Topband: PSK-31

[Garry Shapiro]

Mike:

You said:

"I am not entirely
familiar with the PSK31 specification, but I suspect that
they are using root-raised cosine symbol shaping to
minimize the occuppied bandwidth of the signal."


Although I have not delved into PSK31 deeply, either, I have to take issue
with your raised-cosine statement. It's mostly not about minimizing
bandwidh.

A band-limited baseband system's transient response to a data stream
contains preshoot and aftershoot "tails" (ringing), which spill over to
preceding and succeeding data intervals (symbols). If the data stream is
sampled once per symbol, the tails contribute to the sample and, together
with noise, can cause a decision error. The primary purpose of raised-cosine
shaping is to minimize intersymbol interference due to these  "tails". It
does this not by eliminating the tails but by forcing their zero-crossings
to coincide with the sampling of the symbol in the decoder. This zeros out
their contribution to the sample, ideally leaving only the present symbol
(and noise) as components of the sample.

The rate of rolloff of the raised-cosine does affect bandwidth the closer
one approaches a brick-wall frequency response. But the narrower bandwidth
brings with it larger tails (more ringing) , which works against the primary
purpose.

The scheme only works if the overall shaping (TX, channel, RX) is
raised-cosine (or similar), and the shaping is usually split between TX and
RX--hence the root-raised cosine at each end.

Garry, NI6T