Communications Systems Design - PRFI · 2020. 8. 24. · QAM64 all have the same mean power level, but it is clear that the individual constellation points of the highest scheme (QAM64)

PRFI Ltd.The Plextek Building, London Road, Great Chesterford,Saffron Walden, CB10 1NY, UK

T: +44 (0) 1799 796464 E: [email protected] W: www.prfi.com

This white paper presents an over-view of system level design forwireless communications equip-ment. It will be useful not just tothose involved in the developmentof new communications systems,but also to designers of new prod-ucts using existing systems, oranybody seeking to develop newarchitectures and/or componentsto reduce product cost and/or size.It also describes the benefits andimplications of developing highlyintegrated System on a Chip(SoC) solutions, which dominatein today's high volume communi-cations products.

IntroductionRadio (wireless) communicationssystems have moved a long way in thelast 100 years. From radio broadcast-ing, the first mass market application,to today's ubiquitous Smartphonesoffering an array of applications andcommunication at ever increasing datarates. This march of increasing func-tionality and complexity was enabledby the advent of the integrated circuitand the microprocessor, which togetherhave facilitated miniaturisation andlow-cost high volume manufacture.

Modern communications systems pro-liferate throughout the available spec-trum and use a wide range of channelbandwidths, modulation schemes anddata rates. The main traffic for wirelesscommunications systems is now datarather than voice and the required datarates are ever rising. When designinga communications system the imple-mentation choices depend on a host ofrequirements, some of which can beconflicting.

These include the required functional-ity and performance, the available

technology, operational conditions andlimitations, legislative restrictions,development budgets, unit cost targetsand timescales. The design process iscomplex, requiring an in-depth under-standing of technology, radio systems,digital processing and the ability toaccurately simulate the effects of apractical realisation at a system level.

Applications and RequirementsAlthough consumer demand for wire-less data is relentless, it would bewrong to think that all current develop-ments in communications are aimed atmaximising traffic throughput overwide channel bandwidths. This type ofapplication certainly occupies an enor-mous amount of global developmenteffort, but there are also applicationswhere only a tiny amount of informa-tion needs to be communicated betweensites and on an infrequent basis.

A good example of a modern lowdata-rate application is in the monitor-ing and control of street lighting, suchas that depicted in Figure 1 (courtesyof Telensa).

Sheet Code 0623

Figure 1: Wireless street lighting control system (courtesy of Telensa)

White Paper

CommunicationsSystems Design



Here each street lamp is fitted with asmall two way radio link allowingmonitoring and control by the localcouncil (City authority). The councilcan control when the street lights areturned on, and to save energy late intothe night, when they can be dimmed toa lower level or switched off. The streetlight can report back the state of its bulband ballast, or more recently LEDs, sothat replacement can be undertakenwhen near end-of-life. The amount ofdata that needs to be communicatedto/from each light is small and infre-quent. There is currently strong andgrowing interest in low data rate M2M(machine-to-machine) communicationssuch as this.

Another growth area in communica-tions is high data rate links for personalconsumer applications. High data-rateWLAN and cellular communications isnow the norm and is expected by users.As higher and higher wireless data rateshave become available, new applica-tions evolve to make use of the addi-tional capacity. The volume of data thatmust be transferred is greatly amplifiedin the back-haul network where the livedata from thousands of users needs tobe transferred between the local cellulartower and the cellular operator regionalcontrol centre.

Microwave line of sight links are oftenused for this purpose, operating at highfrequencies where wide channel band-widths are available. In addition tousing wide channel bandwidths, micro-wave links make use of higher ordermodulation schemes to squeeze moredata through each channel.

Satellite systems allow truly globalcoverage. Global Navigation SatelliteSystems (GNSS) provide one-waycommunications of navigational infor-mation worldwide. Two way satelliteservices are provided by other organi-sations and are used by emergency

services in times of disasters as well asby news reporters and those who needcommunications beyond the range ofthe normal cellular services.

Satellite Satellite receivers are verysensitive. The signal levels are verylow by the time they arrive on theground and special attention is paid tominimising their noise figure. Thedown side is that they are more proneto interference from high powersystems on nearby frequencies. Specialprotection is given to the satellitefrequency bands by restrictions onother users nearby to ensure theirradiated power was not excessive.However, the demand for growth ofcellular services has recently put thisprotection under pressure with theallocation of cellular band frequenciesclose to some satellite bands.

Battery life is an important designparameter in many communicationsproducts.

In some cases this simply needs to belong enough to avoid inconvenience(excessive re-charging of mobiledevices) but for other applications(vehicle security devices, remote meterreading devices) a battery life of manyyears is essential for effective commer-cial deployment. Such a requirementmust be considered from the outset andwill be a key parameter in the systemdesign process.

Modulation Schemes and DataRatesThe one resource that all radio applica-tions have in common is the radiospectrum and this has to be shared byall users. As the number of applicationscontinue to increase, it becomes moreand more important that the use of thisresource is undertaken efficiently.

In all cases, the modulation must beappropriate for the application, as effi-cient as possible and constrained to thebounds of the allocated channel to

Figure 2: Combined I-Q constellation diagram



avoid spectral leakage into adjacentchannels which could potentially inter-fere with users there.

There are many different modulationschemes operating in systems aroundthe world today. For microwave point-to-point links, so called "high order"modulation schemes such as QAM64,QAM256 & QAM512 are commonlyused. These offer high data transferrates as each symbol of the modulationrepresents 6, 8 and 9 bits of data respec-tively. Their high data density comes ata price as it places challenging require-ments on the error vector magnitude(EVM), the recovered eye diagram andthe control of the symbol detectionpoints.

To illustrate this, Figure 2 overlays fourseparate modulation schemes on asingle constellation diagram. The fourschemes: QPSK, QAM16, QAM32 &QAM64 all have the same mean powerlevel, but it is clear that the individualconstellation points of the highestscheme (QAM64) are much closer toeach other than are those of the lowestscheme (QPSK).

This makes the higher order schememuch more susceptible to noise and

distortion since the amount of noise ordistortion required to displace one par-ticular constellation point from itsregion into that of another is smallerthan that required for a low orderscheme.

The presence of noise in the receivechannel sets the fundamental limit onthe Bit Error Rate (BER) that can beachieved for a given modulationscheme.

The higher the order of modulationused the higher the BER for a givennoise level. Figure 3 shows a plot ofBER versus Eb/No for a range ofmodulation schemes, where Eb is theenergy per bit and No the noise spectraldensity. It is clear that squeezing moredata through a given bandwidth by theuse of higher order modulation schemeshas an associated penalty in terms ofthe power that must be transmitted to

Figure 3: BER versus Eb/No for different modulation schemes,M=4:QPSK, M=16: QAM16 etc.

Figure 4: Constellation and eye diagram for QAM256 modulation



to determine the symbol's value onlyat the eye opening point.

In addition to corrupting the constella-tion and causing eye closure, distortionin the transmitter can cause the signalto self-intermodulate and producespectral regrowth (leakage) into adja-cent channels. Figure 6 shows thespectral regrowth that PA non-lineari-ties can cause. Here the regrowthexceeds the statutory limits imposedby the regulatory authorities.

This level of distortion also corruptsthe transmitted constellation and eyediagram as depicted in Figure 7 andFigure 8.

Distortion within the receiver mustalso be well contained. Channel filter-ing, if applied too severely, can distortthe modulated waveform, resulting infurther eye closure and constellationdistortion. This directly causes anincrease in bit error rate (BER). Whendesigning a radio system, sufficienttime should be devoted to undertakingsystem simulations to predict perform-ance in terms of EVM, spectral

achieve the same BER (assuming areceiver with the same noise perform-ance). There is also additional com-plexity in the radio design itself.

Figure 4 depicts the constellation andeye diagram for QAM256 modulationwithout any real-world corruption.

In a practical system imperfections inthe transmitter, such as amplitude andphase distortion in the RF amplifiersand the phase noise of the local oscil-lator would cause distortion, the eyeswould close a little and the constella-tion points would spread.

The frequency spectrum of a perfectQAM256 signal is depicted in Figure5 together with the spectral maskshowing the emission limits for amicrowave point-to-point link (greentrace). Root-raised cosine filtering ofthe transmitted data ensures its tightspectral control. It also introducesInter-Symbol-Interference (ISI) awayfrom the centre of the eye, as seen inFigure 4. The receiver must maintainsymbol timing with sufficient accuracy

regrowth, constellation and eye distor-tion and BER sensitivity.

It is possible to simulate the entirecommunications system from bits into bits out. This is a time consumingtask but can be worthwhile particularlyin cases where custom System on Chip(SoC) ICs are under development.Figure 9 compares the simulated BERto measured BER for a low-IF demod-ulator designed by Plextek RFI andimplemented as part of a custom ASICnow in volume production. The closeagreement is a testament to the accu-racy that can be achieved with carefuldesign and simulation.

Modern cellular systems make use ofOrthogonal Frequency DivisionMultiplexing (OFDM) to furtherincrease data throughput. OFDM canbe understood as a two-step process;first the data to be transmitted is splitinto a large number of lower data-ratestreams. Each of these data streams isthen modulated onto a sub-carrier asdepicted in Figure 10, where eachsubcarrier is orthogonal with theothers. Secondly, individualsubcarriers are then modulated with

Figure 5: QAM 256 modulation spectrum (Blue) andtransmit spectral mask (Green) for point-to-point links

Figure 6: QAM 256 modulation sufferingexcessive spectral regrowth due to PA distortions



simulation of the performance ofMIMO OFDM modulated signals.

In free space, with a single directline-of-sight path between transmittingand receiving antennas, the path lossfollows a simple square law as definedin Equation 1, where d is the distancebetween the two antennas, Gtx and Grxare the antenna gains and n =2.

This propagation model would bevalid, for example, for satellite tosatellite communications where thereis one direct propagation path and no

high order modulation to furtherincrease capacity.

The sub-carriers are orthogonalbecause the carrier frequencies arechosen so that there is no cross-talkbetween the sub-channels meaningguard bands between the sub-carriersare not required. The use of OFDMallows a high degree of compaction infrequency giving a very high datadensity per unit of transmission band-width. The advantage of OFDM overhigh data-rate single carrier schemesis its ability to cope with propagationissues such as frequency selectivefading, this is discussed in more detailbelow.

Use of spatial diversity is also increas-ingly common in modern mobiledevices.

This introduces two or more separatetransmission paths between the mobilephone and the cellular mast. TheseMIMO (multiple input multipleoutput) systems increase further thedata throughput to and from the phone.Again, radio system design of theseapplications involves a lot of detailed

additional paths due to reflections fromobjects. For ground based point-to-point links, reflections off the earthmean that the free space model is nolonger valid. This effect can beaccounted for by modifying Equation1 based on antenna height and groundreflection coefficient. For low anglesof reflection (long distance links) thepath loss tends to a 4th law model.

Propagation and Link BudgetsAt microwave frequencies additionalloss is incurred due to rain. An allow-ance for this (rain fade margin) must

Figure 7: QAM 256 constellation corrupted by PA non-linearities

Figure 8: QAM 256 eye diagram corruptedby PA non-linearities

Figure 9: Comparison of simulated to measured BER



be made during the system design. Themagnitude of this allowance dependson the location of the link (and so thetypical weather) and the target availa-bility (percentage of time that thespecified loss must not be exceededdue to rain fall). Statistical data existsto allow appropriate rain fade marginsto be determined.

There is also additional loss due toatmospheric absorption. This is bothfrequency dependent and altitudedependent but only really starts tobecome significant at mm-wave fre-quencies. The oxygen absorption peakat around 60GHz is an example ofatmospheric absorption.

When designing a communicationssystem the link budget must be bal-anced to ensure that there is adequatesignal power arriving at the receiver toensure acceptable Bit Error Rate(BER) for the demodulated data.

The link budget can be improved byincreasing transmit power, increasingantenna gain (transmit and/or receive),reducing noise figure or changingmodulation scheme or rate so that therequired SNR at the receiver isobtained.

For non-line-of-sight propagation,additional losses due to reflection,diffraction and scattering caused byobstructions must be included. Thesimplest way to account for this is toincrease the value of "n" in Equation1 depending on the frequency, heightsof the antennas and propagation envi-ronment. For urban environments,appropriate values of n can varybetween 2, when streets can act likewaveguides, and 6, when significantobstructions exist.

A more complex, but potentially morerepresentative, approach is to modelthe propagation channel and include

this in the system simulation. It ispossible for reflections to cause twocopies of the transmitted signal toarrive at the receiver with a phasedifference of close to 180°. This wouldbe a specific case of flat fading and itgives rise to a large dip in the receivedsignal strength.

An approach to combat this is to imple-ment receive antenna diversity, wherean additional receive antenna posi-tioned in a physically different locationis also available and will be subject todifferent multi-path propagationeffects.

The impact of multipath propagationon a transmission system can beassessed by undertaking a system sim-ulation with an appropriate multipathmodel. A number of internationallyrecognised models have been devel-oped to cover a range of differentpropagation conditions and Figure 11depicts the channel amplitude responsefor two cases of the ITU Indoor OfficeChannel A model. The Indoor OfficeChannel A model is a 6-path staticmodel having an overall 35nsec rmsdelay spread.

The difference between the tworesponses in Figure 11 is simply thatResponse No. 1 will occur at one RF

frequency and Response No. 2 willoccur at another RF frequency. Notethat the x-axis is centred about 0Hz andthis represents the centre frequency ofthe carrier.

The impact of this channel model on a5MBd QPSK signal having a 6.25MHzbandwidth will now be illustrated.Note that the signal bandwidth willoccupy the frequency range of±3.125MHz about the zero Hz centrefrequency. One would intuitivelyexpect Response No.1 to have lessimpact on the signal when comparedto Response No. 2 because the ampli-tude distortion of Response No.1 overthe ±3.125MHz band about zero Hz ismuch less than that for Response No.2 and this is indeed the case.

Figure 12 depicts the eye diagram forthe two responses. As expected, theeye opening for Response No. 1 isgood and shows much less distortionthan that for Response No. 2; theposition of the eye opening in the tworesponses of Figure 12 has been cen-tred.

The corresponding constellation plotfor the two responses is presented inFigure 13 and again highlights thebetter performance of Response 1. Itshows that a channel having Response

OFDM Channel Bandwidth

Frequency

OFDMSubcarrier

Figure 10: Simplified Image of OFDM



quency (Frequency Division MultipleAccess - FDMA) or by applying aunique code (Code Division MultipleAccess - CDMA).

Combinations of these techniques arealso used. Separation in time and/orfrequency is also used to separatereceive signals from transmit signal soproviding full duplex communications.These approaches all have differentbenefits and which is most appropriatedepends on the requirements of thesystem.

Operating Frequency andInterfering SignalsThe radio frequency spectrum is aglobal resource and its use is co-ordi-nated on a worldwide basis by theInternational TelecommunicationsUnion (ITU); an agency of the UnitedNations based in Geneva. The ITUpublishes a table of frequency alloca-tions that represents agreementamongst the 191 member states of theITU. It separates the globe into threeradio regions on geographical grounds:

� Region 1: Europe, Middle East,Africa,the former Soviet Union,including Siberia; and Mongolia;

sub-channel to propagate independ-ently along a multipath channel withvery little amplitude and phase varia-tion for that reduced bandwidth.There will still be amplitude variationbetween sub-channels, but for an indi-vidual sub-channel the aim is to havea minimal variation. This means thesub-carriers can now be recoveredwithout the need for adaptive equalisa-tion of the channel, although it doesrequire a more complex FFT baseddemodulator. It should be noted thatwhilst the need for equalisation mayhave been removed, some sub-carrierswill be transmitted in propagationnulls, frequencies where there is sig-nificant attenuation, and these subcar-riers may arrive at the receiver belowthe noise floor and therefore not recov-ered. The use of appropriate errorcorrection and interleaving of dataacross sub-carriers can be used inmitigation against this effect.

Multiple Access TechniquesTo allow multiple users access to thesame communications system, a meansof separating the different signals fromeach other must be used. This can beby separation in time (Time DivisionMultiple Access - TDMA), in fre-

2 will have more difficulties receivingthe signal. Combatting the effects ofmultipath must be taken into accountwhen a system is being designed.One effective means of combating ISIis to periodically transmit a knownsequence of data (a training sequence).The receiver knows what it expects toreceive when this sequence is transmit-ted and can remove the effects of ISIby adapting the weights of a digitalfilter (an adaptive equalizer) to essen-tially create an inverse model of themultipath interference caused by thereflections in the channel. Withmoving users (and / or objects movingin the local environment) the channeland its multi-path behaviour are con-stantly changing so the filter will becontinually adaptive.

The use of OFDM, depicted in Figure10, also helps to combat ISI. In thiscase the high data rate signal is splitinto multiple (n) sub-channels of alower data rate. Each of these sub-channels is separated in frequency bythe reciprocal of the symbol rate tominimise interference between othersub-channels. The resulting signalcomprises n sub-carriers each with adata rate and modulation bandwidthreduced by a factor n. This allows each

Figure 11: Two instances of the propagation path amplituderesponse of ITU Indoor Office Channel A



� Spectral regrowth in adjacentchannels caused by carrier self-intermodulation

� Spurious transient emissionscaused by transient effects

It is also important that communicationreceivers can operate in the presenceof legitimate communications devicesoperating at other frequencies. Thisrequires care and attention during thedesign process. The main interferencemechanisms experienced in a receiverare:� Insufficient selectivity where

imperfections in a receiver allowan adjacent transmission to corrupta wanted signal. This can be com-bated by additional filtering withinthe receiver.

� Reciprocal mixing - the phasenoise sidebands of the local oscil-lator are sufficiently high to causean unwanted signal to be down-converted into the receive band

� Insufficient intermodulationimmunity - receiver imperfectionsallow transmissions from two ormore signals to interact within thereceiver and generate an on-chan-nel product that corrupts thewanted signal …. even though the

an example of this would be cellularoperators. Yet other parts of the spec-trum are designated licence-exemptand here there is relatively weak regu-latory control of interference levels; anexample of this is the Industrial, Sci-entific and Medical (ISM) bands.

Any communications system mustadhere to the appropriate legislationand the choice of operating frequencyhas a huge impact on the cost, achiev-able performance and design complex-ity.

Radio spectrum is a shared resource,so other users could be operating onnearby frequencies. Whatever theoperating frequency of the systemthere will be strict limits on theallowed emissions at all otherfrequencies and this can be asignificant design challenge,particularly for spectrum close to theoperating band. Problems that can giverise to unwanted emissions from atransmitter include:

� Harmonic emissions� Non-harmonic spurious emissions� Adjacent channel leakage of

carrier phase noise

� Region 2: North and SouthAmerica and Pacific (East of theInternational Date Line);

� Region 3: Asia, Australia and thePacific Rim (West of the Interna-tional Date Line).

Frequency allocations may differ tosome extent between the three regionsand the table is revised periodically atWorld Radiocommunications Confer-ences. Individual countries are respon-sible for managing the radio spectrumin their own jurisdiction and generallyfollow the lead given by the ITU table,but they are free to modify the table tosuit their own needs. Any modificationwill usually be undertaken with the aimof not causing interference to othercountries as these will be assumed tobe abiding by the ITU plan.

Each country controls access to use thefrequency spectrum in their jurisdic-tion by a variety of means. For someparts of the spectrum, the countryregulator will set regulatory limits andissue licenses to control the interfer-ence between users; an example of thisbeing users of private mobile radiosystems. For other parts, the regulatorprovides less restriction and relies onusers to self-control their emissions;

Figure 12: Eye diagram for 5MBd QPSK signal subjected to ITU Indoor Office Channel A multipath model.Response No.1 (LHS) and Response No.2 (RHS)



example where the whole transceiverarchitecture is realised in silicon.Lower volume applications may berealised by a mixture of integrated anddiscrete components.

The move to full silicon integrationover the last twenty or so years hasbeen achieved by adapting transceiverarchitectures to make use of the advan-tages of integration whilst steeringaround the disadvantages. For instanceit is possible to realise integrated ana-logue filters having useful selectivityat low frequencies but not at RF fre-quencies, so receiver architectureshave been developed or adapted tomake use of this. Such architecturesinclude low-IF and zero-IF types andare the preferred approaches for highlyintegrated radios.

Figure 14 shows a block diagram of asingle conversion heterodyne trans-ceiver; an architecture that goes backto the early days of radio when designswere implemented using discrete com-ponents. The receiver mixer convertsthe signal from the transmission fre-

interfering transmissions may bewell separated from the wantedchannel

� Spurious responses - unintendedmixing products translate anunwanted signal into the receiveband

� Blocking - where a very largesignal overloads a receiverattempting to receive a weakwanted signal

Careful transceiver design is requiredto ensure the problems listed above areminimised for the application. Appro-priate filtering must be included, theindividual building blocks must haveadequate linearity, LO and IF frequen-cies must be selected with care, localoscillators must have adequate phasenoise performance, Automatic GainControl (AGC) may be necessary.

Radio ArchitecturesA broad range of radio architectureshave been developed. For high volumeworld wide applications, an aim willalways be to seek the size and costbenefits of an integrated solution. ABluetooth transceiver is one such

quency to a standard IF such as10.7MHz, 21.4MHz, 45MHz or70MHz where crystal or ceramic filtersprovide narrow band channel filteringwith high selectivity. Similar filteringin the transmitter ensures the transmit-ted signal only broadcasts on thewanted channel. Filtering at RF is usedto reject image responses caused by themixing process, provide

protection against blocking signals andreject transmitter harmonics Doubleconversion designs are used in appli-cations requiring significant IF gain.The predominance of discrete filteringmakes the architecture unsuited forintegration onto silicon and so unableto fully benefit from the miniaturisa-tion it can offer. This was the mainreason that new architectures weresought, which were better suited inintegration.

Figure 15 shows a block diagram of azero-IF transceiver. This was one ofthe first developed with full siliconintegration in mind. The receivedsignal is mixed directly down to base-band where low-pass filters now

Figure 13: Constellation diagram for 5MBd QPSK signal subjected to ITU Indoor Office Channel A multipathmodel. Response No.1 (LHS) and Response No.2 (RHS)



cal baseband channels are required tocorrectly produce the output modu-lated waveform. Imperfections in themixer will give rise to leakage of thelocal oscillator signal appearing on theoutput, and to combat this, a DC offsetcorrection (cancellation) would needto be applied.Figure 16 shows a block diagram of alow-IF transceiver; an alternativedesign for full integration. In this casethe received RF signal is mixed downto a very low frequency IF - close toDC - which, as with the zero-IFapproach, allows the channel filters tobe fully integrated.

These “polyphase” filters make use ofthe quadrature IF channel and producea passband response at the frequencyof the low-IF and a rejected responseat its image frequency. The amount ofimage rejection achieved by this archi-tecture is dependent on the degree ofmatching that can be achieved in themixers and IF and the quality of thequadrature LO signal driving themixers. The degree of image rejectionwill be finite and is likely to limit thereceiver selectivity at this frequency.

provide the channel filtering. Low passanalogue filters at baseband are suitedfor integration and can be tailored to adesired response. Two baseband chan-nels (I and Q) are required to preserveboth amplitude and phase informationin the received signal. A drawback withthe architecture is that it is susceptibleto self-reception of its own local oscil-lator with the result that a DC compo-nent (a DC offset) appears in thebaseband alongside the wanted signal.

This DC component needs to beremoved and this can be done by eitherDC blocking with a low cut-off highpass filter or by a feedback arrange-ment that measures the DC offset andcancels it by adding an equal butopposite DC component via a DAC.Another disadvantage to this architec-ture is that it is susceptible to 2nd orderdistortion whereby the envelope of anunwanted interferer is translateddirectly to baseband.

The zero-IF transmitter is well suitedto integration with the baseband mod-ulated signal appearing immediately atfinal frequency. As before, two identi-

An advantage to the low-IF receivercompared to the zero-IF (direct conver-sion) is that the self-reception thatproduced a problematic DC compo-nent can be overcome by the additionof a DC block. A second down conver-sion stage in the digital domain can beadded to obtain a baseband signaloutput.A low-IF transmitter converts a quad-rature modulated signal at a low IFdirectly up to the final frequency.Matching in the low-IF strip and quad-rature is essential to minimise theleakage of the local oscillator andleakage of the image signal to theoutput, however the degree of match-ing needed to meet regulatory require-ments is often insufficient and anexternal bandpass filter may berequired.

This approach can be problematic andso some solutions will use the simplerzero-IF architecture for the transmitterand adopt a mixed low-IF receive,zero-IF transmit design.

Further simplification of an integratedarchitecture may be possible when the

Figure 14: Single conversion heterodyne architecture



tion) is a benchmark for high levelinterference immunity.

� The transmitted EVM and adjacentchannel leakage power willdepend on the quality of the localoscillator, the modulation formatand complexity, the linearity of thetransmit strip amplifiers and theirassociated noise floors, etc.

Some, but not all, applications will beable to take advantage of a softwaredefined radio approach - where thereceived signal is mixed to a lowfrequency and applied to an ADC fordigitisation and then subsequentlyfiltered, amplified and detected in thedigital domain. The adaptability ofsuch a software implementation pro-vides a high level of flexibility to theapplication areas.

System SimulationDesigning and implementing a newcommunications system is time con-suming and expensive. Effective sim-ulation of the whole communications

As in all engineering, trade-offs willform part of the design process, forexample:

� The receiver selectivity that can beachieved will depend on a rangeof factors including the degree offiltering implemented, the qualityof the local oscillator, specificallyits sideband noise and for non-direct conversion architectures, theamount of image rejection that canbe achieved.

� The trade-off between sensitivityand linearity will require carefulspecification of the individualblocks that make up a receiverchain. Too much receiver gain,whilst favouring sensitivity, willalso reduce dynamic range andloss of performance when in thepresence of strong interfering sig-nals. Noise figure is often theguiding benchmark for the ulti-mate sensitivity that can beobtained whilst intercept point (forboth second and third order distor-

operation of the radio system is consid-ered. For example:

� A TDD system has only the trans-mitter or the receiver operating atany one time and this makes itpossible to share one local oscilla-tor between the two parts.

� Some protocols may have periodswhere it is know that there will beno signal to receive, such deadperiods can be used to undertakeself-calibration routines such as:

o Cancelling a DC offset in thereceive IF strip,

o Tuning a banded VCO to thecorrect frequency,

o Minimising the local oscilla-tor feed-though in a transmit-ter,

o Minimising the imageresponse by adjustment of thelocal oscillator I-Q phase andamplitude balance

Figure 15: Direct (zero-IF) conversion architecture



analogue circuitry and is no smallundertaking. For practical reasonssystems are often developed with aslightly scaled back approach, wheresections of the system are simulatedindependently.

If a highly integrated custom IC is tobe developed, either for a new systemor as a cost/size reduction exercise foran existing system, the need to ensureadequate performance through simula-tion is even more vital. The cost ofcommitting to an integrated implemen-tation is high, the options for imple-menting changes after fabrication arerestricted and the timescales and costsof a second iteration cause nightmaresfor many project managers. In thisinstance simulation of the completecommunications system, such as theBER simulation included in Figure 9,are highly recommended.

system prior to implementation isessential. For the RF chain a cascadedspread-sheet analysis of parameterssuch as NF, gain, IP3, IP2 and com-pression was the traditional approachand this is still a valuable tool duringthe development process.

Various system simulator packages arenow commercially available that allowthis approach to be taken a step furtherwith the incorporation of aspects suchas filter responses, phase noise profilesof local oscillators and the effects ofout of band interferers.

To fully optimise the performance ofa communications system it is neces-sary to undertake a complete end toend communications system analysisfrom bits in to bits out, including theeffects of the channel itself. Thisrequires co-simulation of digital and

Concluding RemarksThe development of a new communi-cations system is a huge undertaking.In many cases the use of an existingsystem and/or standard is the mostsensible approach offering shortesttimescales to deployment, lowestdevelopment costs and lowest risk.However, when new application sce-narios or new functional requirementsarise and the expected volumes justifythe development costs a custom com-munications system can lead to per-formance and cost benefits. This whitepaper has given an overview of thecommunications system designprocess and discussed implementationchoices and development require-ments.

Figure 16: Low-IF architecture

Communications Systems Design - PRFI · 2020. 8. 24. · QAM64 all have the same mean power level, but it is clear that the individual constellation points of the highest scheme (QAM64)

Documents