[hpsdr] Help with DSP please

[Bob Cowdery]

Hi all,

Its a long long time since I posted any messages here but I've been
monitoring the list. Over the last several months I've been building,
for my amusement, a server process for HPSDR. I've chosen Google Go for
the development and must say I'm finding it pretty good. However, I've
deliberately backed myself into a corner where I need to develop a
complete DSP processing chain in order to take full advantage of my
architecture. Now DSP is a pretty weak point with me but I'm determined
to master it this time around. I understand some of the principals but
am missing a lot of detail in going from principals to code. SharpDSP is
a good model for me to follow but I don't want to blindly translate the
code without understanding what I'm doing.

Finally, the questions.

 1. Is there a write up at a block level of SharpDSP because I'm not
    sure why blocks are ordered as they are. The purpose of most is
    obvious but I'm a bit confused by the oscillator block which I guess
    is converting to baseband(?).
 2. Demodulation is something I don't fully get. I believe the
    convolution directly demodulates CW/SSB (and from what I remember
    can convert to baseband as well) but for other modes demod is tacked
    on the end of the processing.
 3. When doing the overlap/add method of fast convolution I don't know
    how one calculates how big the FFT should be or  how many taps the
    filter should have in relation to the sample block size and what the
    effect of the choice is. Should a small block size (for reduced
    latency) have a smaller FFT? I've not found a really good algorithm
    that explains this and many I found seem to differ in approach so
    I'm not clear on the correct implementation.

Any pointers appreciated.

Bob (G3UKB)

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openhpsdr.org/pipermail/hpsdr-openhpsdr.org/attachments/20130810/51c39dd1/attachment-0003.htm>