[hpsdr] hpsdr Digest, Vol 3, Issue 12

[Greg Overkamp]

You can get a 3dB improvement in SNR for every halving
of sample rate with simple decimation, assuming there
is no noise-shaping in a feedback loop. Now you have
created an ADC with really good SNR but tiny
bandwidth. Seems like a numbers game to me. If you
want to use the full original bandwidth of the ADC to
tune a band of signals, I don't think you can claim
your radio has 140 dB dynamic range (for example) if
that dynamic range is only good for a bandwidth of
1250 hertz.

Greg
WD9DEX


Robert W McGwier wrote:

>Let us suppose we put a single tone in 1.25 MHz or
2.5 MHz if you are 
>using a pair in quadrature.
>
>You can get close to 100 dB SNR in that full
bandwidth if we believe 
>the specifications. 
>
>Now filter and downsample that to 2500 Hz (say) which
contains the 
>tone.   The noise  will be reduced by a factor of 
1000 BY FILTERING 
>THE 
>NOISE OFF THAT IS OUT OF BAND  through a filtering
and decimation 
>process that saves ticks and gets rid of the useless
junk. 
>
>That factor of 1000 is a fairly large number of dB of
theoretical 
>processing gain and that translates directly to SNR.
>
>All of these numbers are theoretical numbers, based
on flat amplitude 
>responses, etc.  but the prinicipal does hold in
practice.
>
>Bob
>N4HY