If you assume that audio signals are bandlimited to 20 kHz, it is in line with Daugherty and Greiner's sufficient but quite unnecessary condition to prevent slewing.
As you probably know, the plots I have shown are for illustration and they apply to one well known audio opamp. It would be close to impossible to get 55MHz GBW for a power amplifier with Miller compensation, to keep it stable. However, the chart is valid, only for power amp the real GBW would be at least 10x lower, so the plots would be shifted.
That is allways one good question ! ; what is the one audio power amplifier stability adjusted to unity gain in the real life test condition ? ,
IMHO answer is ; have to be unconditionally stable .
IMHO answer is ; have to be unconditionally stable .
It can be done, but it must be a design goal from the beginning. Otherwise, if you design for e.g. Au = 27dB, it does not make much sense to insist on unity gain stability.
I know, for that reason, since some time only a DBT result of properly made test is a proof, to me. I do not care about anecdotal stories anymore, no matter who tells them, how Big Guy is He.
Something else. I am looking for subjects for following columns. I have one on the issue of narrow or wide BW but then from the time domain, as well as one on Damping Factor.
Any suggestions?
Jan
Any suggestions?
Jan
If I may:
Increasing gain-bandwidth product by introducing additional gain stages into the fb-loop ("composite" amps) and how to get them stable (I know it's incredibly complex).
Edit: " how to get them stable" is unrealistic for one article, maybe I should say: strategies to get them stable.
Increasing gain-bandwidth product by introducing additional gain stages into the fb-loop ("composite" amps) and how to get them stable (I know it's incredibly complex).
Edit: " how to get them stable" is unrealistic for one article, maybe I should say: strategies to get them stable.
Both are good. I would suggest DF with respect to complex load of various kind and with respect to complex output impedance of the amp. You know those cheap class D amps with LC filter outside feedback.
Not only just cheap Class-D amplifiers unfortunately 🙁
It's one of the most important factors in a Class-D amp, also in sense of performance.
https://www.ti.com/lit/pdf/sloa242
See page 7 and 8
Jan , my reasoning is that if by one(constant output voltage type) power amp loop gain (or feedback factor) is linear and equal over entire audio spectrum than must be the same with Damping Factor .
A perspective from the time domain can be instructive as relates to distortion generated from non-linear transfer functions. This begins from an understanding that the derivative of dv/dt at any point on a curve results in the determination of the slope when time is reduced to zero. This leads to a visualization that any non-linear transfer curve is made up of an infinite number of straight line gain segments between any two adjacent points on that curve... leading to the further conclusion that signal reduction imposed on any non-linear transfer function results in linear gain when peak to peak amplitudes approach absolute zero time. Seems the foundation of why feedback works by reducing input differences.
Not entirely so, the output impedance may have and usually has pronounced inductive component (even without output coil) and you do not see it in the loopgain plot. Method of injected current into amp output reveals this. DF then decreases at high frequencies.
Sorry... sometimes I can't understand why I can't explain things more clearly. Mom and dad both assured me they were confident I had an IQ well into the double digits...
Transfer functions are drawn such that a signal voltage "A" on an "X' axis can be reflected by the transfer function as an output signal "B" on the "Y" axis. Oftentimes a pure sinusoidal signal on one axis is shown distorted by the transfer function on the other axis. This depiction indicates the mechanism by which a non-linear transfer function generates distortions over a time interval into the frequency domain.
Yet it is that the derivative of any point on the transfer curve that indicates the transfer gain in the movement of a signal about a point. If signal amplitudes are infinitely restricted of amplitude about any point on the transfer curve the output is linear as a function of fixed gain, having a transfer gain according to the derivative at that point.
The mechanism of feedback is a subtractive one. By way of example consider a differential amplifier configured for unity gain and 120dB open loop gain. For a 1 volt signal in and out requires that the difference signal between inputs is 1/1,000,000 or 1uV such that when magnified by the 1,000,000 gain the result is 1V. The conclusion being that the core principal to reducing distortion is to fix an operating point and reduce signal swing by amplifying the out of it and adding feedback to 'reducing input differences'
I see what you mean. I would call it different - there's no reduction of amp block input signal difference, that is, as you say, just Vout/Aol, and nothing the feedback does will change that.
Maybe a way of describing it is that it decreases the overall sensitivity by putting a (sample of) Vout in opposite phase to system Vin so that Vin must be greatly increased for the same Vout.
There's always several ways to look at some physical system. I would prefer to apply Occam's Razor to end up with the simplest view that is correct.
Makes my head hurt less.
Life is simple - How Occam's Razor set Science Free and Unlocked the Universe.
Jan
Maybe a way of describing it is that it decreases the overall sensitivity by putting a (sample of) Vout in opposite phase to system Vin so that Vin must be greatly increased for the same Vout.
There's always several ways to look at some physical system. I would prefer to apply Occam's Razor to end up with the simplest view that is correct.
Makes my head hurt less.
Life is simple - How Occam's Razor set Science Free and Unlocked the Universe.
Jan
Zout generally increases with frequency because it depends on feedback and that decreases with frequency.
So you end up with a 'component' (Zout) that increases in value with freq and that is the definition of an inductor.
Jan