An RF to Baseband Design for a PCS1900 Handset

Liam Devlin, Steve Fitz, Geoff Smithson

The authors are with Plextek Ltd, London Road, Great Chesterford,Essex, CB10 1NY, Engand (+44 1799 533 200)

AbstractGSM is currently the most widely adopteddigital cellular telephony standard in theworld. This paper presents details of thedesign of the RF to baseband section of aPCS1900 cellular handset. Detailed blockdiagrams, circuit descriptions and measuredperformance are presented. The radio has beendesigned to operate with a basebandEvaluation Board (EVB) and the performanceof the radio functioning in conjunction withthe baseband, as a complete handset, ispresented. A casework design for the handsetmechanics has also been developed; it includesan integral patch antenna and allows completez-axis assembly to facilitate volumemanufacture.

IntroductionFirst generation analogue cellular telephonysystems were developed by a number ofEuropean countries during the 1980s. Eachcountry developed independent, incompatiblesystem but it was soon realised that anincreasingly unified Europe would benefitfrom a single standard throughout. In additionto allowing mobile roaming across Europe, asingle standard would produce cost savings onterminal equipment as a result of theeconomies of scale. In 1982, the Conference ofEuropean Posts and Telegraphs (CEPT)formed a group to develop a pan-Europeanstandard: Groupe Special Mobile (GSM).

Phase I of the GSM specifications waspublished in 1990 and commercial servicesstarted in 1992 [1]. At the end of 1997, therewere 70 million GSM users world-wide andthis is predicted to rise to over 100 million by1998 [2]. At the time of writing, there areGSM based cellular telephony systems in over120 countries world-wide [3] and it is apt thatGSM now stands for Global System forMobile communications. There are threederivatives of the standard, the originalGSM900, DCS1800 and PCS1900, the main

difference between the systems, is thefrequency of operation. This is summarised inTable 1.

Standard Uplink (Mobileto BTS)

Downlink (BTSto Mobile)

GSM900 890-915MHz 935-960MHzDCS1800 1710-1785MHz 1805-1880MHzGSM1900 1850-1910MHz 1930-1990MHz

Table 1: Frequency Allocations of GSMVariants

The radio interface for all three bands is thesame. Table 2 summarises the main radiointerface parameters of the GSM standard,additional information can be found in [1].

Multiple Access Method TDMA and FDMAChannel Spacing 200kHz

No. of traffic channelsper frequency channel

8

Channel bit rate 270.8 kbpsSpeech coder bit rate 13kbpsTime frame length 4.6msTime slot length 577us

Modulation scheme Differentiallyencoded GMSK

(BT=0.3)

Table 2: Primary GSM Radio InterfaceParameters

The PCS1900 variant of GSM was developedto utilise the frequency spectrum auctioned offby the Federal Communications Commission(FCC) in 1994, to introduce digital wirelessnetworks for Personal CommunicationsServices (PCS). Legislation governing theauction of spectrum was technology neutral,allowing the market to decide which standardssucceeded. In November 1995 AmericanPersonal Communications (APC) launched thefirst PCS1900 service in the US, just monthsafter obtaining their licence [3]. At the time of

writing, there are 19 GSM based networks inthe USA.

This paper details the design and performanceof the RF to baseband section of a PCS1900handset. Plextek Ltd carried out the design ofthe RF board in conjunction with TTPComLtd, who supplied the baseband EVB andassociated software.

System ArchitectureEver since the advent of the first commerciallyavailable GSM handsets in 1992, the size,weight and cost of the mobile terminal havebeen rapidly decreasing [4,5]. Reducing thecomponent count is the key to reducing theproduct cost. The cost of placing low valuechip components, such as 0603 resistors, canoften be 5-10 times the cost of the componentitself [6]. Increased levels of integration offera means of reducing component count and,perhaps more importantly in a market withsuch a short product lifecycle, allowing muchshorter development timescales [7].

The design presented here makes use of anintegrated RF transceiver. This means thehigh level architecture of the receiver andtransmitter are largely dictated by thetransceiver IC. Whilst this removes somedegree of freedom from the handset designer,providing the system design has been carriedout sensibly, it can save a lot of time. Thisdesign makes use of the Hitachi BRIGHT IC.A functional block diagram of the RF tobaseband circuit is shown in Figure 1.

As in most low cost GSM designs, the sameLocal Oscillators (LO) are used in bothtransmit and receive modes. Two synthesisedLO sources are required, an agile ‘channeloscillator’ which selects the requiredfrequency channel in both the receive andtransmit time slots and a nominally fixedfrequency IF oscillator which is divided toprovide the required IF LOs for both receiverand transmitter.

Figure 1: Block Diagram

÷ 2

T/RSwitch

1930 - 1990 MHzLNA

LumpedRF Balun 225 MHz

IF SAW45 MHzLC Filter

AGC

I-QDEMOD

0o -90o

45 MHz

90o Shift÷ 2

90 MHz

÷ 6

270 MHz

13 MHzTCXO

ChannelVCO

Loop Filter

Dual Synth IC

Loop Filter540 MHzIF VCO

PA

1850 - 1910 MHz

Loop Filter

135 MHz

PhaseDetector

PowerControl LoopFrom Power

Control DAC

90o Shift÷ 2

0o -90o

135 MHz

I-QMODULATOR

I

Q

HDISS111F

I

Q

The receiver uses three stages ofdownconversion to provide analogue basebandquadrature outputs. Analogue to DigitalConverters (ADCs) convert the basebandoutputs to digital and the baseband processorcarries out timing synchronisation,equalisation and error correction. Analternative topology to baseband sampling isto sample an IF signal directly. In this case thefinal mix to baseband is performed digitally.This is often achieved by “sub-sampling”,which allows the sampling rate to be below theIF frequency [5]. One disadvantage with thistechnique is that the ADC must work at ahigher frequency making it more difficult toachieve the required performance.

The main disadvantage with a basebandsampling approach, as used in this design, isDC offset signals caused by leakage of the LOsignal used for the mix to baseband. Any LOsignal leaking into the IF circuits is mixeddown to zero frequency giving a DC offset. Acommon way of significantly reducing thisproblem is to generate the IF LO signal at amultiple of the required frequency and todivide it down on-chip. A divide by 2approach can also be used as a means ofgenerating accurate quadrature LO signals.The signal drives two dividers with the flip-flops in the dividers being triggered on eitherthe negative or positive edges, to give outputsignals at half the input frequency with 90° ofphase difference.

A treble conversion superhet design is apopular receive architecture for GSMhandsets. It allows a first IF at a high enoughfrequency (in this case 225MHz) to allow goodfiltering of the image signal. A secondconversion to a lower IF allows the AGC to berealised at a lower frequency, with a finalconversion to baseband or to a third IF in thecase of IF sampling designs. Doubleconversion architectures are also possible butthe AGC must then be realised at a higherfrequency. Another alternative, which ispossible, is to use a direct conversion tobaseband architecture [8]. This is attractivebecause it reduces the number of RFcomponents used and removes the need forimage filtering. However, the practicaldifficulties of implementing a directconversion GSM receiver should not beunderestimated [5].

In transmit mode the inputs to the transceiverIC are differential I and Q baseband signals.

These are used to modulate a relatively lowfrequency IF signal, which is then used tomodulate a VCO operating at the transmitfrequency in a translational loop architecture.The translational loop works by mixing thetransmit VCO signal with the channeloscillator to generate an IF signal. A phasecomparator is used to compare this IF to themodulated IF with the difference being passedback to the transmit VCO, via a loop filter, asan error signal. This closes a phase lockedloop around the transmit VCO which locks itto the modulated IF signal. Translational looparchitectures are only possible with constantenvelope modulation schemes such asGaussian Minimum Shift Keying (GMSK).One disadvantage with a translational looparchitecture is that an extra VCO is requiredas compared to a direct modulation orsuperhet. architecture. However, it has becomeincreasingly popular for the following reasons:

• The input to the PA is not the direct outputof a mixer with all of it’s associatedspurious mixing products. This eases thefiltering requirements of the transmitter.

• The oscillator driving the PA can have a

relatively high output level. Less PA gainis therefore required, which reduces thethermal noise generated at the output ofthe PA, making it easier to pass the receiveband noise emissions requirement. Thismeans a switch can be used at the antennaport instead of a diplexer filter, which ismore expensive and has higher loss.

• Susceptibility to pulling of the transmit

VCO is reduced because the wide loopbandwidth makes it very agile and able totrack and correct any sudden pullingeffects

Synthesiser Design andMeasured PerformanceThree synthesiser designs are used in thehandset: a channel VCO that must selectfrequencies from 1705.2MHz to 1774.8MHzin 200kHz steps. An IF VCO which is set to540MHz and the translational loop oscillatorwhich must tune over the transmit band from1850.2MHz to 1909.8MHz [9]. Thesynthesiser for the translational loop is on thetransceiver IC, the other two oscillators makeuse a single dual synthesiser IC. VCOmodules were used for the channel and

translational loop oscillators. From a simpleBOM cost point of view this is slightly moreexpensive than designing discrete VCOs butdoes have the advantage that the tuning rangeis guaranteed. This removes the need fortuning of the VCO during manufacture, a timeconsuming and potentially yield hazardousrequirement. A discrete design was, however,used for the fixed frequency IF VCO.

In general, it makes sense to choose as high acomparison frequency as possible since thiswill result is lower close-in phase noise, fasterswitching speeds and lower referencesidebands. As the channel spacing is 200kHzand the oscillator must be able to select allavailable channels, a comparison frequency of200kHz is used. Since the oscillator is used inboth transmit and receive modes, it must becapable of switching across its entire tuningrange within one time slot. A fast VCOswitching time requires a wide loopbandwidth. However too wide a loopbandwidth can increase the total integratedphase noise of the oscillator, which willincrease the phase error of the modulatedsignal. The loop was designed to provide aswitching speed of less than 500µs and aphase error contribution of less than 1° rms. Itis possible to translate this phase errorspecification into a phase noise requirement[10] and ensure the spectral purity of thesynthesised output is adequate.

All three synthesisers used third order loopfilters as shown in Figure 2.

C1

R2

C2

R3

C3

R4

Cv

Charge Pump Output

Vvaricap

Figure 2: Loop filter configuration

Analysis of the synthesiser loop yields simpleequations for calculating the values of thecomponents in the loop filter [11]. These canthen be entered into a maths package, such asMathCad, and the open loop Bode plot can beassessed for gain and phase stability. Figure 3shows the open loop gain Bode plot for suchan analysis, with a predicted loop bandwidthof 12kHz.

1 103

1 104

1 105

1 106

40

30

20

10

0

10

20

30

4040

40

.20 log( )GH( )n

.1 1061000

10n

Figure 3: Simulated open loop gain Bodeplot of channel synthesiser

Analysis of the open loop phase of the channelsynthesiser, plotted in Figure 4, shows a phaseshift of -140° at the unity gain frequency. Thisrepresents a phase margin of 40°, since if thephase shift were to reach –180° the loop wouldbecome unstable. (The phase comparator itselfgives -180° of phase shift so resulting inpositive feedback at greater than unity gain,the condition for instability). If the phasemargin is too high, this is indicative of a loopthat is likely to switch quite slowly.

1 103

1 104

1 105

1 106

180

165

150

135

120

105

9090

180

.360.2 π

arg( )GH( )n

.1 1061000

10n

Figure 4: Simulated open loop phase Bodeplot of channel synthesiser

Plots of the performance of the channeloscillator, measured at a test port supplying alow-level sample, are shown in Figures 5 to 7.

Figure 5: Channel VCO, comparisonfrequency @ -66.1dBc

Figure : Channel VCO, Noise at 20kHzoffset is -81.7dBc/Hz

Figure 7: Channel VCO, Noise at 1kHzoffset is -80.5dBc/Hz

The IF VCO is a fixed frequency design at540MHz. A discrete Clapp type VCO wasdesigned, using a plastic packaged bipolartransistor. Adequate Q could be obtained usinga simple 0805 chip inductor as the resonator.Although the IF VCO does not need to stepfrom channel to channel, a comparisonfrequency of 200kHz was still used. The mainreason for this was that a dual synthesiser ICwas used for the channel and IF VCOs. Ifdifferent comparison frequencies were used forthe two, this would give rise to sidebands atthe difference of the two comparisonfrequencies.

Performance plots of the IF VCO, measuredusing a high impedance probe, are shown inFigures 8 to 10.

Figure 8: IF VCO, comparison frequency @-69.1dBc

Figure 9: IF VCO, Noise at 20kHz offset is -84.2dBc/Hz

Figure 10: IF VCO, Noise at 1kHz offset is -81.2dBc/Hz

Table 3 summarises the measuredperformance of the channel and IFsynthesisers. The switching speedperformance of the channel VCO is evaluatedin the transmitter design section of this paper.

Synth: IF

(540MHz)

Ch.

Level of 200kHz

comparison

spurs (dBc)

-69.1 -67.1

Noise at 20kHz

offset (dBc/Hz)

-84.2 -83.4

Noise at 1kHz

offset (dBc/Hz)

-81.2 -78.1

Table 3: Performance summary of channeland IF synth's

The translational loop synthesiser has a muchhigher loop bandwidth, in the order of 1MHz,as it must pass the modulation spectrum. It’scomparison frequency is the IF frequency onto

which the data is modulated, in this case135MHz. The measured performance of thetranslational loop is covered in the transmitterdesign section below.

Transmitter Design andMeasured PerformanceAnalogue I and Q signals are the basebandinput to the transceiver IC. They are used tomodulate an IF of 135MHz, derived bydividing the 540MHz IF signal by 4. Thismodulated IF signal is applied to a phasecomparator along with a signal derived bymixing the output of the transmit oscillatorwith the channel oscillator. Any differencebetween the two is passed back to the VCO,via the loop filter, locking the VCO to themodulated IF signal. By setting the channeloscillator to 135MHz below the wantedtransmit channel, the output of the transmitoscillator is modulated with the transmit dataand centred on the required channel. Figure11 shows a measured plot of the modulatedoutput spectrum at the antenna port.

Figure 11: Modulated output spectrum

The output power of the transmit oscillatorcan be at a relatively high level, in this casearound 6dBm, which means that less RF gainis required, so reducing the receive bandnoise. Another advantage is that a T/R switchcan be used at the antenna, instead of adiplexer filter to split the transmit and receivefrequencies. Since this was a single banddesign, a PIN diode Single Pole Double Throw(SPDT) was used as it offered low loss andhigh linearity with a +3V control and had abuilt in high pass filter to reduce PowerAmplifier (PA) harmonics. For dual or tri-band designs, broad band, multi-throwswitches with low harmonic generation arerequired [12].

PCS1900, class 1 mobiles are required to haveselectable output power levels from 0dBm to+30dBm, in 2dB steps [9]. This requires apower control loop to sense the output powerand control the PA, accordingly. A MOSFETbased PA module was used, which has arelatively linear voltage versus power controllaw. This makes it easier to realise a stablepower control loop. Figure 12 shows a blockdiagram of the power control loop. A printed -20dB coupler samples the RF output powerand drives a detector diode, forward biasedwith a small current. The diode is a square lawdevice which generates a rectified DC voltagewhich is approximately proportional to the RFpower incident on it. An op-amp amplifies thedifference between the detected voltage and areference voltage from an identical diode withthe same DC bias. The output of this op-ampis compared to the output of the power controlDAC, an integrating op-amp is used tointroduce phase lag and ensure stability of theloop. The output of this op-amp is used todrive the PA control pin.

Figure 12: Block diagram of PA control loop

Figures 13 to 16 show the output power,versus time, of the modulated transmit burst atpower levels of +30, +20, +10 and 0dBm.Theses measurements were made using aspectrum analyser with the frequency span setto zero. The required accuracy for these powerlevels is ±2dB, ±3dB, ±4dB and ±5dBrespectively. This includes all spreads due toproduction tolerances, temperature and supplyvoltage. The reasons for controlling themobile’s output power are to reduceinterference to other units and to conservebattery power.

Figure 13: Transmit burst power set to+30dBm

Figure 14: Transmit burst power set to+20dBm

Figure 15: Transmit burst power set to+10dBm

RF InRF OutPA

Vcc

Vcc

From PARamp DAC

+

-

+

-

Figure 16: Transmit burst power set to0dBm

In addition to controlloing the power levelduring the burst, it is also necessary toaccurately control the ramping profile.Switching the PA on and off sharply willresult in spurious transient outputs. There is atightly defined spectral time mask which mustbe met [9] and the performance of the PAcontrol loop must be optimised to ensure it ismet at all power levels. Figure 17 shows themeasured RF output meeting the requiredspectral time mask, across a burst. Themeasurement was made at +30dBm outputpower, with a 10dB pad before the analyser.

Figure 17: Measured RF output, meetingthe specified spectral time mask at full

power

Measurements of the phase error of thetransmitted spectrum have been made atdifferent output powers levels, using a vectorsignal analyser. Figure 18 shows the phaseerror versus symbol of the transmit burst withan output power of +20dBm. The phase erroris 2.4° and 7° peak, well below the required 5°rms and 20° peak. For higher output powerlevels there is a slight increase in phase error,probably due to pulling of the transmit VCOas the PA is ramped on. Figure 19 shows aplot of the phase error for the maximumoutput power level of +30dBm, with the phase

errors still well inside the specification at 2.6°rms and 13° peak.

Figure 18: Tx. phase error (rms = 2.4°°, pk =7°°), Pout = 10dBm, ch. VCO stationary

Figure 19: Tx. phase error (rms = 2.6°°, pk =13°°), Pout = 30dBm, ch. VCO stationary

The phase error measurements in Figure 18and Figure 19 were made with the receive andtransmit channels selected such that thechannel oscillator does not have to move. Inpractice the channel oscillator must be able tomove from the highest receive channel to thelowest transmit channel in the space of onetime slot. A straightforward way of confirmingthis works is to check the transmit phase errorwith the channel oscillator having to move themaximum required frequency. If the synth.cannot acquire phase lock in time, there willbe an increased phase error at the start of theburst. Plots of this measurement are shown inFigure 20 and Figure 21 for output powerlevels of +20dBm and +30dBm respectively.There is a slight degradation in phase errorcompared to the case of a stationary channeloscillator but the phase error is still withinspecification.

Figure 20: Tx. phase error ( rms = 2.5°° pk =7°°), Pout = 20dBm, ch. VCO max. shift

Figure 21: Tx. phase error (rms = 2.8°°, pk =15°°), Pout = 30dBm, ch. VCO max. shift

Table 4 summarises the rms and peak phaseerror performance of the transmitter withvarying output power level and channel VCOfrequency shift. Not too surprisingly, the worstcase occurs for maximum output power withthe channel VCO moving the maximumrequired. In this case phase errors of 2.8° rmsand 15° peak are achieved. These arecomfortably less than the required 5° rms and20° peak.

Case rms ph.error

pk. ph.error

Pout = +20dBmNo Chan. Shift

2.4° 7°

Pout = +20dBmMax Chan. Shift

2.5° 7°

Pout = +30dBmNo Chan. Shift

2.6° 13°

Pout = +30dBmMax Chan. Shift

2.8° 15°

Table 4: Transmit phase error summary

Eye diagrams of the transmitted output, alsomeasured using a vector signal analyser, areshown in Figure 22, for an output power of+30dBm and in Figure 23, for an outputpower of +5dBm. In both cases the eye is well

open, confirming the low phase errorsmeasured.

Figure 22: Eye diagram, measured withPout = +30dBm

Figure 23: Eye diagram, measured withPout = +5dBm

I-Q diagrams, showing the constellation plots,are shown in Figure 24 and Figure 25. Eachconstellation has constant amplitude andcovers an arc of around 30° in the case of+30dBm output power and around +15° in thecase of +5dBm output power. Thiscorresponds well with the measured peakphase errors across the burst of 15° and 7°,respectively.

Figure 24: IQ constellation plot with Pout =+30dBm

Figure 25: IQ constellation plot with Pout =+5dBm

Receiver Design and MeasuredPerformanceThe receiver is a treble conversion to basebanddesign. Received signals are routed throughthe T/R switch into a ceramic filter, throughthe LNA, through a second ceramic filter andinto the transceiver IC via an RF balun. Theneed for two high selectivity ceramic filters inthe receive path is dictated by the requirementfor the receiver to handle high level blockingsignals whilst receiving a wanted signal of-99dBm [9].

Most bipolar ICs utilise “Gilbert Cell” [13]mixers. These are well balanced and offerconversion gain but have the disadvantage ofhaving a relatively high noise figure andcomparatively poor linearity. The traditionalmethod for improving linearity of this mixertype is by including emitter degenerationresistance. Unfortunately this further degradesthe noise figure of the mixer. Consequentlythe design of this type of mixer is a balancingact between the linearity and sensitivityrequirements of the receiver.

It is necessary to provide adequate RF filteringin the frontend of the receiver to protect themixer from the out of band blocking signalsthe receiver must withstand. In addition to theout of band blocking signals, such astransmissions from other mobiles, there is afurther requirement for the receiver towithstand lower level in-band blocking signals[9]. RF filtering does not help; the receiver’sRF frontend must be sufficiently linear to meetthis requirement. This leads to a compromisein the LNA gain. A high gain LNA willimprove the noise figure, and so sensitivity, ofthe receiver by reducing the second stage noisecontribution from the mixer. Unfortunately ahigh gain LNA increases the level of in band

blocking signals entering the mixer, sodegrading the linearity of the receiver.

The LNA is a single stage design using aplastic packaged bipolar transistor with somelight series resistive feedback to ensurestability. As a stand-alone subcircuit, itexhibits a gain of 14dB and a noise figure of1.5dB. Figure 26 shows the gain of the RFreceive frontend, from the antenna port to theinput of the RF balun, prior to the first mixer.Gain is 11dB and image rejection over 70dB.

CH1 S 21 log MAG 10 dB/ REF 0 dB

START 1.400 000 000 GHz STOP 2.400 000 000 GHz

Cor

PRm

6 May 1998 11:07:14

1 2

3

3_:-62.375 dB

1.540 000 000 GHz

1_: 11.172 dB 1.93 GHz

2_: 10.091 dB 1.99 GHz

Figure 26: Gain response through T/Rswitch, LNA and ceramic filters

After the image filter, a simple lumpedelement balun is used to convert the singleended RF signal to differential. The on-chipGilbert cell mixer converts this signal down to225MHz, which is routed off-chip through thedifferential IF SAW filter, which providesrejection of in band and close in blockingsignals. Correct matching to this SAW filter iscritical to the performance of the receiver. Anymismatch can give rise to excessive groupdelay ripple which will result in Inter-SymbolInterference (ISI).

As communications systems have moved fromanalogue to digital modulation schemes, somore of the receiver functionality has becomeintegrated into the baseband signal processing.This makes evaluation of the RF to basebandsection in isolation, much more difficult. Theperformance can only really be confirmedwhen the analogue and digital sections of thereceiver are functioning together. Someevaluation can however be carried out bylooking at the IF signal prior to the conversion

to baseband, in this case at 45MHz. Figure 27shown the spectral mask of the received signalat 45MHz.

Figure 27: Received spectral mask at45MHz

The phase error of the received signal can beevaluated at 45MHz. However, it should beremembered that even a well matched SAWfilter will introduce some ISI, which willresult in phase error. Part of the job of theequaliser is to account for this, however theequaliser must also account for the ISI causedby multi-path effects through the propagationenvironment. If the ISI caused by the SAW is

excessive, then it will take all of theequaliser’s capability to correct it and it willno longer be able to account for propagationmulti-path. Ultimately, it is the Bit Error Rate(BER) performance of the whole system,analogue and baseband, under the requiredconditions of operation, which is important.The BER of the radio PCB and baseband EVBwere analysed together and met the Bit ErrorRate requirements defined in [9].

Physical RealisationThe design was fabricated on standard 4 layer,1.6mm thick, FR4. Four layers were used andthe board was configured to allow easyadoption into a single board phone. Blind viaswere avoided to keep down the manufacturingcost. Figure 28 shows a photograph of thePCB. The large electrolytic capacitor anddigital interface IC would not be required inany product utilising this design.

Figure 28: Photograph of the PCB

Casework DesignThe mechanical casework design has alsobeen carried out. Figure 29 shows aphotograph of a model of the design andFigure 30 shows an exploded mechanicalassembly drawing. It is intended toincorporate a patch antenna within themechanics of the housing. The design allowscomplete Z-axis assembly, so removing anyneed for clamping or turning of the product.

Figure 29: Photograph of handset model

Figure 30: Exploded assembly drawing ofcasework

SummaryA complete RF to baseband design for aPCS1900 handset has been described. Detailedcircuit descriptions and block diagrams havebeen included. The radio has been tested inconjunction with a baseband EVB, as acomplete handset. Calls to a digital radiocommunications testset have been made and acall to a “live” PCS1900 network is planned.The design will then be available forincorporation into low cost PCS1900 products.

AcknowledgementThe authors would like to thank TTPCom Ltdfor supplying the baseband EVB andassociated software and for their supportduring this development.

References[1] Michel Mouly and Marie-BernadettePaultet, “The GSM System for MobileCommunications”, published by the authors,ISBN: 2-9507190-0-7

[2] “MoU predicts over 100 million GSMsubscribers by the end of 1998”, MobileEurope, April 1998, Volume 8, Number 4,page 12

[3] GSM MoU Association home page,“http://www.gsmworld.com/index.html”

[4] M.H. Norris, “Transmitter Architectures”,Proceedings of the IEE Colloquium on theDesign of Digital Cellular Handsets, London,March 1998.

[5] S.M. Fitz, “Receiver Architectures forGSM Handsets”, Proceedings of the IEEColloquium on the Design of Digital CellularHandsets, London, March 1998.

[6] M.H.Norris, “An Integrated RF ASIC fordual-band GSM”, Proceedings of the IEEColloquium on Multi-Chip Modules andRFICs, London, May 1998.

[7] Liam Devlin, “RF ICs For CommercialWireless Applications”, Proceedings of the

IEE Tutorial Colloquium on the Design ofRFICs and MMICs, London, November 1997.

[8] Jan Sevenhans et al, “An integrated SiBipolar RF Transceiver for a Zero IF 900MHzGSM Digital Mobile Radio Frontend of aHand Portable Phone”, IEEE 1991 CustomIntegrated Circuit Conference, pp 7.7.1-7.7.4

[9] “PCS 1900 Air Interface Specification”,ANSI J-STD-007-1996

[10] Constantine Fantanas, “Introduction toPhase Noise”, RF Design, August 1992, pp50-57

[11] G. Smithson, “Synthesised FrequencySources in Digital Cellular Handsets”,Proceedings of the 1997 Microwaves & RFConference.

[12] J.C. Clifton and L. Albasha, “RF ICAntenna Switch Solutions for GSM DualbandTelephones”, Proceedings of the IEEColloquium on Multi-Chip Modules andRFICs, London, May 1998.

[13] Barrie Gilbert “A Precise Four-QuadrantMultiplier with Subnanosecond Response”,IEEE Journal of Solid-State Physics, Vol. SC-3, No. 4, December 1968

专注于微波、射频、天线设计人才的培养 易迪拓培训 网址：http://www.edatop.com

射 频 和 天 线 设 计 培 训 课 程 推 荐

易迪拓培训(www.edatop.com)由数名来自于研发第一线的资深工程师发起成立，致力并专注于微

波、射频、天线设计研发人才的培养；我们于 2006 年整合合并微波 EDA 网(www.mweda.com)，现

已发展成为国内最大的微波射频和天线设计人才培养基地，成功推出多套微波射频以及天线设计经典

培训课程和 ADS、HFSS 等专业软件使用培训课程，广受客户好评；并先后与人民邮电出版社、电子

工业出版社合作出版了多本专业图书，帮助数万名工程师提升了专业技术能力。客户遍布中兴通讯、

研通高频、埃威航电、国人通信等多家国内知名公司，以及台湾工业技术研究院、永业科技、全一电

子等多家台湾地区企业。

易迪拓培训课程列表：http://www.edatop.com/peixun/rfe/129.html

射频工程师养成培训课程套装

该套装精选了射频专业基础培训课程、射频仿真设计培训课程和射频电

路测量培训课程三个类别共 30 门视频培训课程和 3 本图书教材；旨在

引领学员全面学习一个射频工程师需要熟悉、理解和掌握的专业知识和

研发设计能力。通过套装的学习，能够让学员完全达到和胜任一个合格

的射频工程师的要求…

课程网址：http://www.edatop.com/peixun/rfe/110.html

ADS 学习培训课程套装

该套装是迄今国内最全面、最权威的 ADS 培训教程，共包含 10 门 ADS

学习培训课程。课程是由具有多年 ADS 使用经验的微波射频与通信系

统设计领域资深专家讲解，并多结合设计实例，由浅入深、详细而又

全面地讲解了 ADS 在微波射频电路设计、通信系统设计和电磁仿真设

计方面的内容。能让您在最短的时间内学会使用 ADS，迅速提升个人技

术能力，把 ADS 真正应用到实际研发工作中去，成为 ADS 设计专家...

课程网址： http://www.edatop.com/peixun/ads/13.html

HFSS 学习培训课程套装

该套课程套装包含了本站全部 HFSS 培训课程，是迄今国内最全面、最

专业的HFSS培训教程套装，可以帮助您从零开始，全面深入学习HFSS

的各项功能和在多个方面的工程应用。购买套装，更可超值赠送 3 个月

免费学习答疑，随时解答您学习过程中遇到的棘手问题，让您的 HFSS

学习更加轻松顺畅…

课程网址：http://www.edatop.com/peixun/hfss/11.html

`

专注于微波、射频、天线设计人才的培养 易迪拓培训 网址：http://www.edatop.com

CST 学习培训课程套装

该培训套装由易迪拓培训联合微波 EDA 网共同推出，是最全面、系统、

专业的 CST 微波工作室培训课程套装，所有课程都由经验丰富的专家授

课，视频教学，可以帮助您从零开始，全面系统地学习 CST 微波工作的

各项功能及其在微波射频、天线设计等领域的设计应用。且购买该套装，

还可超值赠送 3 个月免费学习答疑…

课程网址：http://www.edatop.com/peixun/cst/24.html

HFSS 天线设计培训课程套装

套装包含 6 门视频课程和 1 本图书，课程从基础讲起，内容由浅入深，

理论介绍和实际操作讲解相结合，全面系统的讲解了 HFSS 天线设计的

全过程。是国内最全面、最专业的 HFSS 天线设计课程，可以帮助您快

速学习掌握如何使用 HFSS 设计天线，让天线设计不再难…

课程网址：http://www.edatop.com/peixun/hfss/122.html

13.56MHz NFC/RFID 线圈天线设计培训课程套装

套装包含 4 门视频培训课程，培训将 13.56MHz 线圈天线设计原理和仿

真设计实践相结合，全面系统地讲解了 13.56MHz线圈天线的工作原理、

设计方法、设计考量以及使用 HFSS 和 CST 仿真分析线圈天线的具体

操作，同时还介绍了 13.56MHz 线圈天线匹配电路的设计和调试。通过

该套课程的学习，可以帮助您快速学习掌握 13.56MHz 线圈天线及其匹

配电路的原理、设计和调试…

详情浏览：http://www.edatop.com/peixun/antenna/116.html

我们的课程优势：

※ 成立于 2004 年，10 多年丰富的行业经验，

※ 一直致力并专注于微波射频和天线设计工程师的培养，更了解该行业对人才的要求

※ 经验丰富的一线资深工程师讲授，结合实际工程案例，直观、实用、易学

联系我们：

※ 易迪拓培训官网：http://www.edatop.com

※ 微波 EDA 网：http://www.mweda.com

※ 官方淘宝店：http://shop36920890.taobao.com

专注于微波、射频、天线设计人才的培养

官方网址：http://www.edatop.com 易迪拓培训 淘宝网店：http://shop36920890.taobao.com