How much negative feedback is enough?

[ted01]

I've got a pot in place of a fixed value resistor in my negative feedback circuit. What am I doing to the amp when I adjust the resistance from 0 ohms to 10 Kohms? Am I doing something that will cause harm? The original circuit called for a 4.7Kohm resistor. I notice an increase in volume when adusting the pot. Sure wish I understood this stuff better.

Ted

[ic-racer]

I've got a pot in place of a fixed value resistor in my negative feedback circuit. What am I doing to the amp when I adjust the resistance from 0 ohms to 10 Kohms? Am I doing something that will cause harm? The original circuit called for a 4.7Kohm resistor. I notice an increase in volume when adusting the pot. Sure wish I understood this stuff better.

Ted

It would be loudest and most 'out of control' with the highest resistance. Many amps would probably be fine with infinity resistance (open NFB loop) so probably no harm done with you pot at 10k. Going the other way with less resistance causes more NFB and quiets the amp, again causing no harm.

Now that you see what a variable NFB control does, you probably also see that it is not that useful of a 'front panel' control. It winds up just being another volume control. So, now you can see the advantage of a 'presence' control. The presence control alters the frequency response of the NFB signal, thus it is less of a volume control and more of a timber control.

[ted01]

I mounted the pot on the back of the chassis as I kinda guessed that it wouldn't be a front panel type of control. I also noticed what you mentioned...the presence control actually does something now. Originally, I had the output tranny wires reversed, leading to horrific motorboating. I didn't know what I had done wrong, but found that if I removed the feedback circuit entirely it went away. This led me to install a large value pot which I adjusted until the motorboating went away. Course what I really did was effectively remove the feedback cicuit! Well, I finally had some time, and got the output tranny wired correctly and then addressed the feedback circuit again. Originally, I was going to just reinstall the correct value fixed resistor, but got to wondering what would happen if I made the value variable via a pot instead. Just wanted to make sure I wasn't going to burn something up.

Follow up question: How do you guys that design circuits decide what value you use for the feedback resistor? I've looked at a mess of schematics and the value does vary from one amp to another. Is this one of those magic aspects of amp design? Is there some design goal that helps you determine the resistor value?

Ted

[Ears]

Follow up question: How do you guys that design circuits decide what value you use for the feedback resistor? I've looked at a mess of schematics and the value does vary from one amp to another. Is this one of those magic aspects of amp design? Is there some design goal that helps you determine the resistor value?

Ted

To design feedback theoretically "on paper" involves advanced skills and mathematical tools most often aquired at university or college level. These include analysis of circuit behaviour in the complex (frequency) domain (poles and zeros of the system function). You can teach yourself this by first boning up on basic differential calculus, Euler's theorum and then reading REAL electrical engineering texts. It won't be easy.
CAD can be used if the cct can be modelled adequately.
But even trained engineers can opt for the more practical approach of trial and error on the work-bench. Knowing some of the maths isn't helping me get to the cause of the feedback/presence oscillation problems I'm currently having with a commercially built amp on my bench. I think I'll soon be putting my hand up in here for some help too. ''

[Ears]

double post ?

[ted01]

Yo Ears,

I can't count past ten without pulling off my shoes and using my toes! I guess I'll have to opt for the trial and error method...

Ted

[Ears]

Wikipedia has an overview of some of the considerations in FB design http://en.wikipedia.org/wiki/Feedback_amplifier

I suspect the criteria for instrument amps is reasonably simple, its not hi fi, so:

Firstly: closed loop stability, no good without that. So, as you've already started to do the research, work around established values of similar circuits.
Secondly : taste, how punchy/racy do do want the amp to sound.

One word of caution, amps can become unstable with too much negative FB, sometimes at frequencies not discernable to the ear (actually it becomes positive FB at the frequencies concerned). Worst case can result in fried trannies or speaker coils.

I haven't enough practical experience to offer you a cook book method. Would like to hear from others though.

[CaseyJones]

Follow up question: How do you guys that design circuits decide what value you use for the feedback resistor? I've looked at a mess of schematics and the value does vary from one amp to another. Is this one of those magic aspects of amp design? Is there some design goal that helps you determine the resistor value?

Ted

To design feedback theoretically "on paper" involves advanced skills and mathematical tools most often aquired at university or college level. These include analysis of circuit behaviour in the complex (frequency) domain (poles and zeros of the system function). You can teach yourself this by first boning up on basic differential calculus, Euler's theorum and then reading REAL electrical engineering texts. It won't be easy.

Yeah, great. Wanna know how much NFB to use? Ideally ZERO. Givin' a 'Wreck gobs of negative feedback to calm it down makes it less of a 'Wreck. Disconnect the NFB while you're dialing the amp in then give 'er some to hold it steady at full volume.

If you have a 'scope... if it's not makin' supersonic hash you'ved done your job.

If you have an RTA or really well calibrated ears you can do creative things with global NFB, and guess what... there's local NFB available as well.

Final answer... less is better.

[Ears]

Wanna know how much NFB to use? Ideally ZERO. <snip>
Final answer... less is better.

See? It comes down to taste and personal preference and some have stronger views than others!

[CaseyJones]

Wanna know how much NFB to use? Ideally ZERO. <snip>
Final answer... less is better.

See? It comes down to taste and personal preference and some have stronger views than others!

Personal preference? Maybe.

I was told a very long time ago (by a sub-mariner very well versed in every system on "his" attack boat) that the difference between an amplifier and an oscillator is feedback. An amplifier with positive feedback isn't an amplifier, it's an oscillator. An amplifier with negative feedback is still less of an amplifier. NFB can and is used to stabilize an unstable amp, it also smooths frequency response. If you're some golden ears hi-fi nut playin' yer old Mac through some old A-7s maybe that's your bag, me... I don't care about pancake flat frequency response. On the other hand I discussed the use of excessive NFB to add stability to an unstable amp with one of my golden eared hi-fi nut technician associates... he said "Too much NFB sounds like ca-ca". Exact quote, verbatim, that's what he said. So... different angles on the same topic, same answer. Might be sumthin' to it.

[ted01]

Hey CaseyJones,

As usual, I have learned a ton from you. I have also learned that a 1/2 watt 10K pot can't take the power, apparently. It worked for a while, but when I cranked it up, it appears to have fried. Well, for the moment, I replaced it with a 50K pot, it'll probably fry too, but maybe it'll last long enough for me to get an "optimal" value and then I can replace it with a fixed value resistor.

In the mean time, you said "If you have a 'scope... if it's not makin' supersonic hash you'ved done your job. ". Would you be kind enough to elaborate on this? I've got a decent scope, but I could use some guidance in what and where to measure as well as under what conditions. I.E. should I shoot a 1KHz signal from a signal generator to the input of the amp, or actually have someone play a guitar while I am tweaking? Do I put the scope probe across the speaker output? What should I look for? Hope I don't look too ignorant, I've had a ton of experience in things like ham radio and commercial radio transmitters and such, guitar amps are something new and consuming for me.

Thanks to all, as usual, you are the best!
Ted

[jjman]

I would put a resistor in line with the pot so you do not have the ability to reach the "zero ohms" scenario. Maybe a 1k resistor and a 100k pot. I don’t think a high wattage pot would be needed

[LOUDthud]

The usual method I use when tweeking feedback is to feed a power amp with square waves. If there is ringing or overshoot on the front corner, that indicates a possible instability or tendency to oscillate. You need to look at low level as well as high level signals and vary the frequency. Expect some tilt at low frequencies because of the low frequency cut off of the OT. Many times a less than ideal compromise is the best you can do.

The catch 22 of all of this is that the best sounding amps are right on the edge of instability. Amps with good square wave behaviour can sound dull and lifeless. That usually means there is too much feedback. At some point you have to connect a speaker and play. You are not done tweeking until it sounds good. I leave my scope connected and watch for anything unexpected.

One area where feedback helps is when the bass is too loose. A little feedback will tighten it up.

[CaseyJones]

In the mean time, you said "If you have a 'scope... if it's not makin' supersonic hash you'ved done your job. ". Would you be kind enough to elaborate on this? I've got a decent scope, but I could use some guidance in what and where to measure as well as under what conditions. I.E. should I shoot a 1KHz signal from a signal generator to the input of the amp, or actually have someone play a guitar while I am tweaking? Do I put the scope probe across the speaker output? What should I look for? Hope I don't look too ignorant, I've had a ton of experience in things like ham radio and commercial radio transmitters and such, guitar amps are something new and consuming for me.

I just run a simple test to begin with... if there's no signal at the input there should be no signal at the output. It sounds like a no brainer but the scope will tell you things you can only guess. If the amp is unstable you'll see high frequency oscillation you can't hear. You can't hear it because it's above the frequencies your speakers are capable of reproducing. It doesn't take much to be above the range of human hearing. Good thing guitar speakers drop off like a rock above around 5khz, otherwise yer dog would bite yer ass and you'd never know why!

Audio amps are close enough to oscillator circuits that they're capable of producing all sorts of sounds all by themselves, high frequency oscillation, low frequency oscillation ("motorboating"), then if you crank up the sensitivity on yer 'scope you can poke around and find ripple and hum.

A common complaint with homebrews is "the output is distorted and low". Frequently enough that's because the amp is oscillating, if it's bad enough the tubes will redplate. Another symptom is "mysteriously" low voltage with no apparent "alternative" voltage path to ground. If the amp is acting as a big ol' RF oscillator you bet it's gonna drag the voltage down. A few pokes with your probes will confirm you have a problem, another few pokes in the right place with a chopstick will (hopefully) eliminate yer problem. Anyway... a quick fix for an unstable amp is to give it a big shot o' NFB but it's like dumpin' too much salt on yer dinner before tastin' it first.

I frequently dismantle old tube CB radios because the ones I look for contain useful salvage. It would be nice if everything had nice long leads but don't hold yer breath, everything is mounted with just enough lead legnth to go where it needs to go and no more. "Keep it short and tight". That takes care of as much crosstalk and stray capacitance as possible in RF circuits and guess what? It's a good idea in audio amps. I keep a couple pics on hand of utterly ridiculous layouts for a couple laughs at ham fests, there's nothing like too much wire "neatly" tied back to everything else to make for noise and crosstalk. Old radio guys can take one look at it and tell it's wrong, no scope required. There are a couple concepts in layout and lead dress that apparently originated in the twisted psyche of some Website Wonder, must be some kind of "instant expert" 'cuz the old timers will tell you it's just wrong. Oh, well... if enough people say it's right it must be right!

[Andy Le Blanc]

even the old slide rule method for graphicly plotting feedback is a real chore
you have to use a nyquist diagram to plot vector differences and phase angle
and determine the ratios of phase response and magnatude response to determine loop gain....... ouch
but the long and short of it is that there margines of stability that you can listen for and that it does come down to a matter of ratio
one thing to keep in mind as you play with a variable feed back control
is that feed back is about phase differences through the amp, and the
reactive elements (coupling caps) play a significant role so its a good chance
to swap differnt values of caps in the power amp
aiken amps has good piece on ratio