development of a preamplifier for 1 to 1.7ghz with a noise ... · the current state of development...

This article is a revised and expandedversion of a lecture given at the 2012VHF meeting in Bensheim. It showsthe current state of development of lownoise MMIC amplifiers and a sampledemonstrating the procedure for a suc-cessful design. Measurements of theprototype should confirm the designand simulations.

1.

Overview

The development of high quality integra-ted microwave amplifier modules(MMICs) is constantly in flux, leading tobetter performance. Therefore it is intere-sting to build such an LNA because ofthe following benefits:

• Input and output are optimised fora 50Ω system

• The noise figures have fallen tounder 0.5dB in a 50Ω system athigher frequencies (e.g. above500MHz)

• A single device achieves approxi-mately 20dB gain up to 2GHz.

• Little external circuitry necessary.

The sentence "No roses without thorns"applies here. You should know the probl-ems because you can only achieve these

values when you solve a whole range oftasks:

• These MMICs are tiny SMD comp-onents without connecting feet. Adie is used with the dimensions of2mm x 2mm and with 4 connecti-ons on two opposite edges (Pads)plus ground.

• The common ground of the chip isoften not fed to the edge of thepackage but in the middle of theunderside

• The layout design requires veryhigh accuracy (tracks and connect-ion pads on the IC are typically0.2mm with a maximum width of0.5mm)

• The additional SMD componentsfor the external circuit should be0603 size (1.25 x 0.75mm) as far aspossible

The cutoff frequencies of the componentsare so high that the stability below 1GHzand unconditionally up to 10GHz mustbe controlled. The operating point mustbe very carefully stabilised due to thehigh currents (often over 50mA per IC)and the supply voltages decoupled evenmore careful and for extremely broadb-and. The thickness of the board wasreduced, for all applications due to thisextended frequency range, to 0.25mm toprevent undesirable signal modes on thestrip lines. This means that many vias are

Gunthard Kraus, DG8GB

Development of a preamplifierfor 1 to 1.7GHz with a noisefigure of 0.4dB

VHF COMMUNICATIONS 2/2013

90

necessary so it cannot be made manuallywith silver plated tubular rivets.

This project is for use in the 23 cm band(approximately 1300MHz) and intendedto receive METEOSAT at 1691MHz.The following data was used to trawl theInternet for an MMIC suitable for devel-opment:

• Noise figure: maximum 0.4dB • Amplification (S21) approximately

20dB • Absolute stability (k greater than 1

up to 10GHz)

The choice was Agilent Avago, whosesourcing proved absolutely no problem(die components took one week to arriveafter ordering online at Mouser.com) andthe price (€80 for 25 pieces) for the"MGA 635P8" was acceptable.

2.

The development of the circuit

The starting point was the data sheet forthe MMIC together with the downloaded

S parameters and two application notesprovided by the manufacturer for 2500MHz and 3500MHz [1] [2] [3]. Thesame basic amplifier circuit and the samePCB layout are used in all cases. Onlythe component values are adjusted to thedifferent frequency ranges. Fig 1 showsthis circuit and it is not hard to underst-and:

• It uses a supply voltage of VDD =+5V

• Inside the MMIC package is aGaAs pHEMT Cascode amplifier,as well as a bias circuit.

• The operating point is set to thedesired 55mA using pin 1 to conn-ect a resistor (3.6kΩ = Rbias) to thebias circuit and the generated biasis connected to the gate of the firstpHEMT on pin 2 through L1 =6.8nH.

• The other inductor (L2 = 6.8nH)provides the load resistance of thesecond stage.

A major problem of the HEMT compon-ents is stability at low frequencies, theytend to oscillate. A simple trick helps todeal with this: with decreasing frequency

Fig 1: This circuitis used as a patternin the data sheetand two applicat-ion notes; It is thestarting point formy owndevelopment.


91

the additional resistor R1 of about 50Ω isintroduced at the input pin 2 that effectiv-ely prevents such an oscillation.

Explanation:

The reactance of L1 (6.8nH) decreaseswith decreasing frequency but the reacta-nce of C3 (10pF) increases. So at somefrequency only R1 is active on the inputpin.

In addition the very small inductance L26.8nH as a load resistance for the secondstage reduces amplification at lower freq-uencies down to zero. The resistor R2also shown in the circuit with a value of

zero ohm was ignored.

Before the development began this circ-uit was investigated with ANSOFT Desi-gner SV, Fig 2 shows the created simula-tion diagram.

First, the noise figure in dB was simula-ted and at the same time the influence ofthe quality of L1 = 6.8nH was determi-ned (see Fig 3). In both cases studied theNF remains below 0.6dB. However, thenoise figure still needs to be reduced forour purposes in the desired frequencyrange from 1.3 to 1.7GHz to meet therequirements.

Fig 2: The circuitfrom Fig 1prepared forsimulation withANSOFT DesignerSV.

Fig 3: The noisefigure achievedfrom the circuit inFig 2 is unsatisfact-ory and dependsheavily on thequality factor Q ofthe choke L1 =6.8nH.


92

Fig 5: The Sparameters lookvery promising sothere is hope.

Fig 6: This is the simulation circuit after some work (see text).

Fig 4: Because thedeveloper was notcarefully enough,this circuit willoscillate.


93

The stability was investigated up to10GHz and it is evident that this must beimproved (Fig 4). This is only undercontrol up to 8GHz from the simulation.

The S parameters look good - S21 isshown in Fig 5 and it is greater than20dB between 1GHz and nearly 2GHz.

The circuit was adapted to the frequencyrange from 1 to 1.7GHz (Fig 6). This wascarried out by gradually increasing of thevalues for L1 and L2 in accordance withthe simulation results for the noise figureand permanent control of stability.

The desired value of 0.4dB was achievedwith L1 = 33nH / L2 = 15nH and a small

additional resistance of 10Ω in the outputrevealed sufficient stability at 9 to10GHz (fitted close to the output pin ofthe MMIC). The amplification, S21, dro-pped a bit as a result but not below therequirement of 20dB at 1.7GHz. The twocoupling capacitors C1 and C2 on theinput and output were reduced to raisethe lower frequency limit. The outputmicrostrip line (more correct: "GroundedCoplanar Waveguide" with a track widthof 0.59mm, a gap distance of 1mm oneach side and a length of 30mm) couldnot be missed in the simulation and thatresulted in the final simulation diagram.The noise data achieved is shown in Fig

Fig 7: The result ofthe simulation tocontrol the noisefigure is almost adream.

Fig 8: And it lookslike the stabilityfactor "k" gives areally stablecircuit: significan-tly greater than 1up to 10GHz.


94

7 as a result of the simulation.

The required stability (k greater than 1 to10GHz) is now no longer an issue, asshown in Fig 8. The simulated S parame-ters also give no cause for concern (Fig9).

The practical circuit has been modifiedslightly compared to that proposed byAgilent. Fig 10 shows the final version.

The circuit board design could now bedone (material: Rogers RO4350B, 35µmcopper coated on both sides, board thick-

ness = 10mil = 0.254mm). Overall dime-nsions are 30mm x 50mm as shown inFig 11. There is a short microstrip on theleft up to the DC isolating capacitorbefore connecting to the input of theMMIC. The output stripline is 30mmlong and both lines are designed as"Grounded Coplanar Waveguides" with atrack width of 0.59mm and a gap of1mm. The line calculator integrated freeof charge in the ANSOFT Designer SVsoftware provided these values forZ=50Ω.

Fig 9: The Sparameters havenot suffered muchdue to the changes.

Fig 10: This circuitdiagram wasdeveloped for thepractical circuit.


95

The central ground connection on thebottom of the MMIC required its own0.6mm wide ground island with 6 viasthat can be seen in Fig 11. All otherground islands on the circuit board haveenough carefully separated vias. Forthose who do not heed these rules withthe separate islands including many vias,the circuit will very quickly oscillate.Note: all vias have a diameter of only0.3mm

3.

The prototype

The worry with homemade PCBs is that

the silver plated tubular rivets fall outwith a low board thickness of 0.25mm.So a professionally manufactured andthrough hole plated printed circuit boardwas chosen but that is not exactly cheap.

This only needs a Gerber plot to becreated (1 mouse click in the "Target"software) and mailed to the PCB manufa-cturers then everything else runs byitself. Unfortunately there is the so-called"setup costs" and a minimum order quan-tity of 4 PCs due to the minimum size ofboard for the production base load. Thepeople at "Aetzwerk München" [5] werevery cooperative; there was even thecorrect PCB material (RO4350B 0.25-4mm thickness) in stock. But even the"minimum order" costs (including PCBmaterial) was €235 for 4 PCBs

Fig 11: The doublesided PCB, size30mm x 50mm. Seetext for details.

Fig 12: Lookinginto the finishedhousing everythingis small.


96

A comment on the PCB material: usuallyfor this purpose Rogers "RO4003" is thelowest priced and low loss standardmaterial. But it is not "flame retardant",if there is a fire, what is the advice? It hasbeen given the appropriate additives andis now called "RO4350B". This haschanged the electrical data slightly (εrrises slightly, but the losses are higher, at10GHz: loss tangent lt = 0.0027 forRO4003 and 0.0037 for RO4350. At2.5GHz however: lt = 0.0021 or 0.0031).

The assembly process is quite slow beca-use of the 0603 size components it needsa steady hand under a stereo microscope

or an appropriate microscope (see [4]).Without SMD solder paste in a hypoder-mic syringe it also takes longer. It has tobe attached to the solder in tiny amountswith a sharp scalpel. In addition, anarrow temperature controlled solderingtip 0.8 mm wide is required. The MMICis only 2mm square with 4 connectionson each side with a pin spacing of0.25mm so the track and pad dimensionsare tiny as well.

Fig 12 shows a picture of the unit aftersuccessful assembly and fitting into amachined aluminium housing (size = 35x 55mm). The SMB female connector for

Fig 13: When theS21 theory andpractice are thesame the story iscomplete.

Fig 14: S11 isalmost the same aspredicted byANSOFT DesignerSV.


97

the + 5V supply can be seen at the top ofthe picture. The input and output signalsare special SMA connectors with thecentre conductor already flattened (givesthe required low reflection transitionfrom the round inner conductor withinthe socket to the flat microstrip line onthe PCB)

An aluminium plate was placed underne-ath the PCB to lift the now much thinnerboard, with its microstrip lines, up to theflattened centre conductor of the connect-ors to give a good and stress free solderjoint (Tip: this is absolutely necessarybecause it is not possible to bend the pinof the socket; this thin pin and the flatflag will break suddenly and quickly).

4.

S parameter measurements tothe prototype

The S21 measurement was made with thewell known vector analyser (hp8410), theassociated S parameter test set (hp87-45A) and a 20dB attenuator connected infront of the input of the test piece toavoid clipping.

The S21 curve can be seen in Fig 13 andit matches the simulation exactly. TheS11 measurements meander around thesimulation values as shown in Fig 14.

Fig 15: S22 wasbetter thanexpected from1.5GHz.

Fig 16: Measuringnoise figure usingthe well knownprinciple ofdoubling the noisepower output.


98

S22 matches the simulation up to 1.5GHzbut is better than expected (Fig 15). TheS12 measurement was more difficult dueto the low amplitude. But a value ofabout -44dB was determined in the rangefrom 1 to 2GHz that is still smaller thanthe simulation result (between -43dB to -40dB).

5.

The noise

With the very small noise figure of 0.4dB

maximum expected, it was a very diffic-ult to measure anything at all with theexisting local laboratory equipment. Nev-ertheless, an attempt was made and asuitable setup for the frequency f = 1GHzwas put together (Fig 16).

In Fig 17 you can admire what looks likesomething in practice. The "swept" spect-rum analyser is not swept but is set aspure measuring receiver with a bandwi-dth of 3kHz on the frequency f = 1GHz.Then, output level of the noise generatorSKTU was slowly increased until thenoise power at the output of the auxiliaryamplifier doubled. This corresponds to anincrease of the noise displayed on the

Fig 17: Thepractical effort tomake a measurem-ent according toFig 16 is howeverconsiderable.

Fig 18: This curvecan only beclassified as aprecious Christmasgift.


99

spectrum analyser of 3dB (or the displa-yed noise voltage by a factor of 1.41).

Now the residual noise of the amplifier isas big as the performance delivered bythe noise channel and the noise figure NFin dB (or the noise factor as a linearratio) can be read on the display of thenoise generator SKTU.

Several measurements showed values ofthe noise factor somewhere between 1.1and 1.2. That would be a "noise figure"(NF) for the amplifier between 0.4 and0.8dB but because of the very low poin-ter deflection of the SKTU this result canonly be considered as indicative. But it isat least the right size (certainly less than1dB = NF).

The influence of the additional downstr-eam amplifier can be neglected due to thehigh gain of the device being measured(24dB at 1GHz). The cascaded NoiseFigure therefore only shows an increaseof 0.02dB.

For more accurate measurements it relies,as always, on help from friends with amodern noise measuring instrument withhigh resolution and accuracy. One ofthese friends (Ulli Kafka with his comp-any "Eisch Electronic" in Ulm) gave theauthor great pleasure with an email cont-

aining pictures as well as the classificat-ion as "excellent amplifier". This will beimmediately understandable if you lookat Fig 18: indeed the noise figure up to1.7GHz remained below 0.4dB and closeto the simulation. It only remains to say"Thank God" and thank all contributors.

The original measurement of noise andS21 that provided the good impressionwas good, especially when you considerthe "cold temperature specified" (Tcold)of 306.3K with about +33° Celsiusduring the measurement as shown in Fig19.

For the specialist there is an addendumfrom the data sheet for the MGA635P8for I = 55mA at 2.5GHz:

• Output IP3 = +35.9dBm • P1dB_out = +22dBm

If you then start to think about thepossibilities of a two stage version forweak Meteosat signals and receptionwith a patch antenna … there would beenough space on the board...

Fig 19: Here is theoriginal picturewith the NF andS21 together withmarkers. So amodern device givea good result.


100

6.

Addendum

A listener of my presentation in Bensheim recently emailed me with aninteresting proposal in order to determinethese small noise figures with the noisegenerator SKTU:

"Why don’t you use an attenuator with aprecisely known value of 10dB betweenthe generator output and input of themeasuring device? Then you will proba-bly need a noise level of 10.5dB delive-red by the noise generator to double theoutput noise power indicated on thescreen of the analyser. If you then subtr-act the 10dB of attenuation from thedelivered power level you get the solut-ion = the unknown noise figure (In thiscase: NF = 10.5dB – 10dB = 0.5dB). Ihave done this several times in the pastwith success".

He is right, and this method must betried. You are always learning.

7.

Literature

[1] Data sheet and S parameters files forthe MGA635P8 from the homepage ofAvago Technologies

[2] Application note from Avago: "MG-A635P8 GaAs MMIC ePHEMT 2.5GHzlow noise amplifier with superior noiseand linearity performance"

[3] Application note from Avago: "MG-A635P8 GaAs MMIC ePHEMT 3.5GHzlow noise amplifier with superior noiseand linearity performance"

[4] Soldering advice for 0.5mm pitchSMD ICs, Bernd Kaa, DG4RBF, VHFCommunications Magazine, 4/2011 pp232 - 236

[5] Etching factory Munich, Germany;PCB manufacturer website:www.aetzwerk.de


101

development of a preamplifier for 1 to 1.7ghz with a noise ... · the current state of development...

Documents