Reply by Mark July 9, 20082008-07-09
> > are you trying to simply determine if there is a pilot tone present, > > or > > are you trying to actually recover the tone, i.e. recover the exact > phase and frequency of the pilot tone to use it for stereo decoding? > > ----------------- > Just want to detect the presence of the tone, whether it's present or not in > order to control a stereo reception light.
please elaborate.... what is the point of using threshold extension to operate the stereo indicator lamp, if you cannot actually decode the stereo? Mark
Reply by Henry VIII July 7, 20082008-07-07
to the op:

are you trying to simply determine if there is a pilot tone present,

or

are you trying to actually recover the tone, i.e. recover the exact
phase and frequency of the pilot tone to use it for stereo decoding?

-----------------
Just want to detect the presence of the tone, whether it's present or not in 
order to control a stereo reception light.



Reply by Mark July 7, 20082008-07-07
> > > Yes, but can the discriminator be eliminated altogether to achieve even > > better C/N performance? �The great majority of the FM Carson's Rule > > bandwidth is due to the high modulation index audio. �Why deal with all that > > pre-detection bandwidth if we don't care about the audio? �So is it > > realistic to simply detect the pilot's spectral sidebands at RF - no > > discriminator - and demodulate the pilot from them? �Say we can achieve > > filters of a few tens of Hz bandwidth centered on the pilot's Bessel > > components. �That's a noise bandwidth improvement of roughly 30 dB or so > > over conventional discriminator. �Does this approach make sense?
to the op: are you trying to simply determine if there is a pilot tone present, or are you trying to actually recover the tone, i.e. recover the exact phase and frequency of the pilot tone to use it for stereo decoding? Mark
Reply by Jerry Avins July 7, 20082008-07-07
Henry VIII wrote:
> "Vladimir Vassilevsky" <antispam_bogus@hotmail.com> wrote in message > news:xPack.1005$zv7.194@flpi143.ffdc.sbc.com... > > >> What hardware is available? I.e. how big of the DSP you can afford? > > Will be all-new design from scratch. There is no specific DSP cost > limitation at this time. > > >> The practical solution would be optimizing the discriminator and using the >> narrowband filter with a good PLL to pick up the pilot at the >> discriminator output. There could be an improvement of ~10dB over the >> trivial solution just by doing that. > > Yes, but can the discriminator be eliminated altogether to achieve even > better C/N performance? The great majority of the FM Carson's Rule > bandwidth is due to the high modulation index audio. Why deal with all that > pre-detection bandwidth if we don't care about the audio? So is it > realistic to simply detect the pilot's spectral sidebands at RF - no > discriminator - and demodulate the pilot from them? Say we can achieve > filters of a few tens of Hz bandwidth centered on the pilot's Bessel > components. That's a noise bandwidth improvement of roughly 30 dB or so > over conventional discriminator. Does this approach make sense?
No. The sidebands of an FM signal are not in general symmetric, and they vary in amplitude depending on what else is modulated onto the carrier. Once there are other signals sharing the carrier, you need to rethink the whole scenario. Jerry -- Engineering is the art of making what you want from things you can get. &#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;
Reply by Vladimir Vassilevsky July 7, 20082008-07-07

Henry VIII wrote:

> Yes, but can the discriminator be eliminated altogether to achieve even > better C/N performance? The great majority of the FM Carson's Rule > bandwidth is due to the high modulation index audio. Why deal with all that > pre-detection bandwidth if we don't care about the audio?
The efficient bandwidth of the optimal maximum-likelihood FM demodulator is about the bandwidth of the modulation or smaller then that.
> So is it > realistic to simply detect the pilot's spectral sidebands at RF - no > discriminator - and demodulate the pilot from them?
To do that, you have to unscramble the nonlinearly scrambled eggs. Although it seems possible to exploit the short term correlation in the spectrum with the period of the pilot tone frequency without demodulating the whole thing.
> Say we can achieve > filters of a few tens of Hz bandwidth centered on the pilot's Bessel > components.
No use.
> That's a noise bandwidth improvement of roughly 30 dB or so > over conventional discriminator. Does this approach make sense?
No it doesn't. The pilot sidebands are smeared by the other components of the signal. Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
Reply by Randy Yates July 7, 20082008-07-07
"John E. Hadstate" <jh113355@hotmail.com> writes:

> "Henry VIII" <NoBodyHome@SSSpamavoidance.com> wrote in message > news:C1hck.14800$N87.6262@nlpi068.nbdc.sbc.com... >> "Vladimir Vassilevsky" <antispam_bogus@hotmail.com> wrote in message >> news:xPack.1005$zv7.194@flpi143.ffdc.sbc.com... >> >> Why deal with all that pre-detection bandwidth if we don't care >> about the audio? So is it realistic to simply detect the pilot's >> spectral sidebands at RF - no discriminator - and demodulate the >> pilot from them? Say we can achieve filters of a few tens of Hz >> bandwidth centered on the pilot's Bessel components. That's a noise >> bandwidth improvement of roughly 30 dB or so over conventional >> discriminator. Does this approach make sense? > > Not to me. The pilot's spectral components are all mixed up with > components that belong to other features of the signal (as well as > noise).
I'm with you, John. In my opinion, this entire engineering problem (I know exactly what Henry is attempting) is ill-formed. An alternate approach, such as diversity combining, would be more fruitful. -- % Randy Yates % "She tells me that she likes me very much, %% Fuquay-Varina, NC % but when I try to touch, she makes it %%% 919-577-9882 % all too clear." %%%% <yates@ieee.org> % 'Yours Truly, 2095', *Time*, ELO http://www.digitalsignallabs.com
Reply by John E. Hadstate July 7, 20082008-07-07
"Henry VIII" <NoBodyHome@SSSpamavoidance.com> wrote in message 
news:C1hck.14800$N87.6262@nlpi068.nbdc.sbc.com...
> "Vladimir Vassilevsky" <antispam_bogus@hotmail.com> wrote in > message news:xPack.1005$zv7.194@flpi143.ffdc.sbc.com... > > Why deal with all that pre-detection bandwidth if we don't > care about the audio? So is it realistic to simply detect > the pilot's spectral sidebands at RF - no discriminator - and > demodulate the pilot from them? Say we can achieve filters > of a few tens of Hz bandwidth centered on the pilot's Bessel > components. That's a noise bandwidth improvement of roughly > 30 dB or so over conventional discriminator. Does this > approach make sense?
Not to me. The pilot's spectral components are all mixed up with components that belong to other features of the signal (as well as noise). Detection by analyzing the raw RF seems like a non-starter. However, if you feel you must pursue this, you might consider multi-tone filtering, followed by envelope detection of the filters, followed by a neural network to recognize the pilot's spectral characteristics. I probably wouldn't try this because one of the problems with very narrow filters is that when excited by noise, they tend to ring at the frequency they are designed to detect.
> >
Reply by Henry VIII July 7, 20082008-07-07
"Vladimir Vassilevsky" <antispam_bogus@hotmail.com> wrote in message 
news:xPack.1005$zv7.194@flpi143.ffdc.sbc.com...


> What hardware is available? I.e. how big of the DSP you can afford?
Will be all-new design from scratch. There is no specific DSP cost limitation at this time.
> The practical solution would be optimizing the discriminator and using the > narrowband filter with a good PLL to pick up the pilot at the > discriminator output. There could be an improvement of ~10dB over the > trivial solution just by doing that.
Yes, but can the discriminator be eliminated altogether to achieve even better C/N performance? The great majority of the FM Carson's Rule bandwidth is due to the high modulation index audio. Why deal with all that pre-detection bandwidth if we don't care about the audio? So is it realistic to simply detect the pilot's spectral sidebands at RF - no discriminator - and demodulate the pilot from them? Say we can achieve filters of a few tens of Hz bandwidth centered on the pilot's Bessel components. That's a noise bandwidth improvement of roughly 30 dB or so over conventional discriminator. Does this approach make sense?
Reply by Randy Yates July 6, 20082008-07-06
Vladimir Vassilevsky <antispam_bogus@hotmail.com> writes:
> [...] > estimator based on the prediction of the signal. However I doubt if > this has any practical value, especially as the broadcasting changes > to digital.
We've heard a lot lately about the conversion of television to digital, but I haven't heard of a similar thing with broadcast audio (AM/FM). Of course there are things like Ibuiquity's IBOC high-definition radio, but I haven't heard that analog modulations are being retired any time soon. What am I missing? -- % Randy Yates % "Remember the good old 1980's, when %% Fuquay-Varina, NC % things were so uncomplicated?" %%% 919-577-9882 % 'Ticket To The Moon' %%%% <yates@ieee.org> % *Time*, Electric Light Orchestra http://www.digitalsignallabs.com
Reply by Vladimir Vassilevsky July 6, 20082008-07-06

Henry VIII wrote:
> "Vladimir Vassilevsky" <antispam_bogus@hotmail.com> wrote in message > news:Ze3ck.12758$jI5.12353@flpi148.ffdc.sbc.com... > > >>If I remember correctly, you inquired about this problem at some time ago. >>It was about FM stereo pilot tone, is that right? >>Can you describe the problem as a whole; i.e. what are you trying to >>accomplish.
> Yes, that's correct. The goal is to improve FM stereo receiver pilot > detection in the presence of high thermal noise and multipath interference.
What hardware is available? I.e. how big of the DSP you can afford? The practical solution would be optimizing the discriminator and using the narrowband filter with a good PLL to pick up the pilot at the discriminator output. There could be an improvement of ~10dB over the trivial solution just by doing that.
> We want to be able to detect the stereo pilot at C/N levels well below > conventional discriminator threshold, and at the same time eliminate the > limiter which causes small-signal suppression problems when multipath is > present. This will give more balanced receiver performance in environments > where the audio can be recovered but the pilot, due to its much lower > injection level, can't.
It is certainly possible to do the near-optimal detector for the analog FM composite signal. This would be a tracking frequency estimator based on the prediction of the signal. However I doubt if this has any practical value, especially as the broadcasting changes to digital. Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com