NFB and PI overdrive

[BackwardsBoB]

If you have a class AB amp with an LTP PI, and lots of headroom in the PI, so that it's running fairly clean when the power tubes clip, the NFB signal will clip, and the drive to the PI will suddenly increase, perhaps dramatically increasing the grid drive to the already clipped power tube, right?

Is this a good thing? Is there a fix? Should the PI's clipping level be tuned to be only slightly above the level required to overdrive the tubes?

[tbryanh]

If you have a ... right?

This sounds correct. I think this might be the condition where the feedback loop is considered to be broken or effectively open; i.e. no NFB.

Is this a good thing?

For tweedy or blackface amps, this is probably good. For modern amps with high gain preamps, this might not be good or at least not traditional for high gain amps.

Is there a fix?

Maybe a higher wattage amp could be used if power amp distortion is not needed.

Should the PI's clipping level be tuned to be only slightly above the level required to overdrive the tubes?

I have heard that when an amp is set up correctly, the last stage (power tubes) distorts first, then the PI, then the last stage of the preamp and so on until the first stage of the preamp distorts. This is what supposedly gives the amp its touch control. This probabaly only applies to tweedy and blackface amps, etc., not to modern high gain amps.

[Enzo]

I am not sure how the NFB "clips." The NFB is simply the output signal, sampled at a low level and fed back to an earlier stage. If a power tube were to clip, the NFB would send a copy of that clipped signal back to an earlier stage. It wouldn;t stop working though.

And by power tube clipping, do you mean when they start to compress?

[octal]

I think what Bob is trying to say here is "when the output tubes clip, the NFB tries to correct for this and in the process pushes the amp further into clipping." It's kind of the nature of the beast with an amp which uses global NFB. I suppose that if the PI were to also clip on positive swing, it might reduce the tendency to drive the amp into grid blocking/clamping/exaggerated bias shift. I suppose you could accomplish much the same thing by having the preamp start limiting/compressing before the outputs get pushed too, too far into squarewave city.

[Wakculloch]

Tried a little experiment and at the clipping point drive does not suddenly increase, nor does it push the amp farther into clipping.
The drive signal is still good, but no amount of error correction can stop the amp from clipping at the outputs when full power has been exceeded. As the PI is within the loop, feedback will also attempt to correct this.

[Chuck H]

Exactly.

I haven't done any "testing" but the way I figure it whatever amount of square wave is in the NFB loop will be sent to the power tube grids and partially cancel at the output. You can't cancel all the distortion because the NFB loop is limited in voltage. I would think that the PI compressing and moving toward a square wave would have a more significant effect on nulling the NFB loop because the fed back signal at the power tube grids no longer matches the output. So if a particular design drives the PI into hard clipping you would expect a significant reduction in the effectiveness of the NFB loop.

Chuck

[BackwardsBoB]

I like it when the amp starts clipping at the output because full power has been exceeded.

The problem I see is that the LTP phase inverter is a difference amp, with inputs from the preamp and NFB. We give up gain and the output is "linearized". But when an output tube clips (reaches saturation, and can no longer increase its plate current), the NFB signal, a representation of the output, also clips. The input from the preamp continues without clipping, and can far exceed the NFB's voltage. This would cause the gain of the PI to increase, when the power tubes are already saturated, setting up for blocking distortion problems (if you consider it a problem).

Basically, I'm saying that the negative feedback system has a linearizing effect that tries to keep the amp output voltage matching the preamp output voltage with some (possibly frequency dependant) gain. When the power tubes saturate, the negative feedback system should try to correct this "error" by driving the tube harder, which won't help, and could be bad.

Octal gets what I'm saying. Wakculloch's tests are interesting. I need to duplicate them.

[Wakculloch]

The gain of the PI doesn't increase cos the whole point of using feedback is to set the gain of the power amp as a whole, independantly from the individual elements within the loop.
There is no jump in volume when you hit full power as your argument would suggest, just a steady increase in distortion and more power dissipated as the output approaches a square wave.
Not sure i'm explaining this very well, so if someone with more theoretical knowledge is out there, come on in. Time for lunch.

[Chuck H]

I think I get what Bob is saying. If the power tubes are clipping (and therefore at maximum output) there is no more ability to increase voltage in the NFB loop. But there IS usually still some gain to be had in the PI. So the input to the PI will increase not only in volume but in ratio to the NFB voltage applied to the same PI circuit. This in turn pushes the power tubes to try and create more power, but there is no more NFB voltage available to counter this and linearize things. So the power tubes try to draw more grid current and cause blocking distortion.

I don't see how it could create a jump in output though. Maybe I just don't have my head fully around it yet. I expect if the power tube grid circuit COULD supply enough current that an amp would move into class B. But then the NFB loop voltage would also be able to increase, so...

Chuck

Edit: I just re-read octal's post. And if he gets it I guess I still dont.

[Steve Conner]

Yes. This is a well known effect and it causes the clipped edges to be squarer with NFB than without, and the distortion harsher sounding, and blocking distortion to be worse. However with the levels of NFB used in guitar amps it's nothing major.

But in solid-state amps it can be atrocious, especially since some of them have low-frequency poles in their NFB loops that "wind up" and make the output stick to the rail for a while after the overload has gone, resulting in clipping that sounds truly nasty. Another nugget for the "transistor sound" dungpile.

[octal]

Tried a little experiment and at the clipping point drive does not suddenly increase, nor does it push the amp farther into clipping.
The drive signal is still good, but no amount of error correction can stop the amp from clipping at the outputs when full power has been exceeded. As the PI is within the loop, feedback will also attempt to correct this.

This is a well documented effect. Where were you scoping the drive signal, and do you have control grid stoppers at the output tubes? If you have grid stoppers and you're scoping at the 'tube end' of the grid stoppers, you won't see much effect as the grid clamping of the output tubes prevents the voltage from going any higher. You should see something at the PI tube plates. However, the drive signal should increase- it'll look like a sinewave with an extra small hump on top, representing the difference between a sinewave and a sinewave with the top clipped off. This extra spike charges the coupling caps more and drives the output tubes further into grid clamping/bias shift making for harsher clipping.

Here's a vintage article by someone FAR wiser than I am explaining it- hosted by the Aiken amps site. (The article is written from a hi-fi perspective, but a tube amp is a tube amp, more or less.)

http://www.aikenamps.com/Crowhurst_blocking.pdf

[Steve Conner]

Thanks for posting the Crowhurst article, octal! It says exactly what I was trying to say only better.