mikeks said:
Mike, once again, that is correct for the inner loop, which does indeed have a loop gain of zero for the case you mentioned. But there are multiple loops here. I and others are talking about the outer loop. When you do the equivalency stuff in the figure you referenced, that makes the outer loop go away, making the loop gain calculation refer implicitly to the inner loop.
The inner loop has a loop gain of zero and the outer one has infinite loop gain. Both perspectives are equally valid.
Andy, the figures here are all exactly equivalent.
This means that any internal loops generating infinite gain, or, indeed, any conditions for generating positive feedback must be expressed in the final closed-loop expression.
One such condition is |1/K1|>1; this cannot not occur at balance.
This means that any internal loops generating infinite gain, or, indeed, any conditions for generating positive feedback must be expressed in the final closed-loop expression.
One such condition is |1/K1|>1; this cannot not occur at balance.
mikeks said:
Mike,
If you compare the first and last figures here, you'll see that the first figure is a multiple-loop system, while the last one is single-loop. So they are equivalent from an input-output perspective. Likewise, the loop gain of the last figure is the same as the inner loop gain of the first. But because of the block manipulations that have been done, groups of blocks have been coalesced. If any loops ended up being collapsed into a single block or similar, certain information about those loops will no longer be available. One such piece of information is the loop gain of any minor loops that may have been collapsed into a single block.
I don't know how, or even if, I can convince you of this, but here's another attempt. Remember this post? Try taking that sim, and moving the loop gain probe so that instead of being inside the innermost loop, it is, say, right in front of one of the diff amp inputs. Then compute the loop gain using the expression provided. Maybe that will convince you?
andy_c said:
Thanks Andy for one of the most sensible ways to put things in perspective.
The inner loop, resulting of an equivalent form of looking at the error extraction procedure, equates in the limiting case to an infinite gain block. This I noticed 2 years back and devoted a section in my draft to show as an equivalent viewpoint.
In regular negative feedback configurations, we do not cursorily introduce infinite gain blocks as tools, neither insert blocks of "as much gain as practical". At most we could insert positive feedback with this end in mind, but unfortunately the constraints for doing this with net positive results have not been properly put into perspective as far as I know.
Ed. Cherry worked this in course notes for Monahan Univ. and took what I humbly consider an unfortunate path at some crucial point in parameter selection.
Lipchitz - Vanderkooy brushed the issue but went along analyzing particular implementation issues.
Rodolfo
andy_c said:
Hi Andy,
You are absolutely right here in respect of the minor loop. Apologies!!!!!!
Nevertheless, it is apparent, i think you'll agree, that the instability of an embedded positive feedback loop, in this specific case, appears not to have any effect on the system's major-loop-gain, which remains equal to extracted error at balance.
Thanks for that demonstration.
andy_c said:
Oh dear
Precisely!
The ''outer-loop'', in this case, is the major-loop, for which there is no positive feedback at balance.
mikeks said:
No need for apologies Mike - it's just a communication thing
. Maybe we can agree on inner/outer terminology so it will match Rodolfo's paper when it comes out. I'll just change my naming if necessary.I don't know. I wouldn't want to make a snap judgement without looking at it carefully.
One peculiarity with this minor-loop, Andy, is that it generates transmission less than unity at balance although its phase shift is equal to or exceeds 180 degrees.
This probably explains why the loop is stable in fact, and has no effect on the major-loop.
This probably explains why the loop is stable in fact, and has no effect on the major-loop.
Attachments
andy_c said:
I thought natural to refer to the inner loop since it is the shorter one and gets included as a functional block (or infinite gain) when colapsed as Andy suggested. Actually in the real world it is equivalent to an ideal integrator with the same gbw product as the starting stage.
A possible inbalance condition is clipping of the outer loop, the one where the output stage is, breaking balance. To prevent possible DC latchup I use in my apmplifiers a large value capacitor in the inner loop.
Jan (janneman) confessed to have blown at least one woofer because of this.
Rodolfo
ingrast said:
Ouch!
I guess what you and Jan are doing could be referred to, for lack of a better word, as "global" EC, where the EC is around the entire amplifier. Am I correct in assuming that with Hawksford EC (of output stage only) and, say, Baker clamps in the VAS to establish clipping ahead of the EC, that this latchup cannot occur?
andy_c said:
Ouch!
I guess what you and Jan are doing could be referred to, for lack of a better word, as "global" EC, where the EC is around the entire amplifier. Am I correct in assuming that with Hawksford EC (of output stage only) and, say, Baker clamps in the VAS to establish clipping ahead of the EC, that this latchup cannot occur?
I have worked only global EC so I do not qualify for considerations regarding output stage only EC.
Yet I am afraid despite Baker clamps (I do not know the details but imagine is a means to clip earlier), variations in load impedance could in principle provoke clipping in the output stage.
I have not experienced unbalanced conditions even before introducing DC blocking in the inner loop, but I am much more comfortable doing so, and have experienced no problems so far now with several hundred hours of actual listenting. Not even transient induced disturbances.
Rodolfo