When I posted the following, I did not supply the justification why you want to
adjust the noise floor for a largest acceptable SNR loss.
> If P=1 (0 dB excess noise), the denominator changes by 50%, i.e., you have a
> whopping 3 dB SNR loss from the best possible SNR.
>
> If P = 10 (10 dB excess noise), the SNR loss of 0.4 dB.
>
> If P=100 (20 dB excess noise), the SNR loss is 0.04 dB.
After all, you could even adjust the receiver gain so that the sky noise (I
use"sky noise" to mean HF atmospheric noise+man made noise) to 30 dB over the
sound card noise, and result is an even smaller SNR loss, right?
Well, the reason is that all A/D converters and sound cards have a limited
dynamic range.
A typical 16 bit sound card might start with 95 dB of dynamic range (I have
measure some "16-bit" cards in the mid 80s dB).
When you adjust for the sky noise to be 20 dB over the sound card noise floor,
you are robbing 20 dB from the sound card's headroom. When you do that, the
"decent sound card" will now only give you 75 dB of dynamic range.
The picture is not that bleak, though. Even a linear 16 bit sound card is good
enough for pretty much anything a superhet receiver can throw at it. The
reason follows.
You typically use a sound card clock of 48000 samples per second (sps).
However, good software demodulators will have a final bandwidth that is close
to 45.45 Hz. This is a ratio of a little over 30 dB. You will lose some of
this processing gain (lets call it 6 dB) to the decimation filters, but even if
you lose 6 dB, we are still talking about a process gain of 24 dB.
The 95 dB sound card (at 48000 sps) with a 20 dB excess sky noise and proper
filters will therefore have a dynamic range of at least 75 + 24 dB or 99 dB.
So, even if you have chosen 20 dB of excess sky noise, you can see that a good
sound card, when used with a modem that pays attention to this process gain and
has a final bandwidth of 45.45 Hz, will have a dynamic range that is close to
100 dB.
Is 100 dB sufficient? It probably is during an RTTY contest. The reason is
not because there is not a variety of signal strengths, but because keying
sidebands from FSK transmitters is the limiting factor.
You may have noticed from the above that the sound card sampling rate also
determines the dynamic range. Yes, it is completely true. If the modem
software is decimating correctly. A sampling rate of 96 ksps will give almost
10 dB better dynamic range than using a sampling rate of 12 ksps.
Anyhow, the take away is that if the receiver gain is set too low, you will not
be able to print weak signals. The gain should be set so that the sky noise is
at least 10 dB above the sound card noise floor.
I suspect this is the primary cause of people claiming that their analog TNC
copies better than their software modems.
On the other hand, if you set the receiver gain too high, loud signals will
clip your sound card. This situation is even more undesirable than not being
able to copy weak signals. So you don't want to overdo the gain stuff.
Hint: when you are working a DX using split, and there is no QRM from idiots,
crank the gain way up :-).
73
Chen, W7AY
_______________________________________________
RTTY mailing list
RTTY@contesting.com
http://lists.contesting.com/mailman/listinfo/rtty
|