Simulation Analysis of several unique Allison-based output stages.

[keantoken]

I've just realized what should have been, for me, and obvious advantage to NOT driving the Allison through the emitters.

When you drive it through the emitters, all the driver current goes through the Allison Q's and injects harmonics into the feedback loop. But with source/sink drive, the Allison only contends with common mode currents... So theoretically, the current configuration should work better than emitter drive.

- keantoken

[klewis]

Originally Posted by keantoken

I've just realized what should have been, for me, and obvious advantage to NOT driving the Allison through the emitters.

When you drive it through the emitters, all the driver current goes through the Allison Q's and injects harmonics into the feedback loop. But with source/sink drive, the Allison only contends with common mode currents... So theoretically, the current configuration should work better than emitter drive.

- keantoken

keantoken
I had also been looking at the sims to make a clean current source instead of the bootstrap and kept finding sign ificantly higher thd number. So I just finished a cleaned up version 6. But, it's currently oscilating. Every time I try a faster driver, I seem to get oscillations. This too will be solved.

KenL

[klewis]

V6a up and running. Source Sink configuration, 15032/33 drivers 100pf compensation, no oscillation. I think I can reduce the compensation, but, I don't know if it makes any difference in the amp's performance. Best results so far of all test, now the improvements are coming at 1/10,000th at a time Again, I'm up against the noise floor of the LME49811. The 60hz harmonic that I believe was from the bootstrap drive is gone. Still waiting for new sound card to arrive for high power tests. These are done with -24db input.

The interesting thing is the simulation runs best on the 2SC4793/1837 driver pair(these are the _x models), but, the real device won't run on something that fast. So, is it the low Cob of the 2sc4793 or speed or combination of both that causes the real world problem? And why do the slower devices oscillate in simulation? What part of the model is the cause?

KenL

[kenpeter]

OK. I got sim problem with this one, maybe real problems too? I don't know.
If it works, this has the lowest distortion of any circuit EVER, just maybe...

Theory of operation:

Allison corrects output drops to match input drops. What might be at any
current they so please, whatever drive the output transistors may wish.
This voltage reference is flawed. Because drive currents are not steady,
and they must flow directly through the comparator junctions.

Tallison actively corrects input collector currents to match a reference.
Results in input emitters with a nearly fixed currents and voltage drops.
And outputs merely swing to make the input emitter voltage and current
be steady. This is sorta similar what happens in Taylor or White Cathode
Follower.

Now we got Emitter distortion out the way, cascodes can get rid of Early.
And what we have left is a pair of offset followers with near perfect refs.
For error corrections that do not vary substantially with output or drive.
Distortion unbelievably low when sim occasionally decides to cooperate.

-----------

But I can't make the sim to be stable. Even dumbed out the bootstraps
to LT's CCS to see if that was it? Nope... It works great as posted, but
change the frequency, amplitude, or sample rate: It goes straight into
wild oscillations.

What can be done to fix this schematic and/or the sim parameters?

[keantoken]

I'm thinking it's more the circuit than the simulation parameters.

Aha, try this. Large dopey Cdom's work well. 1KHz distortion is the same, 10kHz suffers a bit, but I'm sure that's a price well paid - distortion is below .002% with a 10k source impedance - at 10KHz, 10V pk-pk.

Pretty amazing, for only BJT's (but is it worth the reduced price of similar performing MOSFET Allison variants?). Are you going to try building it?

- keantoken

[keantoken]

Hey Klewis, I thought I replied to your last post! Doh, where'd it go?

I'm not sure why the stability is so different in real life. To an extent trying to get simulation very close to reality is useless because you need to build a prototype anyway.

- keantoken

[PHEONIX]

Hello klewis

What spectrum analyzer software do you use.


Hello keantoken

If you suspect you are having problems with LTspice, you can check the circuit in software called Simetrix which has a excellent reputation for its convergence. You can down load it from Analog Circuit Simulation Software from SIMetrix Technologies - SPICE, a demo node limited version is available. I use it to just for this problem.

Regards
Arthur

[keantoken]

Klewis uses RMAA, which can be downloaded for free.

RightMark Audio Analyzer. Products. Audio Rightmark

I'll look at Simetrix, thanks for the suggestion. It will be interesting to compare.

- keantoken

[kenpeter]

2n perfect! Thanks.

[kenpeter]

Well, actually: 2nF butchers a full output 22V @ 20KHz sine to a
triangle. 1nF looks good and siney at 20K, but not 96K. And i think
we need about 96K sine bandwidth to avoid IMD and TIMD.

But even 1nF don't quite look like the "phase margin?" is adequate.
Whats the rule here? We need attenuate below 0db (unity) line
before reaching 180deg shift? Or is it less than +3db at 90deg?

Though this is with 2K2+220uF+2K2 bootstraps rather than 5mA
CSS. I need to put the simulated CCS back and check again.