[RTTY] Some basic RTTY radio questions

[Kok Chen]

On Mar 8, 2016, at 9:29 AM, Tim Shoppa wrote:

> My preferred way of operating is to disable AGC for everything but
> "ear-saving". I will always, at the very least, turn RF gain down so that
> band noise is not loud. I will always engage one or two stages of
> attenuation on the low bands, in a big contest I might engage the
> attenuator on the high bands too.

Just like with filter bandwidths and AGC, there is also a cold, hard, objective way to set the RF gain.  Please allow me to explain.  

This is long.  If you just want a monkey-see-monkey-do directions, just go to the end.  But, as with everything else RTTY related, the more you know about how things work, the more you will be better at how to set things up.

Recall that the error rate of RTTY demodulators ultimately depends on signal to noise ratio (SNR).   Algorithms and circuits that improves the slicer, and bit clock and character sync extractions, will largely determine the "performance," but when we are given the same algorithms (something a typical user cannot control), it is SNR (something that any user *can* actually control) that determines the error rate.

There are two primary noise sources: 1) sky noise, and 2) "receiver" noise floor.

With the state of modern superhet receivers, the "receiver" noise is the combination of the analog noise (this is where a preamp usually helps) and the least significant bit noise ("quantization noise") from the sound card.  With DDC SDRs (like Perseus, Hermes, and 6000 series Flex), the digital noise comes from the front end A/D converter.

Since the analog noise and digital noise are what engineers and mathematicians call "uncorrelated," or "statistically independent," the noise power (and not noise voltages) adds. I.e., if the analog noise is A nanowatts and the sound card noise is D nanowatts, the resultant system noise is A+D nanowatts.

Similarly, the sky noise, say K nanowatts, is statistically independent from both the system analog and digital noise, so the noise *power* that the RTTY demodulator sees is A+D+K nanowatts.

As an RTTY user, we usually cannot control A (it is a property of the receiver that we bought) or S (a property of band conditions).  So let us lump A+K into R and write the noise that the demodulator sees as R+D.

A+K actually is somewhat in the hands of the user since the ratio of sky noise (K) to the receiver noise (A) is determined by the RF preamp.  But for low bands (as multiple people have already mentioned), the sky noise is so large, that no preamp is required.

We conclude here that the noise which the demodulator sees is simply R+D.  And from the prior paragraph, for most cases, it is simply K+D (sky noise + quantization noise).

Strictly speaking, most people have a gain control between the receiver and the sound card (in most cases today, it is built into the sound card either as a digital control or an analog pot).  So the demodulator noise is really K/P + D where P is the attenuator's contribution.

Notice that a signal (call it S) is also attenuated by P, so the signal to noise ratio is (S/P)/(K/P+D), or simply S/(K+D*P).

Maximizing S/(K+D*P) will maximize demodulator SNR.  S/K is the signal-to-sky noise ratio.

This means the use of as small a value of P (attenuation before the sound card) as possible.  The problem is when P is too small, the sound card will clip too early on even that are not very loud.  

So, we simply want to make P small enough such that we still get a reasonable SNR.

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.

When band conditions are good, 0.4 dB of SNR change is actually quite measurable on good modems.  But when band conditions are poor, a 0.4 dB SNR loss is probably not even noticeable. 

0.04 dB (20 dB excess noise) is definitely a gross overkill, and you will definitely notice weak signal decoding performance with a 3 dB SNR loss even when band conditions are poor.

A sky noise level that is 10 dB above the sound card noise is probably a decent compromise between SNR and sound card clipping.

So, in a couple of sentences, this is what you do:

--------------
The easiest way to determine the gain is to look at the spectrum from the sound card.  

You first turn the gain of the sound card down so much that all you see is the noise floor of the sound card.  

You then turn the gain up and you should see the noise floor eventually start to rise, due to the combined sky noise and receiver noise.  Keep turning the gain up until it reaches about 10 dB over the original sound card noise floor.

Done. 
--------------

(You may need to do this again when you change bands, or when sky noise changes enough with time of day -- it can of course be automated, but I do not know any RTTY modems that automates the process... YET :-).

OK, what if your program does not show noise floor?  You may be able to find a non-ham radio program that displays the spectrum of a sound card.

I have no idea about Windows programs, since I have never owned a Windows computer (I owned a MS-DOS computer back in the dark ages).  From day one, I had built a sound card spectrum display in cocoaModem that is drawn in dB units.

If your program does not have something to display a calibrated noise floor, either in numerical form or graphical form, you can beg David G3YYD to add it to 2Tone.  I know that he already measures things like that.  

73
Chen, W7AY