When something sounds too good to be true, it usually is.

Specifications of sound cards that refer to the "S/N" of the card are
completely irrelevant to use of the card with WSJT -- or indeed just
about any other ham radio software.  The reason is that the
signal-to-noise ratio that matters for MS or EME (with WSJT), or for HF
use of PSK31, is determined far upstream of the device interfacing your
radio to your computer.

A sound card would have to be EXTREMELY poor -- so poor that it would be
unacceptable for "normal" computer uses such as recording/playing music,
etc. -- before its S/N rating would significantly degrade the decoding
ability of WSJT and similar programs.

The noise that the WSJT decoders must cope with is a combination of
cosmic noise, atmospheric noise, and receiver noise; sound card noise is
many tens of dBs weaker, and entirely negligible.

As it happens, there is one way in which poor sound cards can adversely
affect WSJT signals.  It's not S/N, but rather inaccuracies in sampling
rate.  WSJT uses a sample rate of 11025 Hz for both input and output.
All sound cards claim to support this rate, but some do it by
interpolating rather poorly from another sampling rate.

The Rig Expert may be a convenient way to do your radio-computer
interfacing, but it will NOT gain you 10 dB (or even 1 dB) in detecting
weak meteor pings or EME signals.

-- 73, Joe, K1JT

Bob Poortinga wrote:
> Joe Taylor K1JT <joe@Princeton.EDU> writes:
>  
>>As it happens, there is one way in which poor sound cards can 
>>adversely
>>affect WSJT signals.  It's not S/N, but rather inaccuracies in sampling 
>>rate.
> 
> Doesn't sampling jitter also introduce noise?   Wouldn't a highly-stable
> sampling clock produce a lower noise floor resulting in better 
> decodes?

Yes, but for any plausible amount of jitter that noise will cause 
negligible degradation of the WSJT audio signals sent from your radio to 
the sound card input.

The "inaccuracies in sampling rate" that I mentioned are not jitter, but 
rather a sampling rate that is constant but offset from the nominal 
value by a significant amount.  WSJT always requests a sampling rate of 
11025 samples per second.  The actual sample rate can be somewhat 
different.  Some recent sound cards are "off" by as much as 75 Hz, 
sampling at about 11100 Hz instead of 11025.  If uncorrected, this means 
that the WSJT tone spacing will be off by about 0.7% in both time and 
frequency, causing a loss of sensitivity up to about 2 dB.

The next WSJT version to be released will have the ability to correct 
for errors in sound card sample rates.

>>The Rig Expert may be a convenient way to do your radio-computer
>>interfacing, but it will NOT gain you 10 dB (or even 1 dB) in detecting 
>>weak meteor pings or EME signals.
> 
> 
> Thanks, Joe, I was simply looking for a definitive answer in response 
> to the claims of others.
> 
> 73 de  Bob, K9SQL

Happy to help!

There is no reason for any well made sound card to introduce hum or
distortion just because it is located inside a PC.  Hum is almost always
indicative of a ground-loop problem, and can usually be resolved with an
audio isolation transformer between radio and computer.  In addition,
WSJT is relatively immune to hum because 50/60 Hz and the first
half-dozen harmonics are well below the WSJT frequency range.

> Also can low latency ASIO audio drivers that allow high speed cw and
> fast TR transition be also be of benefit ? This seems to be the holy
> grail that the sdr-1000 flex radio forum members are seeking to
> achieve.

No significant benefits to WSJT, which does not need extremely low latency.
-- 73, Joe, K1JT