mikeks said:
Hi Mike, I don't think so.
At this moment, just being back from holidays (been to Crete, yes, nice weather), I've to catch up with mail and lots of forum contributions, so I haven't spend much time on finding the right articles. But I have seen a patent describing an amplifier that contained an inner loop with positive feedback (!). The overall negative feedback made the amplifier stable. A quick look revealed it to be patent 4,467,288 from James Strickland, describing the TransNova amplifier, a concept later on adopted by Hafler.
I think Janneman once told me that he has designed an amplifier with a inner positive feedback loop, but I'm not sure about that.
Steven
Attachments
Steven said:
I am familiar with this patent....and related Hafler designs.....
....but i think you'll find this is merely a commmon-source output stage...(rather than the usual source follower configuration)....
....with local shunt-shunt negative feedback about the output stage...
This is easily seen by merely performing a delta-star transform on Ra, Rb, and Rc....
Contrary to the patent, there is no positive feedback designed into this circuit as far i can see...
andy_c said:
I am not sure about this Andy...only a double integrator produces an unbounded output in finite time.....
On the other hand, the nominally single integration provided by Miller compensation for instance, must exclusively possess poles with negative real parts.....
This is a necessary and sufficient condition for bounded input/bounded output stability...
Steven said:
AAHH yes....i missed that....
Thanks Steven.....
Performing a delta-star transformation with respect to Ra, Rb, and Rc, (see figure below), gives new resistor values Rx, Ry, and Rz...
This shows that a finite amount of +ve feedback, (controlled by the ratio of Rz to Ry), is applied to gain block A1.
Setting Rz=0 for instance eliminates the +ve feedback loop.....
Note however, that since both minor and major loop feedback are derived from the same node, BIBO stability (or lack thereof) of the major loop may only be ascertained by choosing a test point that simultaneously accomodates, (or 'breaks') both minor and major feedback loops.
see section 4 here:
http://www.ece.ucdavis.edu/~hurst/papers/FullyDiffRR,CAS.pdf
Clearly then the system's overall loop transmission must be BIBO stable, notwithstanding the +ve feedback minor loop.
In other words, any propensity to local instabilty in the system must be compensated for elsewhere if the major loop is to be stable.
In this case, the local +ve feedback loop is clearly controlled by the major loop....
What happens then when when the later is momentarily disabled by anomolous operation such as clipping...?
At best you'll have excessively delayed recovery from clipping overload, at worst sufficiently prolonged 'sticking' to the supply rails to activate your DC offset protection....if such a system is employed.....
It is for these reasons that my doubts persist regarding such systems......at least in the context of nominally linear time invariant designs.....
Attachments
Some interesting observations on minor-loop instability...
http://peufeu.free.fr/audio/memory/memory-8-complete.html
In the middle of the page the vexed gentleman observes:
http://peufeu.free.fr/audio/memory/memory-8-complete.html
In the middle of the page the vexed gentleman observes:
andy_c said:
or x(t+1)=x(t)+C where C is any real number > 0.
You can certainly follow that up with either a substraction or a multiplier between (-1.0, 1.0) and make the whole system stable (ie, as t->infinite, x(t) is bounded).
It is a fairly simple example showing that you can have stability in a system where part of it isn't stable.
Here is an early example of positive feedback in an inner loop that is stablized by an outer NFB loop (patent # 3111630 by Wolcott, 1963):
http://patimg1.uspto.gov/.piw?Docid...ageNum=&Rtype=&SectionNum=&idkey=E44F6F40019E
Also, Hawksford output error correction, if misadjusted slightly too high on correction, can have an inner unstable loop that is held stable by the outer global NFB loop.
Don
http://patimg1.uspto.gov/.piw?Docid...ageNum=&Rtype=&SectionNum=&idkey=E44F6F40019E
Also, Hawksford output error correction, if misadjusted slightly too high on correction, can have an inner unstable loop that is held stable by the outer global NFB loop.
Don
andy_c said:
Hi Andy...missed this.....
......by this do you mean an integrator possesses an unbounded output at DC?
...see below for definition of BIBO stability...:
http://www.stanford.edu/class/ee214/handouts/FB.pdf
Yes.......i was thinking in terms of a non-ideal implementation such as exists in an audio amp.....
....for which integration can only be said to occur within a finite band of frequencies....
Then again.....even if ideality were assumed, (with the pole at the origin implying an unbounded output for a bounded DC input), we find that a shunt-shunt integrator of the form used in generic amps. possesses no feedback at DC....making BIBO instability academic.
....for which integration can only be said to occur within a finite band of frequencies....
Then again.....even if ideality were assumed, (with the pole at the origin implying an unbounded output for a bounded DC input), we find that a shunt-shunt integrator of the form used in generic amps. possesses no feedback at DC....making BIBO instability academic.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.