Negative feedback loop in a Silvertone 1482 amp: a possible mod?

[slidincharlie (Carlo P)]

I'd like very much to try adding a switchable NFB loop in my Silvertone 1482 amp (schematic attached).
I am perfectly aware that NFB would change the amp's character, but I definitely want to try it. After all it would be switchable and fully reversible.
There is a mod proposed by a user in the Hoffman Amplifiers Forum (also attached), which feeds the signal from the speaker output jack back to one cathode (pin 3) of the PI tube. I'm headed to a simpler approach based on one single feedback resistor that can be switched on/off via a toggle switch (SPDT?).
Does the 1482's circuit allow for a NFB loop? How should I build it?
Thanks in advance,
--Carlo

[Chuck H]

Adding a resistor from the speaker jack + to the ungrounded end of the 2.2k cathode resistor on V2 would be a typical NFB loop. Probably about 100k. You can experiment with higher or lower values.

EDIT: You'll want to route that NFB lead away from any preamp circuitry. Even then instability is possible because we don't know the phase of the OT and PI relationship. If gain increases or instability occurs you might have to swap your PI leads.

[slidincharlie (Carlo P)]

Thank you Chuck.
I have attached also the schematic of the NFB added by the forum user that I had mentioned in the strating post.
He tied the NFB resistor(s) to the V2 cathode at pin 3. The 2.2k cathode resistor disappeared from his schematic. I wonder if he took it away from the circuit or if he forgot to add it to the diagram.

[Chuck H]

It's in his switching system that he added to the schematic. He is changing values to try and emulate circuits used by other amps. I never recommend this approach because component values don't emulate any circuit by themselves. Conditions surrounding components dictate values. This would be especially true with NFB circuits. I wouldn't be trying to copy anything from that mod link.

EDIT: I just tested those mods on LTSpice and the "Princeton" NFB loop reduces gain more than typical for a guitar amp circuit and the "Harvard" NFB loop actually demonstrates instability in the simulation. This is why no one should arbitrarily cut and paste circuits between different amplifiers. It would be like swapping the beans in your chili for spaghetti noodles and expecting it to taste more Italian

[Helmholtz]

.. and the "Harvard" NFB loop actually demonstrates instability in the simulation.

How did the instability show in the sim?
What's your OT model?

[Chuck H]

How did the instability show in the sim?
What's your OT model?

I simulated the Silvertone all the way to the PI. But that's all I can do because I haven't had luck uploading other tube models yet. So the circuit would be the silvertone with el84's. The OT model is a generic that came from Dave H here on the forum.

The instability is a 32k oscillation modulating the test frequency at an amplitude as high as 6V peak. I've had the opportunity to compare sims done with Dave's models and they're always accurate to what I see on the bench with my el84 amps. So, I know that a 6V6 doesn't have the gm of an el84 so that might prevent this oscillation. But I do think the circuit is less than ideal and likely borders on instability at best. Better to implement a NFB loop from scratch than just paste in a circuit designed for an entirely different amp IMHO.

[slidincharlie (Carlo P)]

Adding a resistor from the speaker jack + to the ungrounded end of the 2.2k cathode resistor on V2 would be a typical NFB loop.

This is the info that I needed.
Re: the other user's mod ("Platefire mod") with the two different NFB loops inserted in the circuit, Platefire himself stated in several different threads in other forums that the mod was very effective and he was happy with the result.
Anyway a single switchable (on/off) feedback resistor is what I'm heading to.

[Chuck H]

This is the info that I needed.
Re: the other user's mod ("Platefire mod") with the two different NFB loops inserted in the circuit, Platefire himself stated in several different threads in other forums that the mod was very effective and he was happy with the result.
Anyway a single switchable (on/off) feedback resistor is what I'm heading to.

If Platefire likes his amp that's all that matters for him.

I'd start with a 100k resistor. You could temporarily wire in a 250k pot as a variable resistor so you can adjust the resistance to your own tastes and then measure the value to choose a fixed resistor. Platefires "Harvard" circuit would be exactly what is proposed using a 56k resistor Which is a lot more NFB than needed IMO and would cause unnecessary loss of gain but use your ears.


EDIT: Oh, and I have my info in a previous post backwards. It's Platefires "Princeton" circuit that caused the instability.

[slidincharlie (Carlo P)]

I have added a 5k linear pot to the stock 820 ohm feedback resistor in my '76 Fender Deluxe Reverb amp. With such small-value range the effect is already noticeable and it's enough for my needs, but it's a totally different beast than the Silvertone.
I will experiment with resistors and report here.
Thanks.

[Chuck H]

I have added a 5k linear pot to the stock 820 ohm feedback resistor in my '76 Fender Deluxe Reverb amp. With such small-value range the effect is already noticeable and it's enough for my needs, but it's a totally different beast than the Silvertone.
I will experiment with resistors and report here.
Thanks.

The way this reads it sounds like you put the pot in series with the feedback resistor. If so it would only allow you to decrease NFB in the Deluxe Reverb. What I proposed above is that you use a pot only for your Silvertone mod. This would allow you to choose any NFB level both lower and higher than typical. Then measure the resistance and use the nearest standard resistor value for your fixed circuit.

[Helmholtz]

The OT model is a generic that came from Dave H here on the forum.
.

Does the OT model contain parasitics like primary capacitance and leakage inductance (as those are essential contributors to phase rotation and instability)?

And yes, EL84s instead of 6V6s makes a lot of difference, as with EL84s the (forward) gain is almost 9dB (factor of 2.76) larger.
The greatly increased NFB ratio ( = gain without NFB divided by gain with NFB) will promote instability.

[Chuck H]

I did reasearch this, in a way. I used a split load for the power tube grid leaks to reduce loop gain to approximate the Silvertone. This didn't solve for the oscillation. What did though is using a Zobel network across the OT primary. There are parameters in the OT model for capacitance but they only offer it across either half wind. Certainly there will be capacitance between either half wind as well but there is no way to put this into the OT model itself. The Zobel should be a similar affect to winding capacitance between either half winding and did correct the oscillation. But only if a resistance in series with the capacitance was used. With only a capacitance the oscillation remained.

EDIT: There is also the aspect of a speaker load instead of a dummy load as well as many aspects about OT specifics that could contribute to a result. My thinking is that you are right to pin the oscillation to the OT model in question. Whether it's a good model or not is secondary to whether it emulates the OT in the Silvertone. When dealing with ultrasonic oscillation these are no small matters. Phase errors in the loop above audible frequency aren't often addressed at my level in the craft but I know a little about them.

[Helmholtz]

I used a split load for the power tube grid leaks to reduce loop gain to approximate the Silvertone.

So you used voltage dividers?

There are parameters in the OT model for capacitance but they only offer it across either half wind. Certainly there will be capacitance between either half wind as well but there is no way to put this into the OT model itself. The Zobel should be a similar affect to winding capacitance between either half winding and did correct the oscillation. But only if a resistance in series with the capacitance was used. With only a capacitance the oscillation remained.

Now that makes sense.
I just couldn't explain an oscillation at a frequency as low as 35kHz with an ideal transformer model (i.e. zero capacitance and zero leakage inductance).
As soon as you add capacitance to one winding of a transformer, all other windings will show a reflected capacitance due to coupling.
E.g. if you place 400pF across one half-primary, the other half will also show 400pF and the full primary will show 100pF.
If you place 400pF across each half-primary, the effective capacitance across the full primary will be 200pF.
I also assume your model uses a coupling factor <1. This corresponds to leakage inductance.
Leakage inductance together with winding capacitance form a resonant circuit.
A resonant circuit can introduce a phase shift of up to 180°, meaning phase reversal, turning NFB into positive FB.
For this reason amps often like to oscillate close to this OT leakage resonant frequency.
(If desired you could add external components in your schematic to improve the OT model.)

A Zobel is a capacitively coupled load resistor. Without the resistor it's no longer a Zobel and won't provide any damping

[Chuck H]

A resonant circuit can introduce a phase shift of up to 180°, meaning phase reversal, turning NFB into positive FB.
For this reason amps often like to oscillate close to this OT leakage resonant frequency.

I've noticed this in the simulations. It became especially important on a particular design I've been putting together that isn't a guitar amp. I recognized the phenomenon and figure I'll need to make fine adjustments on the bench to realize the desired result relative to the specific OT in actual use.

And yes, I turned the grid leaks into voltage dividers. I know this is a blunt effort but I just wanted to see what happened with reduced loop gain.