[RTTY] Decoding troubles after taking a break

[Kok Chen]

> On Jul 16, 2017, at 11:01 AM, James Wolf <jbwolf at comcast.net> wrote:
> 
> This gives me a frequency number of 11576.  Should I still change this
> number to 12,000 or adjust it to a multiple of 300 of which the closest
> would be 11700?

Why not use a sampling rate of 48000, if it is available on your sound card?

First a digression... 

There are two "standards" in the audio world.  

Bell Labs started the first standard, beginning with 8000 samples/s.  This was used in the old BSTJ articles from way back when.

When people needed wider bandwidths, they naturally went to 16000 samples/s.  48000 samples/s was then adopted to pass "hi-fi" music up to 20 kc/s.  The music studios went on to 96000 samples/s and 192 ks/s when technology allowed.

All are multiples of the original Bell Labs 8000 samples/sec.

Sony and Philips mucked it all up in the early 1980s when they introduced the music CD format.  They wanted to squeeze as much as possible into the CD, so they used 44,100 samples/s instead of 48,000 samples/s.  Kinda silly (IMHO, plain stupidity), since that is just an additional 8%.  Folk lore has it they wanted to fit an entire Beethoven's 9th on a single disc.  (Furtwängler's slow Beethoven's 9th lasted 74 minutes, and that may have contributed to the lore.)

For voice bandwidths, where such high sampling rate is not required, 11,025 samples/sec was thus born.

Converting between the "44,100 family" and the "48,000 family" is a terrible thing to do, and always involve some spurii and long latencies.  Thus, it is best to stick with the Bell family.  There is really no reason, unless you are converting audio from your radio into music CD-ROMs, to ever use the Sony/Philips format.

By the way, 45.4545 baud is an even 1056 samples at 48,000 samples/s, and PSK31 is an even 1536 samples at 48,000 samples/second.  These two baud rates do not divide evenly into 44,100 sample/sec.

Now back to why 48,000 samples/sec...

Here is a tidbit that can influence whether you choose 12,000 or 48,000 samples/s.

It comes from a property of digital signals that you can decimate (decimation is the process of low pass filtering and then downsampling) a signal and gain dynamic range through what is called "processing gain." (Google "decimation" and "processing gain" if you want to know more.)  Each factor of 2 in sampling rate reduction gains you up to 3 dB of dynamic range, at the expense of a smaller bandwidth that it can handle.

A perfect 16-bit sound card has about 98 dB of dynamic range (no, not 96 dB, but that's another story :-).   Practical sound cards are seldom more than 6 dB worse.

Lets say you start with 12,000 samples per second, and do the proper decimation down to 1000 samples/second to demodulate an RTTY signal, you gain a factor of 12, or about 11 dB of dynamic range improvement.    

Now, if you use a proper RTTY filter (see http://w7ay.net/site/Technical/EqualizedRaisedCosine/index.html), you can do even better: the equivalent noise bandwidth is actually the same as if you had used a sampling rate of 90 samples/second.  Or a whopping 21 dB of processing gain. 

Magically, your 16-bit sound card now has a dynamic range of 119 dB.

However, if you start at 48,000 instead of 12,000, there is yet another factor of 6 dB.  Starting from 48,000 samples/s gives you potentially 125 dB of dynamic range from a 16 bit sound card.

That being said, you need the noise floor of the sound card to be at least 10 dB over the noise floor from the receiver (receiver + sky noise) so, in practice it is more like 115 dB of useful dynamic range. 

(By the way, this loss does not exist in a direct sampling SDR since you are not adding another sampler (the sound card) in between the receiver and the modem.)

All of this assumes a perfect 16-bit sound card, perfect demodulating filter, brick walled decimation filters, etc.  In practice, you might be 10 dB worse.  

Still, 105 dB of dynamic range is a lot.

You can throw two signals that are different by more than 100 dB at a good software modem, and it will copy the weaker signal just fine as long as the weak signal does not get wiped out by the keyclicks from a loud FSK signal nearby.  High dynamic range is way overrated for RTTY contesting -- the keyclicks are usually the limiting factor, not the dynamic range of the receiving chain.

If your sound card can go to 48,000 sample/s, you may still want to choose that because you have a "free" 6 dB of dynamic range over choosing 12,000 samples/sec.  (Well, not completely free since you do pay a few more pennies for the electric bill.)

73
Chen, W7AY