Actually, this is a fundamental error. Negative feedback never attempts to reduce the error voltage to zero. It merely allows that error voltage needed to realise the demanded output voltage. Note that I use "error voltage" and not "input voltage". The distinction is profound.
Of course the gain of a two stage Miller compensated amplifier drops at a single pole rate above a few Hertz, but its gain remains significant well beyond the audio band, even as its output impedance is reduced.
Nevertheless, I would be interested to examine references where the two stage Miller compensated amplifier is called an "OTA", if you can provide them. Thanks.
Last edited:
And from your posts it is abundantly obvious you're a "newbie" at everything else. Really sad for a man of your advanced "Jurassic" years.
I wholeheartedly disagree.
That sound to me as not accurate description of how an amp
react when NFB is applied.
NFB will tend to reduce the error voltage to 0 as much as allowed
by available loop gain.
That's not what pfigueiredo said. What he said was:
Which is correct. In the ideal case, open-loop gain is inifinite which results in zero error signal when negative feedback is applied. In the real-world case, open-loop gain is "large" and ideally made "larger" (without compromising stability) in order to minimise the error signal.
RHP zeros (...) Usually not a problem because it is above any frequency of interest.
It is possible for a particular combination of reasonable component choices to result in the RHP zero frequency to be sufficiently low to affect the stability a little. This can occur if the LTP emitter resistors are fairly low (to reduce noise probably) compared to the VAS/TIS emitter resistor (used for over-current protection, often).
If your circuit is unconventional then normal rules-of-thumb won't help. But I doubt a RHP at >3GHz will hurt.
That was my conclusion as well, originally stated as "audio amplifiers are minimum phase systems" (where all this discussion started from), then illustrated here.
Dave, you originally stated otherwise, I am still waiting for a shred of proof. Meantime, my statements still hold.
kgrlee, I find your habit of painting the contributors here in various colours, without effectively contributing yourself to the topic (other than references to "jurassic times", and stating your ignorance on various subjects) rather offensive.
I disagree. Read again my original statement, if an amplifier is stable with ZXT796 in the input stage, then bursts into oscillation when substituted with BC546, that's a poor design. Please note, I am not talking about substituting the VAS device(s) or substituting the output stage 3MHz output power devices with 30MHz device.
If the pole formed by Rfb and Cin of the input transistor is a depending factor in the amp's stability, swapping in a device with a lower Cin may actually destabilize the loop as a whole.
As much as I hate to admit
I'm with Mike here.If feedback would succeed to reduce the error voltage to zero, there would be no amp output.
The point is that the difference between Vin and Vfeedback comes out precisely to get the Vout from the amp open loop gain.
jan
As much as I hate to admit
If feedback would succeed to reduce the error voltage to zero, there would be no amp output.
Not if the amplifier has infinite differential gain. Zero times infinity can assume any value. It is clear that as the gain of a differential amplifier in negative feedback configuration tends to infinity, the error signal tends to zero. In the ideal case, gain is infinite and error is zero. Of course this is all an adventure in mathematics; in the real world we can't have infinite gain so just have to put up with minimisation of the error signal, not making it actually zero.
As much as I hate to admit
If feedback would succeed to reduce the error voltage to zero, there would be no amp output.
The point is that the difference between Vin and Vfeedback comes out precisely to get the Vout from the amp open loop gain.
jan
The difference between Vin and Vfeedback is driven by Vin, hence Vin dictates the error voltage. The feedback will indeed try to remove the error voltage between Vin and Vfb when Vin changes.
So I don't understand your first hypthesis that there would be no amp output when the error is reduced to 0 for it is feeding back the output for comparison that enables the feedback system to remove the error voltage.
I agree that if replacement of input stage devices with roughly equivalent devices results in instability, that this is a curious case. However, if it can be shown that the design is stable across all expected variations of the specified device's parameters, I fail to see how it is a problem that the design becomes unstable when the specified device is replaced with a not-specified device.
Zero error is zero effective input signal gives zero output. We colloqially (sp?) say that we want to drive the error to zero, but that implicitly assumes infinite gain. Once you figure in the gain, it should be: drive the error to Vout/OLG.
Edit: we may have a semantic issue here: I use 'error' as the effective input signal of the amp, i.e Vin-Vfeedback.
jan
Actually, this is untrue. It is not desirable to have an amplifier with infinite forward path gain, even in theory.
This is because you would have zero error voltage for an infinite output: the amplifier simply wouldn't work. Thus, infinite forward path gain is neither realistic not desirable, even in theory. This is a very basic misunderstanding of how negative feedback works.
The negative feedback simply allows you to apply that error voltage that would be required to generate the required output voltage. Application of the negative feedback network doesn't change the forward path characteristics of the amplifier in any way.
Mathematically an infinite gain wouldn't work no. In reality we have slewrate and propagation delay implying gain is not infinite for a start. For practicallity we assume an aproach to infinite and zero within realistic bounds.
The differences in view from my end are semantically. After all, "drive the error to Vout/OLG" is what results in the attempt to equalize Vin and Vfb. It's just a different look at the same thing I think.
It's often a common description for an opamp's NFB circuit's action: "The opamp will try to adjust its output so the difference between the + and - inputs become zero."
The differences in view from my end are semantically. After all, "drive the error to Vout/OLG" is what results in the attempt to equalize Vin and Vfb. It's just a different look at the same thing I think.
It's often a common description for an opamp's NFB circuit's action: "The opamp will try to adjust its output so the difference between the + and - inputs become zero."
Last edited:
No. You are wrong.
Take a differential amplifier with gain G(s).
Put it in a 100% negative feedback loop.
Closed-loop gain is given by G(s)/[1+G(s)] and the error signal is given by the input times 1/[1+G(s)].
Let G(s) tend to infinity. In the limit, i.e. G(s) = infinity, closed-loop gain is unity and the error signal is zero. Fact. In this case, what zero (the error signal) times infinity (the differential gain) is equal to depends upon the input signal, i.e. zero times infinity = the input signal.
Semantic debates seem to be the specialty in this thread.
The definition of "limit" will give you the answer. For every real ε > 0, there exists a real δ > 0 such that for all real x, 0 < | x − p | < δ implies | f(x) − L | < ε. That is, when the loop gain increases, the error signal can be made arbitrary small, smaller than any chosen ε > 0. No need to talk about zero error signal.
I agree that since infinite gain is only possible in theory that it is a bit silly to argue about it. But since we are dealing with behaviours that we model as functions of complex numbers, it seems reasonable to use the theories of complex analysis which allow for zero times infinity to assume any value in the complex domain. See Infinity - Wikipedia, the free encyclopedia