ABSTRACT
The ability to support multiple communications channelsper RF band is a fundamental process for many softwaredefined radio platforms. These platforms typically employa channelizer to extract channels from the received RFband for follow-on baseband processing, or to insertchannels into the RF band for transmission. This paperwill compare and contrast three of the more popularchannelization techniques: Digital Down Conversion,Frequency Domain Filtering, and Polyphase FFT FilterBanks. The analysis begins by presenting a basearchitecture for a wideband transceiver, and then exploreseach channelization method within the context of thisarchitecture. These include the computational complexityof the channelization approach, the applicability of theapproach in supporting a given frequency plan, andprocessor selection for the proposed implementation.
1. INTRODUCTION
Wireless transceiver technologies typically break intoeither narrowband or wideband architectures, withnarrowband systems typically supporting only a singlecarrier per RF channel and wideband systems supportingmultiple simultaneous carriers. Narrowband architecturesare often utilized in portable radios, including cellularhandsets and tactical communications systems, wheresupport for multiple channels is typically not necessary,and the interference rejection inherent in the narrowbandarchitecture is important (see Figure 1). Conversely,wideband architectures are often utilized for infrastructuresystems, including cellular base stations, tacticalcommunications gateways, and satellite communicationshubs, where dozens or even hundreds of simultaneouscarrier channels may be active at any one time. Widebandarchitectures are also useful in a number of nicheapplications for military and civil defense, includingsignals intelligence and electronic warfare.A key element of the wideband receiver architectureis the channelization technique that is used to isolate theindependent communication channels contained withinthe wideband signal. This paper will explore several ofthe predominant channelization techniques utilized inwideband transceiver systems, including examining someof the practical aspects of implementation of eachchannelization method in a software defined radioplatform. The paper will focus primarily on receiverarchitectures, but the techniques and analysis presentedare equally applicable for transmission as well.
2. BASE ARCHITECTURE OF A CHANNELIZEDTRANSCEIVER
The base architecture for a wideband transceiver ispresented in Figure 2 [1]. In this architecture, a singlechannelization engine supports multiple channelprocessing elements. On the receive side, the channelizerextracts the channels of interest from the digitized RFbands, and then forwards these channels on to channelprocessing for demodulation and decoding. This processis reversed on the transmit side, with payload dataencoded and modulated in the channel processors andthen inserted into the output signal by the channelizer forretransmission. The number of channel processorssupported by the channelizer is set based on the targetnumber of active carriers operating at any given time.Figure 2: Typical Architecture for Wideband TransceiverThe interface between the channelizer and the RFfront end in this architecture comes from the analogconversion subsystem within the digital transceiver. Twoconversion techniques are typically employed: IFsampling and Zero-IF conversion (see Figure 3). In IFsampling systems the wideband channel into and out ofthe transceiver is centered at a predefined intermediatefrequency (IF). In a Zero-IF conversion scheme, thewideband signal is converted to baseband with inphase (I)and quadrature (Q) channels passed between the RF frontend and the converter devices.Zero IF Wideband ReceiverIF Sampling Wideband ReceiverIQBalance(FPGA)RF ChannelizerReceiver A/DToDemodIF Signal(21.4 MHz,70 MHz,160 MHz)Sample Rate =2*(IF + BW/2)RFReceiverI Channel A/D(0 MHz)Q Channel A/D(0 MHz)ToDemodToDemodToDemodChannelizerToDemodToDemodToDemodToDemodFigure 3: Front End Architectures for Wideband ReceiversIF Sampling typically requires front endchannelization processing to operate at a significantlyhigher rate than a Zero-IF scheme for a given bandwidth.For example, a 70 MHz IF signal with 60 MHz ofbandwidth requires a sampling rate of greater than 200MSPS. This same bandwidth can be supported throughbaseband sampling at 60 MSPS, allowing thechannelization processing engine to operate at a muchlower rate. Note that, in general, dynamic range decreasesas sample rate increases, so identifying a converter devicewith sufficient dynamic range for a target application maybe problematic in an IF Sampling architecture.The Zero-IF technique also comes at a price: in aZero-IF architecture an imbalance in the amplitude orphase of the I and Q arms of as little as a tenth of a dB canreduce the available dynamic range to less than 40 dB [2].Maintaining this level of balance prior to the A/Dconverters is problematic, and as such compensation forIQ imbalances must occur in the digital domain prior tochannelization processing. A number of techniques areavailable to support IQ balancing, some of which may beincorporated directly into the channelizer engine,depending on the channelization approach [3, 4, 5].
3. ISSUES DRIVING THE CHANNELIZATIONARCHITECTURE
Two key issues defining the technical requirements of thechannelizer are the spectral content of the widebandchannel of interest and the types of processing devicesavailable for channelization processing. These issues arediscussed in detail in the following sections.


Download full report
http://spectrumsignalpublications/channel_techniques/Channelization_Paper_SDR_forum.pdf