Another shot at a VAS alternative: Voltage Actuator Stage! Guesss what: we can comfortably keep calling it VAS 
In seriousness, that description isn't bad at all - it describes the primary function of the VAS: generate the desired voltage swing. The stage itself is an actuator controlled by the IPS. This meaning takes no assumption about points of view (voltage/current/impedancec mode, either input or output) and therefore just can't be wrong now can it?
In seriousness, that description isn't bad at all - it describes the primary function of the VAS: generate the desired voltage swing. The stage itself is an actuator controlled by the IPS. This meaning takes no assumption about points of view (voltage/current/impedancec mode, either input or output) and therefore just can't be wrong now can it?
Quote Jan Didden,
"BTW Has anyone here started to read Doug Self's 6th edition? I just stumbled on a great section titled 'feedback intermodulation' (FID anyone?).
He provides, among other very worthwhile things, a solid technical reason why it is indeed a Good Thing to linearise your amp before applying nfb"
. Originally Posted by scott wurcer
Since it is well known for a long time that a non-linearity enclosed in a feedback loop creates new harmonics even from the simplest mathematical treatment, I don't see how this issue could be approached in a way that has nothing to do with previous results.
Didden,
Ah! One of the things Douglas found is that the shape of the initial increase of harmonic levels with increasing feedback, and the subsequent decrease when feedback continues to be increased, is very much dependent on the initial non-linearity of the forward path. (Howzzat for a short, succinct sentence?).
That may be somewhat intuitive of course but the upshot is that a small improvement in forward path linearity even to the detriment of the available feedback factor can greatly decrease this 'Feedback InterModulation'. At least that it how I read it.
But I'd like to hear how others look at this. It's somewhere around page 80 in his book.
Bob,
Having read your section on Negative Feedback what do you think of all this? Intuitively you would think that a more linear amplifier would require less feedback to linearize but it seems to go against the technical arguments about the loop speed for a feedback network and bandwidth if I am getting any of this correct.
This sounds like a partial re-hash of what I referred to as Spectral Growth Distortion (SGD) in my book's Chapter 24, starting at page 502. The matter is basically a derivative of Baxandalls much-abused findings. This has been discussed ad nausium here at DIYaudio. However, if Doug has linked the IM consequences of SGD to a tradeoff of open-loop linearity with amount of feedback in determining overall distortion, that may be a different way of looking at the same problem.
In any case, I and most others I think, have always strongly advocated getting the open loop amplifier as linear as possible before throwing a bunch of NFB around it. Moreover, making the open loop more linear does not usually force you to trade away amount of negative feedback you use. Adding degeneration to the input stage or the VAS, for example, does not take away gain at your disposal for negative feedback loop gain.
There is no need to have less loop speed and bandwidth as a result of making the open loop more linear.
Making the open loop more linear will, however, often mean using more transistors, such as an input cascode, a good current mirror, a 2 transistor VAS, a push-pull VAS, a cascoded VAS, or an output Triple. While these may go against the grain of the simpler-is-better advocates, they do not usually go against amount of loop gain or safe ULGF.
Cheers,
Bob
Degenerating the input stage DOES reduce forward path gain, if the minor loop compensation remains unchanged, and, therefore, the amount of loop gain you can apply.
On the other hand, degenerating the TIS has no first order effect on the forward path gain since the second stage takes a current input and not a voltage. Therefore, degenarating the TIS does not reduce the loop gain of the amplifier.
Last edited:
Not in my book, nor Self's
Remember, you will not normally find a TIS in a no-feedback amplifier.
Cheers,
Bob
Actually, even D. Self acknowledges that it is, in fact, a TIS in his book, although he mistakenly called it a TAS in his fifth edition.
And, no, so-called no-feedback amplifiers are not the subject of this debate. You still haven't read Solomon I take it.
This is a good example to show either statements hold a truth:
If you have an IPS that is too fast for the VAS you have ringing on square waves. This excessive speed is of no use and just upsets the next stage more. You can increase the size of the IPS degeneration resistors to make the stage slower. Just don't make it slower than the VAS, there will be a sweetspot where you retain your speed without ringing on a square. Here you have linearized the IPS more at no cost since you trade excessive gain at the IPS for linearity and "speed matching" with the next active device stage.
When you'd keep increasing the degens at this point, it then affects overall gain/speed.
Last edited:
Over here we call it "color"
Yes; you took a perfectly serviceable language and butchered it.
Nonsense. Cherry already addressed this stuff, and a long time ago; there is nothing to linearize with, beyond a) increasing the bias, so that active devices are in a more linear region (that is, dynamic variations of the bias points are smaller) , b) use distortion cancellation (that is, use circuit symmetry wherever possible, to cancel odd harmonics) and c) use negative feedback
Degeneration linearization effect falls exactly under c) (as a current-current feedback loop). If we can't agree on this elementary fact, it's indeed a waste of time to debate anything else. And please don't come up with the local vs. global NFB nonsense. Negative feedback is negative feedback, period.
I urge you to read the following paper carefully to obtain a good grounding in basic amplifier theory:
http://ece.wpi.edu/~mcneill/524/handouts/solomon.pdf
What he found is that as the amount of open loop nonlineariy is reduced the higher harmonics decrease with increasing rate as their order increases.
We implicitly talked about trading off open loop linearity, open loop gain and feedback ratio as sort of linear, but for the higher harmonics that's not the case.
A halving of the non-linearity will reduce the 'feedback intermodulation' of the higher harmonics much faster: the 3rd is reduced to 1/4, the 4th to about 1/10, the 5th to more than 1/20. So even for a mild improvement of linearity there's an enormous reduction in harmonics.
You really need to read it yourself.
jan
Last edited:
An extra read can't hurt
On the same note, I'd like you to read about a new VAS, notably post #4 and #45 and leave a post about what you think.
Last edited:
You got a good case where the TIS vs. VAS makes a difference. If you look at it as a TIS, then the transimpedance gain is exactly Beta*Rload. Moreover, if the output buffer input impedance is very large, then the gain is limited by the transistor output impedance VA/Ic (VA stands for the Early voltage). Therefore, the TIS gain is Beta*VA/Ic, independent of any emitter degeneration.
No!
Please see Bruno Putzey's excellent (as usual) The F Word
You get it Bob I appreciate your line of thinking.
Edit: I think I should buy your book just because
Everyone who is interested in audio power amplifier design should buy and read Bob's excellent book.
Thanks for the heads-up on the imminent release. Looking forward to getting my hands on a copy. If I may be permitted a moment's immodesty, I particularly like this from page 343 (section on two-pole compensation, viewed on Amazon via "look inside"):
douglas self said:
Harry, Michael,
Thank you for the papers to read. Have downloaded both. Will digest these during the times I get fed up with the stuff I'm currently working on. The thirst for knowledge and understanding is immense at present.
Bob's book is good. It took me from looking at schematics which initially may as well have been hieroglyphics to something I can begin to understand.
Thank you for the papers to read. Have downloaded both. Will digest these during the times I get fed up with the stuff I'm currently working on. The thirst for knowledge and understanding is immense at present.
Bob's book is good. It took me from looking at schematics which initially may as well have been hieroglyphics to something I can begin to understand.
Last edited: