Another error correction output stage.

[Mr Evil]

I've been trying to think of ways to improve output stage linearity, and realized it would be a suitable task for the error correction discussed here recently.

Applying the error correction only around the output stage, inside a normal negative feedback loop, avoids the poor clipping that it otherwise has (it tries too hard to pull the output above the physical limit). Because of the way that feedback is applied, the input impedance of the output stage is actually negative, but that isn't necessarily a problem. Benefits include improved linearity and reduced output impedance.

In the example attached, the input stage is an op-amp for simplicity. Distortion is low across the audio band, and because of the very low output impedance it remains virtually unchanged no matter what the load is (I've tested resistive loads down to 1 ohm, capacitive loads up to 10nF and modelled speakers).

Comments and suggestions welcome! I was wanting the output stage to have gain (by using Sziklai pairs with reduced feedback), but I haven't yet found an acceptable way of stopping cross-conduction.

[anatech]

It would be interesting to see what happens at high freqencies and with impulses. Build it and feed it a differentiated square wave.
'Scope input and output to see phase and amplitude relationships.
-Chris

[sajti]

I think that the error corrector opamp must be the fastest You can get. If no, there will be problem with the delay of the error correction, and it will results more distortion. Very strange distortion....

sajti

[sss]

hello Mr Evil
i made somthing simmilar to this a long time ago (see pic)
i dont know if i drawn it correctly (simplified), i got no time right now
the idea is :
the op amp is a differential amplifier with a gain of one (all resistors are with the same value)
all that it does is simply taking the voltage falling on the Vbe of the output transistor and adds it to the input signal , that way input = output (in theory)
the difference between my circuit and yours is in my circuit the op amp got independant power supply with floating ground , so the input signal is connected to it , i think this makes the circuit much more complicated .

in the circuit i made i was using a good "tuned" instrumentation amp , rather then a differential ,and a compound output stage .

i made that circuit but didnt test it with any kind of distortion metter (because i aint got one ), it was tested with music and it was ok , the output impedance was wirtually zero!!
i connected 0.1ohm resistor in parralell with the output and the music continued to play (the speaker was connected also) and the volume was unchanged!! (the output transistors got pretty hot)

[Mr Evil]

Originally posted by anatech
It would be interesting to see what happens at high freqencies and with impulses. Build it and feed it a differentiated square wave.
'Scope input and output to see phase and amplitude relationships.
-Chris

Simulations suggest that it can have good transient response as long as the error amp is appropriately fast. Measurements I did on the original (two op-amp) circuit showed this, so I expect it to hold true for this output stage too. I will build something like this eventually, but I feel that there is still a lot of improvement to be done first.

Originally posted by sajti
I think that the error corrector opamp must be the fastest You can get. If no, there will be problem with the delay of the error correction, and it will results more distortion. Very strange distortion....

sajti

Yes, it must be very fast. As discussed in the origianl thread I linked to, the error amp must be faster than the main amp (the output stage in this case). I'm still looking for a suitably fast op-amp with wide phase margin. The simulations are done with NE5534, which works well but could be faster. I considered making a discrete differential amp using RF transistors for this, but the gain would be low, negating most of the advantages of the configuration.

Originally posted by sss
hello Mr Evil
i made somthing simmilar to this a long time ago (see pic)...

Yeah, that's very similar except that in your case the op-amp is also in the signal path, whereas in my case it only amplifies the error. Good to hear it worked, so I know I'm on the right track

Originally posted by sss
...i connected 0.1ohm resistor in parralell with the output and the music continued to play (the speaker was connected also) and the volume was unchanged!! (the output transistors got pretty hot)

I think this is possibly the biggest advantage of this sort of design. I know some people complain that amps are not designed with the complex load of a loudspeaker in mind, but if the amp is totally insensitive to load impedance then it doesn't matter what load you design it to work with, it will still work exactly the same with any other load.

[sss]

hi again
i never tested my circuit with global feedback
i see u did , so how's it with a global feedback ?

[Mr Evil]

Excellent. It's a great way to lower distortion, and in many cases it can be applied to an existing amp with only minor modifications. Using an op-amp it's only a few components extra in total.

[sss]

i got no distortion metter so i tryed to simulate the circuit with a simulation software a year ego, but the whole amp and this circuit was too complicated for the software so i left it

[Mr Evil]

I don't have a distortin meter either, so what I did was deliberately increase distortion by adding diodes in the signal path. Then the distortion is clearly visible on a scope and it's possible to make a rough measurement of the amount of improvement. That isn't suitable for measuring all methods of reducing distortion, but it's useful here.

[sajti]

Originally posted by Mr Evil
I don't have a distortin meter either, so what I did was deliberately increase distortion by adding diodes in the signal path. Then the distortion is clearly visible on a scope and it's possible to make a rough measurement of the amount of improvement. That isn't suitable for measuring all methods of reducing distortion, but it's useful here.

The most interesting is the signal on the output of the error correction amplifier.
So, it will work with NE5534, but You have to use low pass filter on the input, to reduce the bandwidht to about 80-100kHz. In this case, there will be no very fast signal at the output stage..

sajti