Beta Nirvana Class A New Amp

20V seems more sensible.
Depends what transformers I have around. 🙂
Been so long since I've built anything I've forgotten what I've got.

You are like me. Always looking to the next thing. 🙂

I'm thinking about measures to get a nice THD profile.
Seem to get better results using the VAS node instead of output for the network so far.
 
If a TO126 transistor is too big. Increase R8 and R9 in your .asc to 100R.
You are then back to a TO92.
Doesn't make any real difference to the compensation or operation.
You still have balance with regards to Q1.

This amp is interesting for me as it is the opposite of what I normally do.
Probably going to build this. I want to try something new (to me).

Edit: It doesn't seem to like clipping to the negative rail. Maybe something to look at or is it just a simulation issue?

Paul
This is a direct snip from LTSpice. Got colour this time.
Here's the requested FFT with the shaping network enabled.
Did you cross check the square wave shape after connecting this network.

After reducing the test frequency to 20kHz it didn't work very well for me.
For test purposes I like to check handling of capacitive loads in parallel with the output, capacitor value {CV} then running the command .step param CV list 10nF 47nF 110nF 220nF 470nF 1uF.
 
Last edited:
Michael, try running square waves sim for longer period (e.g. 500ms).
Usually, even if you some small ringing in the first 100ms, it disappears later on. I think it's due to bootstraps caps or the CCS..
The bigger speaker capacitance, the longer it takes for this amp to stabilize.
Also, we could try square waves with output RL network. It's a fix; obviously would be better if amp could
handle square waves with larger capacitance on its own...
When I compare square waves behavior with >330nF speaker cap in both versions: my latest sim, and Paul's modifications, it's pretty much the same..
Update1: RL output network doesn't help much..
Update2: Original AN is handling square waves + output capacitance better. There is still some ringing, but minor compared to this one.
 
Last edited:
It's a trade off. How much performance are you willing to trade for stability?
You can compensate it for 1uF//8R. Try changing C7 from 10pF to 220pF.
THD has gone up significantly. TMC / TPC could remedy this... THD spectrum may not be the best though.

Mjona, no, I didn't double check the square waves. Agree, they don't look good with network in place. Having looked into this more I much prefer to "shape" using a resistor from VAS out to inverting node of op amp. Limit the loop gain around the MOSFETs.
 
Last edited:
I compensate for 100nF.

We don't listen to them but in case the amp gets some freak signal, it's a good idea to ensure it can cope.

TMC looks to be an option. Bring the THD down and shape from there?
Quick and dirty experiment looks promising.
Got to look closer at the loop stabilise though.
 
Adding 15p cap across feedback resistor (C16 22k) makes 100nF squares look perfect. Didn't check (yet) how it affects stability....
squares.png
 
Here is my adaption of the circuit. Apart from the changes already mentioned I added reverse connected signal diodes in series with the protective zeners.

I have not run a Tian Plot for this. I have noted the view that the parallel capacitor loads I test with are seen as over the top so I have cut back on the values. If anyone uses electrostatic speakers the dominant pole cap can be increased. I have a set of Quad 63's if I ever get around to having them serviced.
 

Attachments

Last edited:
So we have 2 slightly different versions of the schematic. One ('original') from the post #45,
and 'new' one (proposed by Paul).
Differences between both versions (as far as I can tell):
a) simpler compensation scheme for new version
b) 'balanced' Q1 behaviour
c) Ideal OLG plot, with higher gain margin (24), and phase margin close to ideal (88).
d) same capabilities to shape fft profile.
e) pretty much the same Thd
f) pretty much thesame capabilities to handle squares and output capacitance

Both versions can be easily sharing the same PCB (already done).
I'm satisfied with both of them, but I like the newer one better.
I personally don't see the need to experiment further; don't think there is much to improve..
We squeezed as much as possible from this topology and 1 pair of outputs.

Should we settle on one version, or keep it open?
Hugh, can you analyze the new one, and give us your opinion?
All sim files attached (20V + 27V) as zip.

1st version
beta_nirvana.20v.26Feb2022.png


2nd version
beta_nirvana.20v.mcd.04Mar2022.png
 

Attachments

Last edited:
I think you should add the option to use a shaping network from the VAS output. Got better results this way.

Also, an idea, if you add a DC servo you can then bridge R20 and C7 with a resistor to divert the loop gain away from the MOSFETs.
The resistor creates a DC offset but this would be automatically compensated for by the servo. Easy THD profiling on the fly.
But additional complexity... Probably not worth it.

Another couple of small advantages of the new version is one less loop to worry about (Stability) and
the lack of a phase lead cap across the feedback resistor helps to reduce potential RF getting into the amp.

Paul
 
I think you should add the option to use a shaping network from the VAS output. Got better results this way.

I will try to add this option to the PCB. The primary reason to take feedback from OUT instead of VAS - I was unable to simulate
OLG when it was taken from VAS.
Which results will be better this way? Thd was pretty much the same, either way; Bode plot is ideal the way it is..

The resistor creates a DC offset but this would be automatically compensated for by the servo. Easy THD profiling on the fly.

I know you are proposing something different, but here is an 'extra' resistor added for some reason...
Here is the schematic of the op-amp=>follower=>VAS implementation from m the 'Brig' amp from late 70s.
Please note R32 (510 Ohm). I wasn't able to see any difference with or without it. Wonder what it's for..

brig_u_001_amp.png
 
Last edited:
Using the dual than prob as in my .ASC you can see the effect this network has on stability.
My best results were with this resistor and another resistor, acting like a servo to correct the offset.
Can get almost perfect monotonic profiles this way but with an associated large increase in THD.

I'll have a look at that schematic later. Brain not really working right now.
Also, not much time...