FM PLL Demodulation

Yeah, I know. I goofed. Thanks.

I think (not sure) that the spectrum plot is directly from the discriminator output, before de-emphasis, from an FM receiver. The

-6dB/octave is from the transmitter pre-emphasis giving a boost at low frequencies, and attenuating the highs. I was going to reproduce the plot with a receiver on my bench. It's easy enough to do, but I can't find a receiver with a handy non de-emphasized output that I don't have to rip out of a rack somewhere. Maybe tomorrow.

Yep.

Agreed. If a stereo decoder can do it, it should be possible.

"Jeff Liebermann"

The audio pre-emphasis of broadcast FM has a time constant of 50uS (or

It peaks at about +15dB before disappearing into the steep roll of that begins at 15kHz.

.... Phil

Grumble. I just hate it when I'm wrong (and you're right). It's neither pre-emphasis or de-emphasis. The -6dB/octave rolloff is some accidental characteristic of whatever happend to be playing on the station when the photo was taken.

I decided to use my IFR-1500 service monitor as a receiver. That's what it looks like with directly from the discriminator (or whatever the IFR-1500 uses for an FM demodulator. The 19KHz pilot tone is clearly visible, but no slope.

Thanks again.

"Jeff Liebermann"

Bach organ concerto, plus choir maybe ......

... Phil

Nice program. The problem is that between Spectrum Lab 2.7b2 in the original photo, and 2.78b21 in my photos, a huge number of settings were changed or moved. I was lost. So many features, so little time.

On a sunny day (Sat, 24 Nov 2012 17:05:28 -0800 (PST)) it happened George wrote in :

I could be 100% wrong on this, but from memory, if you FM modulate with frequency f, and that is a constant, you get spectral components f apart. Those could perhaps be filtered out with some sort of inverse comb filter? (Inverse combfilter = subtract combfiltered signaltfrom signal or something like that. But does it have a real advantage over demodulating the lot and a LF bandpass filter? I think not, and the last method is simpler.

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.

There is a strong smell of stereo pilot tone here.

For a FM stereo, the modulating signal goes well over 15 kHz, as the difference information is a DSB signal centered on double the pilot tone frequency, 23 to 53 kHz, so there may be more mess to handle.

IMHO, you need first to decode the whole multiplex from the FM before looking for the tone.

Gosh fellers ... I've been tryin' to keep this simple ... :o)

I'm starting with a standard stereo FM broadcast signal, not with any parti cular receiver hardware. It probably WILL turn out to be a custom-designed software receiver. But I'm just at the concept stage now. I just want to get the best possible threshold performance for detecting the 19 kHz tone in faded conditions.

If I detect the tone from within the demodulated composite baseband, I will be limited to the rather high noise floor there which increases 10 dB/dB a t threshold. That will be my limit of weak signal performance.

So my objective is to eliminate that wideband noise and detect the pilot di rectly at RF, taking advantage of the fact that its spectral components wil l be easy to distinguish (as a sinusoid) from the random program audio surr ounding it. Detecting those spectral components (which are just one pair o f Bessell tones due to their low mod index) should be doable in a pre-detec tion noise BW of a couple hundred Hz or so. I hope.

I've never heard of this being done, so maybe I'm way off base, but intuiti vely it seems valid to me at least.

Yes, this is what I meant. Actually, tracking phase rather then frequency. If you combine predicted modulating signal with measured input signal (like Kalman filter), that makes optimal FM demodulator. Those things are known. However, the gain of optimal demodulator over trivial methods is only 1dB or so.

The long term spectrum of random FM signal would be bell shaped curve. I don't think there is a way to deduce modulation from it.

Vladimir Vassilevsky DSP and Mixed Signal Consultant

particular receiver hardware. It probably WILL turn out to be a custom-designed software receiver. But I'm just at the concept stage now. I just want to get the best possible threshold performance for detecting the 19 kHz tone in faded conditions.

limited to the rather high noise floor there which increases 10 dB/dB at threshold. That will be my limit of weak signal performance.

directly at RF, taking advantage of the fact that its spectral components will be easy to distinguish (as a sinusoid) from the random program audio surrounding it. Detecting those spectral components (which are just one pair of Bessell tones due to their low mod index) should be doable in a pre-detection noise BW of a couple hundred Hz or so. I hope.

intuitively it seems valid to me at least.

What good will it do to improve the 19 KHz recovery when it is still usable long after the stereo information is long gone? here have been PLL based stereo decoders since the '80s. National Semicondutors' LM1800 was one that I saw back then.

It was used in a lot of car radios back then.

"George is a Wanker "

Gosh fellers ... I've been tryin' to keep this simple ... :o)

I'm starting with a standard stereo FM broadcast signal, not with any particular receiver hardware. It probably WILL turn out to be a custom-designed software receiver. But I'm just at the concept stage now. I just want to get the best possible threshold performance for detecting the

** There is NO point in that idea whatever !!!

Stereo FM is only of usable quality when an adequate signal level is available and that signal is free of multipath too.

Below that level, the receiver must switch to mono to improve to s/n ratio to listenable.

Pilot tone locking is a NON ISSUE

You stupid PITA f****it.

... Phil

Hi Phil. So nice to hear from you again. If you had read my OP you would have seen "The tone is the only thing I want to recover from the FM signal."

Thanks for contributing.

"George is a Wanker "

" It probably WILL turn out to be a custom-designed software receiver."

Clearly an FM broadcast receiver.

You have trolled this NG once too often.

F*ck off and DIE !!!

If you only want to ride through the fade, why not use a suitable crystal oscillator (VXCO), derive 38/19 kHz from it. Use the 19 kHz to lock to the pilot tone and put a sample/hold circuit between the phase detector and VXCO. When the signal strength drops too low, put the S/H in Hold mode and the VXCO should remain on the correct frequency. When the RF returns after thee fade, the VXCO is still at correct frequency and nearly correct phase, thus relocking would happen immediately, without any bad phase discontinuities.

These days it might even be possible to decode a _strong_ differential (L-R) DSB signal (23..53 kHz) without the pilot tone using a Costas loop.

Please forgive Phil. He is emotionally retarded.

And too ignorant to take his meds. ...Jim Thompson

Mentally decapitated.

But, he's usually correct. At least he usually catches my better screwups. Just apply an IPF (Insult and Profanity Filter) and ignore the rest.

In the not so distant past, I had to deal with an attorney, who's only functional means of communications and debate was to threaten to sue me. The high point in this absurdity was when he threatened to sue me for miscalculating his share of our lunch tab. Otherwise, he was a great attorney and served me well.

Another character from the distant past grew up in an environment that required every 5th word (I counted) be either profanity or swearing. Otherwise, he was perfectly normal. Since his written communications were carefully edited, he was quite functional. Predictably, his wife and kids never picked up his bad habits.

A former boss somehow compensated for his short stature by becoming a compulsive liar. Literally everything he proclaimed was a fabrication or a lie. However, his knowledge of the business was extensive and well worth learning. If I really wanted the truth, I just asked his secretary. So, I filtered out the obvious lies, and paid very close attention to what was left. No regrets.

Then, there was a highly disreputable business owner. He talked like a crook, acted like a crook, proclaimed himself a crook, and even looked like a crook. The problem was that people couldn't believe that he really was a crook. After being cheated a few times, I finally got the message. Once the business rules of engagement were established, doing business with this crook was fairly simple.

I could go on, but my bagel is turning to concrete. In the real world (as distinguished from Usenet), you're going to run into a wide variety of people. Some of them will meet your standards of acceptable behavior. Many will not. If you dismiss those that don't, you're doing yourself a disservice, especially if you run into foreign customs and norms.

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.

and 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.

It could well be. The 19 kHz tone is the stereo pilot frequency and it would be nuts to FM it. However there are Costas loop modulators and demodulators on that pilot sometimes. Perhaps you are looking for something to demodulate SCA in the 60 kHz to 75 kHz range of the bare demodulated audio. Or some other component of the signal?

?-)