The AMNESIS amp: a good amplifier, like a gentleman, has no memory.

[JOSI1]

Current sources
Since I am not familiar with designing amplifier circuits, can anybody tell me
how the current sources (for excample the tail current source) do look like in detail
in terms of a schematic.
Thank you for support.

[padamiecki]

you can use it like Q9-12
btw Cascomp is twice time cascoded and has feedforward EC, so it shoud perform extremely well regarding low memory/nonlinear distortion

[maxlorenz]

Originally Posted by padamiecki

hi Max!
these are my old plays with Peufeu's memdist, the idle is ~90mA
cheers!

Hi Pavel. Thanks very much. I am unable to simulate. Could you please do the following to resurrect your interest:

1) 1K resistor on the emitter of the "second BJT" (PNP) from the CFP (emitter to output node of the CFP-source of the JFET), this way the gain is reduced and the offset is minimal.

2) Can you do the cascoded-cascode VAS, with a 22R to 50R between the first cascoded elements?
This is the way Peufeu planned it. Look at the schematics above.

3) I do not understand your feedback.
Can you do a cascoded-CFP also at the output and have a conventional feedback? When I cascoded the outputs, they became stable.

Edit: look at the interesting PDF attached about Bias circuit.

Hi JOSI1,

Please take a look at:
Current Sources, Sinks and Mirrors in Audio

Cheers,
M.

[padamiecki]

Hi Max!

do not get me wrong, I only wanted to show you a few posts before that the memdist amp is possible (in theory) to build.

The figures are not so bad but due to memdist triples at the input and cascodes it might be marginal stable.

I attached for you the results of the sims after the changes proposed by you, it looks like thd spectrum has more hf harmonics.

2k7 resistor of cfp pair is the value given from the God to get the lowest thd from cfp pair.

I attached before the asc file to tease you to get yourself ltspice, which is for free, and try your own ideas.

I will watch the thread with the interest but please do not expect me to check your ideas with the simulator,

cheers!

[maxlorenz]

Hi Pawel,

Yes, you're right. At this stage of my career I must learn how to use those tools...
The problem is that I use Ubuntu at home...

I place the 1K R in the other BJT's emitter, for lowest offset. With simulators I could find the best CFP option and explore all the possibilities of cascodying different outputs, indeed.

Thanks for your interest.
M.

[padamiecki]

no problem, LTSpice works great through the wine,
tested by me with Ubuntu, Mint, Chatleau!
but only the old version, new VIII has the issues.

[maxlorenz]

OK. Thanks!

[gaetan8888]

Hello

How about using a rush cascode input to reduce the memory distortion ?

Thank

Bye

Gaetan

[padamiecki]

well, LTP is a kind of Rush cascode with the transistors of the same sex,
when the first is emmiter follower steering the second common base via its emmiter,
does it reduce memdist?
I would say yes!
Does it improve SQ?
I do not know!

[mchambin]

Please explain to me what this is all about.
I was expecting to see something about stability of the bias at the output stage of a class aB amp.
I see this thread is mostly about input stage and VAS, I do not understand since those are class A stages known about no temperature effect from audio. May be at very low frequencies where junction temperature time constant could matter.
I must be missing something, buried in many lengthy threads and references.
Please give me the basis on this memory distortion.