How Much Negative Feedback Do You Use

[WA1QHQ]

I am redoing the modulator in my Ranger based mostly on Tim HLR's article in the AM Window. The main difference is I am using a different modulation transformer than what was originally in the Ranger (UTC S-19). I am coupling the feedback off the mod transformer secondary through a .1uF coupling cap and dropping resistor (270K) to the unbypassed cathode of the 12AU7 stage before the phase splitter. If I go much above 12 dB of feedback I see what appears to be high frequecy oscillations riding on top of the low frequency (200Hz) sinewave I am feeding into the modulator, looks like a parasitic oscillation. The modulator is very clean and stable below this level of feedback. The amount of feedback is flat across the audio range with frequency response being determined by the modulation transformer, I am not tailoring frequency response in the feedback loop.

I am assuming the oscillations I am seeing are due to a 180degree phase shift at some very high frequency due to phase shifts mainly in the modulation transformer.

Is 12dB of feedback a reasonable amount or have some of you been able to provide appreciably more than this and still get stability and a wide frequency response?

Any suggestions would be appreciated.

Mark WA1QHQ

[AB2EZ]

Mark

I'm sure that you will get a variety of different views on this.

My advice is not to use feedback at all. Feedback in audio chains is helpful for making a circuit that is already fairly linear (distortion products more than 30dB down; or a well-behaved deviation from a straight line, like Ax + Bx², where the second term is small compared to the first term) even more linear. It is particularly ineffective for dealing with a problem like saturation of the modulation transformer. The feedback will only attempt to push more low frequency current through the modulation transformer (to make up for the missing/distorted low frequency content)... and, in doing so, will simply saturate the transformer even more!

1. Make sure that you replace (one way or another) the interstage coupling transformer. It is the controlling performance (non-linearity and roll off) bottleneck at low frequencies.

2. If all of the low level audio stages are working properly, and if you have implemented the mods that Tim recommends...you should have a very clean (very linear) signal driving the grids of the modulator tubes, with a flat frequency response from at least 30 Hz to 10 kHz.

3. Consider changing the modulator tubes to 6550's (if you haven't already done so). If you bias the 6550's properly (around 30 volts grids-to-cathodes, 300 volts on the screens, which should produce around 100 mA of modulator resting current for the pair of 6550's), you will be able to operate them in Class AB1 (no grid current, even on modulation peaks)... so you won't have to worry about non-linear effects associated with a sudden onset of grid current in the modulator tubes. A pair of 6550's in AB1 will deliver up to 100 watts into 5000 ohms with less than 3% distortion (according to the specification sheets). I'm getting excellent results with them in my Ranger.

4. Make sure that the screen bypass capacitor on the 6146 is no larger than .002 uF*, and make sure that the total of all of the bypass capacitors running between the modulated B+ and ground is no larger than than .004 uF**

*The input impedance looking into the screen varies with the modulation, but is roughly 12k ohms. You don't want the screen bypass capacitor to have an impedance of less than around 12k ohms at the highest modulation frequency of interest (e.g., ~5 kHz)

**The output impedance of the modulator, as reflected to the secondary of the modulation transformer, is roughly 6k ohms. You don't want the sum of the B+ bypass capacitors to have an impedance of less than around 6k ohms at the highest modulation frequency of interest (e.g., ~5 kHz)

With the above precautions, and with your new modulation transformer... your audio should be superb... without using (generally to no avail anyway) any feedback.

Best regards
Stu

 

[WU2D]

Mark,

Feedback is the Devils' playground. Never used it - never will. Next you will be using wirewound resistors.

Still hostage out here in Silicon Valley - I saw a little hole in the wall last night on El Camino Real called Hiphugger Go Go - some kind of radio repair shop I think.

Mike

[K1DEU]

In the 50's I didn't understand how to use it. A lot of fear was normal without knowledge! I quickly noticed that the few that used it in the 50's like Homer Sawtelle in Jafrey, N.H. a friend of W1ECO's (on 3945 Khz daily at 10 AM) who both worked at Harvard underwater sound (Sonar) sounded far better (less distortion) than most. Here are two articles to remove the smoke and voodoo...  73 , John

Articles  After finishing Modulator  http://hamelectronics.com/k1deu/pages/ham/transmitters/am/pages/audio_phase_shift_scope.htm

Implementing network to not have off frequency ultrasonics between 20 to 40 Khz or on frequency motorboating before closing loop
http://hamelectronics.com/k1deu/pages/ham/transmitters/am/pages/audio_phase_shift_scope.htm

Index page  http://hamelectronics.com/k1deu/pages/ham/index.htm

[K1DEU]

To answer your question "20 Db." if we have the extra gain. Yes Mark ultrasonics are common when closing the loop. Paul W1ECO and I suggest to never use the audio negative feedback winding in the ranger modulation transformer, for its few turns couple poorly to the main windings and is so reactive its a compensation nightmare. As there is reasonable inductive leakage between the total primary and secondary, tap the proper outside of the Modulation Transformer primary for feedback takeoff.

Many operate without knowing they have an ultrasonic generator in their modulator and we can notice their Upper and Lower sideband on the air usually appearing with a unstable carrier and their modulation +and - 25 to 40 Khz of their main carrier. After completing work on a modulator simply listen off frequency for any issues. Here is a schematic Paul prototyped in his Ranger for an 20 db. stable feedback example. Because of space constraints Paul and I used the little 80 cent JFET for the phase splitter/inverter 73, John

Words  http://hamelectronics.com/k1deu/pages/ham/transmitters/am/pages/universal_speech_amplifier.htm

Schematic drawn by Bill, W1CKI 
http://hamelectronics.com/k1deu/pages/ham/transmitters/am/pages/universal_schematic.htm

[WA1GFZ]

Mike what town are you in that road rings a bell???

[Bacon, WA3WDR]

I have about 8 dB of negative feedback in my Viking II.

[W2XR]

I have 6 dB of negative feedback from the plates of the 833A modulators, to the first voltage amplifier stage in the audio driver. My entire audio driver/modulator tube line-up is comprised of only triodes: triode-connected push-pull 6J7s, push-pull 6J5s, push-pull 845s, push-pull 833As.

Triode audio power amplifiers require significantly less voltage feedback, when compared to pentode power amplifiers. Pentodes typically require on the order of 12 to 14 dB minimum voltage feedback to ensure a reasonable level of distortion, and sometimes as much as 22 dB of feedback is required for acceptable low-distortion performance. With triode power amplifiers, 0 to 8 dB of negative feedback is all that is required.

73,

Bruce

[K1JJ]

This is a very important and timely topic. In fact, Bruce and I were discussing it on the phone last week....

I'd agree that audio negative feedback is a worthwhile addition to any modulator.

If there is a transformer in the loop, then the level of feedback is usually limited by the xfmr's phase shift. (amplifier taking off) But you can usually get all the NFB you should use, even with a xfmr.  But if there is no transformer, many times there is no instability limit to adding NFB.  Generally most schemes come off the plate of the modulator tube and do not include the final mod xfmr in the loop.

From practical experience: 

I have about -8db NFB around my TRIODE connected 813's going back to a low level 1 volt stage. It makes a tremendous difference in cleaning up the lows and also cleaning up the extreme highs - which had a slight tendency to "spit."  Using the new WA3KLR FET cathode audio driver and NFB, it's now the cleanest rig I have built to date. (not including PDM, of course) The NFB was the final touch.

The 833C TRIODE modulators (4X1 rig) uses a backwards 8 ohm audio transformer to drive them, so I found it more difficult to achieve stable NFB without playing with the feedback caps/resistors and circuit loading values. In the end I use about -7db and it's stable. NFB made a noticable difference in the extreme low end. I can see improvement in the swept audio sine wave and also hear it in the ability to add more deep low end bass to my voice without added distortion. ie, I can add more 3rd-BA 60 hz (false) bass and get away with it without mud - whereas could not without the NFB.

I would also agree that there is a point of diminishing returns and maybe even degrading the sound by going over board with NFB. With the transformerless triode-connected 813's, I could add in up to -20db if I wanted, but found no advantage after about -8db or so. It simply required more audio drive, that's about it. 

BTW, perhaps the exact same conditions apply, but after adding RF NFB to an RF linear amplifier, I saw a noticable decrease in 3rd order IMD... dropped down to about -50db using a pair of 4CX-250's driven by a 6146 in closed loop. I overdid it and saw it clean up tremendously. Compare this with running RF linear 4CX-250's flat out with no NFB. (-32db 3rd at best?) Rule of thumb: Every 1 db of NFB = 1 db of improved 3rd order intermodulation distortion, and also requires 1 db of increased drive to compensate.


73,

T

[WA1QHQ]

OK so the next question... is it better to pull the negative feedback off the primary of the modulation transformer as John mentioned or the secondary as I am presently doing?

[W1VD]

T

RR the same improvement with linear amplifiers. Collins wrapped about 10 dB rf feedback around the final/driver in the 32S3 transmitter and a properly adjusted unit exhibited better than 40 dB third order IMD. Cleanest SSB transmitter I measured at the time (70s).

 

[K1JJ]

OK so the next question... is it better to pull the negative feedback off the primary of the modulation transformer as John mentioned or the secondary as I am presently doing?

Mark,

That's probably why you are seeing oscillations - due to the mod transformer (secondary) in the NFB loop.  It's a noble attempt, though... :-)  If you're now getting -12 db with a mod xfmr in the loop, that's pretty darn good and I would be satisfied even if you back it off to -8db for total stability....

I would imagine it would be great to be able to include the mod xfmr and attempt to buck the phase shift it creates, but I think in the real world most attempts are met with instability if you try to get more than a couple db of NFB.  I have usually had problems including the mod xfmr here, so settle for the xfmr primary tap. (off one modulator plate)

NFB does work with smaller interstage and driver transformers in the loop, (I'm doing it here)  but maybe with the bigger mod xfmrs it's harder to control phase shift? I dunno.

I'd like to hear from others who have tried or are using NFB with the mod transformer in the loop -  and their results, esp with big rigs. 

T

[WU2D]

Mike what town are you in that road rings a bell???

Sunnyvale - and I am staying until Tuesday AM so I will be doing a road trip this weekend - Dunno maybe head to Ocean.

Mike

[W2XR]

OK so the next question... is it better to pull the negative feedback off the primary of the modulation transformer as John mentioned or the secondary as I am presently doing?

This would be the ideal solution, but the issues of phase shift through the modulation transformer and audio driver transformer usually make this very difficult to achieve successfully in practice. Even the later-design broadcast transmitters that eliminated the driver transformer by using a cathode follower arrangement to drive the push-pull modulator grids, did not derive the feedback from the secondary of the modulation transformer.

I personally am not familiar with any commercial broadcast rigs that did take the audio off the secondary of the mod transformer; this was probably not done for the following reasons:

1) The audio in most broadcast rigs is push-pull right out of the audio input transformer. The audio would, therefore, have to be taken from both the modulated B+ side and the unmodulated B+ side of the transformer to provide the feedback voltage for both push-pull phases in the audio driver section. The unmodulated B+ side of the mod transformer is returned to ground through the DC blocking capacitor, and this combination probably forms a very non-linear impedance with attendent additional phase shift issues. I am suggesting here that the phase shift from each side of the mod transformer secondary would be different, creating further instability within the two feedback loops required in a fully push-pull audio line-up.

2) The issue of phase shift through the mod transformer itself, as discussed earlier. In a single-ended/non push-pull audio driver circuit, similar to what I believe you are probably using, this is the dominent issue.

A good reference for these kinds of issues/questions is to look at how the classic plate modulated broadcast rigs of yore handled these things. There were some extremely talented engineers working for the likes of Gates, Collins, RCA, Continental, et al, designing these rigs, and if they could have implemented feedback at the secondary of the mod tranny, you can be sure they would have done so. It would be highly advantageous to do this, as any non-linearities that always exist within the modulation transformer could be corrected to some degree thru the careful application of voltage feedback, thereby improving the overall fidelity of the transmitter.

Most of the plate modulated broadcast rigs that I am familiar with took the feedback right off of the plates of the modulator tubes. This voltage was reduced through a voltage divider, and the  DC component was blocked with a capacitor. The remaining AC audio voltage was then fed 180 degrees out of phase to an earlier voltage amplifier stage within the audio driver to close the feedback loop.

73,

Bruce

[N3DRB The Derb]

Actually, I tend to get negative feedback every-time I get on the air.... 

[K1DEU]

RE; is it better to pull the negative feedback off the primary of the modulation transformer ?

On a common core (such as a modulation transformer) when there are many turns per winding, there is no, in and out. If there is audio phase shift approaching in one winding the common magnetic coupling distortion is experienced in all. 

In plate modulated transmitters (unless very low level) it is almost impossible to decouple RF from the modulated B+ lead back to the modulation transformer secondary without excessive shunt bypass RF, decoupling capacitors, which add greatly to audio phase shift problems. Amplifiers do not like capacitive loads.
Therefore as the audio distortion appears at both the primary and secondary (with reasonable low leakage inductance between windings) of the modulation transformer, its safer and simpler to tap the output waveform from the transformer primary, not the secondary.  John 

[K1DEU]

Re; putting more than band aids on modulators that are not stable using negative feedback.

In the patient (but often boring) design world, after we construct a modulator with an additional 20Db. of gain before closing the loop, we should measure the corners high and low where the phase shift reaches 90 degrees. A perfect oscillator would want positive feedback.  But Oscillators can take off with less. Before we close a feedback loop with reasonable gain we must insure there is little gain above or below these 90 degree frequencys measured, open loop or we have an oskliattor all the time or on voice peaks.

Seat of the pants? design might be.

First close the loop with >extra gain and measure the ultrasonic frequency. Just listen with a receiver for the upper and lower sidebands. They appear as off frequency unstable sub-carriers between 20 to 40 KHz  above and below the main carrier. The frequency difference between the carrier and either sideband is the approximate ultrasonic audio oscillation frequency.  The major phase shift which makes this oscillator is from transformers within the loop.  To stop the oscillation do not drop the feedback loop gain, but simply drop the gain in the amplifier/modulator just before reaching this ultrasonic frequency! Normally a resistor and capacitor series shunt for corner. see below

Low frequency audio motorboating same thing. Reduce the gain before it heads downward to this on carrier audio frequency. Normally a resistor and smaller series capacitor. see below

Place these low level roll off networks in the normal open loop path not in the feedpath line/ladder.

From our article;  Here is a page from the Radiotron Designers Handbook  editor;  F. Langford-Smith, fourth edition 1952   Chapter 7 page 66, 7.59  A+ B   http://hamelectronics.com/k1deu/pages/ham/transmitters/am/images/CHAPTR07_66_0001.jpg   116 Kb  showing  construction of simple networks to reduce the within loop gain to make a stable audio amplifier without motorboat or ultrasonic oscillations. The article  http://hamelectronics.com/k1deu/pages/ham/transmitters/am/pages/negative_feedback_design.htm

Notice how we load the driver transformer secondary and then corner the secondary for high frequency phase shift issues. http://hamelectronics.com/k1deu/pages/ham/transmitters/am/pages/universal_schematic.htm

The UTC S-19 would be too small for me in the Ranger. A gapped transformer should be rated around the DC input approximately. I like the original modulation transformer cross laminated with a DC choke and cap to remove the saturating unbalanced DC from its secondary. After cross laminating (less than 30 minutes) primary plate to plate inductance approximately triples lowering the low end corner significantly. The inductance of the 120 hertz DC choke (used as a audio reactor) may be less than noted. http://hamelectronics.com/k1deu/pages/ham/transmitters/am/pages/using_audio_reactor.htm

Enough; As I do not enjoy typing/one way broadcasting with two fingers. I have a mike and an AM transmitter for interacting with these things on 160 or 75 AM   Thanks,  good night,  John   

 

[AB2EZ]

Looking at the recent posts, it appears that some (not all) of the benefits being described are associated with expanding the frequency response of the modulator to lower frequencies, rather than improving the linearity of the audio path.

As is well known, negative feedback will broaden and smooth out the frequency response of the amplifier that the feedback is placed around, as well is improving the linearity. [To an extent limited by the usual issues associated with the theory of systems that employ feedback… e.g. phase shifts around the feedback loop]

The audio gain vs. frequency characteristics of a modulator of the type we typically use in our AM transmitters is intrinsically fairly flat, except for the roll off a low frequencies and the roll off (and perhaps some ripple) at high frequencies. I.e., if the stages are working properly, there shouldn’t be any significant ripple in the gain vs. frequency characteristic of the modulator* between the low frequency cut off and the high frequency cutoff.

*I am not including such things as the frequency response of an audio processor that is used between the microphone and the modulator.

Therefore, the benefits of feedback in a lot of the cases described in the recent posts (but not all) are primarily to boost the low frequency gain. This can be accomplished, without the use of feedback, by adding low frequency boost with a simple (or, if you prefer, complicated) low frequency equalizer. Likewise, the low frequency response of the modulator could be extended by making some of the inter-stage coupling capacitors larger, by making some of the cathode-biasing bypass capacitors larger, and by avoiding the use of undersized (not enough magnetizing inductance) inter-stage transformers.

Using negative feedback around a large number of stages can be fun to experiment with, and a good educational exercise… but it may be the third best solution to a simple problem of extending the low end frequency response of your modulator.

Putting negative feedback around a push-pull amplifier that is feeding a transformer (without using the output of that transformer) is easier said than done. [As alluded to by Bruce and John).   

To do this properly, you would need to use either:

A) a separate transformer, or
B) a balanced (plate-to-plate) voltage divider, or
C) an equivalent electronic circuit, such as a differential amplifier

to “sniff” the balanced output of the push pull amplifier.

You would need to feed the “sniffed” signal  back to input of a an earlier stage:

either an earlier single-ended-input stage; or (if you are using a balanced voltage divider, or a separate transformer or an equivalent electronic circuit with a balanced output) back to the balanced input of the push-pull amplifier.

All of the lower level stages of your modulator should be extremely linear to start with, because they a operating in Class A with very low signal levels (very little plate current swing compared to the average plate current that is flowing).

Remember, feedback can improve the linearity of a Class B or Class AB push-pull amplifier, but it can’t perform miracles. If your push-pull amplifier can’t put out more than 600 volts of peak audio before it runs out of voltage (i.e. if the B+ power supply is only 600 volts), using negative feedback isn’t going to make it put out 650 volts of peak audio. Negative feedback cannot fix “hard” limiting (among other things it can’t fix)

Repeating what I said in my earlier post… using feedback (or using equalization of low frequencies) to try to overcome the low-frequency saturation effects of an inadequate inter-stage audio coupling transformer is an exercise in futility. You can gain a small improvement, which might make sense if you were working for E.F. Johnson back in 1955, and you wanted to use a transformer with 10% less iron to save a dollar on each unit produced. If the transformer is saturating, either live with its limitations or: replace the transformer with one that can handle the signal levels, at the low frequencies you care about; or replace the transformer with an electronic circuit; or get some improvement by using a resistor to feed the DC to the plate of the tube driving the transformer (for single ended stages, like the one in the Ranger), and by capacitively coupling the output of the tube to the input of the transformer.

Of note, my KW-1 uses 6B4G triodes, in push-pull, as drivers and 810 triodes, in push-pull, as final modulator tubes… and there is no negative feedback employed. The audio reports I receive when using the KW-1 are excellent… and most people can’t tell any difference between how the KW-1 sounds and how my Class E transmitter sounds. Those who can hear a difference have never commented that they like the way one of my transmitters sounds more than the way another of my transmitter sounds.

When comparing my Ranger using its internal modulator (with the usual improvements) to my Ranger using an extremely linear and broadband external modulator… most people can’t hear any difference. Those few listeners who can hear a difference often like the sound of the internal modulator better, because (with its residual non-linearity) it sounds more “natural” to their ears when they are listening on their receiver.

Best regards
Stu
.

[WU2D]

The place where I have sen the greatest benefits in using feedback have been where the components are of the lowest quality or just plain wrong. For example, I use a backward 6.3VCT 5 Amp transformer as a modulation tranny in my mobile. You can imagine what the audio response of such a thing is. Feedback straightens out the response nicely.

Mike WU2D

[K1DEU]

Re; Can we hear the difference on frequency and off freq of a well designed plate modulated transmitter with a low Mu 12au7, 2a3's etc. push pull driver stage and push pull modulators with and without negative feedback. Like the KW-1 or a Johnson desk kilowatt or many other combinations including a strong signal from 100 watt class transmitters. 

Yes we can on frequency if; Our receiver detector and audio train have been modified including negative feedback to our headphones. And if our Balanced Hearing response is within 3 Db from 100 hertz to 5 Khz.

If our own hearing needs repair we can still witness a large difference off channel in third and even fifth order distortion products off channel with and without negative feedback. Especially as many use processing for denser audio.

We can use 7 to 10 Khz channel spacing but not with those who do not observe their off channel products. John

P.S.  Little single ended drivers for Ranger to 200 watt class are not for good ears. I like Stu, AB2EZ's transformerless push pull driver,  but would finish with negative feedback.
http://mysite.verizon.net/sdp2/id13.html

or the little old fashioned, universal push-pull Low Mu 12au7 driver
http://hamelectronics.com/k1deu/pages/ham/transmitters/am/pages/universal_schematic.htm



 

[WA1QHQ]

I don't believe I ever mentioned that I was using an interstage coupling transformer or the Ranger modulation transformer for that matter and I also don't believe I said anything about using negative feedback to extend my low frequency response. Everything is RC coupled right up to the grids of the modulator tubes and the frequency response is flat from 20Hz-20KHz with low distortion.

The purpose of the negative feedback was mainly to get rid of the distortion caused by the modulation transformer (UTC S-19) that I was using at the extremes of it's frequency range.

Wouldn't the distortions due to core saturation and other transformer inadequacies that are seen (especially)at the extremes of the transformer frequency response range be also reflected in the primary winding? If this is the case it may explain why broadcasters would just grab the feedback off the primary, it would still be effective in improving non-linearities induced by the modulation transformer. I can also see where RF on the secondary could get into the feedback loop and play havock this has been one of my concerns.

[AB2EZ]

"Wouldn't the distortions due to core saturation and other transformer inadequacies that are seen (especially)at the extremes of the transformer frequency response range be also reflected in the primary winding?"

If you neglect the effects of leakage inductance and winding resistance (and maybe a few other parasitic effects like coupling capacitance between the windings), then this would be true.

I.e., as John pointed out, and as you are suggesting:  v= N d(flux)/dt; so if both windings see the same flux one would expect the induced voltages to track. (The voltages measured across the input and output windings may not track because of the effects of voltage drops associated with leakage inductance, winding resistance, and other parasitic effects).

Again, the key issue is how much core saturation can you correct with negative feedback before you reach a point of diminishing returns (i.e., where the "correction" signal produced by the negative feedback produces more saturation problems than it fixes... including such things as hysteresis effects, where you introduce not just a non-linear relationship between input voltage and output voltage... but a much more complicated set of non-linearities). These effects may add more problematic forms of distortion (including increased distortion of mid frequencies), even if they appear to make the waveform look better when you put in a low frequency sine wave. I think the answer to that will depend upon the specific saturation characteristics of your transformer, and upon which types of distortion you like better when you listen to your off air monitor.

In the end, the improvement obtained with feedback (if any) will be more subjective (in the ear of the listener) than objective.

Stu

[K1DEU]

The UTC S-19 would be too small for me in the Ranger, even compensated. A gapped transformer should be rated around the DC input approximately such as 50-60 audio watts.
   I would use the original modulation transformer cross laminated with a DC choke and cap to remove the saturating unbalanced DC from its secondary.

   When the constant 72 Watt load (120 ma. X 600 VDC) is removed from the secondary and now carried by the modulation reactor choke there is an extra 72 watts of available core to use for audio ! Now we have more than doubled the original Modulation audio rating and practically eliminated the BH  distortion! After cross laminating (less than 30 minutes time) primary plate to plate inductance approximately triples lowering the low end corner significantly more than any gapped transformer. The inductance of the 120 hertz DC choke (used as a audio reactor) may be less than noted. W1ECO uses apx 15 henrys and 2 Mfd (these items may be shoe horned in) for his Ranger along with 20 Db. of feedback. John

 http://hamelectronics.com/k1deu/pages/ham/transmitters/am/pages/using_audio_reactor.htm

[WA1QHQ]

John,

I considered going the modified Heizing route and relaminating a mod transformer, I have done it with other rigs but the Ranger is just a little too cramped for space without much rearanging of the pieces which I didn't want to do. I originally had a Stancor A-3892 mod transformer in there which is rated at 60Watts and sounded pretty good on the air even without feedback. When I refurbed the Ranger and cleaned up all the cluges I had in it from years passed I took a close look at performance in all areas and one of the things I noticed was that the A-3892 didn't really look as good on the scope as it sounded on the air. I swapped in the S-19 which although rated at half the power of the A-3892 performed with much lower distortion under fully loaded conditions, just goes to show you that UTC really knows how to make a good modulation transformer. If I had an S-20 I would try to shoe horn it in there. By the way really good info on your web site thanks for the links.

[WBear2GCR]

I prefer to think about the use of feedback as correcting amplitude variation. All distortion can actually be described as amplitude variation of some sort.

In the case of the S series transformers, they're not the best in terms of frequency response. If you try to "extend" the HF response of such a transformer you will not really be able to do so, since the iron is drooping in response AND it is also going through phase shift. The result when this is tried in typical audio amp applications, is often extreme ringing on leading edges - and sometimes ultrasonic (parasitic) oscillations.

Imho one solution is to limit the HF response at the input, BEFORE the the feedback is applied - in this way the feedback is not trying to flatten the HF response in a range where it is unable to do so. Remember the feedback is merely comparing what it sees as an input signal to what it sees as the output signal and taking the difference to correct the output signal - no difference, no correction, no feedback!

After that it is also useful to then limit the HF response of the modulator tubes in terms of ultrasonic response. This may take many forms, but the easiest often is a small C or RC between plate and grid. Sometimes an ultrasonic damping (load) network across the primary or secondary winding may be useful too.

I'd also recommend using AB2 drive to the grids of the modulator tubes, although if you happen to be getting suffient drive, it might not be necessary - but the low Z on the grids also tends to eliminate a source of parasitic oscillation and gives much better results when driven into "clipping" than does the AB1 connection, imho.

                      _-_-Wombat Beaver Two Giraffe Cougar Rhinocerous