Understanding the Foster-Seeley Discriminator

[Dukov Ahzrukhal]

Hello everyone, I have recently been working on repairing an old AM/FM radio which works fine on AM but not on FM. The first thing I did was to take a close look a t the schematics, and I can fully understand the AM section. The FM section has an oscillator that had me scratching my head for a while, and an unfamiliar circuit that I believe is a ratio detector. I read about the ratio detector and found that it is a variant of the Foster-Seeley discriminator, so I did some reading on that as well. By now I definitely know enough to fix the radio, but I can't fully understand the ratio detector or even the Foster-Seeley discriminator. I find the Foster-Seeley discriminator quite fascinating, and I would love to know more about it. There are a couple of things about it that are not clear to me. The first thing is that none of the sources I've read mention the leakage inductance of the transformer and how that affects the circuit. Does the circuit rely on leakage inductance? The other thing I don't understand is the coupling capacitor and the inductor. Do those make a tuned circuit as well? I ran a simulation using LTSPICE and found that the center frequency of the circuit is indeed dependent on the coupling capacitor and the inductor, but I'm not sure if that is just because I have no leakage inductance at all in the simulation. I would very much like to read the original patent application for this circuit, but I cannot find it. Let me know if you find any good sources on this topic. I will attach my LTSPICE file here so you can play with it as well.

[ferdieCX]

Hi, I recommend you to read:

Christopher Tibbs, FM Demodulation: Theory of the Discriminator - Wireless World, May 1943
S. Seeley y J. Avins, The Ratio Detector – RCA Review, June 1947
K. R. Sturley, The Ratio Detector: How it works - Wireless World, November 1955

You will find these  articles in  www.americanradiohistory.com

[Dukov Ahzrukhal]

I read the article about the Foster-Seeley discriminator and found it quite interesting. It says that the values for the coupling capacitor and the inductor are unimportant, however in my simulation those values are critical. I set those values as if the coupling capacitor and the inductor were a series LC circuit tuned to the center frequency of the discriminator, and then tweaked the coupling capacitor to compensate for loading on the circuit and the capacitance of the diodes. That works just fine, but I suspect that it is not how the circuit normally works. From what I read in the article it seems like the coupling capacitor is just there to block DC, and the phase delay is done by the tuned transformer. From my simulation I don't really see why the transformer has to be tuned, but if I include leakage inductance then the tuning of the transformer becomes important as well. Should I be relying only on the transformer for phase delay and using large values for the coupling capacitor and the inductor? If so, I will need to learn more about tuned transformers and how to design them.

[ferdieCX]

It is a property of a double tuned transformer, that "at the resonance frequency" there will be a 90º phase shift between primary and secondary windings.
The text in the attached document is in Spanish, but you can just look at the analysis:

Eq (25) and (26) result to be the general expressions of v1 and v2 at any frequency inside the bandwidth
At the resonance frequency, X = (w - wr)/wr is 0  ==>   eq (25) and (26) take the form written below the eq (27)
The j in the expression of v2 implies the 90º phase shift

Here is a very good article in English  https://worldradiohistory.com/UK/Experimental-Wireless/40s/Wireless-Engineer-1944-02-S-OCR.pdf

[Dukov Ahzrukhal]

Good thing Spanish is my native language, I can read that document without a problem. I read the articles about the ratio detector too, and from those papers it seems like the ratio detector also uses a double tuned transformer to get 90 degree phase shift. I spent many hours today essentially trying to reinvent the double tuned transformer but I couldn't figure it out. I can understand why the such a transformer can introduce a phase shift, but I could not figure out how to actually design the transformer. I changed my LTSPICE simulation of the Foster-Seeley discriminator so that the phase shift is done by the transformer and not the coupling capacitor. This seems to work well, but since I could not figure out how to design the transformer I had to guess and check for the inductance values. I'll attach my new  file here.

[ferdieCX]

Hi, I am not very proficient with SPICE, so I cannot help with it.

Double tuned transformers are are usually critically coupled, that means kc = 1/Q

You will find some information in this book, in chap9 page 414 and chapter 26
http://www.tubebooks.org/Books/RDH4.pdf

You will also find an analysis in Spanish about the double tuned coupling here:
https://its.utu.edu.uy/sites/its.utu.edu.uy/files/2019-05/cap_18.pdf

Here there is some more information about the tuned transformers
http://www.jamminpower.com/PDF/Tuned%20Transformers.pdf

[biastee]

Does the circuit rely on leakage inductance?

Yes, it does rely on leakage inductance as the two coils must be lightly coupled (k = 0.3) for the F-S to work properly. This simulation shows how the transfer function or s-curve varies with primary tuning, secondary tuning and coupling coeff.

It says that the values for the coupling capacitor and the inductor are unimportant, however in my simulation those values are critical.

On the other hand, if the two coils are tightly coupled, e.g. k > 0.3 , then the coupling capacitor will influence the transfer function. The image below shows the transfer functions for k = 0.9 as a function of the coupling cap = 3 / 5/ 7 pF. However, these are not real s-curves, even though they may resemble it, as they don't cross 0V at their midpoint
[ Specified attachment is not available ]

[Zero999]

I couldn't read the schematic. It's important to leave space between components, rather than cluttering them all together and move the text, so it doesn't overlap the symbols and can be easilly read.

You need to give the circuit a modulated signal, to see it working, rather than just a plain carrier.

Wikipedia has a short and sweet discription of the circuit.
https://en.wikipedia.org/wiki/Foster%E2%80%93Seeley_discriminator

[biastee]

Do you mean the schematic appended below? Which component or text do you have difficulty reading?



There is no need for a modulated carrier if the frequency is swept. The results proved that the s-curve is successfully simulated.

[Zero999]

Do you mean the schematic appended below? Which component or text do you have difficulty reading?

(Attachment Link)

There is no need for a modulated carrier if the frequency is swept. The results proved that the s-curve is successfully simulated.

I was referring to the LTSpice file attached to the original post, which is very hard to read, so I fixed it.

Original


Fixed