL072

+10V

5

6

R71k

R37

22k

C59

10n

C61100n

8

6 5

9

1

37

R102k2

C53100n

C52100n

6

IC2AD603IF amp

RV2100k

AGC sensitivity

IF Gain

C2110p

R2110k

R2210k

+5V

7

0V to Tx

11 12

1, 13

10, 40V to Rx

3 2

12

6

5

8

9

R2310k

+5V

T2

C50100n

14

4.85µH

L1 T1

+10V

R13

3k3

0V to Rx

0V to TxZD14V7

RFC8

100µH

C62

100n

C181n

Rx Right AF orTx 15kHz from DSP

board Right out

RFC5

100µH

C55

100n

C161n

RFC7

1mH

C121n

Rx Left AF fromDSP board Left out

C151n

'Narrow'AGC

C141n

+12V Rx RFC1

1mH

C11n

+5V

C4010n

+12V Tx RFC2

1mH

C21n

C58

100n

C63100n

C64100n

C191n

C46100n

Tx/Rx IF

8

4

C26220µ

IC7aTL072

1

2

3

Mic input

C45

22n

C101n

+10V

RV547k R16

4k7

Mic Gain

R174k7

Mic AF toDSP board Left input

RFC9

100µH

C34

220n

C171n

R2710k

R33

47k

C2410µ

C2533µ

R3

220R

10, 13

1, 4

IC5

Rx AF toRight stereo PA

RFC6

1mH

C56

100n

C111n

11

C54100n

Rx AF toLeft stereo PA

2

R38

120k

'Wide'AGC

C131n RFC3

1mH

7IC7bTL0725

6

R36

100kC282µ2

C91n

'S' meter

RV410k

+5V

RV310k

FSD

Zero

1mA

To 'S' meter buffer amp

Front panel

+10V RFC4

100µH

C31n

C23220µ

C7010n

C4210n

C7110n

C4410n

VC2100p

VC3100p

VC1100p

C74100n

TR122N3904

R62100k

R1222k

R6310k

Note:-For a low-Z microphone,increase C45 to 4µ7 and add1k from C10 to ground.

Fig 11: IF board circuit diagram. Takes an HF IF feed from a typical bi-directional mixer and post-mix amp and translates it to/from 15kHz. This board mounts back-to-back with the DSP board. All components aremounted on the track side except for the crystal filter FL1 and the SBL-1. The roofing filter is your choice and X1 must correspond. The switches in the receive path are shown as closed for illustration purposes only.

© G3XJP Pic-A-Star Page 1

UILDING THE IF board is detailed thismonth. Next month covers itscommissioning - and integration with theDSP board.

IF BOARD FORMATTHE IF BOARD comprises a traditionalPCB - with two end-plates soldered on toform an H-section as shown in Fig 12. TheDSP board is subsequently mounted on theIF board as illustrated.

This form of construction is not strictlynecessary. You could build the IF boardand DSP board into two separate enclosuresbut this approach was chosen because thesetwo boards are highly interconnected.

The IF PCB dimensions are determinedby the size of (and are just larger than) theDSP board - resulting in generous spacingbetween the functional blocks. The surplusboard area has been allocated around thecrystal filter and the crystal oscillator; theformer so that any reasonable sized filtermay be fitted, the latter to give room for amore sophisticated oscillator if desired.

CONSTRUCTION NOTESTHE PCB is assembled by soldering mostof the components to the track side. Thisapproach makes signal tracing easier andminimises the amount of hole drilling.SMD components were not specified heresince they are not needed but most of thecomponents are in fact mounted SMD-style.

Mask, etch and drill the PCB using theiron-on laser film technique covered in Part3. On the groundplane side, countersink theungrounded holes associated with FL1 andthe SBL-1. Both the coax lead to the SBL-1and C37 are soldered directly to the SBL-1pins - as opposed to PCB track - so drillgenerous clearance holes for these pins. Allother holes are grounded both sides of theboard and are not countersunk.END-PLATE DIMENSIONSThe width of the end-plates is that of the IFboard. The task now is to determine theirheight - which is principally (but notentirely) determined by that of your crystalfilter.

Fit FL1 and then using spacers somewhatlonger than the height of this filter, crudely

trial-mount the DSP board as in fig 12.The height of the end-plates is now that

of this assembly plus at least 20mm for theIF board components. The approximatesum is 24mm for the DSP board plus 20mmfor the IF board components plus 2mm forthe PCB thickness plus the height of yourchosen crystal filter plus 3mm margin. Thelatter two measurements also sum to giveyou the length of the four mounting spacers.Be generous.END-PLATE FITTINGThe end-plates are fitted before mountingthe components since this makes the boardeasier to build - and to handle withoutcontaminating it with finger marks.

Mark the target position of both boardson the inside of the end-plates and thenlooking at fig 13, lay off the position of thefeedthrough capacitors from the IF board.Drill holes for these now, but leave fittingthe capacitors till later.

Clean both sides of the IF board and end-plates immaculately - and apply a light coatof spray flux/lacquer to both sides.

Now seam solder the end-plates to the IFboard with a large iron. If you mount bothat the same time, you will be able to checkon the geometry by eye. Progressivelychecking all remains true and working bothsides of the IF board, use small single tacksfirst of all, then multiple tacks and finallyform neat fillets.COMPONENT MOUNTINGRefer to Fig 13. Tin all the pads except

those under the FST3125s. Mount bothFST3125 chips. Align the chip and tackdown two opposite corners to the largerpads provided. For the remaining pins,offer the iron and solder to the track justshort of the pin - and the solder will spreadalong the board and wet the pins bycapillary action.

Fit the wire link across IC4. Sorry aboutthis, but I just could not design out that onelink.

Cutting their ungrounded leads so thatthey sit just above the board, mount all theother components - except the presetcapacitors. A pair of tweezers is useful forhandling the smaller components.

To surface mount the DIL ICs, cut off allthe pins back to the shoulder except anygrounded pins which pass through theboard. Do not use sockets.

Fit the feedthrough capacitors - which aretypically made off to the IF board by usinga series RFC as a flying lead. Mount RFC1and RFC2 at right-angles to each other inthe vertical plane to minimise mutualcoupling.

Trim and solder all the grounded leads onthe back of the board. Check with acontinuity meter that all grounded track isin fact grounded. Also perform all the usualbasic tests such as checking isolation andintegrity of the power rails.

Mask the preset resistors and give bothsides of the board a final and generous coatof spray lacquer. Finally, fit the presetcapacitors, definitely unlaquered.


B

Componentside of IF board

FL1SBL-1

Component side ofDSP board

EPROM

+8VKEY PTT

IF in

/out

coa

x+1

2VTx

+12V

Rx

Mic

in'S

' met

er o

ut+1

0V

Rx

AF o

utja

ck s

ocke

t

T/RStatus

Seriallink toPIC

NOT TO SCALE

CODEC

Fig 12: DSP assembly illustration and recognition drawing, not to scale. The IF board and twoend-plates are seam soldered to form an H-section. The height of the end-plates is typically 6cm asa minimum but can be up to the full height of the Tx/Rx enclosure. The DSP board is bolted to theback of the IF board. Note that critically, the external connections are brought out at different'levels' depending on which side of which board they connect to - and at different ends dependingon the destination. Two further sides and a top and bottom (not shown) complete the screeningbut are not added until after final commissioning.


C41

R22

C50

C65

R2

R21

RV2

C38

R1

RFC

12C39

C51

C31

R26

C54

C33R11

TR3

C60

R20R32

R14

R37

C57

R25

R30

C32

R28

R35X1

R5D7R

18 C35C36

R19 D10

D8

R34

R6

T2C49

C48

R23R13

TR6

D6R15

R4

R29

C42C70

C16

IC3

C18

C56

TR2

TR1 RV1

C43

C52

C22

R10

C74

IC1

C40

C53

R24

RFC3

C59

RV5

R16

C45

R33

C13C14

RV3

R3

RFC

8

R17

RFC6

RFC

5R31

C55

C58

R36

RFC1

C17

C34

R27

C62

C61

R7

TR4

D5

IC2

C47

T3

L1

D9

RFC4

RFC7

RFC2

FL1

IC5

IC6

IC7

IC4

C19RFC10

T1

C15

+10VC3

C10 Mic in

C2 +12V TxC1 +12V Rx

C9'S' meterout

C11 Right stereo outC12 Left stereo out

IF in/out

C44

C64C63

RFC9

R38

C26C

23

VC1

VC2

VC3

C24

C25C28

C27

C21

PCB dims 5.75" x 3.75" (146 x 96mm) ie just larger than the DSP board

IF Board component location

PCB tracking and hole drilling template. NB:- This image is mirrored

Key:-Mountingholes

Ground, solderboth sides

Screeningpartition

Verticalfeedthroughcapacitor,preferablygrounded bothsides

Componentmounted underboard ie FL1and IC5 only

Feedthroughcapacitormounted throughpartition wall,preferablygrounded bothsides

C37

Wire link acrossIC4

ZD1

C46

C71

R62

R63

R12

TR12

(stereo jack socketoptional)

Fig 13: IF board PCB layout on double-sided board. The reverse side is completely unetched to form a continuous ground plane and screen. Allcomponents with the exception of the crystal filter FL1 and the SBL-1 mixer, IC5 are mounted on the track side. You may need to customise thetracking to suit your crystal filter. The 'holes' are shown on the component layout only to define the tracks should you be producing the PCB by somemanual method. Only components which feed through to the back of the board require actual drilled holes and these are as defined on the trackingtemplate. The tracking template image is mirrored (ie flipped left-to-right) for direct copying to iron-on laser film. The basic drilling size is 0.7mm -with holes for mounting, feedthroughs etc drilled larger to suit. Some internal screening partitions are made from PCB material or brass shim stock.Those around the X1 crystal oscillator need to be particularly RF-tight to prevent oscillator leakage into the IF strip.


HIS COVERS integration of the IFboard with the DSP board - to makethe DSP Assembly. Testing of this

assembly is also covered - as is theadjustment process for the Timer board.

DSP ASSEMBLY - ASSEMBLYTHIS PROCESS STARTS when the IFboard and DSP board are fully built - andthe latter has been tested using the testprogram. The required DC supplies comefrom the Timer Board (or some equivalentarrangement).

Make off all the leads between the boardsas shown in Fig 14. With the two boards atright-angles (but preferably less), trim theirlengths and make off the other ends to theirrespective feedthroughs. Ground the braidsto the adjacent groundplane.

Mate the two boards, and in the process,perhaps trim some excess lead lengths.

Fit diodes D12-D14 - outside the housing- to drop the 10V rail to a nominal 8V.

COMMISSIONINGTHE DSP ASSEMBLY is first proved inisolation and then crudely integrated withsome existing Tx/Rx for verification. Theidea at this stage is to demonstrate hardwarefunctionality, not system performance.BASIC DC TESTINGAs a preliminary, set RV1, RV2, RV5 tomid travel and RV3, RV4 fully clockwise.

On the end-plates, connect up 10V,+12VTx (grounded on receive), +12VRx(+12V on receive) - and the stereo outputs,typically to some domestic amplifier.

For the first few seconds after power-on,a voltmeter on the 'S' meter feedthroughshould show definite activity on a 5V range.Check that the T/R Status line is near +5V.

On the IF board, check all the power railsand then get the X1 oscillator working.Adjust its frequency to the centre frequencyof FL1 + 15kHz.LOADING TEST SOFTWAREConnect the serial cable from your PCCOM port to the DSP Assembly. Also, amicrophone (both audio and PTT). Loadthe test program as previously describedand re-verify operation.

Speaking into the microphone shouldproduce audio from one stereo channel.Adjust RV5 for maximum undistortedoutput - but in any event, no more than 2Vpeak-to-peak on C17.

LOADING OPERATIONAL SOFTWAREReset the DSP board (ie press and releaseS2) and load in the operational software asper the loader on-screen instructions.

Loading is complete when you arelooking at user controls on the screen as inFig 15 - and the DSP board LED is out. Ifthe LED remains - or reverts to - flashing,this indicates a comms failure duringloading.

At this stage the DSP Rx should beoperational. To verify this, feed a sniff ofRF at your IF frequency into the IF in/outcoax. Just tack a few inches of wire to thecoax inner and put it near some suitablesignal source eg the DDS or a GDO. Asyou tune across the IF, you should hear thebeat note - and the LED on the DSP boardshould light in the presence of signal.

Turn the RF gain up and down on the PCto verify that you are in control.

If you speak into the microphone, thisshould also light the LED. Grounding thePTT line should mute the Rx - and the T/RStatus line should go to near 0V.

On the PC, switch to CW. Grounding theKEY line should then produce sidetone.

The T/R Status line may now beconnected to the Timer board and its

operation verified. Under no circumstancesbe tempted to connect T/R Status to someexternal PTT line, say on your transceiver.

Getting to this point is a major milestone.But if any of the preceding fails, stop andcorrect the problem before going further.

INTEGRATION TESTINGTHIS STAGE IS NOT strictly necessary.You could wait until you have a completedtransceiver. But I commend this as thebetter approach - not least because anyproblems will be confined to the new-buildDSP hardware.RECEIVERConnect a short fat ground strap from yourTx/Rx to the DSP Assembly.

Locate a suitable bi-directional 50Ω pointon your Tx/Rx; after the mixer, after anypost-mix amplifier, after any pad is best -but the 50Ω IF port on your mixer willsuffice for test purposes. Patch in the IFin/out coax via a series 100n instead of yourexisting IF strip.

Arrange to be able to switch your Tx/Rxbetween transmit and receive. Turn the AFgain down on your Tx/Rx - and any otherRx gain controls to maximum; and Tx gaincontrols to minimum.


T

PCB ground enclosure

C4 1nC7 1n

C5 1n

C6 1n

IF in/out

+8V

appr

ox

FL0

T/R statusto PIC and

Timer board

DSP code& user

commands

'Wid

e' A

GC

+12V Tx

+10V

'Nar

row

' AG

C

+12V Rx C1 1n

C2 1n

C3 1n

C8 1nKEY

PTT

D12

-D14

1N40

07 Ground plane

Track side

Track side

Rig

ht o

ut

Rx

AF (R

ight

) or T

x 15

kHz

from PICor PC

RS2

32

PF0

PF1

Left

inAm

plifi

ed M

ic

Rig

ht in

Rx

15kH

z

Rx

AF (L

eft)

Left

out

Buffered AGC

Stereo amp

'S' meter

Rx AF (Right)

Mic Mic AFC12 1n

C11 1n

C10 1n

C9 1n

Rx AF (Left)

Ground plane

DSP board

IF board

Fig 14: DSP sub-assembly. The IF board is bolted back-to-back to the DSP board using nuts, boltsand spacers. Both their groundplanes form a screen to isolate the two halves of the box.Feedthrough capacitors are used to route between the two halves of the box - and to the rest of thetransceiver. C11 and C12 may be replaced with a stereo jack socket.


Power up on an LF band and then loadthe operational software as previously.Inject signal frequency plus FL1 centrefrequency into your Tx/Rx mixer - and youshould hear resolved LSB signals from bothspeakers. On a quiet frequency (LED isout), peak VC3 for maximum band noise.Then peak RV1.

Connect a 'scope (DC, 1V/cm) to the 'S'meter output. This should show about 4Von weak signals and progressively less asAGC action occurs. Find a signal givingabout 2.5V and adjust RV2 until it isslightly less. While listening on a noisyband, adjust RV6 until the AGC loop isclearly unstable and hunting - and then backit off until it is smooth. That completes acrude setting up of the AGC system, enoughto verify that the hardware is working.

At this stage, with the DSP Assemblyunscreened, there may be evidence of whitenoise on the higher bands.TRANSMITTER INTEGRATION TESTWith your Tx drive level well down, set TxDrive on the PC to 10. With your Tx/Rxconnected to a dummy load, and preferablymonitoring on another receiver, set up toobserve the Tx output on a 'scope for flattopping etc.

Put your Tx/Rx on transmit. When youground the DSP PTT line, this will put theDSP assembly onto transmit as well. TheMike Gain on the PC should be increased asfar as possible - but only so long as there isno evidence of any clipping, compression ordistortion.

If all is well, bring up the drive on yourTx to its normal setting. Then increase theTx Drive on the PC, ensuring the outputremains clean - up to your normal powerlevel.

Now would be a good time to screen thextal oscillator and add the other screens ontop of the IF board. The fully screenedenclosure is best left until the very end.

PC CONTROL OPTIONTHIS IS A TIMELY opportunity to outlinethe behaviour of the PC control panel. Fig15 shows the screen of the loader after theDSP code and the controllable parametervalues have been downloaded to the DSPassembly - at 9.6KB.

Adjustment and use of the various DSPfeatures themselves follows later. Here, Iam concerned only with the mechanics.Simply key the appropriate number tochange a switch state, the upper-case letterto increase a parameter value - and thelower-case letter to decrease it.SYNTAXThe BASIC control software has beenoptimised for simplicity. That is, only themost basic syntax has been used - and ifyou have ever written any software in anylanguage (very nearly, English will suffice)- then you will have no trouble following it

or editing it. Equally if you want to buildsome controller other than Pic 'N' Mix -either in dedicated hardware or on your PC- then this acts as a model. If you have thebackground to undertake this, then equallyyou will have no issues following the code.CONTROL PARAMETERSThese are held in a separate file,param01.xjp. It is the controller'sresponsibility to handle parameter valuesand to constrain them to be withinmaximum and minimum values - and in anyevent, within an 8-bit byte. The value 255is assigned to any parameter that does notapply in a given mode.

Following any user change the newparameter value is sent to the DSP as threebytes. The first is always a tilde "~", thesecond is unique and identifies theparameter - and the third is the new value.Nothing could be simpler.FREQUENCY CONTROLOne of the virtues of controlling the wholetransceiver from Pic 'N' Mix is that it canhandle the injection offset needed whenswitching between SSB and CW andtransmit and receive. Obviously the PC hasno intrinsic ability to do this, so you need tomake other arrangements - eg operate yourTx/Rx split when on CW.

PIC A TIMEIN THE STAR environment, the solepurpose of the Timer board (see Part 4) is toprovide click and spike free R/T and T/Rtransitions. Hang times for VOX or QSKoperation are controlled by the DSP.

All the switching times may beindependently set between 1ms (very fast)and 63ms (incredibly slow). In general, it isbest to start with the times set to incrediblyslow, and then reduce them progressivelyuntil there are any signs of switching spikeson the transmitted output - or clicks onreverting to receive. Having said that, ifyour Tx/Rx has inherently noisy switching,

a click on reverting to receive is inevitable.The DSP code has a feature for blankingany such click - but it is obviously bestavoided by design.ADJUSTMENT PROCESSThe process for altering the timing is asfollows, starting with all the switches OFF,ie away from the adjacent PIC:-

1 Set the More/Less switch as required toincrease or decrease the time delay. (Moreis towards the PIC).

2 Set the switch(es) for the time(s) youwant to alter to ON, ie towards the PIC

3 Key the T/R Status line down and uponce for each required millisecond ofchange.

The altered time(s) will be implementedimmediately - but not stored. When all therequired changes have been made, put allthe switches to ON, key the T/R Status linedown/up one final time - and all the newtimes will be stored and retained. Asevidence of success, this particular R/T/Rtransition sequence will not occur.Conversely, to abort all changes sincepower-on simply miss out this stagecompletely, power off and wait 20 secondsbefore powering on again.

Finally, set all the switches to OFF.Note that this process can be used to

change several (but not all) of the timedelays simultaneously though you may wishto avoid this practice unless gross changesare required. Note also that the PIC cannotbe programmed via the programmerinterface if the switches are ON.

C:\STAR\STARp1.xjp # to restore default values + save all as defaults * main menu Denoiser ON/OFF 1 to change 0 Filter ON/OFF 6 to change 1 LSB/USB/LSBCW/USBCW 2 to change 0 VOX/QSK ON/OFF 7 to change 0 Noise blank ON/OFF 3 to change 0 RF clip ON/OFF 9 to change 0 Auto notch ON/OFF 4 to change 0 -------------------------------------------------------------------- Denoise beta a/A 25 || Filter width o/O 6 Denoise decay b/B 60 || Tx hang time q/Q 12 RF gain d/D 12 || R->T pre-delay r/R 10 AF gain e/E 40 || T->R blank time s/S 1 Stereo effect f/F 49 || VOX gain t/T 25 Noise blank fine g/G 1 || anti-VOX gain u/U 30 Noise blank coarse h/H 5 || Tx drive v/V 18 Auto notch beta i/I 18 || Monitor level w/W 70 AGC decay time n/N 85 || Mike gain x/X 20 || RF clip level y/Y 10

x/X = less/moreFig 15: The PC screen running under QBASIC. Illustrated are the STAR parameters for the SSBmode. This has since been revised in appearance.


his month we take a break fromconstruction and discuss some of therationale behind the PIC-A-STAR UserInterface (UI). This has had significantimpact on the front-panel layout andultimately on the entire transceiverenclosure.

UNTIL RECENTLYI had followed a philosophy of fixingthe front panel controls with thoseanybody could reasonably need todrive any transceiver - on the groundsthat these controls were essentiallyindependent of the inner workings.That approach bought me 25 yearsdevelopment of four fundamentallydifferent transceivers - all using thesame housing and front panel.

But times they are a-changing!I have come to appreciate that the

opposite approach is more appropriatefor a transceiver with a significantsoftware content. Fig 16 shows, notleast, the consequences to the overalldimensions.

PIC-A-STAR UIA significant amount of effort has goneinto achieving an effective UI. Thechallenge is to avoid the extremes. Onthe one hand, it is easy to end up with asystem whose complexity exceedshuman intellectual capacity. We haveall listened in on those amazing menucomparison QSOs which usually end in"I know I am supposed to hit Return.How hard?". On the other hand,personal preferences do vary and youshouldn't be prevented by the designerfrom adjusting a parameter merely forthe sake of a simpler UI.

The clue to the best approach camevery early.NO PAIN - NO AF GAINDuring the early evolutionary development,there was, of necessity, a period of severalweeks when I had absolutely no adjustablecontrols whatsoever. To increase the AFGain, for example, I had to edit the DSPcode, re-assemble it and download thewhole suite - including this one changedparameter. As you can imagine, I did notbother very often. Actually, it encouragedme to focus on improving the functionality

so that the built-in control systems wouldtake care of the 'variables' without unduemanual intervention.

The background thought here is that as adesign evolves - perhaps over several years- the number and purpose of the controlscan swing wildly. And as technologyevolves, so do the opportunities. Whowould have thought I would need an Auto-notch on / off switch even 15 years ago?And when I designed Pic 'N' Mix, I canassure you the thought that it had the

intrinsic flexibility to control the entiretransceiver never crossed my mind.SO THIS TIME …I have taken the minimalist approach -starting with the observation that allcontrols can be classified under two genericcategories, namely "switches" and"amounts". Thus PIC-A-STAR has two(and exactly only two) correspondingphysical controls, namely:-• a knob which alters "amounts" and


T

DSP Assembly ie

15kHz IF board below

DSP board above

6 Low passfilters

(PCB vertical)

PA

Band Pass Filtersfor 10 bands

Tx predriver

Timer board

Status board

Pic 'N' MixDDS board

PICAdapter

9.5" x 11.75" - x 3.25" high

Stereoaudioamp

Mixer, amplifier,diplexer, switching

Pic 'N' Mix Display board

RF T/R switch

KEY:- Hardware covered in this project

Referenceoscillator

Injectionfilter

Fig 16: A possible transceiver enclosure illustrated at half scale. Mine is fabricated from 2mm double-sided fibreglass PCB stock. This gives excellent access from both sides.


• a keypad which handles the "switches"- as well as specifying which "amount"the knob is connected to.

By "amounts" I mean, for example, amountof AF Gain, amount of RF Clipping andindeed amount of Frequency. By"switches" I probably better mean "choices"eg '80m' instead of '15m' and 'Autonotch on'as opposed to 'Autonotch off'.

Having settled the mechanical issue, thenat any time I can have more or less as many"free" knobs as I like - by simply assigningthem in the software. And at the same time,saving much cash on real pots, knobs,switches - and that real nightmare, theconsequential system cabling.

Now that my STAR is in dailyoperational use - besides changing bandsand frequency - the biggest strain on the UIhas been turning the VOX off when the fastjets go over - and turning the Tx power upand down to suit conditions. All the othercontrols have to be set up correctly, butmost are essentially set-and-forget.

FRONT PANELThe template used to make my front panelis shown in Fig 17. This is designed to lasta lifetime in the sense that I can allocate anyfunction to any switch - including a clusterof related controls as a simple sequentialmenu list. And thereafter, any range ofvalues to the menu items - and so on.

The worst-case change issue is that oneday I may need some new legends on mykeypad overlay. I have never been anadvocate of beautiful home-made radiosversus functional home-made radios (givena finite life-time, you have to choose) - butone non-trivial benefit of this approach isthat the front panel is less than A4 (and USLetter) in size - so I can print off a new oneand stick it on anytime. Upholstery orfoam-backed carpet adhesive is the answerto your next question.

A bar-graph 'S' meter is shown but is notmandatory. A small edge-wise movementcould be accommodated above the keypad,but a 'real' one would need an increase inthe front-panel width to accommodate.

The bar-graph LEDs, six status LEDs andthe keypad all mount on the Status board.

NEXT MONTHSees a start on building that Status board aswell as the PicAdapter board.

Freq

LSB

Save

A=B

Scan

CW

Mem

Rat

eU

SBC

al

12

3

45

6

78

9

*0

# Split

A/B

R x A G C T xR

Fcl

ip /

Spot

Noi

sebl

ank

Filte

r

VOX

QSK

Aut

ono

tchDe

nois

e

On

Off

Dis

play

boa

rd

AB

AB

LSB

CW

US

BS

igG

en

Mic

Rx

Tx

Keys

Keyp

ad b

ody

flush

with

fron

t of f

ront

pan

el

4 x

2mm

nut

s,bo

lts, s

pace

rsSt

atus

boa

rdD

ispl

ay b

oard

7-w

ay ri

bbon

This

leng

th m

ay v

ary

Fig 17: My STAR front panel layout, to scale. In this case a bar-graph 'S' meter has been used.The 9 most frequently used DSP control groups (in yellow) are assigned to the 1-9 numeric keys.


his month covers the circuits of thePicAdapter and Status boards.Constructional detail follows next month.You don't need these boards to firstcommission STAR since you can load andcontrol the DSP software from your PC.

Thereafter, in terms of constructionalsequence, you need the PicAdapter first,which then allows DSP code downloadfrom your PC and subsequent upload to theDSP assembly. Thereafter, you need theStatus board to complete the user interface.

RS232 CONNECTIONSFirst the wires! The required cabling at anyone time is one of the following:-• From PC to DSP - early test• From PC to PicAdapter - load new code• From PicAdapter to DSP - normal useFig 18 shows a simple implementation.The lead with the female connector formating with the PC serial cable should befitted for occasional use - if at all. Certainlythe lead should not be routed via the RFsection of the transceiver to the rear panel -to avoid any potential EMC coupling. Ikeep mine in the drawer and get it out whenI need it.

The link to the PC is needed only to loadnew releases of DSP code whereas the linkto the DSP assembly is used continuously.Normally the lead(s) plug into thePicAdapter board, but for loading andcontrolling the DSP assembly directly fromthe PC, a trivial connector with TX wired toRX can be used for pass-through operationinstead.

PIC ADAPTER BOARDThis plugs into the original PIC socket on

the Pic 'N' Mix DDS board.The circuit diagram is shownin Fig 19.

For compatibility reasons,the PIC, IC9 uses essentiallythe original Pic 'N' Mix codeto provide all the original Pic'N' Mix functionality.However for STAR purposes,it has 4 incremental tasks:-

1. To download newrelease DSP codetogether with controlparameter values fromyour PC - and retainthese in IC10, a serialEEPROM - forsubsequent normal use.

2. At power-on time, toread the DSP code andparameters from theEEPROM - and loadthem to the DSP board.

3. During normaloperational use, tocommunicate anychanges you make tothe control parameters (eg AF Gain) -to the DSP board. The changedvalues are also retained in EEPROMand thereby survive power-down.

4. To drive data out to the Status boardto control the state of the LEDs.

To these ends, minimal RS232 links fromyour PC at 1.2KB; and to the DSPAssembly at 9.6KB are controlled by thePIC.

The PIC also controls an I2C link to theserial EEPROM.

The slower down-link speed from the PCis used to give the EEPROM time to writeeach byte of the DSP code and controlsettings. However reading from theEEPROM is a much faster process, hence

the faster baud rate - which is the modenormally used operationally.

All the connections shown from the right-hand side of the PIC in fig 19 duplicate theoriginal Pic 'N' Mix pin-out and providecode compatibility.

The T/R status line is an input to thisboard (and it also goes to the Timer board).Whether STAR is transmitting or receivingis determined by DSP and the result iscommunicated to the PicAdapter solely toallow 'split' operation. Thus it need not befitted at first test. This line is at logic '1' onreceive, '0' on transmit.

The only other pins worthy of mentionare the Data, Clock and Latch pins whichsimply drive the Status board LEDs.

Because this board is self-contained, itcan be programmed - as an assembly - inthe 18-pin socket of a PIC programmer.

For test purposes this board, onceprogrammed, should provide full normalPic 'N' Mix DDS operation, albeit withslightly longer key presses being requiredthan Pic 'N' Mix users will be familiar with.

Split operation should finally be verifiedwith the T/R Status line connected.

STATUS BOARDThis board carries the bar-graph 'S' meter,the latch/driver for the status LEDs and thepassive connections to - and the mechanicalmounting of - the keypad. You have the


T

RX

35

Serial cablefrom PC COMport plugs inhere

Female connectoreither chassis mounted

- or on a flying lead

TX

C7DSPAssembly

Power-on DSP code,then user commands

Ground

Plug strip

Socket strip

This plugs into either afree socket wired

for pass-through ....PC control of DSP

.... or SK2 on the PicAdapter boardPic 'N' Mix control of DSP

DSP code releasesand default settings

Fig 18: RS232 connections. This can be made up as one composite loom - or as two separate leads.

PicAdapter board in situ in Pic 'N' Mix.


option of not fitting any of these functionalelements should it suit you - and the PCB islaid out so that you can "cut bits off" shouldyou wish. Equally and conversely, theseelements were designed explicitly to beused stand-alone in totally differingsituations should you wish.

The circuit diagram is shown in Fig 20.IC16 is a conventional serial-in, parallel-outdriver. It is identical in function to thosealready fitted to Pic 'N' Mix for band-switching purposes.

IC28, the PIC, exemplifies the pin-outefficiency of the current generation of PICs.Of the 18 pins, only three are assigned to'overheads' - ie ground, power and reset, theother 15 all being available for I/O. Ofthese, one is programmed as an analoguevoltage input (the AGC voltage) - twelve asoutputs driving LEDs - and two are spare.So far, that is.

IC28 is a mere voltmeter which displays'S' units on receive and relative power ontransmit. R73 determines its sensitivity -and I can visualise some non-STARapplications where you may need to tune itsvalue.

20

IC916F870

PIC

MCLR

Tr7BC517

28RB727RB626RB525

24

23RB222RB121RB0

R47

10k

9 OSC1

10 OSC2

RC012 RC113 RC214 RC3

VDD

RB4

RB3

8 19VSS

X24MHz

C73

15p

C72

15p

+5v8

R432k2

IC1024LC256

256kserial

EEPROM

5SDA6SCL

1

2 A1

A0

VSS

4

VDD

I2C bus16 RC515 RC4

C67100n

3 A2

C68100n

13

14

5

4RA23RA12RA0

RA3

1

12

11

10

9

8

7

6

2

1

18

17

4

17 TX

18 RX

R444k7

+5v

Tr8BC517

R4510k

R46

10kRX 5

PL1 plugs into thePic 'N' Mix 18-pin PIC

socket. Note that pins3, 15 and 16 are not

brought through.

T/R Statusfrom DSP assy

Data

Clock

Latch

ToStatusboard

SK2DSP codeand user commands

To DSP assy

DSP codeload from PC

7 WP

6 RA4

7 RA5nc

TX

SK4

SK3

11nc

Fig 19: PicAdapter board circuit diagram. This board plugs into the 18-pin PIC socket on theoriginal Pic 'N' Mix DDS board and upgrades the PIC to a more recent and versatile PIC, the16F870.

17

IC1578L05

VDD

VSS

R722k2

1

+12v

IC2816F627

PIC RA2

2RA3

14

MCLR4

D151N4148

ProgrammerInterface

RB6

MCLR

RB7 RB713

RB612

R77

100R R76

100R

R711k

'S' meter

+5v

C8433µ

C83100

n

C81100n

from DSP Assy

18RA1

15RA6

16RA7

7RB1

8RB2

6RB0

RA0

RB511

nc

S9+12

S9+24

S9+36 RB410

S4

S5

S1

S3

S2

S6

S7

S8

D3

0

1k

D3

1

1k

D3

2

1k

D3

3

1k

RB39S9R74

D22 1k

D23 1k

D24 1k

D25 1k

D26 1k

D27 1k

D28 1k

D29 1k

R75

3RA4R7310k

15

IC164094

Serial inparallelout shiftregister& latch

VSS

8

OE

C82100n

13Q6D20R70

1k

12Q7D19R69

1k

11Q8D18R68

1k

5Q2D16R64

1k

14Q5R65

1k

6Q3R66

1k

4Q1D17R67

1k

Denoiser

Auto Notch

VOX

Noiseblanker

Filter

RF clip/Spot

nc

D21

16

VDD

1 Latch

2 Data

3 Clock

7 Q4

from PIC Adapter

5

Fig 20: Status board circuit diagram. The status LEDs and driver are functionally unrelated to the 'S' meter LEDs and driver but they are co-located around the keypad. Not shown here are some merely passive tracks which are used to make off the 7-way ribbon cable to the keypad. Theconnector shown on RB7 is for future development only.


his month covers a suitable stereo AFamplifier - and concludes the constructionof the PicAdapter board and Status board.

STEREO AMPLIFIERIt is not easy to find a good stereo amplifierwhich works well on a 12V rail. If youhave a scrap car radio with speakers, thatwould provide an instant solution.

Fig 21 shows the circuit diagram of aninexpensive amplifier from the bottom endof the range. Reduce C7 and C8 for moretop response. The PCB is given in Fig 22.

Should you need something with moreoutput, take a look at the TDA2004 orTDA2005. In any event you will need touse decent speakers to get the full benefit ofPIC-A-STAR's audio quality. Severalbuilders have found it difficult to moveaway from 20W per channel into 12"speakers - including me.

PICADAPTER BOARDSee also the photograph last month. Thisboard is pretty tight since it needs to fitwithin the envelope of the original DDSboard. So, as you can see from Fig 23, it issomewhat three-dimensional. The spacingis eminently achievable provided thecomponents are loaded in the correctsequence - which is critical.

Check meticulously for continuity andisolation as you proceed. When insertingthe sockets / plugs, ensure the pin shouldersdo not ground on the opposite side.1) Fit C73 underneath (ie on the

groundplane side).2) Fit IC9 socket and ground pins 8 & 19

on the groundplane side.3) Fit R43, underneath.4) Cut a 9-way SIL header strip. Noting

the larger diameter end fits to thePCB, cut off that end of pin 3 andinsert to make PL1 pins 1-9 ie theinner strip. Solder all track-side pins -and pin 5 to ground underneath.

5) Repeat for the outer strip of PL1 - butcutting off the larger diameter end ofpins 15 & 16. Use a spare 18 pinsocket to ensure alignment.

6) Fit TR8, C72, R44.7) Fit IC10 socket, grounding pins 1, 2, 3,

4 & 7 both sides.8) Fit C68.9) Fit SK2, grounding the centre pin

underneath.

10) Fit TR7, grounding the emitterunderneath.

11) Fit R46, R45, C67, R47 and last, X2.1) If there is any chance of the board

fouling the Pic 'N' Mix Display board,chamfer the copper both sides.

The original 4MHz crystal on the DDSboard may berecovered and reusedelsewhere, though youmay want to postponethis until thePicAdapter is working.

STATUS BOARDThere are no specialconstructional issueshere.

Cut all the IC socketpins back to theirshoulders - except thegrounded ones. Asmall trick forsoldering the socket onthe component side; fitanother socket or ascrap chip into thesocket first. Thisprevents the pins fromwandering as they gethot.

Please note thatshould you wish toprogram this PIC insitu, you may not beable do so with D31

fitted, since it loads the programmer. Forthis reason, avoid giving reports to stationswho are "12dB over S9" until the end of fullintegration and test - when this LED isfinally fitted.


T

4

IC1 TDA2822M orNJM2703

R5

1k2

+12V

1

7

Left outto 8 Ohmspeaker

C9

470µC7470n 8R7

10k

Left in

R15R6

2C1100n

C4100µ

C2220nC5

100µR3

220R

R9

4k7R6

1k2 3

6

Right outto 8 Ohmspeaker

C10

470µC8470n 5R8

10k

Right in

R25R6

C3220nC6

100µ

R4

220R

R10

4k7

From DSPAssembly

Fig 21: A suitable 1W + 1W stereo audio amplifier. Stereo balance and gain are controlled in DSP.

The Status board, viewed from rear. All the components are surface-mounted on this side of the board. The bottom row of connectorscarry short links to the keypad.


R46R44

R47

TR7 C67

TR8

PL1 mateswith originalPIC socket

IC9 (under)IC10

SK2TXGroundRX

Component location PCB track and drilling templateNB:- This image is mirrored

SK4, T/R statusfrom DSP assy

Latch Clock Data

SK3 to Status board

Cutout

afteretching

IC10 insocket IC9 in

socket

PL1(SIL header strip)

Trackside

Cross sectional viewshowing IC mounting

TR7

Unetchedgroundplane

PCB dims:- 1.65" x 1.212" (42 x 31mm)

Key:-solderbothsidesC73

underC72

R43

R45

X2

C68

R43under

Fig 23: Double-sided PicAdapter board PCB. The cut-out is to give access to the existing programming socket. IC10 and IC9 should definitely bemounted in sockets. To achieve the clearance height required you may need to first fit an extra 18-pin socket into the existing Pic 'N' Mix socket - as aspacer. SIL plugs / sockets are used for the remaining leads - with the sockets soldered directly to the track.

PCB track and drilling templateNB:- This image is mirrored

PCB dims:- 3" x 3.1" (77 x 79mm)

Component location(track side, viewed from rear of front panel)

GroundRB6RB7

MCLR

Optionalprogrammer

interface

groundboth

sides

C81

R69

R76R77

R72

R73

R70D15R71

R64

R67

R68

R65R66

C84

R75R74

IC15

IC16

IC28

C82

D16

D17D21

D18D19

D20

D33

D32

D31

D30

D29

D28

D27

D26

D25

D24

D23

D22

LatchData

Clock+12v

'S' meter

fromPic

Adapter

S9+36 S1

Ribbon from DDSDisplay PCB

Links to keypad

De-noiser

Autonotch

VOX

Noiseblanker

Filter

RF Clip/Spot

fromDSPAssy

C83

Fig 24: Double-sided Status board PCB with the ground-plane side unetched. The width of this board (3") accommodates that of my RF front-endwhich sits behind the Status board. All components except the LEDs are surface-mounted on the track side. This allows access to both thecomponents and to the LED pads. The latter is needed to adjust the LED lead lengths for flush-fit to the front panel. The four mounting holes are forthe keypad which is mounted using short spacers. The 7-way ribbon cable from the Pic 'N' Mix Display board is routed between this board and thekeypad and made off to pads / tracks provided on this board and thence via short wire links to the keypad itself. IC16 and IC28 are mounted insockets with all pins - except their respective ground pin - cut back for surface mounting. No connectors are specified. The relevant pads have a 0.1"pitch.

Component location

Left in

R9C9

C10R10R4

R3

R8 C

6C

5

R2C3

R1

C2

C1

C8

C7

R6

R7

C4R5Ground

GroundRight in

Left outGround+12VGroundRight out

PCB dims 2" x 1.1"

IC1

IC1 TDA2822M or NJM2703C1 100n disc ceramicC2, C3 220n dic ceramicC4, C5, C6 100µ 16V electrolyticC7, C8 470n 16V electrolyticC9, C10 470µ 16V electrolyticR1, R2 5R6R3, R4 220RR5, R6 1k2R7, R8 10kR9, R10 4k7

Component list

Top (component side) view

PCB track and drillingtemplateNB:- This image is mirrored

Fig 22: Double-sided stereo amplifier PCB. This is a 'conventional' board with components mounted on the top, the track underneath. Thecomponent side is completely unetched. No connectors are specified, but the relevant pads have a 0.1" pitch.


his month includes a brief reminder of thePic 'N' Mix features - and some hardwareimprovement options. The lattercomplement the performance of the STARreceiver front-end, details of which followlater - but they are of general application forany DDS.

PIC 'N' MIX FUNCTIONALITYFull details of the Pic 'N' Mix features andtheir use were given in RadCom, May 1999.It provides first mixer injection, band andmode switching and a range of usefulfeatures.

DDS KEY SEQUENCESThe keypad sequences are summarised inFig 25. These include the increments forSTAR operation.

See separate user document for latestFig 25: DDS key sequences.

Sequences shown with a leading savethe current frequency in the respectivelocation if the 2-key sequence is precededby the 9 (Save) key.

PIC 'N' MIX CODE CHANGESThe following apply to code shippedexplicitly for use with PIC-A-STAR:-• The opportunity has been taken to

more vigorously de-bounce the keypad.The simple consequence of this is that keypresses now need to be somewhat moredeliberate.• CW offset calibration (previously 34)

is no longer required since the CW offset isnow managed in DSP. The procedure forcalibrating the reference clock and SSB IFoffsets is unchanged. However, thereference clock frequency calibration maynow be loaded directly from your PC. For IF offset calibration you will find ituseful to switch the DSP filters off - so thatyou can hear right down to zero-beat. ACRO connected to the receiver audio outputis also invaluable for seeing the exact zero-beat point. When both SSB offsets arecorrect, there should be a difference ofprecisely 2.7kHz between them.• CW is now received either upper or

lower sideband - so you have the choice oftuning direction. Either in the samedirection as SSB signals on the same band -or always tuning CW in the same directionwhatever the band. The frequency readoutalways shows your transmitted frequency(not least, for legal reasons) and your

receive frequency isdisplaced from this by theCW offset - which is yourchosen and preferred beatnote.• If you are in CW mode

and you key 44 again, thiswill toggle reverse CW. Thisswitches both receivesideband and injectionfrequency to give the samebeat-note, but "from the otherside". It can be useful inclearing QRM and forchecking you are properlynetted since if not, the beatnote will change.• Pic 'N' Mix now has

5MHz (60m) capability. Thelatch output previouslymarked "15MHz WWV" is now active ifany 5MHz frequency is selected and -exceptionally - if you go to 5MHz, thedefault sideband is now USB.

The first 5 memory locations (60-64) areloaded with the UK 5MHz channels. Theseare the correct frequencies for uppersideband operation. If you don't want thisfeature, you can re-program theseallocations with any other frequencies (andre-enter the 5MHz ones yourself any time).• You may now fit an AD9851 DDS

chip which has the ability to multiply thereference clock frequency by 6. 79 togglesthis feature. See later for more detail.• There is a latched output bit labelled

'spare'. This may be toggled by keying 48.It was designed so that you can switch anydevice eg a pre-amp, attenuator, transverteretc - from the keypad. This switch hasbeen used to configure the STAR mixer andpost-mixer amplifier - to be described later.• The output bit labelled 'broadband' 72

is now a spare uncommitted toggle switch.• QSK Split operation has been

improved. Previously it was limited toabout 20wpm.• The frequency display is dimmed after

about 3 minutes of user inactivity - toreduce heat dissipation and to prolong LEDlife. 74 toggles this feature on/off.• All frequencies from 0-29MHz now

activate the nearest band select line.• XIT/RIT operation may be selected

instead of Split. This is still underdevelopment and details will follow.• Part-way through any DDS key

sequence, the # key now aborts it.

PIC 'N' MIX HARDWARESince first designing Pic 'N' Mix five yearsago, some detailed improvements haveevolved. What will never change is therequirement for meticulous (albeit textbook)screening, decoupling and filtering practice- if the DDS spur level is to be contained.This cannot be overstated and is the startingpoint for what follows.

The following modifications produceincremental reductions in DDS spurs and

Part 14 by Peter Rhodes, BSc, G3XJP PIC-A-STARSOFTWARE TRANSMITTER AND RECEIVER

T

R2

10k

C1

50p C282p

C3

82p C5100p

C4

82p C682p

C7

50p

L1680nH

L2680nH

L3680nH

L4680nH

RL2RL1

LO infromDDS

+12V

RL1

RL2TR1BC517

D1

1N4148

R1

10k

LO outto mixer

HF/LF( '1' / '0' )

1N4148

17/15m &12/10m bandselect linesfrom BPF(typical) 1N4148

RL1 and RL2 Omron G5V-1 series SPCO, Farnell part no 466-890L1-L4 RF chokes, EPCOS B78108-T3681-K, Farnell part no 512-450, mounted clear of board by 3mm.All capacitors silver mica or polystyrene +/- 1pf. R1 and R2 are 1206 SMD

Fig 26: 26-40MHz injection filter to strip both unwanted high-order spurs and those at the IFfrom the DDS output - on the higher HF bands only. Not needed (ie switched out) on lower bands.

R41k5

C710n

C6

100n

L1see

fig 29

Existing +8V (regulated & current lmited) via feedthrough

TR1MPSH10

R61k

R7680R

R81k

R10

1k

VC122p

C410n

TR2MPSH10

R9390R

C222p

R2560R

R11100R

R315R

C510n

C1

56p

X1see text

TH110kNTC

R56k8

VR110k

TR3TIP122

C310n

Ref outto

existingC7

R12R7 0.6W

C8100n

-t

Fig 27: Butler oscillator for either 5th or 7th overtone operation -with crystal temperature control.


are easy enough to do to make them allworthwhile. If you are building STAR Icountenance you not to implement any ofthem until you have everything elseworking - and until you have truly attendedto the meticulous bits mentioned above.• Fit a separate 7805 regulator for the

DDS chip carrier assembly. There is asimple track cut under the DDS boardwhich removes the +5V rail from the DDSAssembly. Mount a separate 7805 on therear vertical panel with 100n on both the12V in and 5V out leads - and run a flyinglead from the latter through a ferrite bead -and solder it to the 5V top foil on the DDSchip carrier together with a 100µFelectrolytic to ground.• Fit 1n, 10n, 100n 1206 SMD capacitors

- in a stack - on at least 2 corners of theDDS assembly from the +5V foil to ground.In other words, whatever it takes to ensurethat the +5V rail and ground-plane are atthe same AC potential - AF to VHF.• Add a filter in the LO feed to the

mixer. Figs 26 and 28 show a suitablearrangement for high-side injection withany IF between 8MHz and 11MHz. Thisfilter offers at least 30dB attenuation at theIF and a similar figure at 50MHz - andrising. If all else is right, this producesdramatic results.

It does not need any exceptionalscreening if mounted within the already-screened volume of Pic 'N' Mix. It is shownswitched in by band-select lines, diode-OR'ed. You should wire in diodes for anybands for which your LO falls in the range26-40MHz. However, for evaluationpurposes, it would be prudent to control itwith a simple toggle switch to +5V/0V inthe first instance.• You may simply substitute an AD9851

on the DDS Assembly in lieu of an AD9850under all circumstances ie it is completelyhardware and software compatible. TheAD9851 allows higher reference clockfrequencies - which may be useful if youhave a relatively high IF. This gets youaway from the 1/3 reference clock zone.

In addition it offers a 6x multiplier optionfor the reference clock. So, for example,you could clock it at a mere 30MHz yethave the effective benefit of a 180MHzclock. There are small spur and significantphase noise performance disadvantages forwhich see the AD9851 data sheet; butconvenience issues may predominate foryou. A facility for specifying a 6x clock isnow built into STAR software. You mayinstead want to try clocking your existingAD9850 much faster.

Fig 27 shows a reference oscillator whichwill operate on either the 5th or 7th overtoneof a crystal in the 22-25MHz range - simplyby tuning VC1. Typically the 5th overtoneis used with an AD9850 and the 7th

overtone with an AD9851. In this lattercase the x6 feature would not be invoked.

Also included in fig 27 is an arrangementfor stabilising the crystal temperature, thedetail design of which is due to Harry,G3NHR. It works so well, I have sincestuck (literally) a similar arrangement onmy IF board translation oscillator.

The thermistor TH1 and the TIP122 tabare secured to opposite faces of the crystalcan. Heat-shrink sleeving is highlyrecommended - or failing that, super glue.For smaller crystals, cut off the excessTIP122 tab to reduce height.

Set VR1 to maximum resistance and thenadjust for 100mV drop across R1. Repeatevery minute for 5 minutes and the resultshould be a crystal thermally stable at 35ºC.

In use, the frequency will change rapidlyfor the first 5 minutes after switch-on - butstabilise thereafter. This is not an on-offoven. This is proportional control.

The small PCB, shown in Fig 29, isdesigned as a drop-in replacement - aftersimply removing the components from theoriginal Pic 'N' Mix oscillator.• Fit transformer-coupled output from

the DDS chip. This gives 6dB more LOoutput and further spur reduction. The corefor this transformer is the EPCOSB62152A4X1 available from ElectroValue.(You will need 4 more of them for themixer later).

The primary is 3 bifilar turns 32SWG andthe secondary is 12 turns wound over thetop. The core is mounted in lieu of the100R and 200R resistors on the DDScarrier. Connect instead the primary to pins20 and 21 of the DDS chip, grounding thecentre-tap. Ground one side of thesecondary and take the other via a 100nblocking capacitor to pin 19 on the 28-pin

carrier. If you are not confident your mixerwill stand the extra 6dB, fit a pad on its LOport. (The STAR mixer - yet to bedescribed - is fine.)

To reduce DDS spurs to a highlyacceptable level (ie virtually none) you mayor may not need any or all of these changes.

As with all flexible designs, there areintelligent user choices to be made.

Such is amateur radio!

LO outLO in +12VHF / LFie 5V / 0V

Top (component side) view PCB track and drilling template

PCB dims 1.4" x 0.925"36 x 23.5mm

NB This image is flipped

Small stubs ofPCB formounting tovertical wall

R1, R2, L2, L4, D1mounted under board

Single-sided PCB

Drill all holes 0.7mm

L4 C7

C4

C1 C3

L3

L1

L2

C2

C5

C6

RL2

R1R

2

RL1 D1

TR1

Fig 28: Injection filter component layout and PCB.

Refout

+8V regulated via FT capTop (component and track) view

PCB tracking

L1 is 4t 22SWG wound on 6mm dia mandrel,14mm long. VC1 is polyethylene film. All fixed resistors andcapacitors are 1206-size SMD - except R1 which is 0.6Wwire-ended - and C1, C2 which are ceramic plate

PCB dims 1.75" x 1.025" 44.5 x 26mmMay be made from 1- or 2-sided material

NB This image is flipped

C2R9

R8

L1

VC1

C1R10C7

R11

R3

TR1

R2

C4

C8

becceb

R7

TR3

C5

C3

VR1

R1

b c eX1

R6

TH1

R5

TR2C6

R4

Fig 29: Reference oscillator component layout and PCB.


TAR has more useful facilities thanmost home-brew designs - andgetting a feel for their value may

well determine if this is the project for you.The user interface is also detailed thismonth - with explicit adjustment and useinstructions packaged with the software.

Those who don't have these facilitiesrefer to them as 'bells and whistles'. Thoseof us who do just grin - and keep ringingthem bells and blowing them whistles.Always assuming they are underpinned by arock-solid base receiver performance, thatis.

SSB / CW MODE MANAGEMENTSwitching between SSB and CW - andtransmit and receive for that matter - arenon-trivial design problems if the result isto be user friendly. So some backgrounddiscussion is helpful in understanding whatfollows. See also ref [1]. You may also careto critically compare the behaviour ofcommercial transceivers. If I can findanything friendlier than STAR, I willsimply change the design until it isn't. Ihave, as I write, already invested in thearchitectural infrastructure to make all thispossible.CW OPERATIONIn days of old, especially with 'separates', itwas easy, reliable - but a bit torturous. Youwould zero-beat an incoming CQ on yourreceiver, then zero-beat your transmitter -and finally move your receiver off to get acomfortable beat note.

With modern filters there is a snag. Youcan't hear anywhere near down to zero beat,so this process produces totallyunacceptable errors.

However it does establish two criticalprinciples, namely:- both stations musttransmit on the same frequency - and bothmust operate 'split' if they are to hear a beatnote. CW is inherently a 'split' mode.

With a multi-mode transceiver you musteither explicitly operate 'split' for CW - or

the design must take care of it transparently.PIC-A-STAR is in the latter category. If

you are in SSB mode and hear a CW stationyou want to work, when you switch to CWmode the received pitch will not alter andyou will not have to retune. And vice versaif starting out from CW mode.

While on the topic of CW, take a look atthe photograph of STAR's transmittedwaveform. You won't find better.SSB OPERATIONWhen you change sideband, neither yourindicated nor actual frequency should alter.This is common currency nowadays. Thereare some but not many occasions when thismatters - given that we don't usually operateon the 'wrong' sideband. If you operate viaOSCAR or into a transverter or on 60m, it iscritical.

This point also reinforces a principlewhich should be obvious - namelythat if a given feature is critical to aminority interest, provided it is notimposed on the majority - thenthere is no good excuse for notproviding it.

Pic 'N' Mix also gives you theoption to switch between high- andlow-side injection. Since thisimplies a sideband inversion, thesoftware puts in an equal andopposite change of sideband - andyou end up on the same netfrequency.

But be aware that many band-pass filters are optimised forinjection from a preferred side. TheSTAR front-end is optimised foruse with high-side injection; andthe optional filter in the LO linealso assumes high-side injection on thehigher bands.PTT AND KEY BEHAVIOURPIC-A-STAR uses the simple conventionsthat on CW, the microphone audio isignored - and on SSB the key is ignored.

If on CW and you merely key, you willproduce only sidetone. This is for CWpractice since to actually transmit youeither need to switch QSK on - or holddown the PTT line for non-QSK operation.

If on SSB, then you must either switch onVOX - or hold down the PTT line - beforeanything will happen.

AT POWER ON TIMEThe display shows your chosen start-upfrequency, but flashing. You then have four

structurally different options:-• Upload DSP code from Pic 'N' Mix to the

DSP Assembly. This is normal every-dayoperational use. The Status board LEDsflash strangely so you know something ishappening.

• Run the DDS without loading DSP code.This is mainly a diagnostic mode.

• Enter a DDS reference clock frequencyfrom your PC. A useful utility.

• Download a new (or your first) DSPsoftware release - via the Internet to yourPC - and thence to Pic 'N' Mix. This latterprocess takes several minutes. During thisperiod the incoming bytes are counted onthe display - albeit faster than the eye canfollow - much like money on a petrolpump. Unlike petrol it gives you a warmfeeling you are getting good value - andthe fact that it is counting at all signifies

that it is working. Once the new code andthe default control values are all in, youcan then proceed to upload them to theDSP Assembly thereafter.

KEYPAD / DISPLAY MODESThere are now two main modes, namely"DDS mode" and "DSP mode". The formerwas outlined last month. The latter lets youtune all of the STAR DSP controls.

Fig 30 shows a suitable keypad overlaywith the DSP legends in yellow; the DDSones in white or blue. But before we get tothat ….SPLIT OR XIT/RITThis is a new sub-mode choice for DDSuse. One of these is always enabled - andyour choice is retained at power-down.Both give you the potential to transmit and

Part 15 by Peter Rhodes, BSc, G3XJP PIC-A-STARSOFTWARE TRANSMITTER AND RECEIVER

S

FreqLSB Save

A=B R=T

ScanCW Mem

RateUSB Cal

1 2 3

4 5 6

7 8 9

* 0 #SplitA/B R/T

Rx

AGC

Tx

RFclip /Spot

Noiseblank

Filter

VOXQSK

Autonotch

Denoise

On / Off

Mute

Reset

Quick

Hold

REVXITRIT

Fig 30: Keypad allocations and corresponding statusLEDs. This is to scale and may be used as a keypadoverlay. Note that this reflects recent developments.

David, G4HMC's STAR CW waveform,photographed off-air on 80m from my STAR.


receive on different frequencies. See alsoref [2] for a general discussion.

Split mode is unchanged from Pic 'N' Mix- but with enhancements. It operates on thetwo independent VFOs, 'A' and 'B'.

Conversely, XIT/RIT operates on eitherone of the VFOs - and that choicedetermines the initial Tx and Rx frequency.But thereafter the Tx and / or the Rxfrequencies can be independently changed -and retained throughout an XIT/RITsession. Meanwhile, the 'other' VFOremains uncontaminated - and available.

What are the differences between Splitand XIT/RIT? Split can be cross-bandand/or cross-mode and 'rests' on your Rxfrequency when off. Conversely, theXIT/RIT tuning range is the current band,current mode. It 'rests' on your Txfrequency when off - thus providing an RITon/off capability - which Split doesn't giveyou.SPLIT AND XIT/RIT USEBesides pure transceive when switched off,both modes have the following options:-• Tune only the receive frequency while

your transmit frequency remains fixed (ieRIT). If you call CQ and a station answersoff-frequency, this option (in either mode)is the answer. However, in a net with onestation off-frequency, XIT/RIT mode isbetter since you can turn XIT/RIT offwhen the offending station is nottransmitting.

• Tune only the transmit frequency whilecontinuing to monitor your receivechannel (ie XIT). To call a DX stationwho is operating split, use in either modeto tune your Tx quickly to the specifiedDX listening frequency - while notmissing a word on your receive channel.

• Tune the transmit frequency whilemonitoring what is about to be yourtransmit channel (ie REV XIT). Use thisto check for a quiet spot before calling.

You can do any of the above -independently - in any order using merelyone key-press to define your choice. Onreceive, the frequency displayed will be thatwhich you are changing - and on transmit,always your Tx frequency.

Further utility options let you swap thetwo VFO frequencies - or initialise them asthe same. Likewise for the XIT/RIT

frequencies. Initialisation is done for you inXIT/RIT mode should you change VFO; orband; or frequency by more than 2.5kHzwhile on pure transceive - on the groundsthat any difference must then be irrelevant.DDS / DSP MODE SWITCHINGAs supplied, the DDS mode is permanentlyengaged and you would be unaware thatthere is any other. This is deliberate in orderthat you can check out the DDSfunctionality after first commissioning -without distraction.

Once you are happy here the DSP modebecomes available following the first timeyou download DSP code from your PC toPic 'N' Mix. The remainder of thisdiscussion assumes that this has happened.

In normal use, STAR 'rests' in DDS modeand displays frequency. The key toswitching to DSP mode is the duration ofthe first key press. A quick press on a keyactivates DSP mode; whereas a longer pressinvokes the "business-as-usual" DDSfunction. The resultant displays are quitedifferent so it will only take you a fewminutes to get the 'feel' ingrained.

The DSP functionality is given thispriority because most DSP functions areneeded quickly in real operating conditions.For example, turning the auto-notch onwhen that tuner starts up is a moreimmediate issue than, say, changing bands.

Certainly a 'quick press' need not betentative, merely not overtly sustained.Once the mode has been determined by theduration of the first key press, the durationof subsequent key presses is unimportant.

The very deliberate exception is thebottom row of keys - which act to give DDSfunctions - and which therefore have noDSP functionality as the first key-press.

Once you are in DSP mode if you neitherpress a key nor alter a value for about 3seconds, PIC-A-STAR will revert to DDSmode. You can prevent this by pressing the# key - which toggles holding the displayin DSP mode. This is invaluable whensetting up the DSP control settings.

USING THE DSP MENUThe DSP controls are grouped to form amenu as shown in Fig 31. This menu detailwill change over time, but not the intrinsicstructure. Project without end, right?

MENU GROUPINGSThe menu is ordered with the morecommonly used controls near the 'top' ofeach menu group - and rarely used onesnear the 'bottom'. In practice, some of theselatter items can be regarded as presets.

There are no sub-menus, so the system isinherently limited to 99 controls in 9 groupstimes 2 modes - a limitation I for one canlive with.

See separate user document for latestFig 31: DSP menu structure.

The menu is intrinsically SSB/CW modalin the sense that if you are in SSB modethen you simply can't get at controls whichare unique to CW - and vice versa. Thesecontrols are annotated 'SSB only' or 'CWonly'. Some controls share one commonvalue for both modes and are annotated'both'. Two controls, namely 2.2 and 8.1have different values per band.

Conversely, all other menu items that arenot peculiar to mode have different settingsstored for SSB and CW. For example,different AGC time constants, Denoisersettings and so on can be set up, varied -and retained independently for SSB andCW. This applies also to the on/off switchsettings.

All this is designed to foster a "set-and-forget" philosophy.HOW TO PIC FROM THE MENUImmediately after you (quickly) press a 1-9key, the display will switch to DSP mode.For example, a dab on the 7 key will show'7.1 13' where 7.1 denotes the first controlin that menu group, namely VOX/QSKhang time - and 13 is its present value. Seealso the photograph.

If you want to move on to the secondcontrol in the same group, press the 7 keyagain - and so on. If you want to move backup through the group, press 0. If you wantto move to a completely different menugroup, simply press the corresponding key.CHANGING VALUESAfter you have selected a control, if youwant to change its value, turn the knob -clockwise for more, anti-clockwise for less.

There are maximum and minimum valuesfor each control - and the rate of change isproportional to the range.ON / OFF SWITCHESTo switch a DSP feature on/off, press thecorresponding key (1, 4, 7, 3, 6, 9) followedby . The adjacent LED will change toprovide visible status thereafter.

For example 7 will switch VOX on/offif in SSB - and QSK on/off if in CW mode.

In fact, irrespective of which menu itemin a group you are addressing, the keywill toggle the associated switch and youwill revert immediately to DDS mode.

STAR display in DSP mode


MUTE2 near-mutes the receiver and suspendsVOX operation. This is the 'panic' buttonfor unexpected interruptions eg when thephone rings. Any subsequent key press orknob turn restores your pre-panic state.QUICK SWITCH OPTION8 toggles the quick switch facility. Whenengaged, any of the 1, 4, 7, 3, 9 keys - whenpressed - simply toggles its respective DSPswitch. Because you are thereby notpresented with the values for those menugroups, you would not want to use thisoption until those groups are set up.Conversely, once the values are tuned andyou have gained familiarity, I think this isthe mode of choice. It is for me, anyway.RESETTING CONTROL VALUES5 resets the control values (both SSB andCW) to those that you last downloadedfrom the PC - with the exception of RFGain and Tx Drive - whose latest per-bandvalues are retained. All the DSP values areremembered across a power down, but notswitch settings - which initialise to off butwith the DSP filter on - and in SSB mode.

DSP FEATURES DESCRIPTIONA few words are in order for some of themore esoteric features you may not havemet before. In roughly menu order:-DENOISERThis is rather more a comfort feature than aperformance one. It acts to reducebackground white noise - and whenworking well, is not unlike squelch on FM.It is especially effective on CW and veryuseful if just monitoring a quiet (albeitnoisy) channel. The 'right' combination ofsettings is somewhat subjective and canoccasionally vary from one signal toanother - and certainly by mode. It is bestwith the RF gain turned up and with longerAGC hang times. Experiment!

Both the Denoiser and Autonotch (seelater) are essentially as implemented inDSP-10 by Bob Larkin. See also refs [3] -[6] for the pioneering work and the theory.RF GAINThis comes right at the front of the DSPreceiver chain and is used to set the SNRfor differing conditions. It should normallybe turned well up so that AGC action

produces constant audio output - and thebest possible SNR. This also contributes toclean VOX operation.STEREO EFFECTThis gives body and presence to signals andwarrants a decent stereo audio amplifier andspeakers. Some folks report an increase inreadability on weak signals. Personally, Ijust love it! For me, it completelytransforms the listening experience.STEREO BALANCEValues above 100 decrease the rightchannel output; those below 100 decreasethe left channel output. Another scratchypot saved.NOISE BLANKERAs opposed to the Denoiser which acts onwhite noise, this acts on impulseinterference - eg ignition noise, thermostats,electric fences and the like.AUTO NOTCHThis removes an interfering heterodyne -and in many circumstances, several. Itworks best on pure tones and especiallylower pitched ones. It has exactly one use inCW mode, namely for monitoring keyclicks ie what's left after removing the tone.One very popular commercial transceivershows up here every time.

Auto notch is applied after the filter bankand is outside the DSP AGC loop - to avoidstrong-signal overload of the DSP.CW OFFSETThis is your preferred beat note and may bepre-set (when you download from your PC)to 5, 6, 7, 8 or 900Hz. The centre frequencyof the CW filters is changed to match.

If you are interfacing with other than Pic'N' Mix, you will need to adjust your Tx/Rxmixer injection frequency by mode. So, forthe record, the following are the exact DSPIFs (in kHz) used by PIC-A-STAR:-

LSB Tx = Rx = 16.35LSB CW Tx = 16.35, Rx = Tx + CW offsetUSB Tx = Rx = 13.65USB CW Tx = 13.65, Rx = Tx - CW offset

There is also a Reverse CW option and ifset:-LSB CW Tx = 16.35, Rx = Tx - CW offsetUSB CW Tx = 13.65, Rx = Tx + CW offset

SIDETONE FREQUENCYNot to be confused with CW Offset, this isthe tone you hear when sending CW. TheQSK experts tell me it can be useful to havethis at a different pitch from an inboundsignal - so you can intuitively tell thedifference between you and the stationbeing worked when using fast break-in.

To enhance this effect, the sidetonecomes from one speaker only, the inboundsignal from both. The pitch may beexcessively varied between 10Hz and2.54kHz in 10Hz increments - this extendedrange being useful should you need it todouble as an instant audio signal generator.CW TONES 1 OR 2This allows 2-tone testing. The two tonesare 700Hz and 600Hz. If you have not useda 2-tone test for linearity checking before,be aware that the duty cycle is very high -so use only short or pulsed bursts.AGC HANG TIMEThis can be set anywhere between veryshort and very long. I understand that mostpeople usually only change this per QSOfor CW work - and so you can vary the CWsetting without altering the SSB setting. Avalue of 0 turns DSP AGC off.

While actually changing frequency, thehang-time is set to short to avoid annoyinghangs after tuning across large signals.FILTER WIDTHThis allows you to set the receiver filterwidth by turning the knob (or on/off using6 ). There are currently six filtersprovided, three each for SSB and CW(though any one can be used in eithermode). See Figs 32 and 33. The status LEDis tri-colour and corresponds to wide,medium, narrow - or off.

Turning the filter off is a useful way ofchecking if a signal has stoppedtransmitting or has just slipped out of thepass-band, since when you switch the filterback on again, it reverts to the previousfilter width.FILTER DEPTHThis concept was inspired by yet anotherconversation with Bill Carver, W7AAZ. Intraditional analogue terms, it allows acontrolled leak past the filter. For CWoperation it is in many ways more usefulthan controlling the filter width. In use,

0dB

-25

-50

-75

0Hz 500 1000 1500 2000 3000 0Hz 500 1000 1500 2000 3000 0Hz 500 1000 1500 2000 3000-100

160Hz - 3.1kHz 250Hz - 2.7kHz 300Hz - 2.4kHz

Fig 32: Plot of wide, medium and narrow Rx SSB filters. Other filters in the DSP receiver give a further theoretical 30dB of ultimate stop band.The pass-band ripple in all cases is less than 0.2dB. Because STAR is not a mere audio add-on, these widths are actually achievable - and useable.


having tuned a wanted signal to the centreof the pass-band, you simply increase thefilter depth (ie the stop-band rejection) untilthe QRM is reduced to any level you feelcomfortable with. Putting it another way,you come up to periscope depth to find atarget - centre it up - and then go downagain so the nearby destroyers can't get you.VOX AND QSKThe hang time is the duration spent ontransmit after you have stopped speaking(or keying) before PIC-A-STAR reverts toreceive. If you have a relay-free T/Rsystem, then there is no need to set otherthan a very low value here. With relays,any greater setting will minimise thenumber of rattling occasions. It is adjustedto allow the trailing edge of yourtransmission to pass - before switching toreceive.

The Rx-Tx pre-delay is the time yoursignal is delayed by DSP to allow for relaysettling when switching to transmit. It isadjusted so that the leading edge of atransmission is not truncated - just.

The Tx-Rx blank time is the duration ofDSP receiver blanking immediately afterreverting to receive. It should be set to theminimal value possible, consistent with noobjectionable click coming from thereceiver after the transition. With a fullSTAR configuration, this is simply zero.

The above three parameters areseparately set for SSB and CW. For SSBonly, VOX and anti-VOX gains may also beset. See Part 4 for further discussion.TX DRIVEThis is the power setting control. When ontransmit, the 'S' meter reading correspondsto Tx drive level - and is therefore modal.MONITOR LEVELOn SSB this control sets the level at whichyou monitor your own voice - after all VOXprocessing and filtering. If, for example,you turn the Rx-Tx pre-delay up high, youwill - somewhat disconcertingly - hear your

very delayed voice. And you will hearleading or trailing edge truncation if thetiming is not set up properly. Foroperational use, however, the level shouldbe kept low to avoid any feedback; orworse, confusion of the VOX software.

On CW, this control sets the sidetonelevel; and you do not hear a delayed signal -since this would play havoc with yoursending.MIKE GAINIn conjunction with the Mic Gain preset onthe IF board (RV5), this should be set toprovide adequate input to the softwareVOGAD. The latter will hold the audioamplitude at a substantially constant leveleven in moments of excitement.BASS AND TREBLE BOOSTThese act independently to tailor thetransmitted audio profile.RF CLIPPINGI have always found this the most effectiveform of SSB processing - as opposed toaudio compression. It increases the averagepower while holding the peak power steady.In mechanical engineering terms, itincreases your transmitted signal's power-to-weight ratio. So it also increases thestrain on your power supply, linear andATU.

Use it only sparingly and when necessary(and not because it is there), bearing inmind that any form of processing - bydefinition - introduces distortion.CW SPOT LEVELA 'spot' tone (equal in frequency to yourCW offset) may be injected into thereceiver output. As you net onto anincoming CW signal you will hear it beatwith the 'spot' tone - and when they are onthe same frequency, you are indeed netted.

This control alters the minimumamplitude. However the amplitude is alsoincreased automatically with the strength ofthe incoming signal. This is done because it

is easier to beat two notes of similaramplitude - especially when the 'spot' toneamplitude tracks the inbound keying.

FINALLY ….Check with me for the latest changes,please. Project without end, right?

Next month, we start on the front-end, thedeterminant of that rock-solid base receiverperformance

REFERENCES[1] In Practice by Ian White, G3SEKRadCom Feb 2003[2] HF by Don Field, G3XTT RadCom June2003[3] 'A DSP-Based Audio Signal Processor'by Johan Forrer, KC7WW. QEX Sept 1996[4] 'Low-Cost Digital Signal Processing forthe Radio Amateur' by Hershberger, D.QST Sept 1992[5] 'DSP—An Intuitive Approach' byHershberger D. QST, February 1996[6] 'Using the LMS Algorithms for QRMand QRN Reduction' by Reyer SE andHershberger D. QEX, Sept 1992

0dB

-25

-50

-75

0Hz 500 1000 0Hz 500 1000 0Hz 500 1000Fig 33: Plot of the Rx CW filters. This is the bank centred on 600Hz - and those on differentcentre frequencies are otherwise similar. Their widths are approximately 200Hz, 500Hz and750Hz but this absolutely depends on where you measure them. This filter shape gives less ringingthan a brick wall. Other filters in the receive path give a further theoretical 30dB of ultimate stopband.

he next four episodes cover the design andconstruction of the STAR front-end, namelythe mixer/ amplifier and band-pass filters.This development is of general application.

HOW GOOD?When it comes to receiver front-ends, thereare those who would scale the highest IP3mountains; and those who explore thedepths of classic simplicity andminimalism.

Both pursuits are valid and fascinating intheir own right. I am typically to be foundsitting on a fence (at sea level?), aware thatbetter performance is always achievable -but unclear if it is of real operational value.Aware also that the law of diminishingreturns cuts in exponentially when it comesto cost and complexity.

Conversely, faced with the delightfulquality emanating from the STAR DSP, itwould have been remiss not to provide itwith a proportionate front-end. "Howgood?" is, as ever, the question.

I much enjoyed "HF Receiver DynamicRange: How Much Do We Need?" in QEXof May/June 2002 by Peter Chadwick,G3RZP not least because it was written inpractical and tangible terms. See also TT inRadCom of February 2003 for a summary.

I conclude from this paper:-• Phase noise performance is critical

especially when there are many unwantedstrong signals in the pass-band.• The strong-signal Dynamic Range

requirement is not horrendous - if you areprepared to shift the DR up and down tosuit conditions. That is, sometimes youneed good sensitivity; sometimes you needgood strong-signal performance. Rarely, inpractice, can you use both. So I don't thinkI can use all the dynamic range offered by,for example, the CDG2000 design [1].

This argument absolutely assumes youare not blessed with a specific point-sourceproblem such as a strong nearby transmitter.Normally propagated signals are assumed toapply. In real life - in Europe - the abilityto handle 40m at night is the pragmatic test.

Pic 'N' Mix has intrinsically good phasenoise performance. The trick is to avoiddegrading it (eg by adding a crude PLL) toget round the issue of DDS spurs.

The strategy of moving the DR up anddown is equally compatible with the need toextract the maximum possible range from a

finite number of bits in down-stream DSP.STAR achieves this with the ability to

dynamically reconfigure the receiver gaindistribution.


T

C36

100n

C35

100n

R3450R

TR5BC517

C9100n

Balance

C222µ

IC374AC86Squarer

1

2

3

4

5

6

7

14

13

12

11

10

9

8

C7

560p

R8

270RR356R

R2610k

R2710k

R2

56R

C33100n

R2922k

RV1500k

C34

100nLO injectionfromPic 'N' Mix

1

2

3

4

5

6

7

14

13

12

11

10

9

8

C8

100n

R10

2k2

R16

2k2

C19100n

RFC3

IC6 FST3126MRF/IF Switch

1

2

3

4

5

6

7

14

13

12

11

10

8

9

12VTx

IC4FST3126MH-modemixer

+12V

C13100n

T3

T4

T5

RF in/outfrom BPF

IF in/outfrom

IF board

C322µ

C422µ

C11100n

12V

(see text)

KEY:-

wire link

5VIC5 FST3126MRF/IF Switch

1

2

3

4

5

6

7

14

13

12

11

10

8

9

C18100n

C21100n

C12100n

TR1J310

FB

C22100n

R422R

R11

2k2

Rx ModeR12

2k2 D2

1N4148

C11n

C15100n

R182k2

R202k2

R172k2

R192k2

R212k2

R2410k

R91k

C27

100n

c

a

b

d g

h

f

e

T2

1t

ba 4tT1

dc

C20100n

R222k2

R132k2

R232k2

C14100n

Tx

bestIP3

best IP3or Tx

D11N4148

5V

R25

10k

bestNF

R3047k

C31100n

C29100n

C30100n

C28100n

C17

100n

R142k2

C26100n

C5330p

R3350R

L10.66µH

C6330p

L20.66µH

5V

C10100n

D4

1N4148D3

1N4148

R31

47k

R28

10kTR6BC517

R15

2k2

C16100n

IC2LM78L62ACZ

RFC1

TR2J310

FB

C23100n

R522R

TR3J310

FB

C24100n

R622R

TR4J310

FB

C25100n

R722R

12V

5V

5V

5V

5V

R3247k

RFC4

12V to BPF

g ds

Diplexer,see text

6V2IC1ZR78L05C C32

100n

RFC2

R1

56R

R36

22R R35

47R

R37

22R

Fig 34: Dynamically configurable H-mode mixer - and quad J310 amplifier, better known as the'Magic Roundabout'. The 'Rx Mode' line is set to +5V ie logic '1' for best intercept (IP3) - or tologic '0' for best noise figure (NF). 12VTx - when taken to +12V - selects the transmitconfiguration. The mixer requires fundamental frequency injection between -10 and +10dBm.

THE MAGIC ROUNDABOUTWith early STAR, I used a mere SBL-1mixer both with a bi-directional 2N3866post-mix amplifier - and with a bi-directional J310. They both 'work'.

But in 1998 Colin Horrabin, G3SBIshowed that there are no technical excusesfor not using an H-mode mixer [2] - andGiancarlo Moda, I7SWX showed that witha fast bus switch, it can be truly affordable -except - in my view - for those veryexpensive transformers.

However, I can't turn away theopportunity of an H-mode mixer withhome-brew transformers since although thisdegrades the intercept somewhat, there isstill plenty in hand. And no great cost.

In fact there is so much in hand that Icontemplated the heresy of fitting an RFamplifier before the mixer. Or should Isettle for less sensitivity and instead fit anIF amplifier after the mixer? I wanted bothoptions; and after doing the performancearithmetic, I was convinced I needed bothunder different operational circumstances.Typically, on 10m I want the sensitivity, on40m I want the higher intercept.

So, pragmatically, I decided to make itconfigurable - so that I indeed have bothoptions and can compare and contrast themunder differing real-life conditions - at thetouch of a button.

Thus I can switch between 'best NF' and'best IP3' modes to suit the prevailingconditions - with the key related benefitthat I can have enough RF gain so thatDDS spurs are below the band noise.

This approach takes care of T/Rswitching also - and the whole concepthas become known in STAR circles asthe Magic Roundabout.

The circuit diagram is shown in Fig 34and the switching arrangements aresummarised in Fig 35. The switchreferences are common to both.

CIRCUIT DESCRIPTIONThe LO injection is squared up and madesymmetrical by IC3. Critically, thisremoves the even harmonics. RV1 isbest adjusted for minimum DDS spurii on10m. In 'best NF' mode these should be

very hard to find. IC4 is a conventionalfundamental-injection H-mode mixer. Seealso [2].

TR1-TR4 comprise a quad J310 amplifierused either as an RF or IF amplifier. I firstsaw the feedback arrangement in"Introduction to RF Design" by WesHayward, W7ZOI; and the use of multipleFETs (to raise the intercept) in an IFamplifier design by Bill Carver, W7AAZ.

With an 8dB pad (R35-R37) in 'best IP3'mode only, the system gain is essentiallyconstant in either receive mode. Ontransmit, the J310s are always used asan IF amplifier - irrespective of thereceive mode.

IC5 and 6 with TR5 and TR6control the roundabout switching.I7SWX IMPROVEMENTSFig 34 uses the original squarer andmixer from [2] - but in private corres-pondence with Giancarlo in early2003, he suggested two improvements.You may wish to incorporate either orboth. I certainly have and commendthem.

Fig 36 shows changes to the squarerwhich improve the symmetry and'squareness' of the switchingwaveform. Also, IC3 now does notrequire (nor gets hot in the absence of)LO drive.

Some people (including myself)have reported unstable lumps of RFenergy apparently emanating from theoriginal squarer - and with thismodification I have seen / heard nofurther evidence of them.

Giancarlo omitted the balancearrangements; but I prefer to retainthem for the STAR application. TheLO injection level requirement isbetween 0dBm and +10dBm.

Fig 37 shows Giancarlo's two-transformer mixer. This is to bepreferred in principle since when itcomes to improving the mixerintercept, the only really good ferritetransformer is an eliminated one.

REFERENCES[1] The CDG2000 HF Transceiver by ColinHorrabin G3SBI, Dave Roberts G8KBB,George Fare G3OGQ in RadCom June-December 2002[2] Technical Topics, RadCom, Sept 1998

a

MixIF

port

RFport

c

b

d

f

ge

amp

LOinjection

h Pad

Fig 35: Magic Roundabout block diagram.For 'best NF', switches c, g and e are closed.For 'best IP3', switches a, b and h are closedand the pad improves the intercept.For 'transmit', switches d, f and a are closed.

COMPONENT LISTResistors 1206 SMDR1-R3 ............................................................. 56RR4-R7, R36, R37 ........................................... 22RR35 ................................................................. 47RR33, R34 ....each 2 off 100R stacked to give 50RR8 ................................................................. 270RR9 ......................................................................1kR10-R23..........................................................2k2R24-R28..........................................................10kR29 ..................................................................22kR30-R32..........................................................47kRV1 .............................500k multi-turn preset potCapacitorsC1 ................................................. 1n feedthroughC2-C4 ................................ 22µ radial electrolyticC5, C6 ........ 330p ceramic plate (see next month)C7 ........................................... 560p ceramic plateC8-C36 .......................................100n SMD 1206

SemiconductorsD1-D4 ......................................1N4148 or similarIC1 ............. 5V regulator, 150mA eg ZR78L05CIC2 .......................6V2 regulator, LM78L62ACZIC3 .................................................74AC86 SMDIC4-IC6 .. FST3126M (IC4 could be FST3125M)TR1-TR4........................................................J310TR5, 6 ..................................... BC517 DarlingtonInductorsL1, L2............0.66µH on T25-2 (see next month)RFC1-RFC4......6t thin enam wire on type 43 FBFB............... small ferrite bead on J310 drain leadT1-T6 ... wound with 32SWG self-fluxing copper

wire on EPCOS (was Siemens) B62152A4X1binocular core (available from ElectroValue)

T1, T6...........................details follow next monthT2 .................. 4 bifilar turns (wire wound at 5tpi)T3-T5 ............4 trifilar turns (wire wound at 5tpi)

IC178L05

Balance

IC374AC86Squarer

1

2

3

4

5

6

7

14

13

12

11

10

9

8

1

n

510

RR35 6R

R26

10kR2

710k

C33100

n

R29

22kRV1500

k

C34

100

nLO injectionfrom Pic 'N' Mix

C11100

nC32100

n

RFC2

1

n1

n510

R

2 k2

2k

2

+12V

Fig 36: Modified squarer by I7SWX. The changesfrom fig 34 are shown in blue. Please note that neitherthe changed nor the incremental parts are included inthe component list.

R1

56

RR2

56

R

1

2

4

5

6

7

14

13

12

11

10

9

8

C8

100

n

R1

0

2k2

R1

6

2k2

IC4FST3126MH-modemixer

T3

T6

C322µ

KEY:-

wire link

RF port IF port

LO port

C35 100n

3

Fig 37: Two-transformer H-mode mixer by I7SWX.Details of transformer T6 (which replaces T4 and T5)will be provided next month. It is wound on the sameferrite as the other transformers.


ompletion of the Magic Roundaboutconstruction is this month's mission. Referto last month for the circuit and optiondetails. The complete receiver performanceis also discussed here.

ROUNDABOUT CONSTRUCTIONThe PCB component layout and artwork isillustrated in Fig 38. It is again made usingthe iron-on process. The board is double-sided with the underside unetched exceptfor small pads to mount R11 and R12 - andto make off the Rx Mode and 12VTx lines.

Thus the underside of the board providesa ground-plane as well as screening. R11and R12 are mounted in the thickness of thePCB as 'feedthrough resistors'.

There are a number of wire links on theboard for which I apologise. These derivefrom the need to make this board as smallas possible to minimise the risk of pick-upand radiation.

The RF Port requires a DC blockingcapacitor. For STAR, this is fitted on theBPF board. Do not omit it!TRACK OPTIONSThe board is shown tracked for all options.If you are definitely fitting the 2-transformer H-mode mixer, then you shouldremove all the tracking between T4 and T5and T3 (illustrated in blue) - leaving only asmall pad to connect a wire from T6 to theIF port diplexer. To 'remove' the track

before etching, simply fill and join up theground - with an indelible pen.Alternatively, after etching, you can cut thetracks feeding the diplexer - and then bridgeall the unwanted track to ground. The latteris easy and reversible so gives you moreoptions for experimentation.SQUARER OPTIONSThese are shown in blue on fig 38 with thecomponent values taken from fig 36 lastmonth.

This is a classic example where there isabsolutely no way this board could be madecommercially since there is no suitabletrack layout. For one-off purposes,however, the two 1n wire-ended capacitorsfit beautifully - as shown - across the top ofIC3. If you want to retain the originalmixer configuration, simply replace themwith wire-links; omit the 'blue-outlined'components; and revert to the values in fig34 last month.CONSTRUCTION SEQUENCEThe holes in the grounded areas of theboard are for links through to the ground-plane. These should be fitted first andsoldered both sides.

Then mount all the SMD components;then the discrete devices with the exceptionof TR1; then the wire links (ideally usingthin self-fluxing wire) and finally thetransformers.

Note the gap between TR2 and TR3 is togive space for the feedthrough to the BPFboard - via RFC4. Slip a small ferrite beadover each J310 drain lead before soldering.

IF PORT DIPLEXERL1/2, C5/6, R33/34 form a diplexer on themixer IF port. L1, L2, C5 and C6 shouldeach have a reactance of about 50Ω at theIF frequency - using the nearest preferredcapacitor value. The values given are for10.7MHz. L1 and L2 in this case are each14t of 32SWG wire wound on a T25-2 corespread over about 2/3 of the circumference.This derives from an AL value of 34µH per100 turns for this core.

Before fitting, connect each coil inparallel with its resonating capacitor, pass asingle turn through the toroid and looselycouple it to a GDO; and dip it at your IFfrequency.MIXER TRANSFORMERST3, T4 and T5 are illustrated as MCLtransformers and the tracking is appropriateshould you want to use these.

If using the home-brew 3-transformermixer, the EPCOS ferrites mount verticallyas shown for T2.

Mixer balance is critically dependent onthe transformers all being the same. To thisend, make up enough trifilar wire in onelength for T3-T5. Rather than winding oneend of the wire through the corecontinuously, wind alternate ends.

If building the 2-transformer version,then see fig 37 last month for the circuitdiagram and Fig 39 for the mechanicalresult. This may indeed turn out to be thedefinitive test of your understanding of 'thephasing dot convention' but if you followthe steps below you can build it by rote.


C

C2C3

T1

a

c

T4

LO port

+12V

RF port

C1FT

IC2

L2

ct

RxMode

R11

IC6

R12

D1

D2

TR5

RV1

12VTx

IF port

Wirelinks

R8C7R26

R29R3

C10

D3

R10C8

C13

C11C

32

R16R2

R28

TR6

C34 IC1

R23

C17

R13

C30

C14R15

R17

R18

R24 C16

C18

C26

R9

R30

R22

C21 b

R34

C5

C6

T2C15

RFC

1

RFC

3

C29

R31

R14

R25

TR4TR3TR2

R32

C4

C19

D4

PCB dims 3" x 2.5" 76.2 x 63.5mm

NB

This image is flipped

RFC

4

R1

2k22k2

C33

R27

510 1n1n

IC3

IC4

L1

+12V Feed to BPF

C9

R19

R20 C20

C31

IC5

C28

C12

C27C36R35

R36

R37

T5

RFC2

R21

TR1

R4

C22

R5

C23

R6

C24

R7

C25

d

R33T3 C35

Fig 38: 'Magic Roundabout' component layout and PCB artwork. This board is designed to mate mechanically with the BPF assembly described nextmonth. Components shown in blue are for the improved squarer. Track shown in blue may optionally be removed (see text) for the 2-transformer H-mode mixer.


T6 is wound with 5 parallel strands ofself-fluxing wire - untwisted. Cut 5 lengthsof wire to some 20cm long. Solder one endof all of them together to retain them. Thentrim them all to exactly the same length andsolder the other ends together.

Now wind 4 turns on the core undermodest and continuous tension - passingalternate ends through the core.

Cut off the surplus wire equally andinitially to approximately 5cm. Tin all 10ends. Using a continuity meter, locate onepair and twist them to form the IF port feed.That was easy!

Now locate 2 more 'pairs' and cross-connect a start/end (and twist themtogether) to form the centre-tap. Repeat forthe remaining 2 pairs.

Locate T6 on top of IC4 and trim the 4leads that are made off to the track to thesame length. Those going to pins 6 and 8pads define that length. Tin and solderthem to their pads as per fig 39.

Then make off the IF port leads to theend of the diplexer.

Now wind the trifilar transformer T3.Solder the centre-tap to the track first, thenthe RF port feed. Finally trim off the flyingleads to T6 equally; and as short asreasonably possible; and solder them up.J310 TRANSFORMERST2 is a conventional bifilar transformer withno complications. The input transformer T1needs a little explanation. See Fig 40.

The T1 feedback turn is madefrom a length of miniature coaxbraid. Form the U-shapedprimary turn - with plenty ofexcess lead length. Pierce andspread the braid to fill thetubular holes in the core - fromboth ends of the core. Make offthe braid leads to the PCB track.

Wind the 4 turns - inside thebraid - out the back of the core,across and back through thebraid on the other side - andsolder to the track. As far aspossible, then tease the braidback round the turns on the non-lead end (as W4ZCB says, so

nobody can see how it was done).TR1 may now be fitted.

INCREMENTAL TESTINGDepending on your personal style andconfidence, you may want to functionallytest this board progressively. By lifting oneend of some RF chokes you can selectivelyenable parts of the circuit. 100n wire-endedcapacitors should be used to couple RF inand out of each circuit element under test.

You can test the J310 amplifier is indeedamplifying by placing it in the down-lead ofsome receiver. Equally the squarer andmixer - with suitable injection - can betested as a crude converter.SYSTEM INTERFACEYou could control the Roundabout with asimple switch and a status LED. Arguably,you could do worse than driving it from theSSB select lines, ie 'best NF' on the USBbands, 'best IP3' on the LSB bands. But60m is exceptional and complete userchoice is desirable until you see how itworks for you. It depends not least on whatantennas you have. I control it from the'spare' output on Pic 'N' Mix - with a 1k5resistor in series with a tell-tale LED acrossthe line. On STAR, this line is toggled bykeypad sequence 48.FINISHING OFFWhen all is working well, trim RV1 tominimise DDS spurs on a high band, eg10m in 'best NF' mode. When connected toa dummy load, you may still be able to hearsome. But when connected to even amodest antenna, then although the bandmay be essentially closed, the ambient bandnoise should mask them to the point thatyou will need to try very hard to find them.

Finally, enclose and screen the wholeboard and re-trim RV1.

STAR PERFORMANCEThese performance measurements weremade by Harold, W4ZCB usingprofessional test equipment. They werecorroborated by myself using both testequipment borrowed from I7SWX - and myown. My thanks to both because these arevery important numbers.

All measurements relate to the completeSTAR line-up - ie including the band-passfilters which follow next month. Thesehave, by design, decreasing insertion losswith increasing frequency. So to get thecomplete picture you need to consider theperformance on each band. Fourrepresentative bands are shown, the restbeing somewhere 'in between'. Since thedesign features a 'best IP3' or 'best NF'mode, that adds a further dimension.

The mixer is Giancarlo's 2-transformertopology driven by the modified 74AC86squarer. MDS was measured in a 3kHzbandwidth; IP3 at 20kHz tone spacing. 'BEST IP3' MODE

80m MDS -123dBm IP3 +33dBm40m MDS -122dBm IP3 +35dBm20m MDS -124dBm IP3 +31dBm10m MDS -127dBm IP3 +28dBm

'BEST NF' MODE80m MDS -127dBm IP3 +30dBm40m MDS -123dBm IP3 +30dBm20m MDS -127dBm IP3 +28dBm10m MDS -130dBm IP3 +25dBm

AGC RANGEAGC holds the audio output constant within1dB for a 100dB change of signal. You canplace this range anywhere on the amplitudescale by adjusting the RF Gain, but from-95dBm to +5dBm would be typical.OBSERVATIONSNote that excess sensitivity is not providedon the lower bands where it could never beused. Instead, it is traded for superiorstrong-signal performance. For comparisonwith commercial transceivers see TT ofDecember 2002 - Table 1, the IP3 columnin particular. On a stormy 40m night, whichwould you rather own? And much more tothe point, be proud to own!

d

a b c

(a) View fromconnectionsend

(b) View fromnon-connectionsend

Finally tease this braid back around the wire

Fig 40: Transformer T1 detail. The paddesignations correspond to those of fig 34.

The Magic Roundabout of Alan, G3TIE

IC4

IF Port(L1 & R33)

RF Port (C12)

3

6 8

11

T6

T3

100n

Fig 39: Details of T6 / T3 connections to eachother and to IC4. This is not strictly to scale,but does show the correct relative positioningof the components. T6 and T3 will end upseparated by about 6mm in practice.


he next two episodes cover thedesign and construction of a band-pass filter - for 10 bands - which iscompatible with the rest of STAR. I

felt a need to improve on my 3rd Methodfront-end (and add 60m) - and could findnothing suitable in the literature that met myrequirements.

This development is of general interestand application.

DESIGN AIMSLike everyone else, I think I want narrowfilters with no insertion loss, superb IP3performance - and acceptable cost. Idefinitely want a finished size that does notimpact the overall dimensions of mytransceiver.

For a fact, you can't simultaneouslyoptimise all these parameters, so this design- like all others - is a careful compromise.

The prime function of any BPF is toadequately reject the image frequency - andthis is achieved by this design (with a littlehelp from an ATU and / or your LPF on thehighest bands) provided your IF is 9MHz orhigher - and you use high-side injection.

You could use different filter topologies.The mechanical construction is for threeinductors (see Fig 41). That is the onlypractical constraint. The filter capacitorsare soldered directly to the coil terminalsand there is plenty of room for lots of themin different configurations.

PERFORMANCE REQUIREMENTSSince on the higher bands Noise Figure iseverything I need low insertion loss aboveall else. Say 1.5dB. The consequence iswider filters which have the significantbenefit of spanning more than one band.

The other great benefit - and integral tothe whole front-end design strategy - is thatI need the highest possible signal levelgoing into the mixer so that any DDS spuriiare below this level on the higher bands.

As you move down in frequency, NoiseFigure becomes increasingly unimportantand greater insertion loss is a positivebenefit, adding directly to the mixerintercept. It also helps to keep the poweroutput flat on transmit if like me, it tends todrop off at the higher frequencies. Phrasedbetter, I don't drop off, but the power does.

COST CONSIDERATIONSOn the cost front, diode switching is thecheapest - but unacceptable for strong-signal performance. Good relays (and youwould not want to use bad ones) are veryexpensive given the quantity involved. Isettled on integrated bus switches and usethe FST3126. These are a close relative ofthe FST3125 as typically used in the H-mode mixer, the difference being that theswitch control logic is inverted andcompatible with direct drive from the Pic 'N'Mix band-switching latches.

As I have configured them, the ONinsertion loss is less than I can measure atwell under 0.5dB; the OFF isolation isbetter than 90dB; and IP3 is better than40dB. They are very inexpensive.

I settled for Toko coils because they arereadily available and also inexpensive - butI used the larger cup-core inductors wherepossible ie on all bands up to and including20m.

As a gross alternative, you could usefixed torroidal inductors and trimmercapacitors if the Toko coil Q or IP3performance are issues for you.

CIRCUITDESCRIPTIONThe filters (see Fig42) are all 0.01dBChebychev designs.They are switched byidentical switches ateach end (IC3 andIC4). For eachswitch, two sectionsare paralleled toreduce 'on' insertionloss, one sectioninverts the controllogic and one sectiongrounds the filterwhen 'off'. For the

T/R switch (IC1), three sections areparalleled on receive and one is used ontransmit.

L1 and C16 form a series-tuned IF trapand C16 should be chosen to resonate withL1 at your chosen IF frequency.

A spare filter position is provided forexperimental purposes eg differenttopologies, different frequencies.

The filter capacitor exact theoreticalvalues are given in fig 42 and the nearer youcan get to them, the better. I obtained alarge bag of assorted polystyrene capacitorsand arrived at the values to within 1pf (asmeasured on my DVM) with never morethan two in parallel - by measuring theactual values within the tolerance range.

COMPONENT LIST

FOR THE OVERALL BPF ASSEMBLY:-Resistors 1206 SMDR7, R9 .............................................................1kR8 ............................................................0R linkR10, R11 .......................................................2k2CapacitorsC7-C13 ..................................... 100n 1206 SMDC14 ..............................................1n feedthroughC15 .........................................100n disc ceramicC16 ..........................................................see textInductorsRFC1, RFC2......................... 100µH axial chokeL1 ........................Toko BTKANS-9445HM 3µ3Integrated circuitsIC1.................................................... FST3126MIC2............................................. 78L05 regulator

PER BPF FILTER BLOCK:-Resistors, 1206 SMDR1-R6, R12 ...................................................10kCapacitorsC1-C6 ....................................... 100n 1206 SMDFilter capacitors ..........Polystyrene or silver mica

for values, see fig 42.Inductors for filtersall Toko coils, 3 off160m........................................154ANS-T1017Z80m..........................................154ANS-T1012Z60m..........................................154ANS-T1014Z40m..........................................154ANS-T1014Z30m......................154ANS-T1012Z but see text20m..........................................154ANS-T1007Z17m/15m .................................. TKAN-9448HM12m/10m ............................. BTKANS-9450HMIntegrated circuitsIC3, IC4 ............................................ FST3126MMiscellaneous .................small ferrite bead, 2 off

For example, the 40m and 20m blocksrequire 6 capacitors of near 20pf. The exactvalues were found from a small selection of20pf and 22pf 5% capacitors.


T

IC3

Tokocoi l

Tokocoi l

Tokocoi l

Coi l mount ing tabsNB adjacent tabsshare a common hole

Brass shim screen

RF bus-bar wi reconnected wi ththrough-board l inks

IC4

Brass shim screenmakes spr ing contacton bot tom cover p late

Single-s ided PCB,copper th is s ide

Fig 41: Cross-section of band-pass filter enclosure. The main board issingle-sided with critical incremental screening on the top provided bybrass shim. The same material is used to form a central spine shieldrunning up the middle of the filters in the lower compartment. It isbent over to make spring contact with a bottom cover plate - madefrom PCB stock.


PERFORMANCE SUMMARYUp to 20m the filters give >100dB imagerejection. By 12m/10m this has fallen tosome 50dB so you could benefit from someincremental filtering provided by low-passfilters or an ATU or even a beam.

Insertion loss is around 5dB on the lower

bands, falling to 3dB on 20m, 1.6dB on17m, 1dB on 15m, 1.5dB on 12m and adelightful 1dB at 28MHz rising to a mere1.3dB at 29.7MHz.

30m is exceptional because of theproximity to my IF. Here a 5dB in-bandinsertion loss rises to 18dB at 10.7MHz and

55dB at 9MHz. Depending on your mixerbalance and trap tuning, this could bemarginal with a 10.7MHz IF.

My compromises, your call!

C 8

100n

IC4FST3126

IC1FST3126

IC3FST3126

IC2 78L05

R 5

10k

Rx in

C14 1n

3 2

1

11 12

13

5 6

4

8 9

10

C 3100n

R 210k

R 110k

+5V14C 2

100n

7

C 1100n

F B

C a

C b

Cc

C d

C e

R 410k

R 310k

2

6

12

9

R 6

10k

5

8

1

10

C 4100n7

3

13

11

4

C 6100n

+5VC 5100n

La Lb Lc

2

9

12

5

8

6

1

4

7

3

13

11

10

C 7100n

R 7

1k

R 8

0 R

C11100n

C10100n

14

14+5V

C12100n

C 9

100n

Tx out

End p late

C16

see text

L1 3 .3µH

Filter block - 160mLa, Lb, Lc 50.1µHCa, Ce 150pfCc 162pfCb, Cd 2080pf

Filter block - 80m La, Lb, Lc 16.7µH Ca, Ce 127pf Cc 144pf Cb, Cd 1083pf

F B

Filter block - 60m La, Lb, Lc 25µH Ca, Ce 39pf Cc 41.5pf Cb, Cd 760pf

Filter block - 40m La, Lb, Lc 25µH Ca, Ce 21pf Cc 22pf Cb, Cd 565pf

Filter block - 30m La, Lb, Lc 16.67µH Ca, Ce 15.5pf Cc 16.1pf Cb, Cd 391pf

Filter block - 20m La, Lb, Lc 7.15µH Ca, Ce 19pf Cc 21pf Cb, Cd 282pf

Filter block - 17/15m La, Lb, Lc 1.67µH Ca, Ce 44pf Cc 58pf Cb, Cd 193pf

Filter block - 12/10m La, Lb, Lc 1.25µH Ca, Ce 32pf Cc 41pf Cb, Cd 141pf

Filter block - spare

R 9

1k

R10 2k2 R11 2k2

+12VRx +12VTx

IF trap T/R switchSide plate Side plate

R12each10k

C15100n

R F C 1

100µH

R F C 2

100µH

C13100n

160m 80m 60m 40m 30m 20m 17/15m 12/10m spare +12VTo / f rom mixer

End p late

Fig 42: Band-pass filters circuit diagram and mechanical overview. All the filters have identical circuit diagrams. A band-select line is set to +5V toengage a filter block and to 0v to isolate it. This is compatible with direct interfacing to the Pic 'N' Mix band-select outputs. These lines aredeliberately routed around the side-plates and not across the filter. For illustration, the 160m filter is shown 'on' and the T/R switch is 'on receive'.


his month concludes the band-passfilters with the layout in Fig 43 andthe PCB artwork in Fig 44. There

are no special constructional issues.

FILTER ADJUSTMENTIn the first instance, put the entire filterassembly in series with the antenna of anexisting receiver and check that all the filterswitches work and the coils peak.

There are sophisticated ways of tuningthese filters, but if all coils are peaked mid-band and then the two end coils are peakedat the two band edges, you will not be farout. A refinement of this is to put Pic 'N'Mix in wobbulator mode [58] across thesegment of interest; and while on low powerCW, transmit into dummy load and tweakfor a flat pass-band.

ACKNOWLEDGEMENTSThese filters were designed by Harold,W4ZCB to whom I am much indebted. Hein turn credits ELSIE, a filter designpackage by Jim Tonne, WB6BLD - whichdid all the sums. I have subsequentlybecome a fan of this software. Delightful!

Harold also independently measured theswitching performance.


T

C 2

R1

C 3

R2

C1

CeC

d

La

C bLb

Lc

Ca Cc R4 R

3

C 5

C 4

C6

R 9C10

C12

C 9

R8

R 7C 7

C8

C11

R 6IC4

IC3

IC1

F B

F B

F B

F B

F B

F B

L1

R10 R11

Rxin

Txout

+12VRx +12VTx

+12V

IC2

C14(feedthrough)

C13

R F C 2R F C 1

C15

To/from mixer

R 5

End-plate

Side-plateSide-plate Filter board

End-plate

Location of filterboard on sideand end plates.

R12(one per band

select line)

RF in/out bus-barsin red (on opposite side of board)

C 2

R1

C 3

R2

C1

Cd

La

C bLb

Lc

R4 R

3

C 5

C 4

C6

R 6IC4

IC3R 5

C 2

R1

C 3

R2

C1

Cd

La

C b Lb

Lc

R4 R

3

C 5

C 4

C6

R 6IC4

IC3R 5 Ca Cc

Ce

Ca Cc

Ce

these resistors are fittedthrough - and in the

thickness of - the board

these resistors are fittedthrough - and in thethickness of - the board

C16

Screen part i t ion

NB this section of trackis grounded bysoldering to end-plate

Fig 43: Component layout for band-pass filters. Three filters are shown, the remainder being constructionally identical. More space has beenallocated to the 160m and 80m filters to accommodate larger filter capacitors. The components are all mounted on the track-side except for the Tokocoils, the RF in / out bus-bars (22SWG tinned wire) and C15. The bus-bars are connected to the IC switches with through-board wire links. Notethat the middle Toko coil is rotated by 180°° relative to the outer two. Unused coil pins are cut back so they do not appear on the opposite side.


PCB dims:-Filter board 2.90" x 5.60"Side plates 1.65" x 5.60"End plates 3.00" x 1.65"Complete assembly:-3" x 1.65" x 5.757"assuming 2mm thick PCB

NB This image is mirrored

Drill theseholes for tight

fit on 1206resistors

Drill several holes (notshown) through each side-and end-plate - and intimatelyconnect the inner and outergrounded faces.

Fig 44: Band-pass filters PCB layout for 9 filter blocks. This assembles into an H-section brick (for the want of a better term). The filter board issingle-sided - and SRBP if you want to save on drill bits; the side and end-plates are double-sided. Ensure opposite sides of all the double-sidedboards are intimately connected. The outside faces of both end-plates have simple oval pads to make off the feedthrough resistors to the band-selectlines at one end - and the 12VTx and Rx lines at the other. The artwork for this is not provided since some elementary removal of the masking sprayafter drilling and prior to etching will achieve the desired and non-critical result.


his is the concluding episode of thearticle - but the project lives on. Soif you are awaiting the end before

you start, this is your beginning! In summary, the hardware design feels

completely stable so I have more timeavailable now to support other builders -and to continuously enhance the software.Which remains the objective.

MODULARITY IS EVERYTHINGFrom the outset this project has beenmodular by design. Both in the sense thatseveral people have substituted differentblocks of hardware and software in theirSTAR; and conversely have used STARblocks in other applications.

To summarise the whole article with anemphasis on the modularity dimension:-

Part 1 - a discussion of STARhardware integration possibilities.

Part 2 - a discussion of softwareoptions and flexibility.

Part 3 - a generalised process forproducing precision one-off PCBs.

Part 4 - a T/R changeover timer thatwould suit any Tx/Rx.

Parts 5 & 6 - details of a DSPprocessor assembly that could be the basisfor any DSP project. With modulardaughter boards to allow future upgrades.

Part 7 - a completely repeatable gener-alised process for mounting SMD ICs withlots of closely spaced (eg 0.5mm) pins.

Parts 8 & 9 - a bi-directional IF stripthat could be readily adapted for use in anyhome-brew design.

Part 10 - a PC based loader andcontroller that could be adapted for anyAnalog Devices 218x DSP project.

Part 11 - one of a number of possiblephysical implementations. A glance at thephoto shows you that every builderexercised completely different options here.There are no two the same.

Parts 12 & 13 - an adapter to replacean 18-pin PIC to give greater I/O and moreprocessing capability generally. And abargraph 'S'-meter.

Part 14 - a spur reduction filter for anyDDS. And a stable reference oscillator.

Part 15 - a useful DSP shopping list, atthe least. Check out the competition!

Parts 16 & 17 - a general purpose bi-directional mixer/amplifier with config-urable topology. Of the strong and silenttype.

Parts 18 & 19 - a universal front-endthat could drop into almost any existing HFTx/Rx.

So you don't have to be building a STARto find something of interest here.

Equally and oppositely, you don't have tobuild it all to benefit from STAR DSP.

But conversely, if you just want to buildan error-free Rx/Tx design that worksbeautifully, now is the moment.

HALF A MILLION POUNDS?Yes, at the most conservative consultancyrate, this is the cost to date of mydevelopment time. That of the Beta teamwas not recorded and I would estimate it tobe about the same again.

For your interest, this splits downapproximately 15% on hardware develop-ment, 65% on software development, 10%on creating and maintaining the projectdocumentation - including this article - and10% on offering explicit help to others.

By comparison, the materials cost iscompletely buried in the noise - andcertainly significantly less than the price ofa new commercial Tx/Rx. Strangely, thecost of ink-jet ink was the single largest lineitem for me. Horrendous!

WHAT DOES THIS DO FOR ME?This 'time to materials ratio' is about asgood as you can get in any hobby since thetime is in fact free - and yet spending it isalso the source of all the pleasure, rewardand satisfaction.

At the very least when I am having aQSO, I know that it is truly me having thesatisfaction of that QSO - and not merelymy acting as the surrogate operator onbehalf of some RF design engineer in a far-away land. If you have never experiencedthe difference, you have missed out on oneof the unique thrills of the hobby.

And you get that pleasure with everysingle QSO.

Part 20 AND LAST by Peter Rhodes, BSc, G3XJP PIC-A-STAR SOFTWARE TRANSMITTER AND RECEIVER

T

TheConstellationBeta.

Starry-eyedand leglessafter theirmuch-acclaimedperformancein theaccordionbandcompetition.


WHAT DOES THIS DO FOR YOU?Well, all the fruits of this development areavailable to you at no charge provided onlythat they are for your personal use.

To obtain all the software - including thesource code - follow the process given inPart 2. For all the PCB artwork, anyenhancements and all ongoing support,simply join the Group (see WWW).

You don't need to understand DSP. Atleast not to use my STAR code and get yourTx/Rx going. Thereafter, it is entirely bothup to you - and down to you.

DIY DSP FILTERSThis is the last technical topic I want totouch on - and that, briefly. There is notmuch prescriptive that can be said here.The trick is to search the web for a programthat will generate coefficients for FIRfilters. From time to time these areavailable in the public domain. Thesecoefficients then need converting to aformat suitable for loading.

"Experimental Methods in RF Design" byHayward, Campbell and Larkin is also auseful source of understanding - pages10.13 to 10.19 in particular. This tells youhow to do it - and provides the software onthe CD to let you.

I suggest the best way to prove theprocess in the first place is to see if you canbuild the existing medium-width SSB filter(FL5), plug it in and prove to yourself thatyours is no different.

To this end, all the Rx filters arepackaged individually and discretely. Thespecification for the existing FL5 is:-

Sample rate = 8kHzRemez equiripple - but not mandatoryOrder = 198, ie 199 coefficientsPassband = 320Hz - 2284HzLower stop = 177HzUpper stop = 2400HzOnce you have replicated this and proved

the process, you can rapidly produce filtersfor any specialised applications eg RTTY.

STAR BUILD SEQUENCEMost receiver designs start from the antennaand logically follow the signal flow to theloudspeaker. This project has taken (moreor less) the opposite approach. This is toencourage you to build the more tricky bitsfirst.

I think it is a better strategy anyway sinceonce you have some sort of noise comingout of the speakers, you can work backtowards the antenna using the completedelements to test the new build. I commendit to you.

Once you have the STAR Rx working,then the few incremental components to getthe Tx going can be taken from any of themany HF designs. You need a driver, PAand low-pass filters - commensurate withthe amount of power you want to run.

CORRECTIONSAt the time of writing there are no knownunpublished errors in this article - andhence to the hardware build. But assomebody famous once said, we still awaitthe unknown known errors.

This is down to the diligence of the Betateam. And to RadCom which - on thisproject - published the original masters ofthe engineering drawings (Figs 1-44) fromwhich we all built.

All of these published drawings werehand-crafted using a drawing package withno engineering intelligence.

This is because I have never foundaffordable layout software that can do abetter job than I can - and I enjoy thechallenge. But this approach does increasethe risk of errors - and it is a credit to theteam that we have not had a singlesignificant one. Yet! In any event, for thelatest state of play, please see WWW.

On the software side, with some 4,000lines of DSP assembler and about the sameagain for the PIC code there have been oneor two interesting learning opportunities.But the great logistics attraction here is thatyour updated radio is never more than an e-mail attachment away.

ACKNOWLEDGEMENTSThere are lots of these since this was andcontinues to be a truly collaborative andinternational effort. In summary and in nospecial order:-

THE ORIGINAL BETA TEAMThe photograph shows the original teamthat still today builds, evaluates, tests andcontinuously suggests improvements - forboth the hardware and software.

Left to right; top row:- Alan G3TIE,Harry G3NHR, Les GW3PEX, PeterG3XJS. Bottom row:- David G4HMC,Eddie G0SEY, Peter G3XJP, Bill W7AAZand Harold W4ZCB with Harold's STAR.

INFRASTRUCTURE & UTILITIESMy thanks to David Tait for his latest andgreatest TOPIC (PIC programmingsoftware). Jim Tonne WB6BLD for ELSIE(filter design software). Analog Devices fortheir DSP utilities and code fragments.

INSPIRATION & ACTUAL HELPLee G3SEW for much useful discussion onthe UI in the early days. Gian I7SWX foruse of his test equipment and mixer design.Bill W7AAZ for many, many ideas andencouragement. Harold W4ZCB for thedesign of the front-end filters, for muchperformance evaluation - and hisunwavering enthusiasm. (It took him 3weeks to work DXCC with his new STAR).

Keith G3OHN, PaulG0OER, Mike G3XYG, JimG3ZQC, Michel ON4MJ,John G6AK - for muchbuilding and testing and thebenefit of their diverse skills

and wide-ranging experience.Fran for the proof reading. She is starting

to lose her value here - since she isdeveloping a dangerous understanding!

George Brown M5ACN, the TechnicalEditor of RadCom for steering all this intoprint with his oft-tested but never-phasedsense of humour.

And last, but by no means least, ourthanks to Bob Larkin W7PUA for sharinghis original DSP-10 work, the DSP chaptersin "Experimental Methods in RF Design",the adaptation of the STAR boot code, hisadvice and suggestions - and the ultimateinspiration for the whole STAR project.

THE LINEARITY CONUNDRUMWhy does a STAR sound so good both onreceive and transmit? There can only everbe one answer. It is because it is linear.

Because STAR is an IF processor it has abuilt-in head start over any AF add-on. Thelatter may indeed be better than nothing, butif the DSP has neither control of the AGCnor of the detection process then thedamage has already been done - so to speak- before there is any chance to benefit fromsubsequent DSP.

Thereafter, most of the 'star quality'derives from the inherent nature of digital(as opposed to linear analogue) signalprocessing.

With analogue processing the quality isdetermined ultimately by device linearity.Any non-linearity - and all devices havesome - results in intermodulation distortionproducts which can fall within the pass-band. And to a varying extent they grate onthe human ear - which is particularlysensitive to their presence.

With digital processing the nature of anydistortion is quite different. It arises fromrounding errors, quantisation errors andlack of arithmetic precision generally.

So long as the system is not grosslyoverloaded (in which case it would fallapart in a big way) these errors do not resultin discrete in-band IMD products.

Rather they result in a general noise floor- of trivial amplitude. And in any practicalHF radio communication system this noiseis indistinguishable from and is buried wellbelow the band noise. There are simply noconventional IMD products to hear.

So the answer to the conundrum is this -and it may not be instinctive. Digital signalprocessing is inherently more linear …..than linear analogue processing.

This is the technical rationale for the PIC-A-STAR project. And the ultimate rationalefor the project? You may have guessed itby now. Promoting the interests of amateurradio.

WWW join the interest Group athttp://uk.groups.yahoo.com/group/picastar/

pic-a-star - · pdf filetrol the transceiver - and a basic utility is ... pic-/ pic-a-star...

Documents