[hpsdr] Penelope output S/N ratio ?

[Steve Bunch]

John,

On Mar 20, 2010, at 11:43 PM, John Petrich wrote:

> ***** High Performance Software Defined Radio Discussion List *****
>
> Phil,
>
> Your comment on Penelope's output S/N ratio got me thinking.
>
> In your post to Chris, under the "Pennywhistle completed at last"  
> thread you made reference to Penelope's output S/N ratio.  What  
> exactly do you mean "output S/N ratio?  In the CW mode I see a clean  
> coherent signal with no noise pedestal down to at least -70 dBc.  On  
> SSB, I'm guessing, that there is a residual broadband audio noise  
> about -32 dBc or less with the microphone disconnected.  Maybe the S/ 
> N ratio that you are talking about is not readily measured using  
> classical spectrum analysis methods?

When generating that coherent signal with a D/A converter, you have  
(in a 14-bit D/A) only 2^14 = 16,384 unique values to use to "draw"  
your output waveform.  That level of accuracy could yield the order-of- 
magnitude 70+dB of S/N range you see (theoretically, 14 bits would  
yield closer to 80dB, but there are some other sources of error we'll  
ignore, e.g., the D/A isn't quite capable of accurately representing  
all 14 bits).  If you reduce the range of numbers being fed to the D/A  
converter to reduce the amplitude of the output signal, you will be  
using fewer of the available distinct values, and the inaccuracies in  
representing the signal, as a proportion of the total signal, go up.   
Imagine approximating a sine wave using side-by-side rectangles with  
only a distinct number of heights -- the more bits you have, the  
smaller the jump in height from one rectangle to the next (to continue  
the analogy, the faster you sample, the narrower the rectangles get).   
This inaccuracy in height manifests as "noise", i.e., signals that  
will be generated in addition to the desired signal.  To follow this  
to the logical extreme, consider what happens as the D/A converter  
length is reduced down to a few bits -- you get a very coarse  
approximation to the desired signal, and an increasing fraction of the  
output is undesired signals.  (Much of this isn't "noise" in the sense  
of "random noise" -- it is completely predictable from sampling  
theory, but it is, in any event, an undesired part of the output).

So the effect of feeding smaller numbers to the D/A to reduce the  
amplitude is that you would see the desired output signal reduce in  
magnitude as desired, but the "noise" caused by sampling errors  
wouldn't go down along with it, so your "output S/N ratio" would be  
degraded.  The more of the full range of a D/A you use, the larger a  
fraction of the capability of the D/A you paid for will show up in the  
output.  Each roughly 6dB of amplitude reduction by reducing the  
values going into the D/A converter throws away a bit of resolution,  
basically, and with it a similar degradation to your best-achievable S/ 
N ratio.  So running the D/A converter at near-full-scale and using an  
analog attenuator after it is a cleaner way to drop the signal  
amplitude by more than a few dB, and also helps protect the next  
stages from major overdriving if you happen to make an error setting  
the output level in your SDR software.

Steve, K9SRB

> In response to your post I added enough input attenuation to my PA  
> such that Penelope is now delivering 300+ MW at the desired PA  
> output level.  Prior to the addition of the attenuation, Penelope  
> was putting out 76 MW for that same PA output.  Haven't really  
> looked at the output spectrum yet.
>
> A penny for your thoughts,
> John Petrich, W7FU
> _______________________________________________
> HPSDR Discussion List
> To post msg: hpsdr at openhpsdr.org
> Subscription help: http://lists.openhpsdr.org/listinfo.cgi/hpsdr-openhpsdr.org
> HPSDR web page: http://openhpsdr.org
> Archives: http://lists.openhpsdr.org/pipermail/hpsdr-openhpsdr.org/


 1269187126.0