Armstrong 224 FM Tuner - AGC circuit

It's a bit like mean level agc from the sync separator in a TV!

None if the IF valves have any bias. The important one is the EF80, which sees the largest input.

On +ve half cycles, the grid and cathode act as a diode which conducts, clamping the +ve peak to earth, thus all the signal is now -ve going.

This -ve voltage appears across R13 and is decoupled by C23 before being fed back via R11 to the grids of the two EF89s.

The meter isn't part of the agc process at all. It is a centre zero meter used to ensure accurate tuning.

You might have RF AGC on an FM set. But the IF ideally wants to NOT have AGC, but be designed to have "limiting" action.

AGC is really very important on AM. On FM you want to remove "AM" content altogether.

You might have RF AGC on an FM set. But the IF ideally wants to NOT have AGC, but be designed to have "limiting" action ...

... Except that, in this case, the schematic clearly shows the opposite ...!

Terry,

Thanks for explaining the AGC - I just couldn't figure out how the negative voltage was produced. I appreciate that the centre tuning meter has nothing to do with the AGC circuit, but was just wondering if the AGC action would mean that the meter deflection wouldn't be much different, whether the station signal was a weak or strong one.

Michael,

Is AGC on FM really that bad? If it reduces gain on stronger signals, won't it also help to reduce spurious noise?

Cheers,

Alex

FM has to have some kind of AGC for very strong stations otherwise the IF will drive into clipping and produce loads of unwanted harmonics.

It's a bit contra-intuitive

google FM limiter action

LOADS of good stuff. Some FM IFs are designed to ALWAYS clip. Harmonics and overload only matter before the IF. Harmonics on FM IF are irrelevant.

Remember a good FM detector is sensitive to phase noise / frequency shift. Not amplitude variation.

... I appreciate that the centre tuning meter has nothing to do with the AGC circuit, but was just wondering if the AGC action would mean that the meter deflection wouldn't be much different, whether the station signal was a weak or strong one ...

Alex, think of it like this: if the agc is doing its job properly, all the signals reaching the discriminator will be, within reason, at the same level, no matter whether the signal hitting the front end was a strong or weak signal.

Therefore, as you say, the meter deflection will be independent of signal level.

What the meter does it to show the DC offset that appears at the discriminator if the station is slightly off tune. This has to be decoupled to some extent in order to stop it producing a thump when tuning thru a station but it is useful to see it on an indicator for ease of tuning.
You can pick up the same point on small battery portables with a DMM if you are interested.

Some very useful replies - thanks everyone for your help.

AGC arrangements for valve FM tuners and receivers do seem to vary quite a bit. As well as (usually delayed) RF agc to prevent front end overload, receivers with ratio demodulators and without limiters required agc to keep the audio output reasonably constant. Ratio detectors limit AM, but do not maintain constant output over the longer term. Also, with distributed IF selectivity, as was usually the case with valve receivers, I think that there was a desire to keep the earlier IF stages linear so as to avoid limiting out the selectivity.

Sometimes very “fast” agc was used to IF stages to provide, as it were, “inverse” AM modulation and so help with the limiting process. The Quad FM is often quoted as an example of this, although this circuitry applied only from S/N #8677, the 2nd iteration of Series B. Here tapped down but unfiltered AGC voltage from the 6AU6 limiter grid was fed back to the 6BJ6 1st IF stage, but not to the 6BH6 RF stage. In previous versions the AGC voltage was filtered and fed back to the 6BJ6 RF stage, but not to the 6BH6 1st IF stage. The Quad FM circuit was evidently a learning curve from the G. Horn original, and I suspect that the designers found that in practice, more AM rejection was advantageous whereas RF overload was not common in the early FM days, and when it did happen, and easily dealt with by inserting an aerial attenuator.

In the Armstrong 224 case, one might reasonably infer that the objective of applying AGC over two IF stages was to keep them linear and so allow the EF80 to do all of the limiting work. Also, one can also wonder whether or not there was an element of happenstance. The EF89 might have been chosen because it was apparently easier to deal with from a minor response-skewing feedback perspective than the EF85 or EF80 (although here the 2nd IF stage looks to be neutralized). As far as I know there was no sharp cut-off counterpart to the EF89 that might otherwise have been used. Thus, with remote cutoff valves in situ anyway and AGC voltage readily available from the limiter grid, Armstrong elected to use it mostly because it was there.

One also finds AGC applied to IF stage suppressor grids in FM receivers; there are early Armstrong (FM56) and Chapman examples of this practice.

And the limiter grid could also be used as an AM demodulator, as was the case with the Jason JTV/2 TV Sound/FM tuner. I guess the component values were chosen for good AM linearity. I think that the original Quad FM concept back in 1953 also made provision for reception of the then Wrotham VHF-AM transmissions, which might explain why the 2nd IF stage is a high gain amplifier not a combination amplifier/limiter.

The arrival of lumped IF selectivity (usually in the form of ceramic filters, but occasionally L-C combinations, as in the Revox A76) meant that the whole IF gain section could be limiting as well as amplifying, with AGC reserved only for the RF stage(s) if needed. Of course, this happened in what might be called the second phase of the solid state, at which time FETs and ICs were also being deployed. The landmark RCA CA3089 IC is an example; it provided the post-selectivity IF gain, limiting, and demodulation, and amongst other auxiliary functions, RF AGC.

Regarding tuning indication, the UK makers seems to have chosen either the centre-zero type (such as meter or twin neon) driven from the DC output of a Foster-Seeley discriminator or balanced ratio detector, theoretically zero at centre-channel, or the signal strength type (meter, magic eye) driven from the AGC line, the limiter grid or from across the ratio detector limiting capacitor. The former could give a relatively precise centre-of-channel indication but no indication of signal strength. The latter indicated signal strength, with centre-of-channel indication requiring peaking of the IF response, which strikes me as bit of a compromise. Somehow the late 1940s American idea of a magic eye that could do both functions (6AL7GT) seemed not to make it across the Atlantic. Maybe Philips/Mullard just did not want to do it. There were exceptions, though, as with the magic eye with AC injection used on the original Leak Troughline, but not carried over to the Mk II and later versions. Armstrong evidently varied its valve FM tuner practice according to model; centre-zero meter for the 227 at interest here, but conventional magic eye for some others.

Cheers,

Steve

Great summary

Excellent!

Some extremely useful information Synchrodyne, thank you.

This has encouraged me to look a little further at the circuit I have. I believe that it uses a Foster-Seeley discriminator, and from what I have quickly read, this was sometimes favoured in Hi Fi applications because the output is capable of being more linear (ie less distorted) than a ratio detector. Comparing the Armstrong discriminator to example Foster-Seeley circuits on the internet, Armstrong have not included a choke between the centre of the two load resistors (R18/19) and the centre tap of the secondary (L23). Any thoughts on why this was not included in the Armstorng circuit and any implications for the stereo decoder/amplifier sections which follow?

Alex

No problem, and thanks for the kind words.

Generally I think that Foster-Seeley discriminators were preferred for hi-fi work in the valve era because of their better linearity as compared with most other forms of FM demodulators relatively easily realized with valve technology. (Ratio detector, slope detector, nonode, gated beam valve, locked-oscillator, etc.) But once solid state FM tuners reached the point of competing with valve designs in terms of signal handling, etc., in the late 1960s, very wide band ratio detectors seemed to be the order of the day until the quadrature demodulator took over. (UK examples include the Rogers Ravensbourne (1968) and Ravensbrook (originals), the Leak Stereofetic (1969) and I think the Sugden R21/R51 (1970?). The Quad FM3 (1971) was quadrature from start, as was the Armstrong 600 series.) I am not quite sure why wide band ratio detectors were easier to implement with solid sate than with valves, though. Maybe lumped selectivity was a factor as well, in that the ratio detector IFT was not required for an IF bandshaping as might be the case - a little anyway - in a distributed selectivity design.

But your detailed discriminator question goes beyond my layman’s knowledge and understanding, so it has been necessary to look at the books.

Starting with the usual suspects, Witts (7th) and Henderson seem not to cover the Armstrong variation of the Foster-Seeley, sans RFC, although they do cover the shunt diode variant, which also does not use a RFC. Tibbs & Johnstone (2nd) do not cover it in its Foster-Seeley section. But luckily Keller does mention it as a circuit suitable for wideband use because the diode load resistors also damp the IFT primary. (I’ll scan and post the page when I get a chance.)

But on the basis that the information for which one is looking is not always where you expect to find it, I took another look through the books at interest. Witts unwittingly (groan) shows such a Foster-Seeley circuit in its section on AFC. Tibbs & Johnstone wait until they present a comparison (mostly about relative linearity) of the Foster-Seeley, the Philips nonode (EQ80) and the gated beam valve (6BN6). Here they point out that a different Foster-Seeley circuit is shown, with the diode load resistors damping the primary.

So I’d say that Armstrong chose its Foster-Seeley circuit to be of the wideband type for the reasons given by Keller and Tibbs & Johnstone, which I think would make it better suited for use with a stereo decoder.

Of course, after reading through the quoted references, what was previously a puzzle now seems to be at least partially understandable.

Then after all of that I looked in Langford-Smith (1997 reprint). Much information there, and lo and behold, the Foster-Seeley form used Armstrong is presented as the primary form, with the RFC-variety as an alternative. That makes me wonder what was the original form developed by Foster and Seeley of RCA, which I think was originally intended for AM receiver AFC purposes.

Please let me know if you want scans of any of the above items.

Cheers,

Steve

... That makes me wonder what was the original form developed by Foster and Seeley of RCA, which I think was originally intended for AM receiver AFC purposes ...

Funny, isn't it, the way a chance remark like that can throw the brain completely off target ... That instantly brought to my mind images of the Russian radiograms marketed here in the 60s under the Rigonda name.

These had AFC on both AM and FM which drove a small motor connected to the tuning drive! It was a strange sensation if you tuned through the Medium waveband to feel the tuning knob tug under your hand at every carrier you passed as the AFC tried to lock on to it!

Even without the low price, these machines really outperformed most home grown products with particularly good sound quality but looked a bit out of place from a styling point of view - the main cabinet stood on legs about 3' long and the wood, when matt finishes were universal, had a fantastic gloss finish.

The long legs would probably suit us old 'uns as it made changing the records less of a back aching task but they did look rather unstable - later batches were supplied with much shorter legs ...

The SABA Automatics have AFC on AM & FM too if you leave it on automatic/Search.

Click on 2nd thumbnail to see the Motor. I think uses a separate EABC80 and ECL80 for the Search and AFC functions. SABA are also unusual because of the variable coupling adjuster to change spacing of IF coils and thus bandwidth.
http://www.radiomuseum.org/r/saba_meers ... 125_s.html

Some versions had the four search/seek buttons on a plug in cabled remote. Of course you can't see the scale from your arm chair to know the station it found.

I think that the original version of the Ekco "radio-brain" motor drive radios used AFC as well. Steve, my technical knowledge is strictly limited, but I would be interested in learning more about discriminators - particularly practical applications in FM broadcasting - is there one particular reference book you would recommend I look out for?

I do have one further question about the tuner. When I tune to any particular station, I notice that the best reception (ie least distortion) is when the tuning meter is off-centre. I have checked the drive voltage to the meter to confirm that the meter is not faulty and indeed the voltage from the discriminator is typically about -0.5v for "best" reception (the max is + or - about 3v as you tune across any station). If I adjust the tuning for 0v at the discriminator output, I notice that the AGC voltage is significantly below its peak for any particular station. Is it reasonable to conclude from this that the discriminator is no longer accurately aligned with the rest of the IF? I've obtained a copy of the service data for the tuner, and unfortunately the alignment procedure requires access to a signal generator with with FM modulation, and mine is AM only. The procedure for aligning the discriminator is to inject 10.7MHz (with FM modulated audio tone) into the grid of the EF80 and peak the two discriminator tuning slugs for maximum AC audio at the discriminator output. The procedure also shows oscilloscope traces from the discriminator when a sweep generator is connected to the same input. Again, although I have an oscilloscope I don't have a sweep generator. I guess the answer is probably "no", but is there any other way to reasonably accurately align the discriminator coils with the limited test gear I have? If necessary, I will look for a suitable sig gen for FM work if anyone can recommend something simple and reliable?

Alex

... When I tune to any particular station, I notice that the best reception (ie least distortion) is when the tuning meter is off-centre. I have checked the drive voltage to the meter to confirm that the meter is not faulty and indeed the voltage from the discriminator is typically about -0.5v for "best" reception (the max is + or - about 3v as you tune across any station). If I adjust the tuning for 0v at the discriminator output, I notice that the AGC voltage is significantly below its peak for any particular station. Is it reasonable to conclude from this that the discriminator is no longer accurately aligned with the rest of the IF?

Alex, I would say that that it is not an reasonable diagnosis but it may not be the correct one.

Imbalance between the two OA79s or between R18 and R19 could cause a similar effect. (I assume that the meter indicates zero when the set is switched off.)

A leaky C27 could also cause problems but, if it is a good quality Silvered Mica component, unlikely ...

As has been said here on many occasions, do not touch any tuning cores unless all other possibilities have been exhausted!

As a last resort, check the tuning cores a look for any sign of tampering. If they have hexagonal holes and you do not have the correct trimming tool, under no circumstances touch them as they are virtually guaranteed to crack if you use the wrong tool, which then puts you in really serious trouble ...!

... unfortunately the alignment procedure requires access to a signal generator with with FM modulation, and mine is AM only ...

Not necessarily a problem ...

The procedure for aligning the discriminator is to inject 10.7MHz (with FM modulated audio tone) into the grid of the EF80 and peak the two discriminator tuning slugs for maximum AC audio at the discriminator output

Interesting, as the schematic you posted shows only one!

The slug of interest, if you really need to touch one, it is the secondary (adjust for zero output at 10.7MHz) ...

Terry,

Thank you for your detailed advice. There is no sign that any of the cores have been touched and I have previous experience of cracking an IFT core which I don't want to repeat! Well, I'm in a bit of a quandary because I would like the tuner to perform as well as possible, but I appreciate that there is a risk of wrecking it. So for now, I think I will just check the components you have highlighted and not attempt to adjust any cores. In terms of checking the two OA79s, is this simply a question of disconnecting one lead each and checking their forward voltage drop on a multimeter? How similar should the readings be for a reasonable match?

Alex