AB-dynamic

[anli]

Idea: http://gaydenko.com/ab-dynamic/schematics01-idea.png
Using R5 as sensor, M1 trys to keep current through M2.

Schematics:http://gaydenko.com/ab-dynamic/schematics39.png
Goals:
- linear and high amp input resistance ("ALPS-ready", no any preamp needed)
- high PS pulsing rejection
- thermal stability without any thermally coupled parts
- minimization of dynamic thermal distortions ("memory distortions")
- crossover distortions elimination
- DC-coupling with source and load (DC amplifier)

Regulations:
R15 - quiescent current
R3 - output offset

Typical spectrum: http://gaydenko.com/ab-dynamic/spectrum39.png
Transfer function derivation (above pane shows output voltage and mosfet currents): http://gaydenko.com/ab-dynamic/transfer39.png

Current project state: thinking about PCB
Publication aim: your criticism

[anli]

And to add an intrigue - there are mosfets currents: http://gaydenko.com/ab-dynamic/currents39.png

[janneman]

Andrew,

This is pretty intriguing! I guess the Iq is aroundf 50mA? Have you tried to run it with larger current, what would the transfer function look like then?
Do you know the open loop freq and phase response?

Jan Didden

[anli]

Originally posted by janneman
Andrew,

This is pretty intriguing! I guess the Iq is aroundf 50mA? Have you tried to run it with larger current, what would the transfer function look like then?
Do you know the open loop freq and phase response?

Jan Didden

As shown at the schematics, Iq is about 210mA. With too low Iq derivation isn't so smooth around 0V. I think, Iq about 100-150mA will be acceptable. Will see.
dVout/sVin doesn't change with more amplitude of linear input signal: "all can happen already happened" inside +/-11V range :-) Only a clipping at ~19V is outside the picture.

As for frequency prediction, I have found simulated results rather differ from reality for gnfb topologies. Moreover, you can see rather "strong" hf correction. At any case I'm not a fanatic of using instrumental amps for listening to a music

[anli]

Aha... Silence... The idea is perfect, and the realization via the schematics is beyond any criticism...

... or it is impossible to read a schematics beyond standards...
... or all gurus are at their vacations...
... or ... WHAT?

[CBS240]

Hi Andrew

Interesting way to drive a quasi fet output stage. Very asymmetrical, even for quasi. There are members here who chastise asymmetrical topologies, but not me. I believe choice of components is more important, depending on goal of the circuit. The input j-fet is cascoded and keeps 5-6 Vds. This is good, and no need for matching because there is only one. I do have one question, do you think there would be need to have an active turn off device for M2 instead of R7 even though it appears to never turn completely off, but operates as class AB, or some type of totem pole buffer to drive the gate capacitance? How is capacitive loading for a sqare wave look at that 2KHz? And 10KHz?

[anli]

Originally posted by CBS240
Hi Andrew

Interesting way to drive a quasi fet output stage. Very asymmetrical, even for quasi. There are members here who chastise asymmetrical topologies, but not me. I believe choice of components is more important, depending on goal of the circuit. The input j-fet is cascoded and keeps 5-6 Vds. This is good, and no need for matching because there is only one. I do have one question, do you think there would be need to have an active turn off device for M2 instead of R7 even though it appears to never turn completely off, but operates as class AB, or some type of totem pole buffer to drive the gate capacitance? How is capacitive loading for a sqare wave look at that 2KHz? And 10KHz?

I have tried CC-follower between R7 and R18 - at least in simulation all the amp seems to be closer to oscillator :-) Moreover, I have two amps with 560 and 680 ohm resistors at similar places with the same mosfets - both works fine.

As for square waves, at first, there isn't any difference which frequency we use to test. A rise or fall time period does mean only. At any case, the amp is sufficiently fast to work with sound - just from experience.

As I have mentioned above, I don't think we must use instrumental amplifiers for music listening. Say, -120db, 100MHz, 100V/uS or such numbers mean nothing. At first message I have listed few demands I consider to be important for audio amplifier, if we say about a schematics only, omitting realization (PS, PCB and so on) details. I can add monotonically falling spectrum as another demand - while increasing harmonics number or lowering signal level. Plus small spectrum dependency from frequency.

[padamiecki]

Originally posted by anli
Aha... Silence... The idea is perfect, and the realization via the schematics is beyond any criticism...

da,
we were waiting from 2005 on your update,
so do not be surprised of the silence, everybody hard thinks.

It is hard to me to make some interesting comments,
please explain what parts should be matched (if any),

congratulations!

[anli]

Originally posted by padamiecki
da,
we were waiting from 2005 on your update,

Sorry, my fault... But... what do you mean?

Originally posted by padamiecki
please explain what parts should be matched (if any),

Regulations were mentioned above. For two channels, probably, input fets must not be too different. A fet - current source must be able to generate ~4mA. Currently I don't see something else. Just want to repeat, the project isn't soldered yet - just design. But parts are ordered.

[padamiecki]

Originally posted by anli

Sorry, my fault... But... what do you mean?

what fault? I am far from indicate someone's fault, there was no fault at all.
first time you showed similar project long time ago, and promissed finalize it. From your description I read that the sound did not make fatigue during listening and was fully detailed. I just could not be patient to wait for so long! As I remember you had even PCB for this project! This was good training for me waiting for your ready project.
Maybe before you wrote about another amp?

Originally posted by anli
A fet - current source must be able to generate ~4mA.

This is exacly what I asked!

Anyway, the design is clever, I like it!

Thanks!