FM PLL Demodulation

I have a wideband FM signal primarily modulated by random voice and music, but the baseband also contains one pure sinusoidal tone at a known fixed fr equency and at a relatively low modulation index. The tone is the only thi ng I want to recover from the FM signal.

The total RF signal occupied bandwidth is much greater than needed for the tone alone. I'd like to narrow the receiver BW for improved threshold. Ca n I do this using a PLL optimized for the tone frequency?

Reply to
George
Loading thread data ...

No, you can not. PWM is nonlinear modulation. There is no simple way to accomplish what you are trying to do. You have to recover entire signal. With optimal processing, you can achieve only a modest gain of somewhat 0.5 .. 1 dB in SNR.

Vladimir Vassilevsky DSP and Mixed Signal Consultant

formatting link

Reply to
Vladimir Vassilevsky

That sounds very much a CTCSS tone used in commercial 2way FM radio, order wire signaling systems, site alarms, etc. Search for "PL decoder" or "CTCSS decoder". In a past life, I used to design these boards for retrofit into various manufacturers radios.

No need to build one as finished products are available:

Yes, it can be done with a PLL but I suspect that you would not want to do it that way. There's going to be quite a bit of noise and intermod in the baseband from the random voice and music. Simple PLL's tend to false on this noise and are slow. I had better luck with switched capacitor and twin-T filters than with PLL. However, that may be because the various PL tones are about 2 to 4 Hz apart and the typical spec is for less than 300 msec response time. Your system may be different.

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
 Click to see the full signature
Reply to
Jeff Liebermann

PWM? Surely Vladimir meant PM (phase modulation, which for his purposes is the same as frequency modulation) or FM.

At any rate -- yup, he's right. FM is a nonlinear process, it'll take that one low-amplitude tone in the audio and smear it over the whole occupied bandwidth of the signal. If you're trying to recover the tone cheaply then your best bet is to use a cheap FM demodulation method then recover the tone from the audio. If you're trying to recover the tone at a much weaker signal than normal then you're probably doomed.

--
Tim Wescott 
Control system and signal processing consulting 
 Click to see the full signature
Reply to
Tim Wescott

Well, there are methods to improve situation with FM, but they are hardly practical and the gain achievable is below 1dB. First, one can take the spectrum of RF signal and look for periodicities with the tone frequency. Another way is recover entire signal using optimal Kalman filtering and LPC as a model for modulation. But, despite heavy computation, not much of improvement could be achieved over trivial FM demodulation methods.

Vladimir Vassilevsky DSP and Mixed Signal Consultant

formatting link

Reply to
Vladimir Vassilevsky

Speaking of FM, I heard rumor that it was possible to deconvolve the message from the spectrum alone, using Bessel functions -- just not that you would want to, since it's so much more complicated. I've never seen a reference to this, anyone got a hint?

Tim

--
Deep Friar: a very philosophical monk. 
Website: http://webpages.charter.net/dawill/tmoranwms 
 Click to see the full signature
Reply to
Tim Williams

Your PLL isn't the PLL in the original question. You are talking about PLLing the demod signal, while the original poster wanted to PLL the tone right from the FM signal itself. As pointed out, FM in not linear. [BTW if you do the math, FM bandwidth is infinite.]

I think the CTCSS detection these days is just LPF and a counter. Easy to do since CTCSS is SF, not MF.

Reply to
miso

I wouldn't think that 1dB of improvement is going to help you much unless your situation is really odd.

A honkin' big gain antenna pointed in the optimal direction and a lower- than-normal noise receiver front end may help somewhat, but even that is situation-dependent, and hardly cheap.

--
My liberal friends think I'm a conservative kook. 
My conservative friends think I'm a liberal kook. 
 Click to see the full signature
Reply to
Tim Wescott

Exactly. I got that and tried to answer what I think he was trying to ask, instead of what was actually written. I'm guilty of reading between the lines as charged. If the OP ever returns, we might find out if my guess is correct.

I see no reason why he would want to demodulate the entire FM bandwidth in order to recover a single discrete tone. He probably already has an FM receiver for dealing with the "random voice and music" which presumably has a functional FM demodulator. Unless he's building the entire receiver from scratch, which is unlikely because he's only asking about an FM demodulator, I don't see any reason to add a 2nd demodulator to the existing receiver. Decoding the audio output should be sufficient.

To avoid having to measure to infinity, we use an occupied bandwidth measurement. For FCC NBFM, it's Carson's Rule where the occupied bandwidth is out to where 2% of the total power is in the sidebands (i.e. 1.0% per sideband): For 3GPP (TS 34.121 5.8) it's out to where 1% of the total power is in the sidebands (i.e. 0.5% per sideband).

Well, CTCSS is essentially obsolete, replaced by various digital PL or DCS (digital coded squelch) mutations. These daze, the PL (CTCSS) decoders are all done with a low pass filter and a period counter. It can be in a PIC or as part of uP that controls the radio. Most radios offer PL and DCS decoding, using the same chip and methods. Fancy radios will also encode and sometime decode DTMF, MDC1200, 2 tone paging, POCSAG, and whatever else moves the marketting department.

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
 Click to see the full signature
Reply to
Jeff Liebermann

OP here. I was hoping to extract just the tone from the original FM signal without having to demodulate the whole composite baseband first. That wou ld avoid thresholding of the main carrier and allow detection of the tone m uch lower into the noise.

Conceptually, imagine the carrier modulated by just the tone alone at first . The mod index is only around 0.3, so the tone is represented essentiall y by just the first Bessel pair and the carrier. If I could recover just t hose components at RF I should be able to demodulate the tone. In principl e that takes three narrowband receive channels rather than one wasteful wid eband one.

Now imagine adding the high-index random audio. That's going to smear the tone components across the entire Carson's Rule BW. But if I could track t he original three components (they are the only coherent components in the band) then I should still be able to demodulate the tone. That's what I wa s hoping to do with the PLL demod with loop filter that is tuned to the ton e frequency.

Or am I looking at this wrong?

Reply to
George

age

d

nce

If the tone is really low-level, most of its energy is going to be in a single pair of RF sidebands anyway. It's reasonable to recover a narrowband tone from a WBFM signal in many cases, IMHO, if you aren't pressed for every last dB of SNR.

-- john, KE5FX

Reply to
jmiles

Some numbers and details would be helpful.

  1. Do you have an existing receiver?
  2. If so, what is the IF frequency and bandwidth?
  3. Is there an IF output you can grab?
  4. What is the approximate tone frequency?
  5. If the FM modulation analog, PWM, FSK, BPSK, etc? If it's NOT analog, then please ignore my suggestions.
  6. Any fast response requirements?
  7. What do you mean by tracking three components? I thought from your original description that there was only one tone and that it didn't move. If you're trying to follow a moving tone around, please ignore my suggestions.

The mod index is only around 0.3, so the tone is represented essentially by just the first Bessel pair and the carrier. If I could recover just those components at RF I should be able to demodulate the tone. In principle that takes three narrowband receive channels rather than one wasteful wideband one.

components across the entire Carson's Rule BW. But if I could track the original three components (they are the only coherent components in the band) then I should still be able to demodulate the tone. That's what I was hoping to do with the PLL demod with loop filter that is tuned to the tone frequency.

You're looking at it correctly if you want a universal solution. What I see if a conventional FM receiver, with a fairly wideband IF and demodulator. A common FM broadcast receiver would be a good example.

What I would do is borrow some of the IF signal, and filter it through a narrow band IF crystal or LC filter that's approximately 5 times the tone frequency (based on Carson's Rule). For example, if the tone is

100Hz and the IF is 10.7Mhz, the filter bandwidth would be about 500Hz. That's a Q of 21.4 so LC should work.

Presumably all the random voice and music is above 100 Hz. That may be a bad assumption if the random voice and music extends into the frequency range of the tone. I suspect that's unlikely as nobody wants to listen to a tone mixed into their music.

After the crystal or LC IF filter, add your favorite FM demodulator, crystal slope detector, or PLL demodulator for the 10.7MHz. You should see mostly 100Hz audio coming out which you can detect with either a 100Hz PLL, period counter, or commercial PL decoder. Don't bother with de-emphasis if you're using pre-emph and de-emph. It will just emphasis any jitter, hummmm, and noise below 100Hz.

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
 Click to see the full signature
Reply to
Jeff Liebermann

rst. The mod index is only around 0.3, so the tone is represented essenti ally by just the first Bessel pair and the carrier. If I could recover jus t those components at RF I should be able to demodulate the tone. In princ iple that takes three narrowband receive channels rather than one wasteful wideband one.

he tone components across the entire Carson's Rule BW. But if I could trac k the original three components (they are the only coherent components in t he band) then I should still be able to demodulate the tone. That's what I was hoping to do with the PLL demod with loop filter that is tuned to the tone frequency.

The signal is standard commercial analog FM broadcast with 15 kHz audio cha nnel cutoff. The tone is 19 kHz. The receiver hardware will be custom des igned to whatever is necessary. There may be a lot of receivers built even tually in production.

I'm hoping the PLL demod can be tuned to respond to just the 19 kHz tone an d lock to it in the presence of the random audio modulation. (Although the tone is above the audio modulation, at RF it will be buried in the higher- order Bessel spectral components.) The narrow noise BW of the loop filter (~100 Hz?) should give a better overall receive threshold than that of the conventional ~150 kHz FM receiver.

I'm afraid I must be missing something here.

Reply to
George

------------ No possible. BTW, what you are trying to do is very typical line of thinking. It doesn't work that way. You are not first person who tried to recover just a pilot tone from combined FM signal and extend the sensitivity.

Vladimir Vassilevsky DSP and Mixed Signal Consultant

formatting link

Reply to
Vladimir Vassilevsky

Bessel functions apply to simple case of pure sine wave modulation. They won't help to demodulate arbitrary signal. However it is quite simple to do optimal FM demodulation in frequency domain. This could be a good exercise in DSP for beginners.

Vladimir Vassilevsky DSP and Mixed Signal Consultant

formatting link

Reply to
Vladimir Vassilevsky

first. The mod index is only around 0.3, so the tone is represented essentially by just the first Bessel pair and the carrier. If I could recover just those components at RF I should be able to demodulate the tone. In principle that takes three narrowband receive channels rather than one wasteful wideband one.

tone components across the entire Carson's Rule BW. But if I could track the original three components (they are the only coherent components in the band) then I should still be able to demodulate the tone. That's what I was hoping to do with the PLL demod with loop filter that is tuned to the tone frequency.

channel cutoff. The tone is 19 kHz. The receiver hardware will be custom designed to whatever is necessary. There may be a lot of receivers built eventually in production.

lock to it in the presence of the random audio modulation. (Although the tone is above the audio modulation, at RF it will be buried in the higher-order Bessel spectral components.) The narrow noise BW of the loop filter (~100 Hz?) should give a better overall receive threshold than that of the conventional ~150 kHz FM receiver.

Why not just use a stereo decoder chip, with an output to indicate stereo? You can pick the signal off the detector or discriminator, before the de-emphisis network.

Reply to
Michael A. Terrell

Ok, you're not going to answer my questions and prefer to spoon feed us information while we keep guessing. With all due respect, I don't have the patience to interrogate you for basic information, especially when you won't even bother to confirm my guesses.

I think I've mentioned this to you in previous similar exercise, where you also provided the absolute minimum amount of ambiguous information in your question and in followups. Please excuse the repetition. If you want to get sane answer on usenet, you will need to provide:

  1. What problem are you trying to solve? A one liner is sufficient. In this case it would be "I'm trying to design an FM broadcast band receiver that will provide an indication when a 19KHz pilot tone is received".
  2. What do you have to work with? What existing hardware, software, programming facilities, devices, do you have available?
  3. If for troubleshooting, what have you done so far, and what happened.

Ok, It's now a standard FM receiver, probably with a 10.7MHz IF. What you're looking at is something like this: Note the -6dB/octave rolloff from the de-emphasis network. By time you get to 19KHz, there's not much left from the original modulation to trash 19KHz. Note the highly visible 19KHz line. Looks like about

-6dB down from peak audio level. That should be easy to decode.

As I previously mentioned, with pre-emphasis, there's no garbage at

19KHz to worry about. If there are any harmonics of the audio, or intermod mixes, fix your receiver as such distortion will never be tolerated in a typical FM broadcast receiver.

Correct. The narrower the bandwidth, the higher the sensitivity. See the typical stereo decoder chip for details. Tell me why you need extra sensitivity when the above GIF file showing the audio spectra shows rather fabulous SNR?

Think about an all digital solution. Most FM receivers are all digital these days, including the cheap receivers found inside headphones, cell phones, and music players. I'm not really very familiar with them, but some of them have a 19KHz pilot tone decoder that lights up a "stereo" light on the front panel. With luck, you may not need to design very much.

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
 Click to see the full signature
Reply to
Jeff Liebermann

Care to elaborate?

If you track the frequency a little bit at a time, of course, you can get excellent accuracies from the individual FFTs (or whichever method you prefer of determining the frequency for that period), or do a sliding FFT, or a sliding average of the results, or whatever. But true "frequency domain" means the entirety of the signal (which could be millions of samples for a practical case, without downconversion that is).

Tim

--
Deep Friar: a very philosophical monk. 
Website: http://seventransistorlabs.com
Reply to
Tim Williams

"Jeff Liebermann"

** Huh ??

There is a +6dB/oct pre-emphasis applied to the audio modulation, comes in at around 3 to 4 kHz and the receiver un-does that with a matching de-emphasis in the audio circuit.

Nothing of it becomes visible on that graph.

** The 19kHz pilot tone is modulated at close to 10% of peak audio level or about +/- 7.5 kHz deviation.

PLLs can be made to lock onto signals that are below noise - so if the OP filters the signal from the FM detector to a narrow band cantered on 19kHz, he should have no problem locking onto it with a PLL.

... Phil

Reply to
Phil Allison

I did some Googling and found that that 19KHz pilot tone is suppose to be about -26 dBV(rms) or 10% of the modulation. I took a quick look at the local FM stations on my antique HP8554L spectrum analyzer. Seems about right.

Also, I screwed up a little in the above paragraph. The -6dB/octave roll off is due to transmit pre-emphasis.

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
 Click to see the full signature
Reply to
Jeff Liebermann

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.