Linear supply regulator design doesn't seem to work, am I missing something?

[EBRAddict]

Hi, newbie here with a power supply regulator question. I've built this regulator from Art of Electronics 3rd ed, around page 608 depending on your copy. It is supposed to dual-track from 0V to 25V.



For most of the range it works very well. The + side seems to work full range. However, the tracking at low output voltages doesn't work on the negative side.

I observed that when the potentiometer is set to 0 ohms, pin 3 (+ input) of op amp A1 is -1.24V as expected. However, since the negative supply for the op amp is also -1.24V, and since this isn't a RRIO device it won't output -1.24V. Instead it outputs around +0.3V. Op amp A2 is an inverter and ideally would set the adj voltage on the second regulator to 1.24V causing it to also output 0V. But instead the max output voltage of A2 is -0.3V + Vdropout. I modeled the positive section in TINA with approximately the same result.

Link to OPA27 datasheet http://www.ti.com/lit/gpn/opa27

I feel like the negative supply for A1 should be -5V on the bottom rail and 36V on the top similar to how A2 is powered by +5 on the top rail and -36V on the bottom.

What am I missing?

[StillTrying]

I agree, that op amp's output can't get to within ~2V of its supplies.

A possible bodge if you don't have a -5V.
3 silicon diodes in series with the top of R5, isolate A1's power -Ve and connect it to the junction of R5 and the diodes. That should lower A1's -Ve rail by another 1.9V.

I'm sure you know that +/-36-38V is pushing your luck on both the op amps and regulators.

[basinstreetdesign]

Nope, you're not missing much.  Sticking a 3-5V zener diode in series with R5 where StillTrying suggests for diodes should help also.  This sounds like a boneheaded design error on the part of whoever came up with the original circuit, IMHO 

An, yes, +/- 38V in is very close to the theoretical worst-case limit of +/- 40V for these regulators but they should survive if you can guarantee that the top of the ripple on the input waveform never violates that +/- 40V even with a slightly high line voltage (+10%).

I notice that U1 has a protection diode between pins 1 and 2 in case the Vout drops for whatever reason, but U2 does not.  Its probably worth it to put one there too, between pins 1 and 3 with cathode end on pin 1.

[jitter]

Ehm, no. The LM317/337 are floating regulators. The actual spec says Vin-Vout < 40 V. So at 38 V in and 25 V out, it only "sees" 13 V.
And yes, that does mean that you can use these to regulate down far higher voltages. I know of one circuit that drops 125 V to 110 V using an LM317 and it works perfectly fine as Vin - Vout is only 15 V.

Or another example:


As for the opamps PSU, the worst will be A2. In the worst case scenario it sees 43 V, which is 1 V lower than its absolute max. spec. Still good enough, I'd say.

[StillTrying]

Ehm, no. The LM317/337 are floating regulators.

But they're not floating in this circuit, if the input is the max 38V, then when the output in adjusted down to 0V there's 39.25V across the regulator.

I notice that U1 has a protection diode between pins 1 and 2 in case the Vout drops for whatever reason, but U2 does not.

There's no C1 to rapidly discharge on the -Ve side, just the op amp's output ~25mA so no need for a diode.

I wouldn't bother with the op amps at all, I'd just use a good quality dual gang pot, the +/- tracking would be close enough for my use.

[jitter]

Yeah, you're right for OP's circuit.

[Kevin.D]

Using 2 Expensive precision (25uV Vos) Op's like this is a bit wasteful.
From a (TI) LM317 data sheet (V reference min-max variation 100mV) so between
two unpaired similar regs when Vset are equil, expect a Vout offset upto +-/100mV .
 If it's really precision tracking you want (else why choose OPA27) then because of other variables like a small output impedance (Ti's LM317 has ~.005 Ohm meaning a 1mV change in Vout per 100mA load current) and then thermal drift (0.7% Vout) over it's temp range so any differential load/temp between your two regs also affects tracking. The way to do it then is by sensing and comparing at the Vout's and so remove all those offset's from the equation . (remote Vout sensing if using long output leads).

Regards

[EBRAddict]

Thanks for the reviews. I started this regulator following the textbook assuming it was a good design, obviously there are some shortcomings in the design and my initial understanding of how it operated. Edited: I was going to sub a TL0X2 at some point but it wouldn't handle the supply voltage extremes.

I also noticed that C1 should be bipolar/non-polar since the voltage can swing -1.24V to +25V at the anode.

I'm running Vin +/- 28V so it should be within regulator tolerances. I will probably tap -5V with an off-board 7905 for the A1 lower rail and hardwire it to the socket. I've already got the existing board mounted in an enclosure so I'll work around the issues.

[MosherIV]

Hi

Well done for being so thourough and diligent. BTW if your goal is to have a dual output tracking PSU, it can be done in a much simpler way:
You do not need the op-amps in the circuit. You just pull the reference of the LM317 to -1.25V and the LM337 to 1.25V and use a dual pot (replace R2 and adds a similar pot to the ref pin of the LM337) to control both for a far simpler circuit. The design already shows you how to get the -ref to the LM317, you can do a similar thing for the LM337
The dissadvantage of doing it this way is that there will be a small difference between the + and - rails but for most simple applications, this is not a problem. This will give a simple 0 to 25V dual supply.

The next thing that is worth having, it an adjustable current limit

Good luck and have fun