Discrete OPA627/687?

[scott wurcer]

In checking the data sheet after a comment here I realized how simple this circuit might look in a fully complementary process. Not much optimization yet, I don't think these values are unity gain stable and it appears to phase invert when slammed into the plus rail (could be my models). I simmed it with some fairly ordinary npn's and pnp's and SK170/SJ74 models and got -150dB in the 3V rms into 600 Ohm condition at G=11 off of the datasheet. I ran this at 25mA though so as not as JC would say "choke it like a chicken".

Something here for the "real' op-amp guys too AOL >1e6 and the usual single pole behavior so the gobs of feedback are available.

I added the VAS that "chases its own tail" so this whole circuit lights up with only the input stage current. Truly no use of ground inside the circuit for bypass of bias rails or cascodes. The FET cascodes in the output might be better as bipolars but that adds a need for a place to tie the bases.

This looks like a nice candidate for a little amp on a card that could pop into many of the OPA627 phono circuits for some BA style listening fun.

No IP (that I know of) here so anyone is welcome to have fun if they feel inspired. No stones please. If this is already out there, sorry, I didn't make a search.

[john curl]

Looks good, Scott! Now, let's make these things, and the consumer audio world will be at your door.

[jacco vermeulen]

Mr Wurcer,

Thank you for posting the schematic, did i already mention you have much nicer Santa legs than Scott Calvin ?

Mr Curl, how about a TPC thread ?

[scott wurcer]

Originally posted by john curl
Looks good, Scott! Now, let's make these things, and the consumer audio world will be at your door.

The seventh was about -200dB!

[syn08]

Originally posted by scott wurcer
In checking the data sheet after a comment here I realized how simple this circuit might look in a fully complementary process.

Very interesting Scott, this is something I was looking for in a previous post, thanks :-)

Now I may need to redraw the schematic but:

- how does this circuit start (bias wise) at power up?
- it seems like it's latching to the rails, following an overload condition?
- Q1...Q4 is this practical for a discrete implementation?
- same for Q71/Q90 and Q88/Q91

Any noise estimates?

Am I wrong or chances to integrate such, due to the lack of a compatible complementary process, are slim...?

BTW, OPA687 is gone for good, OPA847 is the replacement, but that's not FET input.

[scott wurcer]

Those diode connected devices could be just diodes or maybe just left out. Q1-4 just balance the voltage drops fastidiously and could be dropped. The latch up needs to be addressed, there might be a simple clamp that works. The quad of FET's is self starting, in fact in the old days an EPI FET was often used for startup. The diodes across the second stage stop one obvious misbehavior. You are also right in that complimentary IC FETs are a fantasy, the P would probably be a curiosity at best (or visa versa of course). The N's are native to our SOI proceses while P's are native in the old days.

Ramping the supplies and saving all the node voltages and currents gets it to converge the next time with supplies turned on all the way.

[syn08]

Originally posted by scott wurcer
The quad of FET's is self starting, in fact in the old days an EPI FET was often used for startup.

Yep, sorry, I forgot for a sec it's JFET input. Of what I can recall, I think uA723 was using an EPI FET to bias the circuit at power up (and then bias the reference).

[GK]

Yikes!

Perhaps I'm missing something, but I can't see this thing working with real life discrete parts without a CMCL.
Those current source/mirror loaded LTP have so much DC gain with the slightest imbalance/component mis-match the thing will never bias up right.

Cheers,
Glen

[john curl]

I'll tell you this, Scott. If I were to make a high feedback, moderate open loop gain amp, I would choose your circuit. Another related approach is to make a 12dB/octave internal rolloff and get the open loop response way up. I did a design for Sound Technology that had a 100KHz open loop bandwidth and max feedback. This topology might not go that far, but 10KHz would be interesting, and then max feedback. I am just pointing this out for discussion. Still, I am proud of you, as you have seen the light. Dick Sequerra says hi.

[scott wurcer]

Originally posted by G.Kleinschmidt
Yikes!

Perhaps I'm missing something, but I can't see this thing working with real life discrete parts without a CMCL.
Those current source/mirror loaded LTP have so much DC gain with the slightest imbalance/component mis-match the thing will never bias up right.

Cheers,
Glen

Glen, If you notice the outside halves of the differential VAS have an excape path for current out the emitters of the darlington to the rails. Yes the amp has to become differential to single ended somewhere or it's like paralleling two op-amps and having the input offsets fight. Offsets then make these two currents slightly unequal while the loop forces the currents into the high gain node to be equal (within Ib matching etc.)

I definately will try building one since I have all the parts around. If successful I'll bring a couple to the next BA.