Discrete Opamp Open Design

[coluke]

Originally Posted by fitzfish

Have you ran a curve trace of both to see how closely they are complimentary?

Nope - only Id matched, but I wonder if close matching is worth the effort at all: I mean, if throwing in 4 randomly selected devices you get 0.0005% THD at the expense of a slight DC unbalance...

L.

[ChocoHolic]

Originally Posted by Joachim Gerhard

Choco, i like the extension you made. If it sims well i would like to build a prototype.

Joachim,
I started to sim, but had no models of reasonable JFets.
After seeing silly results as a consequence of using silly types I decided to check for models of the 2SJ74 and 2SK170.
I have to ask for patience and cannot make promises on a time schedule.
Busy week ahead...

[dimitri]

George Wilson presented an all NPN and NJFET op-amp in >>1968<< that had 200V/usec slew rate and 50MHz BW.

I never saw any followup

[RNMarsh]

Originally Posted by scott wurcer

I found it interesting that George Wilson presented an all NPN and NJFET op-amp in >>1968<< that had 200V/usec slew rate and 50MHz BW. This WAS potentially good for audio but I never saw any followup.

Yup, the same Wilson as in the current mirror.

Maybe we could do it for him. RIP. Do the followup for audio.

[RNMarsh]

I have never, in my life time, had a hard time getting low thd in an amplifier. It just isnt that hard. If you have a symetrical and balanced circuit, it doent take much gnfb to have very low thd. I think i showed that with a measured .0005% thd + n with modest gnfb. Decades ago I had <.002% at 22 v p-p on the very first line stage I ever designed. That one had only 20db gnfb.

So, remind me again -- I am not sure what we are doing now for a discrete opamp design? What is it that we can do discrete better than an IC opamp? -Thx - RNM

[counter culture]

linuxguru

I think there's something wrong with your sim. The descending slope on the right might be caused by a LF drift in the caps, it's not always there anyway... The skirts round the fundamental don't look right either.







Spice directives for simulation courtesy of keantoken.

2V in, 6.4V out (peak) into 16 ohms, as shown. < 2 PPM THD. It's not the least number of components, but you can get these transistors for $0.03 ea. Surely that's got to count for something.

[elektroj]

Bases of Q11-Q14 are connected to the wrong end of R6.

[counter culture]

Damn. Thanks.

You'd think it would make more difference to the sim. I moved it and got 0.00016.

[linuxguru]

Thanks, Dimitri. As shown, Wilson's topology has a minor biasing problem for the 3rd-stage darlington, but that's easily fixed with a dual-diode in the emitter. The CCB can be grafted on for the O/P stage, but the darlington VAS won't be good for sonics, at any rate.

[Chris Hornbeck]

Originally Posted by RNMarsh

So, remind me again -- I am not sure what we are doing now for a discrete opamp design? What is it that we can do discrete better than an IC opamp?

As Scott has reminded us, the existing discrete opamps for audio run +/- 24 VDC rails for +24dBu output into 600 Ohms. No sense doing less. Class A output at that level isn't difficult either.

Some things not always clear in monolythic opamp performance are high frequency common mode distortion levels with high (maybe 10K Ohm or so) source impedance, clipping overload recovery behavior, and sensitivity to power supply imperfections. A truly hot dog design would address these things too.

Thanks,
Chris