[RTTY] Sound Cards
chen at mac.com
Wed Jan 20 11:23:38 PST 2010
I had earlier sent this to Rick and copied the reflector, but since it
hasn't appeared, here it is again.
> I'm wondering if there is any advantage or performance improvement
> to be had with MMTTY and other RTTY programs by using a 24-bit PCI
> card ? Tests done? Charts etc?
IMHO, it is all about (blocking) dynamic range. Roughly speaking...
If you are using a narrow I.F. filter where the demodulator is only
seeing a single RTTY signal, you really need only about 70 dB to 80 dB
of dynamic range (since the rig's AGC is handling the rest) and almost
any of today's 16 bit sound card will do that for you. How much you
need depends on the dynamic range of the audio from the receiver to
the sound card. From what I'd measured, the microKeyer II's 16-bit
codec has a 96 dB of blocking dynamic range and also has no IMD above
the noise floor for the full range. Even cheap 16-bit sound cards
that I have measured have dynamic range of 80 dB before any IMD sets
in (and for single signal RTTY, that is not a bit problem, in any case).
On top of that, if you have some means of riding the gain of your
receiver, and are willing to keep adjusting knobs, a software
demodulator probably won't even need more than 30 or 40 dB of dynamic
Now, if you are using a wide filter, i.e., you are using 2.4 kHz
passband and the sound card is doing all the work to reject adjacent
channel interference, etc, for the sound card not to be the limiting
factor of your receiving chain, you need perhaps 10 dB more of sound
card blocking dynamic range than your receiver's BDR. That extra
headroom is needed because you don't want to work with just a single
bit to decode a weak signal.
A weak signal under clean propagation that only moves a single bit of
the sound card will suffer some 2.9 dB of SNR loss, which can mean the
difference between perfect copy or horrible copy. (Anyone really
interested can look up the Van Vleck Noise Linearization theory, also
called the "Arcsine Law" in Probability books.)
Again, if you are willing to ride the gain in this case, the dynamic
range of your sound card only needs to be as large as the dynamic
range of all the signals in the audio passband (plus the 10 dB or so
headroom in case the signal you are trying to copy is one of the
On the other end of the spectrum [sic], the sound card is almost
always the limiting factor of the dynamic range of an SDR that uses a
Tayloe sampling mixer. You will want as much dynamic range from a
sound card as you can muster. The Flex-5000, for example uses the
Asahi AK5394 codec, which is spec'ed for a dynamic range of 123 dB.
(That is why you will see the blocking dynamic range of the Flex-5000
quoted at around that number).
In addition, if you really want to take advantage of SDRs, you will
want to use very wide passbands, and you will want your sound card to
also handle 192 ks/s sampling rate. Many of them don't.
An affordable and good sound card for this application is something
like the E-MU 0404 if you use USB 2.0, or the Edirol FA-66 if you use
FireWire. There are better "sound cards," but you can be talking
prices near or over the list price of the original ST-8000 (:-) and
they often come with 16 or more channels, which you don't need unless
you are experimenting with phased arrays.
The "direct-sampling" SDRs such as the Perseus are a different breed
than the Flex and SoftRocks, they use "process gain" to achieve a high
dynamic range. (I.e., if done properly, each time you reduce the
sampling rate by 2, your dynamic range increases by 3 dB.) However,
since they start off in the region of 100 Ms/s sampling rate, there
aren't any codec available with the kind of dynamic range as the Asahi
chips have at 192 ks/s. When decimated down to the passband of a CW
signal, they can be better than the Flex-5000. (Take a look at the
measured dynamic range numbers between the Perseus and the Flex-5000
at Sherwood's list, for example.)
That being said, you can use process gain to achieve more blocking
dynamic range not just with the direct sampling SDRs but with
SoftRocks and even "regular" modems. Decimation and process gain is
just another weapon that DSP designers have at their disposal.
In conclusion, the "short" answer is: if you are using a narrow I.F.
filter that only lets through a single RTTY signal, don't even sweat
it with today's sound cards (just pick the cheapest :-). If you are
using a wide waterfall with a standard superhet type receiver, then
you need a good sound card since the software is the one that is
filtering away the strong QRM. If you are using the SoftRock, the
quality of your receiving system is going to be almost completely be
determined by the sound card.
Just MHO, of course.
More information about the RTTY