The main approach for solving the multipath problem would be a long symbol duration and some interleaving for error correction. The symbols should be longer than a typical delay between two paths.
Since the phase measurements on a typical hf channel cannot be very sensitive because of noise, the symbol would be distributed on several simultaneous carriers (that is the approach used e.g. in DRM digital shortwave broadcast and in the european DVB-T television standard - and FreeDV without the interleaving). Long symbols plus interleaving would introduce some delay into the communication. FreeDV uses (I believe) 14 carriers (differential phase keying) and roughly 1500 bits/sec, resulting in a symbol length of 10 ms which is in the order of magnitude of a typical multipath delay (as reported by VOACAP propagation software or by the DRM channel analysis graph in Dream) Since FreeDV does not use any interleaving for error correction, the only delay is the symbol length (plus some processing time). I guess the individual carriers would need a signal/noise ratio that is similar to one of the PSK (maybe PSK125?) modes in order to be reliably detected. The problem is, the multipath delays on the channel (in time) correspond to a sum of sine functions in "frequency space" - if two paths are about the same strength, they could completely "interfere out" on some frequencies causing individual carriers disappearing in the noise - I have no idea what FreeDV does if one carrier is completely missing. This same problem occurs with analog signals. Much of the information transported by the voice consists of the relative phase/amplitude relations of the pitch (fundamental frequency of the voice) and its higher harmonics (this is basically the largest fraction of the information the FreeDV Codec converts into a bitstream). If your channel destroys this information by taking many (or the most significant) of the higher harmonics out, you can not read the audio any more. If the individual harmonics are not burried in the noise, it should however be possible to restore amplitude/phase information by applying the "reverse" of the transfer function of the hf channel to the signal in Fourier space. Maybe, a kind of "equalizer" that not only aligns the frequency bands but also the relative phases could be used. The main problem is the large number of parameters you would have to align (two per aditional path - one corresponding to the delay plus one corresponding to a relative ampl itude), if more than two paths are present. However, I am no expert in DSP, and not sure if this approach would result in a "causality problem"... Greetings Ralf, DL6OAP ______________________________________________________________ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html
