Please help improve Self's Blameless 50W amp

[jpc2001]

At one time, the popular diyAudio take on the blameless was the "Honey Badger". I'm surprised nobody mentioned that one in this thread already. It's a blameless with TMC and optional cascoded IPS.

Yesterday I started a thread about two improvements on the Badger and other blamelesses.

 

[jpc2001]

I think your logic regarding -100dB distortion is somewhat flawed. It's true that we hear nothing below 0dB (more or less) but that's in an INCREDIBLY silent room.

I recently did a single blind, side-by-side before-and-after test of an amplifier mod that dropped simulated distortion from -105db to -115db. The listener picked out the difference and preferred the lower distortion amp. I don't have a big sample size on this; it's not scientific.

That surprised me. I would have thought that beyond -100db any further improvement is overkill. It seems improvements are still valuable though. My guess is that there are two effects:

- Real amps make more distortion than their simulations. (Why? Does spice model nonlinear parasitics like Early effect and BJT Cob as if they are linear? Not sure.) So the improvement in the real amp may have only been from -80db to -90db in this case, something on those lines.

- Even if the noise floor in a quiet room is (say) 25 or 30db, we can hear sounds quieter than this. Can you make out a refrigerator in the hallway? a passing truck? If you recognize anything, the noise has a spectrum and you are hearing features below the aggregate SPL.

For inaudible error, my guess is that we should design for -90db real world distortion floor. That probably requires simulating for -110db or better.

 

[Shaq888]

I don't think anything is wrong with the kink, Shaq888 is using transmission line to simulate long loudspeaker cable and that was wrong.

Why do you think a speaker cable is not a transmission line?

In 4.2.2 describes how to compensate minimum phase, when Bn(fc) = 0.
This is the baseline case that can be extended to non-minimal phase.
The minimum phase can be an arbitrary amount, while the non-minimum phase is limited to less than one radian.
So I would say it is to compensate mostly minimum phase, with an extension for a limited amount of non-minimum phase.
Therefore it seems exactly applicable here, in my view.

It can be applied to a minimal phase system, why not, but...

What is the point if phase margin is sufficient? Why to do this using local feedback (including output stage) rather just adding phase lead cap in the feedback divider? Beside that, with an existing output filter it is just like shoot in the own foot.

If you are interested then my comments will be in Damir's thread.
David

Will have a look, thanks.

What specifically is wrong with the kink?

Because with such a strong coupling with a load (essentially the output filter doesn't do its job), which consists element of transmission line, it creates area with dangerously low gain margin. While not really helping in any other department.

 

[Dave Zan]

...It can be applied to a minimal phase system, why not...

It can be applied to practically any linear system, minimum phase or not, provided the excess phase is not more than 1 radian. That should cover any practical audio power amplifier. So it's applicable here.
Just wanted to establish some common fundamentals.

What is the point

The technique can provide the maximum feedback, and therefore lowest distortion, for any particular level of stability.

Why to do this [with] local feedback...rather [than] phase lead cap in the feedback divider?

I also think the phase lead capacitor in the feedback divider is better.
But this is funny because I think your reasons are faulty!
The variation of fT is an issue that needs to be checked.
But both methods, Damir's local feedback or the phase lead cap in feedback divider, have the OPS in the loop, so should be similarly sensitive to the variation.
Provided they are otherwise similarly compensated.
And, of course, provided I haven't made a mistake.

...it is just like shoot in the own foot.

Certainly the Bode step has more possibilities to make mistakes, indeed like a rifle, more effective if accurately used but definitely can shoot own foot.
Dominant pole is a caveman stone weapon, very simple and not so easy to hurt your own foot, but still possible 😉

Will have a look, thanks.

Actually, I did not notice that you had recommended the "phase lead in overall loop technique" until after I wrote my post in Damir's thread.
I was distracted by the discussion about the variation in fT of the OPS between you and Damir.
It seems a "Cup half full" "NO, cup is half empty" type of discussion, it is a pity that you talk past each other.
You talk of the issues with the specific circuit in this thread, Damir talks of the broad principle and his own circuit.
Meanwhile, no one noticed that the fT variation is the same for either technique.

Best wishes
David

 

[Shaq888]

The technique can provide the maximum feedback, and therefore lowest distortion, for any particular level of stability.

I know the theory. I'm talking about the specific schematic, where it wasn't really optimizing anything in this regard.

I also think the phase lead capacitor in the feedback divider is better.
But this is funny because I think your reasons are faulty!
The variation of fT is an issue that needs to be checked.
But both methods, Damir's local feedback or the phase lead cap in feedback divider, have the OPS in the loop, so should be similarly sensitive to the variation.
Provided they are otherwise similarly compensated.
And, of course, provided I haven't made a mistake.

My objections are based on several things found in this schematic. Perhaps I would not object the presence, and may be even the way this step implemented, if amp output would be properly isolated from load/cable on those freq. (while I still think it does not do any good with given values) 🙂

And yeah, ft is a killer 🙂