Please sanity-check my current feedback mode for Alesis RA-100.

[Kaz Kylheku]

Please don't multiply feedback values by 100, it does not work that way.

This is a good discussion topic. Can you explain why, for sampling the voltage to be fed back, a more resistive voltage divider can't used which has the same ratio between the resistors? Broadly speaking about amplifiers in general, what are the issues to watch out for with doing that. I'd still like to try it. If it doesn't smoke, it's good!  :duh I should be able to tell if the amp becomes obviously more gainy. If it oscillates, I will hopefully be able to kill it before it kills the speakers.

(One obvious issue might be that the feedback input may have a lower impedance than expected, which could makes the more resistive voltage divider less tall, so to speak. I.e. the same issue as with the potentiometer being too resistive is just relocated somewhere else, but now causing a real problem: less negative feedback, destabilizing the amp.)

Although the power amp is *some* kind of "big Op Amp", it's not a TL071 or similar Fet input one by a long way.

Yes, no FET input there. You cannot count on megohms of input impedance.

[phatt]

Quite obviously some of this is over my head ,,,but my guess is also quite simple.
This amp is going to go poof fairly soon if he keeps fiddling with it.

Hint,,, don't mess with DC coupled power Amps unless you know *EXACTLY* what you are doing.
I took out at least a dozen Power TR's before I got wise. 0:)

From memory a $5 power transistors will blow about 100 times faster than  a 5 cent fuse.
Cheers Phil.

[Kaz Kylheku]

Experiment settles it: too much overall resistance in the feedback path is very harmful.

Firstly, the diminished current in it causes the feedback path to pick up hum (bad S/N ratio).

Secondly, frequency response is impacted quite severely: highs are rolled off.

Third, although there doesn't appear to be a noticeable rise in gain, the amp's biasing appears to be destabilized. It normally runs cold, but with this change, its output transistors warm the heat sink. I was watching for that and shut it off immediately.

[J M Fahey]

Well, it sort of answered itself.
The Lin structure power amp (99.9% of all modern amps) is built out of 3 basic blocks:
* an input differential stage, which compares (substracts) the input signal with a sample from the actual output signal which is driving the speaker (negative feedback), generating an error correcting signal (very intelligent chap, this Lin)
* a very high gain stage, basically a single transistor, which provides most of the amp open loop gain, plus the full rail to rail voltage swing
* since you already have all the peak to peak or rail to rail your power supply can give you, you only need to amplify the *current*, from a few mA into a couple to tens of Amperes, as needed to drive the speakers fully.
* the input differential stage has 2 inputs ; one of them gets the input signal, the other gets the feedback signal, which is the output signal attenuated by a couple resistors, plus some capacitors if needed.
* the input impedance of said feedback input is usually not very high, say 10 or 20KOhms.
A classic feedback net has usually the input to ground resistor between 5K and 100 ohms, so it's affected very little by the NFB input impedance, but if you multiply everything by 100x .... you now know what happens.
Also any stray capacitance, say in the order of 100pF which could previously be ignored, now becomes a highs killing monster or the contrary, an unstable nightmare.
In a FET input TL071 (as an example), the input resistance on either input is around 1000000000000 ohms. No, it's not a typo, one million megohms, there's 12 zeros there.
http://www.datasheetcatalog.org/datasheet/stmicroelectronics/2296.pdf
Even so, if you use a feedback loop of, say, 20Meg/1Meg , I'm *sure* you'll lose highs because even the tiniest real world stray capacitance you'll get there, will kill all your sparkle and then some.
But now you know that, of course.