0.14 Mix’n’match receiver design  (Page 4/9)

How was this ordering of components chosen? The authors have consulted with, worked for, talked about(and argued with) engineers working on a number of receiver systems including HDTV (high definition television),DSL, and AlohaNet. The ordering of components in Figures  [link] and [link] represents an amalgamation of ideas from these (and other) systems. Sometimes it is easy toargue why a particular order is good, sometimes it is a matter of preference or personal experience,and sometimes the choice is based on factors outside the engineer's control. For instance, the company might have a patent on a particular method oftiming recovery and using any other method might require royalty payments.

For example, the carrier recovery algorithms of Chapter  [link] are not greatly affected by noise or intersymbol interference (as was shownin Exercises  [link] and [link] ). Thus carrier recovery can be done before equalization,and this is the path we have followed. But it need not be done in this order. For instance, in the QAM radio of A Digital Quadrature Amplitude Modulation Radio , available in the website, the blocks appear in a different order. Another example is the placement of the timing recovery element.The algorithms of Chapter  [link] operate at baseband, and hence the timing recovery in [link] is placed after the demodulation.But there are passband timing recovery algorithms that could have been used to reverse the orderof these two operations.

Stage two: selecting components

Choices for the second design stage are relatively set as well. Since the sampling is done at a sub-Nyquist rate $f_s$ (relative to the IF frequency $f_I$ ), the spectrum of the analog received signal is replicated every $f_s$ . The integer $n$ for which $f^{†}=|f_I-n{f}_s|$ is smallest defines the nominal frequency $f^{†}$ from which further downconversion is needed.Recall that such downconversion by sampling was discussed in [link] . Using different specifications, the ${\mathcal{M}}^6$ sampling frequency $f_s$ may be above the Nyquist frequency associated with theIF frequency $f_I$ . Indeed, changing parameters such as this allows an instructor to create newtransmission “standards” for each class!

The most common method of downconversion is to use mixing followed by an FIR lowpass filter.This will be followed by an FIR matched filter, an interpolator–decimator for downsampling,and a symbol-spaced FIR equalizer that adapts its coefficients based on the training data contained in the transmission.The output of the equalizer is quantized to the nearest 4-PAM symbol value, translated back into binary, decoded(using the (5,2) block decoder) and finally turned back into readable text.

Given adequate knowledge of the operating environment (the SNR in the received signal, thecarrier frequency and phase, the clock period and symbol timing, and themarker location), the designer-selected parameters within these components can be set to recover the message. This was, in fact, the strategyfollowed in the idealized receiver of Chapter  [link] . Said another way, the choices in stages one and two arepresumed to admit an acceptable answer if properly tuned. Component selections at this point(including specification of the fixed lowpass filter in the downconverter and the fixed matchedfilter preceding the interpolator/downsampler) can be confirmed by simulations of theISI-free ideal/full-knowledge setting. Thus, the upper half of [link] is specified by stage two activities.