'active feedback' for distortion cancellation

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I was looking through some old stuff and found this idea from Barrie Gilbert (1978). He always liked to promote what he called active feedback. The idea being to put an identical input stage in the feedback path and to the first order distortions will cancel. Really careful design and you can get .02% or so with a couple of volts dynamic range. One neat feature is you have diff or SE input of either phase. I don't know if this has been kicked around here before, I didn't see it explicitly in the cancelling odd harmonics thread.
 

Attachments

  • amp.gif
    amp.gif
    5.7 KB · Views: 971
Hi Scott,

This looks like an elaboration of the opa627/637 design rationale....

...with the exception, of course, of the diff. pair to which the minor resistive feedback loop is returned.....

By invoking elementary signal tracing, it becomes evident that this last diff. pair works to reduce the error signal applied to the output diff. stage (represented by the op. amp.)...

I am assuming said error signal is a voltage...at least nominally...?

Looks interesting... :scratch2:
 
aren't the diff pair inputs labeled backwards?

i'm not sure i can see how to use the circuit yet but with repect to input diff pair distortion reduction i would think the designer of the AD797 should be convinced of the ability of high loop gain to "linearize" the input diff pair

3rd harmonic distortion is porportional to the square of the input diff signal - 10x more loop gain (ref to some ordinary design) will give 100x less 3rd harmonic
 
try burning candle on the other end

Seeing as how multiplication of non-linearities is associative (disregarding phase effects), any reason we couldn't unravel the amplifier at the OTHER end? ie, take feedback from before the output stage, run it thru a duplicate output stage on the way back to the input summation. This way the signal gets pre-distorted to correct for the actual output stage that is after the feedback pickoff. The output stage is usually considered more problematic than the input stage for distortion.
Off course a few complications come to mind, such as an accurate mimic for the speaker load. But then if we run the speaker thru both output stages in series, they will have the same load currents.

Don
 
Thanks Scott Wurcer, for bringing this AFA (Active Feedback Amplifier) idea of Barrie Gilbert. I've been searching this AFA since I read the "Are OpAmps Really Linear" article by Barrie Gilbert, but never find it.

How to implement this idea, if the input is not balanced, like in ordinary single RCA input power amp?

And will this idea rise the OL gain (I like low OL gain)?
 
Yes, I might have switched sign (50/50 you know). Well here's more elaborate circuit. I was thinking maybe a gain of 5 line amp driving a 50k constant impedance volume control. The feature I was most interested in was the true high impedance differential input that rejected common mode and could be differential or single ended of either sign. Even this simple circuit sims out at .015% second and 3rd and above at <.002% at 10V p-p out.

Just some old food for thought. And yes the main idea is in the alternative input stage ideas seeking some of the properties mentioned in Barrie's articles. Output stage issues are still in a feedback loop.
 

Attachments

  • lineamp.gif
    lineamp.gif
    12 KB · Views: 767
AX tech editor
Joined 2002
Paid Member
scott wurcer said:
Yes, I might have switched sign (50/50 you know). Well here's more elaborate circuit. I was thinking maybe a gain of 5 line amp driving a 50k constant impedance volume control. The feature I was most interested in was the true high impedance differential input that rejected common mode and could be differential or single ended of either sign. Even this simple circuit sims out at .015% second and 3rd and above at <.002% at 10V p-p out.

Just some old food for thought. And yes the main idea is in the alternative input stage ideas seeking some of the properties mentioned in Barrie's articles. Output stage issues are still in a feedback loop.


Scott,

Is that how you guys did the AD830, or is that strictly bipolar inside?

Jan Didden
 
AX tech editor
Joined 2002
Paid Member
I really like those fun things also, I always was intriguid by the AD830 and its successor, IIRC AD8130?

Anyway, the major drawback for active feedback as I see it is that the input requires a substantial input differential voltage. In a 'normal' opamp the input differential is nulled by the feedback loop and you end up with just a couple of mV or so across the input, and non-linearity of the input stage is less of a problem.

But if I built an amp with say a gain of 10 with active feedback I need a full volt input to get say 10V output. Now the input stage non-linearity is a much more significant factor in the whole performance.

Jan Didden
 
Yes, but you can do pretty good with the cancellation. The circuit I showed has 20dB of improvement in thirds and if you tweak down the current in the right pair a little the seconds can be nulled perfectly and the circuit has .007 degrees diff-phase over the whole +-1 volt input range.
 
lumanauw said:
What do you think about this? Does it has any merit or any drawbacks?


Dear All!

Pls remember that in such designed circuits you are still leading speaker's coil impact to the audio signal (all global nfbs do so) and the signal will be distored because of inertness of coil's mechanism, that induces-back-eletrical-signals. :cannotbe:

These type solid state designs are good only for resistance load, not for complex (inductive+capacitive+resistive) loads.

I guess that you dot agree,
what a pity!
 
mikeks said:


Hi Scott,

I reckon you've got diff. pair inputs reversed again...in your revised schematic? :(

Moreover, the current mirror output in the first incarnation differs significantly from that in the second... :scratch2:

OOPS! I was only worried about switching things to get positive feedback. Since the inputs are true differential I figure you can get either phase at the output, by simply switching the inputs (one of the features of this circuit).

The second circuit was one I might actually build. I simulated exactly that circuit with fairly ordinary pnp's and some Siliconix dual FET's that I have 100's of from the days when we made those hybrid op-amps.

A power amp built like this would be tough. Even though the input nonlinearities mostly cancel the input stage has to still work with the desired different input voltage. A bipolar diff-stage still only does 10's of millivolts without degeneration.
 
fooling around with a "kitchen sink" bjt diff pair sim i get <-120 dB for any distortion component in the output current with 2K degen R and 1:1 2KHz + 20Khz,@ 1 V ea for 2V pk amplitude

cfp input. input cascode, compound cascode feeding a cascoded feedback current mirror

takes a few more transistors but local feedback linearizes bjts to beyond fets performance
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.