diyAudio

diyAudio (http://www.diyaudio.com/forums/)
-   Solid State (http://www.diyaudio.com/forums/solid-state/)
-   -   opamp + buffer feedback loop clarification (http://www.diyaudio.com/forums/solid-state/126181-opamp-buffer-feedback-loop-clarification.html)

Archwn 10th July 2008 08:36 PM

opamp + buffer feedback loop clarification
 
2 Attachment(s)
Hi everyone,

Thank you for your answer on my previous topic about virtual ground / split supply. I really appreciate your help.

I have another one here. It's the feedback loop of the opamp + buffer normally used for the high current headphone amplifier.

From the buf634 and lm49600 datasheet they recommend type I schematic. however I have seen some design like type II so I would like to know if there is any benefit of doing this.

And also, if the gain of the buffer = 1, would it be better if I put it totally outside feedback loop?

Thank you in advance for your answer. :)

Wavebourn 10th July 2008 08:58 PM

It depends on a buffer.
I prefer type III (like type II without R5, and with a cap in series with R6)
If the buffer adds enough of phase shift to cause instability a small cap across R6 is needed.

AndrewT 11th July 2008 04:11 PM

Hi,
type II is very similar to Walt Jung's.
Have a good read of his advice on composite opamp stages.

unclejed613 11th July 2008 08:36 PM

type II is only needed if the buffer is a lot slower than the op amp. if you aqre using a buffer with the same or higher bandwidth buffer, use type I. go to www.linear.com and look up Application Note 47 for an in depth explanation of what happens with fast op amps and slow buffers.

Christer 11th July 2008 09:13 PM

Some op amps might still need a small feedback capacitor in case I, though. See for instance recommendations in datasheet for BUF634.

unclejed613 11th July 2008 11:50 PM

LT's app note talks about that in detail. "overcompensating" a fast op amp to keep a slow buffer from oscillating. i had an idea for a power amp using an op amp as an input stage and a discrete VAS and output stage. the problem was that most of the op amps i tried caused oscillation. no amount of compensating the VAS could cure it, since the internal compensation of the op amp dominates the compensation of the whole amp. i finally "rolled my own" op amp so i could have control over the compensation, and other circuit parameters. i am currently redesigning the input stage again to include a method of adjusting open loop gain so i can have a variable output impedance.

nelsonvandal 12th July 2008 06:46 AM

I thought the idea with mulitloop, is to have just enough feedback in the "internal" loop to get a flat frequency curve in the audio spectrum (with some margins, like 100 kHz) for the opamp before global negative feedback is applied.

In the original article by Walt Jung, I think he used an AD823 as input stage and AD815 as buffer. AD815 is very much faster than AD823, so compensation for a low speed buffer can't have been the issue.

AndrewT 12th July 2008 08:14 AM

application note 47 is not listed with the others.
But do a search and Linear finds an47.

Christer 12th July 2008 11:18 AM

Quote:

Originally posted by nelsonvandal
I thought the idea with mulitloop, is to have just enough feedback in the "internal" loop to get a flat frequency curve in the audio spectrum (with some margins, like 100 kHz) for the opamp before global negative feedback is applied.

In the original article by Walt Jung, I think he used an AD823 as input stage and AD815 as buffer. AD815 is very much faster than AD823, so compensation for a low speed buffer can't have been the issue.

If the 2nd stage is just a buffer, there is no point in local feedback around it, as it has gain 1, so the interesting case is where the buffer also has gain (typically an op amp too). There are probably many good reasons for local feedback loops, but here are two:

1. Consider the simplified case of two identical single-pole op amps with open loop gain A and corner frequency f. A single such op amp has a unity gain frequency of Af and a a phase shift approaching 90 deg there. There is no problem at all with stability when applying feedback. Now cascade two such op amps. The composite open loop gain becomes A^2 (that is A*A). The corner frequency is the same, but since we now have a double pole there the unity gain frequency is still Af. However, because of the double pole we also approach 180 deg phase shift at unity gain, which can be troublesome when applying feedback. Then considering that real op amps have more than one pole, even if they are usually designed to have a strongly dominant one, means even more problem with phase shift. Using local feedback loops for the two op amps can give us more control over the phase shift.

2. The differential voltage at the input of an amp with feedback depends on the open loog gain of the amp. The higher the OLG, the lower the differential voltage. That is good for linearity, but bad for noise. If you cascade two open loop op amps and then apply feedback over both, you get a very low differential voltage at the first one, requiring a very-low-noise device. If we instead apply local feedback to both op amps, we get a higher differential voltage at the first one. That is, we get less noise, but loose some linearity.


Your point about getting a flatter frequency response may or may not be relevant depending on whether you think that matters or not. The most important thing is to have enough OLG at 20 kHz so feedback is effective. Some argue that it is also beneficial that the OLG is flat up to that point, but that is debated.

unclejed613 12th July 2008 02:24 PM

actually there is a purpose for including a buffer in a feedback loop. buffers are not completely distortion free, and including them in the feedback loop reduces any distortion that they might introduce in the system. remember, power amp output stages are buffers with unity voltage gain (in practice, just below unity, about 0.99) and tend to be slow buffers at that (with a power bandwidth below 1 Mhz in most cases). however, having a nested feedback loop can also have it's uses. the amp design i mentioned previously also suffered from poor clipping behavior. it would "hang" on the rails for about 50 microseconds, a sign something was saturating and not recovering immediately. when i used a nested feedback loop, limiting the op amp gain to 100, that cured the "hangover" problem (and i'm sure such clipping behavior would have caused a nasty headache after a few minutes of listening). since this amp will find some application as a guitar amp (guitar amps tend to spend a lot of time in their clipping region), the amp must clip as cleanly as possible.


All times are GMT. The time now is 04:05 PM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio


Content Relevant URLs by vBSEO 3.3.2