0.14 Mix’n’match receiver design  (Page 8/9)

Finally, after tuning each component separately, it is necessary to confirm that when all the pieces of the system are operatingsimultaneously, there are no excessive negative interactions. Hopefully, little further tuning will prove necessaryto complete a successful design. The next section has more specifics about the ${\mathcal{M}}^6$ receiver design.

No soap radio: the ${\mathcal{M}}^6$ Receiver design challenge

The analog front end of the receiver in [link] takes the signal from an antenna, amplifies it, and crudely bandpass filters it to (partially) suppressfrequencies outside the desired user's frequency band. An analog downconverter modulates the received signal (approximately)down to the nominal intermediate frequency $f_I$ at 2 MHz. The output of the analog downconverter is set by an automatic gaincontroller to fit the range of the sampler. The output of the AGC is sampledat intervals of $T_s=850$  kHz to give $r\left[k\right]$ , which provides a “Nyquist” bandwidth of 425 kHz that is ample for a102 kHz baseband user bandwidth. The sampled received signal $r\left[k\right]$ from [link] is the input to the DSP portion of the receiver in [link] .

The following comments on the components of the digital receiver in [link] help characterize the design task:

• The downconversion to baseband uses the sampler frequency $f_s$ , the known intermediate frequency $f_I$ , and the current phase estimatesto determine the mixer frequency needed to demodulatethe signal. The ${\mathcal{M}}^6$ receiver may use any of the phase tracking algorithms of Chapter  [link] . A second loop may also help with frequency offset.
• The lowpass filtering in the demodulator should have a bandwidth of roughly 102 kHz, which will cover the selected source spectrumbut reject components outside the frequency band of the desired user.
• The interpolator/downsampler implements the reduction in sample rateto $T$ -spaced values. This block must also implementthe timing synchronization, so that the time between samples after timing recovery is representative of thetrue spacing of the samples at the transmitter. You are free to implement this in any of the ways discussed inChapter  [link] .
• Since there could be a significant amount of intersymbol interference due to channel dynamics,an equalizer is essential. Any one will do. A trained equalizer requires finding the start of themarker/training segment while a blind equalizer may converge more slowly.
• The decision device is a quantizer defined to reproduce the known alphabet of the $s\left[i\right]$ by a memoryless nearest-element decision.
• At the final step, the decoding from blockcode52.m in conjunction with bin2text.m can be used to reconstruct the original text. This also requires a frame synchronization that finds andremoves the start block consisting of marker plus training, which is most likely implemented using a correlation technique.