[hpsdr] PowerSDR / Thetis / Friedrichhafen

[Warren C. Pratt]

Hello Georg,

You probably know that I wrote the WDSP library that we use.  Doug, 
W5WC, and I have worked together to write the new ChannelMaster.dll that 
is in Thetis - this is basically all the rest of the datapath code 
connecting to the network and surrounding the DSP.  So, I suppose that 
makes me as guilty as anyone of not writing more documentation.  I agree 
that we should have more documentation, and, given the time, I would 
actually enjoy writing more.  In fact, I have it on my list to update 
and expand a document about WDSP that I have provided to Hermann, 
DL3HVH, John, G0ORX/N6LYT, and a few others, that has helped them 
integrate the library into their console applications.  I sincerely hope 
to get that done this year; however, at this moment, I am busy with some 
Thetis work.  I have also tried, from time to time, to recruit other 
people to write documentation about how to use our features and console 
applications.  However, I suppose they don't view this as the most 
enjoyable way to spend their time.:-)

The "best practices" of software engineering would suggest that 
developers should plan and document their work from the top level down.  
They should then keep that documentation up-to-date as they discover 
problems and make changes.  This would provide some amount of "standard" 
documentation.

However, for the most part, I, and I suspect some of the other 
developers, don't do that.  I do have a fairly rigorous structure that I 
follow for new modules in WDSP.  So, when I write a new module in that 
same form, a number of things are already defined. Beyond that, some 
algorithms seem straightforward to me and so I just sit down and code 
them.  Some of the others, like our new Spectral Noise Reduction, NR2, 
required reading a lot of technical papers and coding and, sometimes,  
experimenting with several different versions of algorithms.  However, I 
tend to just read several papers, decide what I want to try, and then 
code it in the "WDSP form".   There are some things that I do work 
through on paper before coding.  For example, the new Spectral Noise 
Blanker involves quite a bit of matrix formation and manipulation.  I 
did work through the formation of the matrices on paper before coding 
that. However, the notes never evolved beyond being quite cryptic.  When 
I work this way, I rarely have any issues other than typographical 
errors or minor things that I simply forgot to do.

Probably some of the best algorithm documentation I have done has been 
the presentations I've prepared for Ham Radio Friedrichshafen. In 2014, 
I gave two presentations, both available online, one on WDSP basics and 
the other on PureSignal.  For Ham Radio 2016, I plan to present on 
Advanced Noise Blanking and Noise Reduction. Preparing these 
presentations requires a lot of time; however, I enjoy doing it and I 
look forward to presenting.

Well, that's a long message.  I'm as guilty as anyone of not documenting 
more.  As I can find time, I'll try to do more. However, for me at 
least, to a large extent, it's a tradeoff between developing and 
documenting.  The other developers may choose to express differing views 
on their efforts.

Hope to see you in Friedrichshafen!

73,
Warren  NR0V