Slew rate distortion in a 1970 Fender Vibrolux reverb.

[Helmholtz]

I could find little info on slew induced distortion (SID) in tube amps.
It's typically discussed in the context of opamps.

SID tends to change the shape of a sine signal into a triangle above some frequency and voltage amplitude.
Lowering frequency or amplitude gradually restores the sine shape.
(Any change of a sine signal means non-linear distortion and thus added harmonics - as opposed to the effect of linear filters, which don't change the sine waveshape.)

That's exactly what I'm seeing with this amp: https://docs.google.com/viewer?a=v&p...ZmMjg4ZWExYjE3

At higher output and above around 3.5kHz a sine signal starts to morph into a triangle.
The reason seems to be the 2nF caps wired between the power tube grid and ground.

Assuming a PI source impedance of around 20k, the caps will cause a LP filter with a 4kHz corner frequency.
Means that open loop gain drops by 6dB/octave above 4kHz.
But NFB counteracts, so measured closed loop -3dB frequency is about 9kHz.

All this can't explain the observed sine distortion.
But there's another effect caused by these caps:

Changing the voltage across a cap requires a current. The faster the rate of change (dV/dt), the higher the required current.
The dV/dt of a sine is highest at the zero crossings.
The current has to be delivered by the PI, but by design there's a peak current limit.

A 3.5kHz sine voltage having a peak value of 35V across a 2nF cap has a max. dV/dt of about 0.8V/µs and requires a peak current of 1.5mA.
I estimate the max. peak current the PI can deliver to be around 2mA and a considerable part of that will be absorbed by the 47k grid leaks.
Current compression is likely to start at lower values.

The current limiting results in a max. possible dV/dt across the caps, and that is the slew rate limit (as the slew rate scales with voltage level, the slew rate at the amp's output is lower than the slew rate at the power tube grids).

Signals that exceed that slew rate get distorted.

As cap voltage and current are out-of-phase by 90°, the distortion starts at the zero crossings of the voltage, where current is max.

NFB cannot increase amp slew rate.

Slew rate limiting is not a frequency response effect.
It is not caused by a limited frequency response (which doesn't cause distortion) but rather by the inability of some internal stage to charge a load capacitance fast enough.

Lifting the caps considerably improves sine wave reproduction.

I think Fender used the caps as a brute force means to avoid any risk of stability issues.

There are certainly more elegant/less invasive methods to achieve that goal.

I'll see what it takes.

[Chuck H]

Brute force was what they had at the time. The lead dress in amps that use the power tube grid cap to ground weren't typically as eloquently wired as the earlier versions. Perhaps it had to do with cutting expense and labor expertise or no better idea by any CBS designers. They would have been well off to have someone with your skills in the design department. It's probable they didn't. Which could explain why so many post CBS design changes don't satisfy the amp owners and the "black face" legend got started. The earlier bf versions may not have simply been keeping things rudimentary. Perhaps it was recognized that too many peripheral corrections can have detrimental affects on how graceful the amp performs over a wider range of circumstances.

I know much of the design staff for Fender that made the post CBS changes was already under Fenders employ. Or so I've read. But their efforts to make some changes were stifled. Once CBS owned the company they trusted these engineers. And why wouldn't they? So some changes were made and thought to be improvements. though maybe not with enough consideration for the application of the product at hand. Which was a guitar amp.

[pdf64]

Over on TGP, Jeff Gehring has reported on stability problems in Fenders with the 022848 OT https://www.thegearpage.net/board/in.../post-32062697
That may be at least part of why the brute force stability measures were implemented.
Some frequency and phase plots of your OT may be revealing

Regarding the engineers at Fender, I dunno about the late 60s era following Leo’s departure, but by the early - mid 70s, Ed Jahns seems to have been the main EE. My understanding is that he was a respected figure in the EE sector, though I fear that as a guitar amp designer, his rep will forever be tarnished by the pull boost on the 70W and 135W models. His later 75 and 140 models are a lot better than those though.

[Helmholtz]

Over on TGP, Jeff Gehring has reported on stability problems in Fenders with the 022848 OT https://www.thegearpage.net/board/in.../post-32062697
That may be at least part of why the brute force stability measures were implemented.
Some frequency and phase plots of your OT may be revealing

Thanks for the interesting link.
But I don't agree with some of the conclusions.

Fender used those 2nF grid capacitors in a number of SF models having different OTs, including the Twin Reverb.

The 022848 was used in quite a few (mostly BF) models without needing those caps. Some had a 100pF between the PI plates, others didn't.

This OT sounds good and measures fine. In fact I liked every Fender amp that had it. Replacing it would be the last thing I'd consider.
About 20 years ago I did listening tests and measurements with some originals and replica samples, helping the TAD guys to get their replicas right.
The 022848 has high inductance and low capacitance, only leakage inductance is relatively large.

Without the caps and with a resistive load the amp is stable, but shows some ringing on square waves.
If there's instability with speakers, I'll try a textbook approach, i.e. lead compensation in the NFB path.
E.g. a 1nF cap across the series NFB resistor.
But there are many options.

[tschrama]

To add and summarize to your description of slewrate limiting: Slewrate distortion is clipping in the current domain. Somewhere a current is max'ed out working into a capacitance, resulting in a max'ed out and constant dV/dt. There is no response to an increased drive voltage, so there is no open-loop gain to counter its effect with global NFB. In traditional voltage clipping, dV/dt is max'ed out and constant at zero, in slewrate dV/dt is max'ed out and constant at some maximum value. It is similar to ordinary clipping, just at a different dV/dt.

When I first saw those 2nF cap I couldn't believe my eye's. If you would more traditional values between the PI anode resistor and measure the openloop gain/phase response, I wonder what you are left with for additional stabilization needs.

[Helmholtz]

If you would more traditional values between the PI anode resistor and measure the openloop gain/phase response, I wonder what you are left with for additional stabilization needs.

Today I saw (and heard) a stabiity issue at high output with speakers (2nF caps removed). No problems with resistive load.
A cap between the PI plates of at least 150pF stops the issue.

[Helmholtz]

Update:

I found that 150pF between the PI plates is not sufficient e.g. with different speakers.
Had to increase to 300pF.

Other attempts to improve stability like capacitors across upper or lower NFB resistors did not help - rather made things worse.

This experience again shows that textbook frequency compensation for resistive load is rather useless, as the speaker impedance changes loop gain and phase response at HF.

[Chuck H]

This experience again shows that textbook frequency compensation for resistive load is rather useless, as the speaker impedance changes loop gain and phase response at HF.

Not to mention that this will be dependent on the specific speaker to a small degree. I guess most modern "guitar speakers" shouldn't be too dissimilar from common full range "loudspeakers" from the early to middle tube amp era. Since that's what was loaded into those amps. But maybe they are?

I guess I'm wondering if the "textbook" stuff might work better with different (not guitar specific) speakers.?.

[Helmholtz]

Not to mention that this will be dependent on the specific speaker to a small degree. I guess most modern "guitar speakers" shouldn't be too dissimilar from common full range "loudspeakers" from the early to middle tube amp era. Since that's what was loaded into those amps. But maybe they are?

I guess I'm wondering if the "textbook" stuff might work better with different (not guitar specific) speakers.?.

I can't comment on full range speakers, but the original speakers (2x Oxford 10") were fine with 200pF, while the 12" Jensens in a Twin cab required 300pF.

Don't get me wrong, I'm not sayig textbook compensation is generally useless.
But it's typically done for resistive load only, which doesn't assure stability with speakers.

It probably works better with SS amps having low open loop output impedance.
But tubes have high output impedance and require an OT.
OTs typically have several resonances above 30kHz and the speaker impedance interacts with them.

Applying textbook methods would require full analysis of the amps open loop gain and phase response with designated speaker load up to several hundred kHz.

[Helmholtz]

For the sake of completeness I also tried a conjunctive filter (aka Zobel or Boucherot network) across the OT primary, consisting of a 4k resistor in series with a 1nF or 2.2nF cap.

Result: No stability without the PI cap.

[g1]

Can you comment on any audible manifestations of what you see on the waveform?

[Helmholtz]

Can you comment on any audible manifestations of what you see on the waveform?

You mean how the instability manifests?

[g1]

Yes, if you are just listening, what do you hear and how noticeable?

[Helmholtz]

With a 400Hz sine fed to the PI input, I increase the signal level slowly until the sound breaks into a nasty crackling distortion. At the same time/ level the HF oscillation shows on the scope around the 400Hz peaks. It looks like the scope pic in the link pdf64 provided (post #3).
It always reliably starts at the same signal level - as long as I don't change compensation.
When I see it I back down quickly to avoid stressing the OT.

Another method is using a square wave signal. With this an instability already shows at low signal level as a ringing that last for the whole top of the square wave.

[g1]

Sorry, I'm not trying to be difficult and not being very clear. What I was wondering about was when playing with a guitar at typical Fender clean settings.
In particular I'm wondering about how the affect on the sound would correspond to what people complain about Fender SF not sounding as good as blackface.
Is this something that would make someone think 'this amp is broken', or rather think 'people are right, silver face don't sound as good'.