If you can find a differential amplifier that gives some degree of access to the inputs of the internal opamp, like LT1991 (but hopefully faster), then the common mode is no problem. See attached, values are fairly arbitrary, just a topological example. R8 and R7 reduce common mode to very managable values, with the tradeoff of increased noise gain, thus lower bandwidth.
Now the common mode is dealt with, this circuit is still limited by the ouput voltage of the IC, this can be solved with a simple discrete gain stage (gain of ~2-10) between the output of the IC and the junction of R2 and R3.
P.S. The matching of R7 to R5 and R3 and the matching of R8 to R2 and R4 is as important as the rest of the matching, so something like the LT1991 with multiple trimmed resistors is needed. Something advertised as selectable gain.
Attachments
Nixie pointed out that the comments below should be in this thread and not the Bip vs Mosfet one.
Attached below is some of the stuff I've been looking at wrt to bipolar error correction. I am getting some pretty intersting full power THD-20 results in simulation using an op-amp and bootstrapping the supplies to the output stage. If I understood the circuit and discussion around Glen Kleinschmidt's proposal (post circa 590 in the bip vs mosfet thread), the whole output stage is enclosed in the error correcting amplifier loop but the trick used to overcome op-amp supply limitations was to bootstrap the supplies to the output. In the attached circuit, the output is driven from the VAS and the error correction signal summed at the base of the output stage pre-drivers, so the approach is a little different to his and more in line with EC topology. 20Khz Output open loop distortion into 8 ohms is around .005%, and importantly, into 2 ohms around 0.01%. In simulation, this kind of performance can normally only be had by running well into class A and does not hold up well into heavy loads. I figure that if this got enclosed in a global -ve feedback loop, THD-20 could go down by another 20db or so. The ratio of the bidge resistors in the summing network is quite critical - too high and the distortion reduction declines and too low and the EC loop oscillates - but this seems quite in line with what others have mentioned in this thread.
I was very intersted to read about the possible op-amp power supply interactions and jcx's analysis.
I have a 'test bed' amplifer (Ovation amp - its in another thread so you can take a look there if you want to) up and running so will try this scheme out on it in a few weeks.
Bob/jcx maybe you care to comment
rgds
Attached below is some of the stuff I've been looking at wrt to bipolar error correction. I am getting some pretty intersting full power THD-20 results in simulation using an op-amp and bootstrapping the supplies to the output stage. If I understood the circuit and discussion around Glen Kleinschmidt's proposal (post circa 590 in the bip vs mosfet thread), the whole output stage is enclosed in the error correcting amplifier loop but the trick used to overcome op-amp supply limitations was to bootstrap the supplies to the output. In the attached circuit, the output is driven from the VAS and the error correction signal summed at the base of the output stage pre-drivers, so the approach is a little different to his and more in line with EC topology. 20Khz Output open loop distortion into 8 ohms is around .005%, and importantly, into 2 ohms around 0.01%. In simulation, this kind of performance can normally only be had by running well into class A and does not hold up well into heavy loads. I figure that if this got enclosed in a global -ve feedback loop, THD-20 could go down by another 20db or so. The ratio of the bidge resistors in the summing network is quite critical - too high and the distortion reduction declines and too low and the EC loop oscillates - but this seems quite in line with what others have mentioned in this thread.
I was very intersted to read about the possible op-amp power supply interactions and jcx's analysis.
I have a 'test bed' amplifer (Ovation amp - its in another thread so you can take a look there if you want to) up and running so will try this scheme out on it in a few weeks.
Bob/jcx maybe you care to comment
rgds
Nixie obliges!
An externally hosted image should be here but it was not working when we last tested it.
Bonsai said:
I hate to re-open the debate on what constitutes error correction in the nature of Hawksford as opposed to what is largely conventional negative feedback, but unless I'm looking at it wrong, this circuit looks like conventional negative feedback to me.
Bear in mind, I've always said that the Devil is in the details and the proof is in the pudding, and it looks like the simulation results show that this circuit really performs well. I just would not call it error correction.
The reason is this: it is true that in the HEC scheme you can manipulate the block diagram so that you see a positive feedback loop with gain = 1 inside a larger negative feedback loop. It is true that you can say that this inner positive feedback loop is nothing more than a block of theoretically infinite gain. It then follows that one can in principle say that this block can be replaced by an ideal op amp. It then follows that one can say that in the real world EC is never perfect anyway, so you can replace the ideal op amp with a real op amp. By the time you have reached this last manipulation, most would seem to agree that you now have conventional negative feedback. It may work extremely well depending on the details of the implementation, but it is no longer what I would call HEC. Where in this series of manipulations it lost its character of HEC one can debate.
In looking at this circuit here, I do not see what I normally look for in defining a circuit as HEC. For example, I cannot find a positive feedback loop. Normally I look for a circuit function that essentially straddles the gain block being corrected to establish a difference error signal. I then look for that error signal to be injected PRIOR to the straddled gain block. That is not the case here. In this circuit, if one believes that the error signal is the output of the op amp, one can see that it is injected back at a point inside of the straddle. It looks like the op amp here is just making the whole thing look like a unity gain voltage follower, since the positive input to the op amp is the true input signal and the negative input to the op amp is essentially the overall output signal.
I certainly realize there is more than one way to skin a cat, and that this discussion is certainly subject to semantics. I just wouldn't call this arrangement error correction. Others might call it error correction. Still others might not call the original HEC error correction, but just NFB. To each his own. If you can build a real-world amplifier with this that is fully stable and can produce THD-20 below 0.001% out to full power, it probably makes little difference what you call it. Bear in mind that one is not finished until one does in fact enclose the whole thing in a global feedback loop that is stable and that can accomplish that goal.
Cheers,
Bob
EC or negative feedback?
I guess an explanation like above can hardly be bettered in any way.
Just to throw a dime, the conceptual difference among conventional negative feedback and error correction begins to blurr when one leaves the ideal model to introduce at least a first order approximation to the real world.
The summing block performing error correction must of necessity be bandlimited, and substitution for a first order network in turn ends up to be equivalent to an integrator. A good OpAmp for its part behaves (open loop) much like an integrator due to first order dominant pole compensation, meaning an error correction scheme using an OpAmp as summing block is not much different than a regular negative feedback topology where the single feedback loop encompasses both the OpAmp and the output stage.
This does not detract from the fact that, from a conceptual point of view, one thing is to insert and infinite gain block, and quite another is to insert a nulling network. This last insight and the implications in terms of local performance on global performance allows for an entirely different approach not readily apparent in conventional negative feedback terms.
Rodolfo
I guess an explanation like above can hardly be bettered in any way.
Just to throw a dime, the conceptual difference among conventional negative feedback and error correction begins to blurr when one leaves the ideal model to introduce at least a first order approximation to the real world.
The summing block performing error correction must of necessity be bandlimited, and substitution for a first order network in turn ends up to be equivalent to an integrator. A good OpAmp for its part behaves (open loop) much like an integrator due to first order dominant pole compensation, meaning an error correction scheme using an OpAmp as summing block is not much different than a regular negative feedback topology where the single feedback loop encompasses both the OpAmp and the output stage.
This does not detract from the fact that, from a conceptual point of view, one thing is to insert and infinite gain block, and quite another is to insert a nulling network. This last insight and the implications in terms of local performance on global performance allows for an entirely different approach not readily apparent in conventional negative feedback terms.
Rodolfo
If anyone interested this is the site with many original Hawksford publications:
http://www.essex.ac.uk/ese/research/audio_lab/malcolms_publications.html
CHEERS!
http://www.essex.ac.uk/ese/research/audio_lab/malcolms_publications.html
CHEERS!
AKSA said:
Both negative and positive feedback can be seen in
http://www.diyaudio.com/forums/showthread.php?postid=1136690#post1136690
Bob, Ingrast
thanks for the insight. I was probably a little hasty in applying the 'HEC' term, and looking at the circuit again, it is indeed NFB . As you point out, there may be more than one way to get to 0.001 THD-20 and at the end of the day, I'd like to get as close as possible to that.
I'll report back on what I find when I try this out in a few weeks.
Thanks
thanks for the insight. I was probably a little hasty in applying the 'HEC' term, and looking at the circuit again, it is indeed NFB . As you point out, there may be more than one way to get to 0.001 THD-20 and at the end of the day, I'd like to get as close as possible to that.
I'll report back on what I find when I try this out in a few weeks.
Thanks