So I've got ~1.45V overhead from Vin to Vout(max) on the regulator so there may be dropout.

There is no question that there's a dropout voltage. These numbers are meaningless without you telling us what the load current is.

This is the dropout voltage for typical parts. If your parts came from Ali Express or similar, they might be rejects.

 

Actual meas through a 1kΩ to gnd is 20.5mA which says I need 1.6V and I am at 1.45. So I either need to increase the Vin (at the limit now) or? decrease the load? Minimum curve is for 20mA so...

 

For the vast majority of opamp circuits, the split power rails do not have to be tightly matched in absolute value. As long as the input and output stages have enough headroom to produce the peak voltages you want, everything above that is just margin. For example, let's say you want an audio output that is 20 Vpp. The supplies have to be at least 10 V plus whatever the spec sheet says for the minimum operating headroom (let's say 2 V for each rail) plus another couple of volts for margin. So +/- 15 V, +/-10% supplies should be fine for a 20 Vpp signal out of a NE5532, but so would +14 V / -17 V.

In case you have not yet committed, I strongly recommend against switching supplies, even for driving linear post-regulators. The techniques to manage the conducted and common mode noise are well known, but why bother?

Also, consider at least 0.5 A output current capability for each rail. Not so much for large circuits, but small circuits where you need a little extra oomph, such as adding a couple of outboard transistors to an opamp to drive a small speaker, small DC motor, LEDs, etc.

ak

 

Actual meas through a 1kΩ to gnd is 20.5mA which says I need 1.6V and I am at 1.45. So I either need to increase the Vin (at the limit now) or? decrease the load? Minimum curve is for 20mA so...

If this is a one off build, you need to determine what the maximum dropout voltage will be for the maximum current you ever plan to use. If you intend to operate below room temperature, you need to factor in the rise in dropout voltage.

If you plan to build more than one, you need to do a worst case design. And, it appears that TI doesn't provide that information. They've really gone downhill recently. Their datasheets used to be high quality. Now they have 10 pages of garbage at the end. We can look that stuff up if we wanted it; no need to put it in *every* datasheet.

 

Definitely a one-off. The idea was to free up my 2 linear 0-30V 3A linear PSUs when powering Op Amps. Having 20Vo is a bit of overkill since the max for most Op Amps is 18V max. And the min doesn't have to go to zero either. It could be bad parts as I did not test these upon arrival since I typicaly (before now) havn't used -V. So I have veered into uncharted territory for me.

 

I'm going to back up a bit and test this lot of LM337Ts that I have.

 

Time to rat myself out... It's no wonder the negative voltage wasn't changing. Somewhere along the way swapping out regulators I put a 317 in for the 337. Found and fixed! Now my only issue is a ~2V difference between + and -. Maxs out @ +20.4 & -20.66 and down to +0.12 & -2.43. Realistically I think I'd be happy with a min 8-10V as long as they were close, within a couple of tenths. Note to self- on negative voltages the capacitor negative does NOT go to the ground rail. Blew 2 in a row screwing around with them... And I really did know better...

 

I played mix and match with the ~10 ea of the regulators and was able to get the voltages to within 0.4V of each other across the range of +/-6-20V. Found 3 out of the 20 obviously bad. The deviation doesn't change from max to min so I am satisfied the regulators are closely matched. Now I will have to tweak on the 100Ω and see what happens. More to do tomorrow...

 

but so would +14 V / -17 V.

In that case, my 0.4V is perfect as is.

I strongly recommend against switching supplies

Not even a consideration. Eliminated that possibility very early on.

Also, consider at least 0.5 A output

I had a Triad VPL28-180 and overloaded and burnt it up (was only good for 180mA) so I upgraded to the VPL28-2000 (only other one DK had in stock) which is a bit of overkill but is rated for 2A @ 28V, which the rectifier current is rated for, so the limiting factor is the regulator. Should handle more than 500mA. Ordered another 180 but thinking I might just leave the 2000 in or at least design for it.

 

Put a multiturn pot in and works nicely. Now comes the fun part, converting to PCB layout... Got a small board picked out and will block it in to see if it all will fit or need to go larger.

 

Not to be snotty or anything, but your first opamp project could be a dual tracking power supply.

Just sayin' ...

ak

 

I looked at what OBW0549 had and I was a bit overwhelmed at the sheer number of components. Unfamiliar with the dual op-amp he used but looked it up and scanned over the PDF for it and found it interesting. I know about regulators, or at least the 317, and am comfortable there. Just starting on op-amps so not sure I'm ready to tackle them in a PSU. Wouldn't mind seeing a good example to see how it's done though... Good project idea for later.

 

If you are talking about post #21, it is not a basic design; it has some extra features.

The overall topology is very straightforward, and is the most common one for a dual output, symmetrical tracking supply. For the positive output regulator, a 5.0 V voltage reference (D3) with an adjustable output (R2) drives a fixed-gain non-inverting amplifier (U1A) with an extra-beefy output transistor (power darlington Q2). The amplifier has a fixed gain of 3.2, set by R5 and R3, for a maximum output voltage of 16 V. The output voltage adjustment (R2) affects this value directly.

The negative regulator is a unity-gain inverting amplifier (U1B), again with a high-power output transistor (Q4). Its input signal is the output of the positive regulator. The two resistors that set the gain are R8 and R9. As with all inverting opamp circuits, the input signal is inverted about the voltage at the non-inverting input; in this case that is GND. R8 = R9, so the inverting gain is -1. Thus, a 9.3 V positive output is mirrored to a -9.3 V negative output. If you overload the positive output only and it droops, the negative output will track it. BUT - if you overload the negative output only and it droops, the positive output will stay at the programmed value.

While many bench supplies have an adjustable output current limit, this circuit has fixed-value current limiting as used in many audio power amplifiers. R7 and R12 are current measuring shunt resistors. The base-emitter forward voltages of Q1 and Q3 act as comparators with a 0.6 V threshold. When the current through R7 produces a voltage greater than 0.6 V, Q1 starts to conduct and shunts base drive current away from Q2, reducing the output voltage. This mini-loop is within the overall voltage control loop, so R7 does not affect the output impedance of the supply. R4 protects the opamp output stage from seeing a virtual short circuit if you short the supply output to GND and Q1 comes on hard. Another way to look at it is that Q1 and R4 form a variable voltage divider, controlled by the output current.

There also is some reverse-polarity protection and frequency compensation.

R7 and R12 might be typos. At 10 ohms, the output current is limited to 60 mA, a very small value for the TIP output transistors rated for 5 A continuous. Changing them to 1.0 ohm each has no affect on any other part of the circuit.

The LT1638 is an excellent part, but just about any single or dual opamp will work in this circuit as long as it can operate directly off the peak voltages on the bulk filter caps. Since high open-loop gain, noise performance, and crossover distortion are not important in this application, an LM358 is a common choice. A couple of 741's will work just fine, but be prepared for flak from the purists. Note that the LT1638 is a rail-to-rail part, with very low headroom requirements. Older bipolar parts will have higher headroom requirements, so 18 V at the filter caps will not produce 15 V at the output.

Also, powering the reference directly from the bulk filter guarantees a small amount of ripple on the output that increases with load current. For better performance, use an R-C filter, emitter follower, or low-power three-terminal regulator to clean up and regulate power to the reference independent of the power stages. Another trick is two zeners in series, such as 6.2 V and 5.0 V, one as a pre-regulator for the other.

ak

 

I know about regulators, or at least the 317, and am comfortable there. Just starting on op-amps so not sure I'm ready to tackle them in a PSU.

It's easy to use an opamp with an LM317 to make a tracking negative supply. It's probably in some voltage regulator datasheet.

 

R7 and R12 might be typos. At 10 ohms, the output current is limited to 60 mA, a very small value for the TIP output transistors rated for 5 A continuous.

No, R7 and R12 are indeed 10 ohms and that's a deliberate choice: the analog stuff I build rarely needs more than about 50 mA total operating current-- and often, less than 10 mA-- so the 65 mA current limit was chosen.

I'm devastated that you don't like my schematic; I thought it a sublime work of art!

 

At this point, I am pretty well committed to finishing this design. I would like to consider an op-amp design once I am a bit more familiar with them. Blocking out the protoboard now. Pretty much have the input rectifier w/ smoothing caps and the output filtering done. Still playing with the regulator and adjust layout.

 

Actually still waiting on a few parts so no rush.

 

I looked at what OBW0549 had and I was a bit overwhelmed at the sheer number of components.

Granted, there are a lot of parts there-- which is why I prefaced that post with "For whatever it's worth." Probably not a good starter project, especially if you're just starting out with op amps.

AK's description of how it operates is excellent.

 

Still waiting on parts to finish this design/build but in the meantime, I found these small linear PSUs on eBay. https://www.ebay.com/itm/Continuous...Power-Supply-DIY-Kit-Transformer/293513256925

Price is about the cost of the transformer alone. Did make 2 modifications. Added an inline fuse to the power cord and a multiturn POT for easier voltage adjustment. One is spot on voltage with my bench meter and the other is 0.1V low. They max out @ 15.5V. Still need to tape up the line in wires where I couldn't shrink wrap to protect them.

 

No, R7 and R12 are indeed 10 ohms and that's a deliberate choice: the analog stuff I build rarely needs more than about 50 mA total operating current-- and often, less than 10 mA-- so the 65 mA current limit was chosen.

I'm devastated that you don't like my schematic; I thought it a sublime work of art!

I like your op amp power supply. small wall powered supply, it could also be easily modified and dedicated to a battery project. The
choice of TIP122, 127,LT1638 over top, good cmrr, input npn pnp input stages. other applications op amp will do fig 4 current source and adj 4-20mA transmitter, LM336-5 Vref excellent ... not beneficial to put a limit on the number of parts, the thinking is wow a great bipolar supply.
Thanks for sharing. The Power Supply below submitted by OBW0549.
https://www.analog.com/media/en/technical-documentation/data-sheets/16389fg.pdf