mikeks said:
Mike
I don't appear to have a JLH power amp design from WW 1983 in my index (http://www.tcaas.btinternet.co.uk/jlharticles.htm). If your date is correct, could you please let me have the relevant details of the article so that I can add it to the index.
Thanks
Geoff
Note however....
..that positive feedback within the global negative feedback loop is sometimes recommended by some authors in theoretical treatise....e.g:
http://grove.ece.iastate.edu/docs/Papers/20020507.pdf
...i remain unconvinced, as in the example given, the authors blithely refuse to consider the effect clipping or otherwise disabling the global feedback loop may have on their carefully worked out analysis......
..that positive feedback within the global negative feedback loop is sometimes recommended by some authors in theoretical treatise....e.g:
http://grove.ece.iastate.edu/docs/Papers/20020507.pdf
...i remain unconvinced, as in the example given, the authors blithely refuse to consider the effect clipping or otherwise disabling the global feedback loop may have on their carefully worked out analysis......
Like any technique in science, there will be critics of that theory or method. Myself I like Multi-Loop Nested feedback in Audio Amplifiers. I use this technique Exclusively with IC op amp and Buffer composite Applications. The technique I use was popularized by Walt Jung from his Tools and Tips articles in EDN the most relevant is http://www.elecdesign.com/Articles/ArticleID/7206/7206.html http://www.elecdesign.com/Files/29/7206/Figure_02.gif Schematic.
The next article describes the real problem as I see it and this is the limited open loop bandwidth of the Gain stage in this instance an IC OP-AMP. Most of these have the open loop gain start to decline at 10-100 Hz. Others will have lower open Loop gain and feature a wide bandwidth. These types of devices can have a flat gain vs. frequency to 10 KHz or more and I have seen a few flat up to 100KHZ. I agree that with these devices and a suitably fast output stage that Multi-loop loses a lot of its appeal. However if one was to consider that most of these fast devices are not practical in audio and as noise is concerned you still want the benefit of a High feedback factor then Multi-Loop becomes attractive. Myself I would much rather have a constant 01% THD from 20-20,000 Hz than a THD vs frequency that is say 0.0001% from 20-100 HZ rising to 0.01% at 20 KHz. I believe that the Ear perceives this Rising distortion vs. frequency as added Brightness because the harmonics are artificially raised in relation to the Fundamental starting at some predefined corner frequency
http://www.elecdesign.com/Articles/ArticleID/7207/7207.html
Walt Jung is not the only person nor is he the first to advocate Multiple gain stages each operation at a reduced gain via local feedback and then enclose this in an overall feedback loop Malcolm Hawksford also advocates this approach in his article on Fuzzy distortion. http://www.elecdesign.com/Articles/ArticleID/7207/7207.html
This is a short 10 page article and I refer you to page 8 figure 7 for his Basic Multi-Loop Feedback architecture.
The next article describes the real problem as I see it and this is the limited open loop bandwidth of the Gain stage in this instance an IC OP-AMP. Most of these have the open loop gain start to decline at 10-100 Hz. Others will have lower open Loop gain and feature a wide bandwidth. These types of devices can have a flat gain vs. frequency to 10 KHz or more and I have seen a few flat up to 100KHZ. I agree that with these devices and a suitably fast output stage that Multi-loop loses a lot of its appeal. However if one was to consider that most of these fast devices are not practical in audio and as noise is concerned you still want the benefit of a High feedback factor then Multi-Loop becomes attractive. Myself I would much rather have a constant 01% THD from 20-20,000 Hz than a THD vs frequency that is say 0.0001% from 20-100 HZ rising to 0.01% at 20 KHz. I believe that the Ear perceives this Rising distortion vs. frequency as added Brightness because the harmonics are artificially raised in relation to the Fundamental starting at some predefined corner frequency
http://www.elecdesign.com/Articles/ArticleID/7207/7207.html
Walt Jung is not the only person nor is he the first to advocate Multiple gain stages each operation at a reduced gain via local feedback and then enclose this in an overall feedback loop Malcolm Hawksford also advocates this approach in his article on Fuzzy distortion. http://www.elecdesign.com/Articles/ArticleID/7207/7207.html
This is a short 10 page article and I refer you to page 8 figure 7 for his Basic Multi-Loop Feedback architecture.
ppl said:
Nothing intrinsically wrong with nested feedback loops old chap.......in fact i love them.....
http://diyaudio.com/forums/showthread.php?s=&threadid=14482&highlight=
...But these have to be carefully designed, otherwise you may have a wonderfull oscillator.......
ppl,
Having listened to class A BJT amps with (I believe) relatively flat THD all the way up to 100k or so and comparing them to FET amps with a raising curve at 1k or so...
(pause, breath... long sentence..)
has led me to a similar conclusion as you. But... to my ears the FET amps with higher dist. does not sound bright but slightly soft and "muschy/foggy" with reduced energy in high frequency attacks/transients.
/Peter
Having listened to class A BJT amps with (I believe) relatively flat THD all the way up to 100k or so and comparing them to FET amps with a raising curve at 1k or so...
(pause, breath... long sentence..)
has led me to a similar conclusion as you. But... to my ears the FET amps with higher dist. does not sound bright but slightly soft and "muschy/foggy" with reduced energy in high frequency attacks/transients.
/Peter
Pam this is also my observations also and i belive that since FET is a Sq law device that this raises the evean order harminics lending to a warm soggy top end however the BJT with its pedomintly odd order distortion component and following an Exponential curve ted to sound Bright.
Oh yes instability can result if all the poles and zeros are not dotted so to speak
Oh yes instability can result if all the poles and zeros are not dotted so to speak
AKSA said:
Ah, but academia was so much fun back then! I guess I'm just a frustrated academic, 'cause I wish I had stayed with it. The only time I've ever been literally pursued by women was my last semester in grad school, when I taught an EE course that was required for many non-EE engineering majors. Man! That was back when jogging shorts and heels were in
. Whoever the guy was that started making fun of women who wear jogging shorts and heels, I'd like to go back in time and give him a piece of my mind. Something like "Shut up or they'll stop wearing them you idiot!!!". Too late though.How did I get on this subject? Oh well. We now return control of your TV set to you...
mikeks said:
...there just might be a workaround to stabilise the minor-loop afterall...
...working on it...
In a perfect world, transfer is completely defined by p(s)/q(s) no matter what the internal topology is used to achieve it including whatever hierachy of nested loops.
As I suggested in the "how many ways to stabilize" thread , it is perfectly legitimate to start with unstable transfers and upon closing the loop, check with root locus analysis whether there are regions of stability.
In the real world one must consider the model accuracy, operation beyond design limits and so on, which can disrupt the stable behavior.
The matter is not what topology is inherently stable or not, but rather which are the risk margins for a particular implementation.
Rodolfo
PS. Stabilization of inherently unstable systems with feedback is old art in control systems engineering. I can recall of an experimental forward swept wing fighter evaluated by NASA some decades back. It should simply blow apart in flight were it not for multiple redundant feedback stabilization control.
And the test pilot knew about engineering too ...
mikeks said:
The X-29 i believe?
After spending billions of dollars, they 'discovered' the thing's response had a RHP pole that simply could not be compensated for.....and had to shelve the project...
True, it was the X-29. As for the unmanageable pole, I don't know but neither was somobody (as far as I know) killed test flying it to the outer extrmes of the envelope.
As for unstable designs, all modern fighters are inherently unstable and depend on fly-by-wire (that is, computers take care of control surface command given flight regime sensor input plus pilot input).
The rationale is to achieve a superior agility with respect to inherently stable designs.
Rodolfo
mikeks said:
What brings about the issue I pointed at before, there is always the possibility for a particular combination of design, load, and input signal, of eliciting an unwanted unstable response.
So the problem is not inherent to the "inner-loops-or-whatever" intrinsic stability, but to the individual combination of circumstances, including exitation.
In theory a perfectly linear polynomial system can always be compensated (as long as it behaves as such).
Rodolfo
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.