3INTEGRATIONOFRF SYSTEMSON A CHIPPeterJ.MoleNortel NetworksLondon RoadHarlow, Essex, CM 17 9NA, U.K.3.1RFISSUESBeforeconsidering the details ofimplementation ofan RF system on a chip,it is worthspendingsometimediscussingtheproblemsthatanyradiosystemmustovercomeiftoachievean acceptableperformance.The importantaspecttorememberis that theradiosystem never operates in isolation. It is not sufficient to think only of the transmitter,the link and the receiver, though many problems lie in this simple chain alone. It is alsoimportant to realise that the transmitter can interfere with other links, and a receivermay be unduly sensitive to unwanted, but entirely legitimate, signals.3.1.1ReceiverconcernsIn any system, a band is defined in which the receiver may receive signals.The re-ceiver must be ableto reject signals outside theband without loss of performance.These signals maybe very large(e.g.TV transmissions)and quite capable of over-loading a sensitive receiver. If the receiver is overloaded, amplifier compression willreduce the receiver gain and hence the ability to detect weak signals, or the non-lin-earities in the receiver that are excited by the overload will allow unwanted signalsto intermodulate.This may result in a distortion product falling onto the wanted sig-23
3 INTEGRATION OF RF SYSTEMS ON A CHIP Peter J. Mole Nortel Networks London Road Harlow, Essex, CM 17 9NA, U.K. 3.1 RF ISSUES Before considering the details of implementation of an RF system on a chip, it is worth spending some time discussing the problems that any radio system must overcome if to achieve an acceptable performance. The important aspect to remember is that the radio system never operates in isolation. It is not sufficient to think only of the transmitter, the link and the receiver, though many problems lie in this simple chain alone. It is also important to realise that the transmitter can interfere with other links, and a receiver may be unduly sensitive to unwanted, but entirely legitimate, signals. 3.1.1 Receiver concerns In any system, a band is defined in which the receiver may receive signals. The receiver must be able to reject signals outside the band without loss of performance. These signals may be very large (e.g. TV transmissions) and quite capable of overloading a sensitive receiver. If the receiver is overloaded, amplifier compression will reduce the receiver gain and hence the ability to detect weak signals, or the non-linearities in the receiver that are excited by the overload will allow unwanted signals to intermodulate. This may result in a distortion product falling onto the wanted sig- 23
24CIRCUITSANDSYSTEMSFORWIRELESSCOMMUNICATIONSnal and effectivelymasking it.It is therefore imperative that these signals are heavilyfiltered before the input to the receiver.The receiver must respond to any signal in the system band, however, it must alsoreject unwanted in-band signals without suffering from overload.Such unwanted signals may be significantly larger than the wanted, consider the situation when you aretrying to make a cellular phone link to your service provider's remote base-station but,unfortunately,you arephysicallyadjacent to a second service provider's base-station.ThesystemspecificationprovidesguidelinestothelevelsofsignalswhichcanexistFig.3.1 shows the relativelevels for signals in the DECT cordless phone system.Inthis system,wherecommunication channels areallocatedtoboth time andfrequencyslots, the frequency channels are separated by 1.728 MHz.A weak signal, (defined intheDECT specification as-73dBm)musstbereceivableinthepresenceofasignal13dBstrongerintheneighbouringchanneland34dBstrongerinthenextneighbour-ing channel.To achieve this reiinterferers,atunablefilterisrequired.Inpracticethisisachievedbsignaltoafixedintermediatefrenequency andthen filtering.the factthatit is more difficultThenetofilter the neighbouring channeUntilthesignalscanbefiltered.however.theradiomusthetheinterferersdonotcreateoverload.Unlikeout-of-bandblockingsignin-bandinterfererscannotbefilteredpriorto the sensitiveinputofthereceiverSignal1.728MHzPower34dB13dBFrequency1MHzFigure3.1Adjacentchannel interferersforDECTsystem.The frequency translation that is required in a receiver to move the radio frequencyinputtotherequiredintermediatefrequencyforfilteringisachievedwithamixer.Amixereffectivelymultiplies theRFinput signal witha chosen local oscillator (LO)signal.Thus we can ideally represent the RF input by sin(wrt)and the local oscillatorby cos(wzt).Thus the output can be written as:sinWrtcoswLt=0.5(sin(Wr+wL)t+sin(or-WL)t)From this wecan seeimmediatelythat an unwantedoutput isproducedat the sumfrequency,butthis is easilyfilteredbecause ofthefrequency difference.Amore subtledefect is thattwo inputfrequencies(wz±wrr)can produce signals at theintermediatefrequency (wrr).Thus there are two frequencies we are sensitive to, the wanted and the
24 CIRCUITS AND SYSTEMS FOR WIRELESS COMMUNICATIONS nal and effectively masking it. It is therefore imperative that these signals are heavily filtered before the input to the receiver. The receiver must respond to any signal in the system band, however, it must also reject unwanted in-band signals without suffering from overload. Such unwanted signals may be significantly larger than the wanted; consider the situation when you are trying to make a cellular phone link to your service provider’s remote base-station but, unfortunately, you are physically adjacent to a second service provider’s base-station. The system specification provides guidelines to the levels of signals which can exist. Fig. 3.1 shows the relative levels for signals in the DECT cordless phone system. In this system, where communication channels are allocated to both time and frequency slots, the frequency channels are separated by 1.728 MHz. A weak signal, (defined in the DECT specification as –73 dBm) must be receivable in the presence of a signal 13 dB stronger in the neighbouring channel and 34 dB stronger in the next neighbouring channel. To achieve this rejection of neighbouring interferers, a tunable filter is required. In practice this is achieved by mixing the signal to a fixed intermediate frequency and then filtering. The specifications reflect the fact that it is more difficult to filter the neighbouring channel than more distant ones. Until the signals can be filtered, however, the radio must be designed so that the interferers do not create overload. Unlike out-of-band blocking signals, in-band interferers cannot be filtered prior to the sensitive input of the receiver. The frequency translation that is required in a receiver to move the radio frequency input to the required intermediate frequency for filtering is achieved with a mixer. A mixer effectively multiplies the RF input signal with a chosen local oscillator (LO) signal. Thus we can ideally represent the RF input by and the local oscillator by Thus the output can be written as: From this we can see immediately that an unwanted output is produced at the sum frequency, but this is easily filtered because of the frequency difference. A more subtle defect is that two input frequencies can produce signals at the intermediate frequency Thus there are two frequencies we are sensitive to, the wanted and the
25INTEGRATIONOFRESYSTEMSONACHIPimage frequency. It is therefore essential that any signal present at the image frequencyisfilteredout infrontofthemixer.Inpractice,toachieve efficientmixing,itisnormal thattheLO signal isclosetoa square wave.Thus the LO signal will contain the third harmonic of the Lo (about9 dB lower in signal amplitude)which is also multiplied by the input signal. Thisimmediatelymeansthatsignals at(3wz±wr)willalso appearattheoutput as anintermediate-frequency signal.Other harmonics will behave similarly.Again inputsignals at thesefrequencies must be removed prior tomixing.Another defect of the LO signal is that it suffers from phase noise (or jitter). Thismeans the LO is not a pure tone but a spectrum of signals centred on the desiredfrequency.Thus the mixer will respond to signals close to the wanted signal with asensitivityfalling off with the phase-noise spectrum.This problem is known as recip-rocal mixing. Thus, to ensure that the receiver can reject neighbouring channels, it isimportant to provide an LO signal with lowphase noise.These issues are illustratedinFig.3.2.WantedImageLOleakage3rdharmonicPhasenoise/spursonLOLOFigure3.2Non-idealitiesinamixer.3.1.2TransmitterconcernsThetransmitter must efficiently produce enoughoutput powerto ensurethat the sys-tem has sufficient range whilst ensuring that power is not transmitted into neighbour-ingchannels.Itis importanttounderstand howpowercanbetransmittedatunwantedfrequencies.Firstly,harmonicswillbegeneratedinanefficientpoweramplifier.Thesecan usually be filtered out before they reach the antenna, but it must be rememberedthat any filter will also attenuate the wanted output, hence reducing the overall effi-ciency and increasing the current consumption from the battery.Non-linearity in thepower amplifier will spread the spectrum of the signal into neighbouringbands.Thisbehaviour is known as spectral regrowth. It is worse with modulation schemes that are
INTEGRATION OF RF SYSTEMS ON A CHIP 25 image frequency. It is therefore essential that any signal present at the image frequency is filtered out in front of the mixer. In practice, to achieve efficient mixing, it is normal that the LO signal is close to a square wave. Thus the LO signal will contain the third harmonic of the LO (about 9 dB lower in signal amplitude) which is also multiplied by the input signal. This immediately means that signals at will also appear at the output as an intermediate-frequency signal. Other harmonics will behave similarly. Again input signals at these frequencies must be removed prior to mixing. Another defect of the LO signal is that it suffers from phase noise (or jitter). This means the LO is not a pure tone but a spectrum of signals centred on the desired frequency. Thus the mixer will respond to signals close to the wanted signal with a sensitivity falling off with the phase-noise spectrum. This problem is known as reciprocal mixing. Thus, to ensure that the receiver can reject neighbouring channels, it is important to provide an LO signal with low phase noise. These issues are illustrated in Fig. 3.2. 3.1.2 Transmitter concerns The transmitter must efficiently produce enough output power to ensure that the system has sufficient range whilst ensuring that power is not transmitted into neighbouring channels. It is important to understand how power can be transmitted at unwanted frequencies. Firstly, harmonics will be generated in an efficient power amplifier. These can usually be filtered out before they reach the antenna, but it must be remembered that any filter will also attenuate the wanted output, hence reducing the overall efficiency and increasing the current consumption from the battery. Non-linearity in the power amplifier will spread the spectrum of the signal into neighbouring bands. This behaviour is known as spectral regrowth. It is worse with modulation schemes that are
26CIRCUITSANDSYSTEMSFORWIRELESSCOMMUNICATIONSnot constant amplitude.Intermodulation is usuallyat itsworst inthepoweramplifieritself, but intermodulation throughout the transmitter must be considered.Broadbandnoiseinthetransmittereitherduetonoiseinanaloguecircuitelementsorduetoquantisationnoiseifthesignal isgeneratedfromaD-to-Aconvertermustbeconsidered.The latter, in particular,will require careful attention to band filtering toensure low spurious emissions.A low-phase-noise local oscillator is also required tokeepthenoiseemissionstoacceptablelevelsIn a mechanism similar to phase noise, spurious signals present on the local oscil-lator will also give rise to spurious emissions on the output signal.Spurious signalscan arise from thefrequency synthesis process or from unintentional couplings to theVCO.On single-chipsystems,theidentificationofthesecouplings canbeparticularlydifficult, but attention must be paid to single-ended CMOS input signals which couplestrongly to the substrate and also to high-power outputs.An example oftheformer isgiven in [1]. These effects are illustrated in Fig. 3.3.WantedSignalSignalPowerHarmonicofwantedPhase NoiseSpuriousand intermodsignalsignal spreacSystem NoiseBandFrequencyforsystemFigure3.3Non-idealitiesinatransmitter.3.2RADIOARCHITECTURESWith the issues discussed in the previous section, the advantages of different architec-tures can be discussed both in terms of performance and suitability for integration.3.2.1ReceiverarchitecturesThe double-superheterodyne architecture (Fig.3.4).This is probably the mostcommonlyemployed architecture in current wirelesssystems.Theout-of-bandblock-ing signals are reduced by an RF bandpass filter placed immediatelyafterthe antennaThe signal is then amplified byan LNA,whichmust havea sufficiently low noiseto al-lowdetectionof weak signalsbutmustalsohave thedynamicrangeto handlein-bandinterferers.Thebandpassfilterisusuallyinsufficienttoreducesignals attheimagefrequencytothesystemnoiselevel,and soasecondimagefilterisinsertedpriortomixing.To ensurethat the image issufficientlyfarawayfrom the wanted signal toallow effective filtering, a relatively high first intermediate frequency must be chosen(for1to2GHzRFsystemsanIFof100-200MHziscommon)
26 CIRCUITS AND SYSTEMS FOR WIRELESS COMMUNICATIONS not constant amplitude. Intermodulation is usually at its worst in the power amplifier itself, but intermodulation throughout the transmitter must be considered. Broadband noise in the transmitter either due to noise in analogue circuit elements or due to quantisation noise if the signal is generated from a D-to-A converter must be considered. The latter, in particular, will require careful attention to band filtering to ensure low spurious emissions. A low-phase-noise local oscillator is also required to keep the noise emissions to acceptable levels. In a mechanism similar to phase noise, spurious signals present on the local oscillator will also give rise to spurious emissions on the output signal. Spurious signals can arise from the frequency synthesis process or from unintentional couplings to the VCO. On single-chip systems, the identification of these couplings can be particularly difficult, but attention must be paid to single-ended CMOS input signals which couple strongly to the substrate and also to high-power outputs. An example of the former is given in [1]. These effects are illustrated in Fig. 3.3. 3.2 RADIO ARCHITECTURES With the issues discussed in the previous section, the advantages of different architectures can be discussed both in terms of performance and suitability for integration. 3.2.1 Receiver architectures The double-superheterodyne architecture (Fig. 3.4). This is probably the most commonly employed architecture in current wireless systems. The out-of-band blocking signals are reduced by an RF bandpass filter placed immediately after the antenna. The signal is then amplified by an LNA, which must have a sufficiently low noise to allow detection of weak signals but must also have the dynamic range to handle in-band interferers. The bandpass filter is usually insufficient to reduce signals at the image frequency to the system noise level, and so a second image filter is inserted prior to mixing. To ensure that the image is sufficiently far away from the wanted signal to allow effective filtering, a relatively high first intermediate frequency must be chosen (for 1 to 2 GHz RF systems an IF of 100–200 MHz is common)
27INTEGRATION OF RFSYSTEMS ONACHIP图工RFbandImageChannelGainFilterFilterFilterMixto basebandQuadratureLO2LO1Figure3.4Thedouble-superheterodynearchitecture.The mixer must still handlethe complete dynamic range of the in-band signalAfter the mixer, a SAW filter can be used to achieve the channel filtering.At thesefrequencies the SAW filter is small but usuallyhas a large in-band loss when completechannel filtering is tobe achieved.The output drive of themixer must thereforeboostthesignal level toallowforthisloss.Once the interfering channels have been attenuated, the signal can be boosted toa high level (it can be limited if a constant-amplitude modulation scheme is used)The signal is then reduced to baseband frequencyfor demodulation.It is of coursepossible to split the channel filtering between thetwo intermediatefrequencies.Thiswill requireagreater dynamic range in the second mixer.This architecture requires the synthesis oftwo local oscillators, and their frequen-cies mustbe chosen so that spurious responses fromthe radioarekeptto a minimum.This aspect of frequency planning, which will not be discussed in more detail here, isa well-understooddesign process which requires considerable care and experienceThis design requires several external filters and therefore does not lend itselfto easyintegration as the pin count increases.Moreover,thefilters are usually single endedthough this is not essential-and hence achieving isolation between pins becomes anissue.In particular,the channel filter will often need to provide50dB of attenuationat keyfrequencies, thus implying thatgreater isolationmust be achieved between thepins and with respect to signal ground ifthe filter response is not to be degraded. Theimagefiltercanbeeliminatedifanimage-rejectingmixeris used.Thiswill preventthe need to come offchip after the LNA and makes an LNAplus image-reject mixer auseful integrated building block.Thedirect-conversion architecture (Fig.3.5).Thedirect-conversion receiver,be-cause of its simplicity,appears to offer the best opportunityfor integrated systems.Someexamplesofitsuseinwirelesssystemsdoexisttoday,butitisnotassimpleinpracticeOnce again, an RF bandpass filter is placed at the input.The LNA's output ispassed into the mixer.The LNA must handle the same dynamic range as for thesuperheterodyne architecture and it must have enough gain to lift weak signals abovethe noise of the mixer. The mixer however, now converts directly to baseband. Thusthesignal is its own image,and channel filtering cannowbecarried out by low-pass
INTEGRATION OF RF SYSTEMS ON A CHIP 27 The mixer must still handle the complete dynamic range of the in-band signal. After the mixer, a SAW filter can be used to achieve the channel filtering. At these frequencies the SAW filter is small but usually has a large in-band loss when complete channel filtering is to be achieved. The output drive of the mixer must therefore boost the signal level to allow for this loss. Once the interfering channels have been attenuated, the signal can be boosted to a high level (it can be limited if a constant-amplitude modulation scheme is used). The signal is then reduced to baseband frequency for demodulation. It is of course possible to split the channel filtering between the two intermediate frequencies. This will require a greater dynamic range in the second mixer. This architecture requires the synthesis of two local oscillators, and their frequencies must be chosen so that spurious responses from the radio are kept to a minimum. This aspect of frequency planning, which will not be discussed in more detail here, is a well-understood design process which requires considerable care and experience. This design requires several external filters and therefore does not lend itself to easy integration as the pin count increases. Moreover, the filters are usually single ended— though this is not essential—and hence achieving isolation between pins becomes an issue. In particular, the channel filter will often need to provide 50 dB of attenuation at key frequencies, thus implying that greater isolation must be achieved between the pins and with respect to signal ground if the filter response is not to be degraded. The image filter can be eliminated if an image-rejecting mixer is used. This will prevent the need to come off chip after the LNA and makes an LNA plus image-reject mixer a useful integrated building block. The direct-conversion architecture (Fig. 3.5). The direct-conversion receiver, because of its simplicity, appears to offer the best opportunity for integrated systems. Some examples of its use in wireless systems do exist today, but it is not as simple in practice. Once again, an RF bandpass filter is placed at the input. The LNA’s output is passed into the mixer. The LNA must handle the same dynamic range as for the superheterodyne architecture and it must have enough gain to lift weak signals above the noise of the mixer. The mixer however, now converts directly to baseband. Thus the signal is its own image, and channel filtering can now be carried out by low-pass