No. Not only it reduces distortion, but it changes spectral content of distortion. Global feedback converts low order harmonics to high order harmonics, though low in amplitude. This has been proven many times in math, Jan please do an intensive search to find and realize this.
You have to be very careful with signal levels, as your distortion is arising from clipping diodes. There may also be a potential complication as you are applying global feedback to the same node as the distortion-producing local feedback. If the global feedback injects enough current to keep the + and - opamp inputs together then the local feedback has nothing to do and so does not create distortion either.
This is a facinating conversation, although I am still not sure that anything useful has emerged from it yet. Unfortunately I will be busy for the next few days so won't be able to contribute. Maybe this investigation should be in a separate thread?
This is a facinating conversation, although I am still not sure that anything useful has emerged from it yet. Unfortunately I will be busy for the next few days so won't be able to contribute. Maybe this investigation should be in a separate thread?
Which raises the issue if I have two stages and use 10 db of local feedback on each is that the same as using 20 on them globally.
I think it is showing up as that would be close but not the same.
However if I use 5 db of local feedback on one stage and 15 on the other properly used there can be a significant improvement!
You can't possibly still cling to the fantacy of in-harmonic components.
Try Volterra, harmonic balance based analysis. Both find real large signal solutions with no assumption on small signal linearization.
Scott,
You know perfectly well I am referring to papers and articles that seem to show up everywhere.
I still prefer to do it in hardware to show actual results. The issue at hand is what are reasonable typical topographies for audio circuitry that have enough distortion to show practical results.
ES
You asked a simple question. There are mathematical frameworks to solve this class of problems. I don't particularly care if "they show up everywhere".
Yes that is one reading of what I wrote. When I do safety lectures it used to be important to make sure that folks followed the instructions intent and not some interpretive variant. The new method is much more successful and easier. You suspend anyone violating any safety rule for a first offense and fire them for a second. Although OSHA still requires the weekly safety briefing getting rid of the interpreters and just plain stupid really reduces the number of accidents!
So do you want to do the math for the general case to determine the optimium gain structure for any given distortion model?
That's my thread. I just wanted to say "Hi". I'd like to participate in this discussion but it's 1000 pages to read!! OMG!!
Creates new (not present in the spectrum before the feedback is applied).
Please include non-feedback designs. Any disturbance of time constants from equilibrium will generate these effects. They will prove to be of no importance at all IMHO.
Okay let me start with the simple stuff. For most electrical engineers PIM is Passive Intermodulation Distortion. A big issue for the cell telephone base station designers.
Here you mean Phase Intermodulation Distortion. The common explanation is that in the typical amplifier design you have a single voltage gain stage that has in addition to the more well know problems a bandwidth that is modulated by the actual voltage level. This is cause by variations in the Miller effect capacitance varying with the Vce voltage.
Is there agreement on this or expansion required?
Here you mean Phase Intermodulation Distortion. The common explanation is that in the typical amplifier design you have a single voltage gain stage that has in addition to the more well know problems a bandwidth that is modulated by the actual voltage level. This is cause by variations in the Miller effect capacitance varying with the Vce voltage.
Is there agreement on this or expansion required?
- Status
- Not open for further replies.