Discrete Opamp Open Design

[bcarso]

Originally Posted by rsavas

Top assembly drawing.
Took me equivalent of < one working day to get this far, so it ain't hard to duplicate. Suggest to put some copyright's notices on the material, if we want to protect the information? I have not seen any copyright notices on material on this site!
Seems that all the jfet mfg's can agree on a standard pin out, in SOT23, so can substitute parts quiet easilly.

Looks like a nice job. I love how large the SOT-23 packages appear compared to the complementary duals

[grhughes]

Thanks Brad!

[scott wurcer]

Originally Posted by donpetru

I have attached below an audio preamplifier that can be the basis for achieving an operational amplifier with discrete components.

All inputs are appriciated, but with all due respect this is pretty far from a broadly applicable circuit. With 1K output resistance, split feedback, and here set up for phono there would be major changes needed.

[scott wurcer]

Originally Posted by qusp

$40-50USD for standard values in 0.01%; for what you will get in exactly the custom value you want from Texas components in 0.02% for ~$10-12....

I was thinking more of pennies for the resistors.

[scott wurcer]

Here is a first shot at a blow by blow description of my op-amp idea. No guarantee that it is clear to everyone so important clarifications can be added.

[jcx]

splitting the Ccomp, bootstrapping the middle to the driver output gives 30 dB more loop gain to 20 KHz; costs < 10 degrees phase margin

heck it even looks like flat loop gain to 10 kHz

from the below Ltspice sim, simple Middlebrook loop gain V souce in series with feedback divider

split Comp, doubled pF, added 500 Ohm to Q15 emitter

green trace: original single pole loop gain;
yellow: split, bootstrapped Ccomp 2-pole comp

Originally Posted by fas42

To keep the ball rolling, here's a quickie knock up of the schematic from #1568 for those who are averse to fiddling with the versions already posted. Appears to be OK, haven't checked everything, I'm sure people will point to any problems ...

The PSRR looks much, much better of course. One thing though, can't push the output standing current up to 8mA easily for some reason ...

Frank

[scott wurcer]

Originally Posted by jcx

splitting the Ccomp, bootstrapping the middle to the driver output gives 30 dB more loop gain to 20 KHz; costs < 10 degrees phase margin

This is good, we tend not to use these techniques because they can effect fine scale settling. For audio this might not matter.

[bcarso]

Finally set up to screen some today off a strip and got through 19 pieces. The tempcos are all fairly small and the measurements didn't take long to settle. The bench ambient was about 75 F.

Vds was fixed at 4.00V except for the drop across a current sampling resistor in the drain of 10.0 ohms. 200 ohms was in series with the gate to prevent RF oscillations. Idss was measured, then the reverse bias on the gate to bring the current down to 100uA.

Mean value of Idss: 15.693mA. Standard deviation 1.567mA.

Mean Vgs for 100uA Id: -634.3mV. Standard deviation 43.9mV.

I also looked at Id with the gate floating (actually tied to the source through the 200 ohms and 330pF). This value corresponds to the operating point where the drain-gate leakage current is the same magnitude as the forward gate-source current. It ranged from a high of 25.2mA to a low of 17.8mA.

So, for differential applications matching is necessary. The good news is that otherwise the temperature coefficients are small.

[bcarso]

Originally Posted by scott wurcer

This is good, we tend not to use these techniques because they can effect fine scale settling. For audio this might not matter.

I've been looking at settling in sim lately as a matter of course. I can't easily simulate thermals, which are there if you look hard enough, but it's interesting to see how close intuition is to what sims say.

I wonder what the 990-style compensation does to settling behavior?

[Wavebourn]

Originally Posted by scott wurcer

This is good, we tend not to use these techniques because they can effect fine scale settling. For audio this might not matter.

I would say, for audio wider flat loop gain matters more, and +30 dB on 20 kHz is a huge advantage.