WiMAX continues to grow in popularity, and the solidification of the802.16 standard has ensured that thetechnology will be around for the long haul. As the fixed 802.16dstandard becomes more readily available, the industry's focus turns tothe mobile aspect of WiMAX inthe 802.16e standard.
The focus is initially on customerpremises equipment (CPE) where high-volume distribution isthebusiness model. WiMAX basetransceiver station (BTS) solutions are following closely tocomplete the system. The BTS is a lower-volume (and higher-cost) systemwhere the RF architecture's performance generally takes precedence overcost.
Certainly, there is not an absolute right architecture for theseWiMAX systems. Each option has advantages and disadvantages that thedesigner must weigh against the system's goal.
The first consideration for the RF architecture is to define the signalinterface between the digital baseband processor and the RF circuitry.
|Figure1. For most applications, the choice of RF chain conversionarchitecture is between a superheterodyne and direct-conversion radio.This choice determines the types of devices used and the requirednumber of filters.|
The two choices for the signal interface are low-IF and I/Qinterfaces. The digital baseband processor may force the designer'sdecision as it may have only one interface option.
Other processors may provide a choice between a low-IF and an I/Qinterface. Since the digital processors for CPE usually contain thedata converters, the interface to the RF circuit is predefined.
The BTS will have more flexibility as the data converters areexternal devices that can be selected to match the chosen architecture.
The low-IF interface uses a single signal path for the transmitterand receiver to the data converters. The choice of the low-IF frequencymay be limited by the RF chipset circuitry or delegated by the samplingrate of the data converters.
This sampling rate must be at least twice that of the highestfrequency in the signal bandwidth. Eliminating the spurious outputgenerated from the converter is imperative to meet the WiMAX standard.
TheI/Q interface uses twosignals in quadrature that connect to the data converters. The signalsare usually at baseband, although it is conceivable to have an I/Qinterface with the signal centered at some low IF.
An I/Q interface at low IF is not generally used for CPEapplications, but can have a useful place in BTS applications wherehigh-performance data converters are available.
Because the signal is split into its quadrature components, thesignal bandwidth of each path is reduced to half, thus the samplingrate of the data converters is not as stringent, though two converters(or a dual) are required.
While the low-IF interface can simply use a mixer to convert theinput signal to a higher frequency value, the I/Q interface requiresthe use of a modulator or demodulator.
The term modulator will be used henceforth to refer to both themodulator and demodulator, as the characteristics are similar betweenthe devices.
The modulator internally uses two mixers driven with two LO signalsin quadrature. Although the modulator is more complicated than thesimple mixer, it has a key advantage in that the LO signal and theimage frequency are naturally suppressed.
The amount of suppression is determined by two key parameters: DCoffset balance and quadrature balance. The DC offset balance between thein-phase (I) path and thequadrature (Q) path determines theamount of carrier feed-through.
Specifically for the transmitter operation, suppressing oreliminating the carrier is crucial since it is very close to thedesired signal.
The amplitude and phase balance between the signal paths willdetermine the suppression of the image signal or unwanted sideband.Similar to the DC offset conditions, suppressing the image frequency,since it is usually close to the desired signal, is very important forregulatory compliance.
The parameters are generally very sensitive, and it isn't uncommonto have contingencies for fine-tuning the DC offset, amplitude andphase balance between the I/Q paths due to variations between lots andover frequency and temperature. The calibration and adjustment of theseparameters over environmental conditions is a critical component of thearchitecture and implementation of the I/Q interface.
RF chain conversion
Once the interface has been established, the designer must choose theappropriate RF chain conversion architecture. For most applications,the choice is between a superheterodyne and direct- conversion radio.This choice determines the types of devices used and the requirednumber of filters. Outside the chipset itself, the filter cost is thehighest in the radio.
The superheterodyne is a dual-conversion approach that converts theinput signal to an IF and then converts it again to the appropriate RFchannel. With this architecture, the IF is static and allows the use ofa highperformance SAW filter.
The final conversion mixer uses a tunable LO to place the signal atthe desired output channel. The IF SAW filter eliminates most of theunwanted spurious outputs from the DAC or first conversion mixer beforethey can reach the transmitter amplifiers and PA and get broadcastthrough the antenna.
On the receive side, the SAW filter is used to block adjacentchannel interferers and jammers from significantly affecting thesensitivity of the radio. The RX blocker performance is a key parameterin ensuring WiMAX compliance.
The direct-conversion architecture converts the input signaldirectly up to the desired RF channel; no intermediate frequency isused. While the superheterodyne radio can use either a low-IFconversion or I/Q modulator to convert to the intermediate frequency,the direct conversion radio must use the quadrature modulator.
Since there is no highly selective SAWfilter in that architecture, the modulator must be used tosuppress the image frequency and the carrier component.
The LO frequency is tuned to the appropriate channel to convert theinput signal to the desired RF channel. On the receive side, where theblocking characteristics are critical, the direct down-convert radiomust have good dynamic range and suitable baseband filtering to handlethe adjacent channel interferers and narrowband jammers withoutsignificantly degrading the sensitivity of the radio.
Another key parameter for direct-conversion radio is transmit outputnoise specified in dBc/Hz. Without the IF filtering, all the noise fromthe modulator hits the PA and the antenna. Thus, the dynamic range mustbe high and the output noise low to meet WiMAX and regulatoryspecifications.
The direct-conversion architecture uses a minimum number of filtersand synthesizers, which makes it an attractive architecture forlow-cost initiatives, but puts significant performance demands on themodulator and demodulator device. Care must be taken that theappropriate device is capable of meeting all regulatory standards tomake it a viable approach.
The CPE is a high-volume product that must meet not only the WiMAXspecifications, but also the low-cost targets and manufacturingrobustness to be successful in the market.
As previously stated, the selection of the baseband processor is thekey parameter and its interface will likely dictate the low IF or I/Qinterface chosen. Most of the CPE baseband solutions available todaycan be configured for either low IF or I/Q interface and there is not asignificant limitation for this choice. The selection of the conversionarchitecture is a more relevant matter.
For the next wave, there is a push to maintain the performancecriteria associated with existing deployments of fixed WiMAX whileincreasing integration and lowering cost for mobile applications.
This technique can use a direct-conversion architecture with anexternal power amplifier(PA) andan external LNA, or a superheterodyne architecture with integrated PAand LNA in the RF front-end chip. Both architectures can succeed inreducing costs, increasing integration and reducing the chipset to twoor three devices, thus saving board space.
Direct conversion must be able to satisfactorily maintain DC offsetbalance and quadrature balance, and have suitable dynamic range tohandle the blocker signals to be effective. The superheterodyneapproach will use an additional SAW filter that will help meet spuriousand Rx blocker requirements. With proper implementation, the additionalfilter and synthesizer needed for the approach will be offset by theintegrated performance of the PA.
Cost and board area will likely be about the same between the twoapproaches, where the deciding factor will be the PA performance andcost. For BTS systems, the emphasis is more on performance than on costand size, although there still is an interest in low cost since WiMAXis a new deployment.
The initial WiMAX systems used an open-loop PA that would deliver2-4W modulated to the antenna. The PA strategy for this approach was touse a sufficiently high power PA final transistor (or combination oftransistors) to achieve the desired modulated power out by sufficientlybacking off the amplifier enough to get suitable error vector magnitude(EVM) performance.
While this approach works, it is very inefficient and requires largeheat sinks and active cooling. Nextgeneration BTS systems will bedriving higher powers and the back-off linearization approach becomesunfeasible.
Designers will then start investigating other linearizationtechniques. Initially, crest factor reduction (CFR) will beused to reduce the peak-toaverage ratio(PAR) of the incoming signal in the digital domain before theconversion to analog.
Because of the stringent in-band signal integrity requirement of theWiMAX signal, this technique will likely only provide 1.5-2dB of PARreduction before the inherent EVM degradation of CFR becomes toosignificant.
For more significant improvement, designers will use digitalpre-distortion (DPD) linearization. This technique will modify thedigital input signal in such a way that by the time it goes through thenonlinear power amplifier, the unwanted intermodulation products aresuppressed. This approach is widely gaining favor in the cellular PAmarket and is now being investigated for WiMAX systems.
DPD has two special architectural requirements. The firstrequirement is for a feedback path after the PA to send the outputsignal to the DPD processor for adaptive adjustment of thelinearization coefficients to maintain good linearization over outputpower and environmental condition changes.
This requirement comes free withtime division duplex (TDD) WiMAXsystems, as the receiver is already in existence and idle during thetransmit cycle. The second requirement relates to the bandwidth of thepre-distorted signal. The DPD signal will encompass the thirdandfifth-order products that are used to suppress the ones generated bythe PA.
This results in an input signal that is five times as large as thedesired signal. The SAW filters used in the superheterodyne radio thatare only as wide as the desired signal will not sufficiently pass theDPD signal.
Furthermore, the long group delay and group delay variation over thebandwidth properties associated with a SAW filter have a detrimentalaffect on the DPD adaptation algorithm.
As such, the direct up-convert approach is the most viable approachusing DPD. The direct up-convert approach with DPD puts significantstrain on the performance of the quadrature modulator and the DAC thatdrives it.
For reasons of symmetry, the BTS receiver may also use a directdownconversion architecture. However, the stringent blockingrequirements may dictate the use of narrowband filtering used in thesuperheterodyne approach.
The BTS designer may opt for a hybrid architecture approach wherethe transmitter employing DPD uses a direct up-convert modulator andthe receiver uses a dual-conversion with an option to bypass the IF SAWfilter when the receiver is used as the DPD feedback path. Thistechnique uses the best advantages of each architecture for the entiresystem.