What signal characteristics in an SDR receiver signal path should be evaluated or computed to determine whether the audio output should be squelched or not? What sort of computed result should be compared against a scalar squelch "knob" setting?

I assume that for FM signals, the software should look at some characteristic in the demodulated audio data stream rather than in the IQ data stream for this determination. Is this correct?

(I am looking for algorithms, not any pre-coded solutions or circuits, as my goal is to write my SDR app in 100% Swift 3.)

  • $\begingroup$ So, I'm kinda busy right now, but: there's more than just the demodulator output for squelching, and I'd say, no, that is not correct (in general; there's very good Squelchers working on the demodulated signal) $\endgroup$ Feb 16, 2017 at 18:12
  • $\begingroup$ (Also... I'm not a fan of reinventing the wheel... Why implement something that hundreds of people have solved before?) $\endgroup$ Feb 16, 2017 at 18:12
  • $\begingroup$ Are you only considering analog FM? $\endgroup$ Feb 16, 2017 at 18:13
  • 1
    $\begingroup$ @MarcusMüller : Re: reinvent? I hope you not trying to discourage people (on an Amatuer Radio site no less) from designing and building their own radios! (or learning enough to do so, whether tubes, transistors, or lines of Python or Swift, etc.) $\endgroup$
    – hotpaw2
    Feb 16, 2017 at 18:49
  • $\begingroup$ ha! got me! I surely don't want to discourage people to build their own radios; I just wanted to encourage you to be more courageous of stealing existing approaches; and in the software world, just replicating the same code in a different language is, and that is really a personal opinion of mine, not much better than just linking an existing library. But you're right, my wording was pretty bad, especially for the audience here (and twice since I'm talking to you, who's far from a newb). $\endgroup$ Feb 16, 2017 at 21:03

1 Answer 1

  1. The simplest squelch, applicable to almost all modes, and the one provided by most amateur radios' “squelch knob” as far as I've experienced, is power squelch: measure the signal power (in SDR, $I^2 + Q^2$), average over some short time interval, and compare against a threshold. This is done before the demodulator as the demodulator removes amplitude/power information.

    The disadvantage of this technique is that the needed squelch setting varies with conditions: it must be higher than the noise floor (which varies depending on frequency and local conditions, yet lower than the minimum (allowing for fading/picket-fencing) carrier power of the expected signal.

  2. According to Wikipedia (and I think I've heard about this before from another source), there is a FM-specific technique for squelch:

    A typical FM two-way radio carrier squelch circuit is noise operated. It takes out the voice components of the receive audio by passing the detected audio through a high-pass filter. A typical filter might pass frequencies over 4,000 Hz (4 kHz). The squelch control adjusts the gain of an amplifier which varies the level of noise coming out of the filter. The audio output of the filter and amplifier is rectified and produces a DC voltage when noise is present. The presence of continuous noise on an idle channel creates a DC voltage which turns the receiver audio off.

    That is, this technique operates after the demodulator and simply looks for the high-audio-frequency noise which results from the absence of a consistent carrier, making the assumption that the expected FM signal will have different frequency content.

    In software, this would consist of a high-pass filter → an envelope detector → averaging and comparator to control squelch.

    This technique should operate equally well no matter what the RF environment. If it requires any “knob” at all, that would be the threshold for deciding when the noise is too much, which can probably be set generously high, but if not, it still depends only on the audio characteristics of the transmitted signals and not the RF characteristics.

  3. A transmitter may encode information in the signal to allow precise squelch. This is common on FM amateur radio repeaters. The most commonly used method is CTCSS (a.k.a. PL) in which a so-called “subaudible” audio tone is added to the signal. The receiver mutes when the tone is absent.

    Because different tones can be used, this allows ignoring signals from other transmitters, provided that the intended transmission is stronger than the other (e.g. two repeaters with all their users significantly nearer to one than the other).

  4. It seems to me that there ought to be some technique which operates on the signal before demodulation to detect whether it contains more or less a single dominant frequency (as expected for a FM signal) as opposed to a “flat” spectrum of white noise (as expected for no signal), without depending on an assumption about the absolute level of the noise floor. But I am not familiar with any specific designs or how to implement them.

  • 2
    $\begingroup$ Your first bullet point is an example for your third. Also: variance/kurtosis-based estimators, periodogram-based estimation, autocorrelation-based estimators. Heck, there's people that do this on raw IQ with deep neutral networks, and it probably works! $\endgroup$ Feb 16, 2017 at 18:19
  • $\begingroup$ @MarcusMüller No, by the third I meant something that does not contain an assumption about the minimum absolute power of the signal or the maximum absolute power of the noise floor, as the first does. $\endgroup$
    – Kevin Reid AG6YO
    Feb 16, 2017 at 18:21
  • $\begingroup$ OK, I see your point. $\endgroup$ Feb 16, 2017 at 18:23
  • $\begingroup$ By the way, detecting presence of FM is definitively one of the first larger applications of SDR, dating back to the Cold War, when signal intelligence was all about automatically detecting, categorizing and analyzing enemy comms. Thus, I'd presume there's a hell of a lot of research, but it might have become a bit hard to come by good books these days $\endgroup$ Feb 16, 2017 at 18:25

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .