[hpsdr] Some things worthy of your attention

[James Courtier-Dutton]

Robert McGwier wrote:
> With the 200 kHz wide A/D's available in 24 bit packages such as the 
> AKM5394A that is being used in Janus,  you will find it darn hard to 
> ever get the instantaneous dynamic range that is achievable by these 
> narrower band IF's  available from the QSD (and its variants) where you 
> get something like 115 dB SFDR and NO NEED FOR ANALOG AGC IN FRONT OF IT 
> in the narrow band voice grade channel applications.  No multi hundred 
> MHz A/D can give this kind of performance because of the near/far 
> problem so far as I have been able to find.    You can get increased 
> dynamic range (not instantaneous) by getting processing gain through 
> downsampling/filtering.    We are beginning to get interesting enough 
> parts that we might come close to this with these 16 bit (very 
> expensive) A/D's running at high bandwidths (such as Mercury, etc.). 
> 
> In the end,  the criticism of the SDR-1000 is misplaced.    Getting 
> performance that rivals much more expensive radios out of a design done 
> in a garage over a couple of years whilst talking to amateur radio 
> breakfast buddies on every Saturday about it and then helping to foment 
> the revolution we have seen it do and starting a major open source, open 
> hardware initiative along with proponents of GnuRadio is not something 
> to be dismissed without comment.  To me,  the SDR-1000 is one of those 
> little miracles that continue to make life interesting.
> 
> If Gerald and Flex were to continue this form factor into future 
> radios,  then I would say you have an argument. 
> 
> Bob
> 
> 


I have a E-MU 1820m that has that same ADC. I wrote a Linux ALSA driver 
for it and I measured a noise level, with the balanced inputs 
disconnected, that made the resulting waveform accurate to about 20bits, 
the rest being noise. At this resolution, I think that the RF mixer 
would be at a higher noise level than this, so I don't see the point of 
much better than 20bits. I think that is also what the Janus developers 
found.

With baseband ADCs running at 130Mhz now working to 16bits, probably 15 
or 14bits real due to dithering, they are at least getting closer to 20bits.
Baseband ADCs have other advantages over the SDR-1000 RF mixer, so there 
a arguments in both directions.
1) QSD has higher number of bits ADC, but have image rejection problems 
and noise introduced by the QSD clock and are cheaper
2) Baseband ADCs have potentially better filters, fft bins can be made 
smaller because there are more samples taken. Less detector noise as the 
ADC sample and hold circuits are less noisy. These are more expensive.

James

 1168005772.0