+2V rail with an opamp

A, the opamp will probably not be happy with 10000x the rated capacitive load. B, output will vary with load. C, is the standard

Nah, it’s only unstable when the internal compensation pole is too close to the output load pole. There’s another region of stability where the load capacitance dominates.

You can verify that by looking at the supply current. I’m using TCA0372s for four rails in a forthcoming project, because (as JL pointed out) it’s stable with any alpo capacitor on the output.

Its negative PSR stinks, but it’s very good for positive rails.

Cheers

Phil Hobbs

RRIO opamps are typicaly stable with big, low-esr caps. Of course they are happier with alums or tants, with a bit of ESR. Circuit A just rings a bit with a load step. The load capacitance is in fact a bunch of bypass caps on a power plane.

We're going with circuit B with a 15 ohm resistor. The 2V is a shared reference but has no significant load. Might sim the bead just for fun too.

It's a 4-channel tachometer circuit where it was handy to reference all the signals to +2.

Here's the board. It has about 35 parts per square inch, using both sides.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

Circuit B with a 1K bead rings forever, but it's very nice with a 5 ohm resistor.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

What is wrong with using a LM317?

it'll only source

We could add a dump resistor to a vreg; the load is light.

But we have the opamp and resistors on the BOM already, and the opamp thing is small and precise.

And a 317 is not c-load stable, so we still have a problem.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

What do you mean by that, only loaded a by a capacitor?

The output R to adjust input to make 2 V is already a load.

I have used the LM317 with very low AND with very high loads.

This circuit I once found online:

If you only use wet electros or solid tants, 317s are fine, although noisy. Ceramics and low-ESR alpos can make them sing like the birdies.

JL has a trick for that, namely a 10nF-ish cap across the upper resistor of the volume divider.

That increases the noise gain, which helps stability. You can also parallel a ceramic and a similar-valued wet electro on the output, which makes a lead-lag characteristic.

Cheers

Phil Hobbs

Your pics are unreadable.

It just ocurred to me that I can use a 2.048 volt shunt reg, an LM4040, and one pullup resistor to make a nice +2V rail. LM4040 is c-load stable. We have that part in stock.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

One tantalum across the ceramics will stabilize a 317 nicely. But all that makes for a lot of parts. My board is insanely dense already.

LM317 is ancient, older than most of my engineers.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

afaik the 1117 is much worse of the two

Yes, but dropout is lower. It can make 3.3 from +5.

There are a lot more modern parts around these days. C-load stable LDOs.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

yeh, afair the 317 is a darlington NPN emitter follower, the 1117 a Sziklai "PNP" common emitter

I have always used tantalum caps, or electrolytic caps.

Eyes?

Max 15 mA is that enough?

Microchip has some voltage reference chips too:

There were so many things, like ADCs and DACs and analog switches, that were inhernetly unpolar that it made sense to offset the signal levels +2 volts.

Somebody, TI I think, makes a "rail splitter" chip too, but we don't have it in stock. Adding a new chip to inventory is a nuisance.

The LM4040 family of shunt regs is geat. Lots of people make them. And you can float them most anywhere and flip ends.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

The LM4041-ADJ is super useful too. Unlike any other reference except IIRC the LM385, it looks like a PNP transistor—it senses the voltage between ADJ and anode, so to get minimum voltage you tie ADJ to anode. (The others all work the other way up, so to get minimum voltage you tie ADJ to K.)

That makes it possible to use all the usual or unusual positive-rail shunt reference tricks on the negative rail.

Its max voltage is sort of low, but a cascode fixes that.

Cheers

Phil Hobbs

cathode,

The circuit is from Fig.9 Low Cost 3A Switching Regulator, Nat Semi App Note 181. It's a sort of 60ish design, hysteretic switcher, relatively low frequency 10-30 kHz range, load dependent of course, noisy and ripply output, no overcurrent protection since it bypasses the '317. It's not a very good candidate for a reference rail.

LM4040 is the answer for the application. You can use enough bypass to get 80dB attenuation of worst-case load-step voltage coupling onto your rail, at which point the shunt regulator can take it down even more. You can bias it with enough current to handle light loads. Main thing is low parts count, especially versus the double differential feedback required to stabilize capacitive loaded OAs.