Bipolar benchtop power supply, and why are they so hard to find.

[David Hess]

I made my own, many years ago when I was a student. I used transformer with two 15V secondary windings, which I connected in series, with the centre tap going to 0V on the schematic.

That is the straightforward way to do it.  It is possible to do without the operational amplifier as shown below.  EDN published another example.

Driving the virtual ground to the feedback network as shown in the EDN example has the virtue of maintaining tracking symmetry from both sides.  I might combine that with the example shown below.  I think I have seen another design which uses two operational amplifier to cross connect the feedback networks achieving the same thing.

[Zero999]

I made my own, many years ago when I was a student. I used transformer with two 15V secondary windings, which I connected in series, with the centre tap going to 0V on the schematic.

That is the straightforward way to do it.  It is possible to do without the operational amplifier as shown below.  EDN published another example.

Driving the virtual ground to the feedback network as shown in the EDN example has the virtue of maintaining tracking symmetry from both sides.  I might combine that with the example shown below.  I think I have seen another design which uses two operational amplifier to cross connect the feedback networks achieving the same thing.

That should be better than my design. It should also be more stable. Others have said that my op-amp based design might oscillate but I've not managed to make it do that on the bench, even though no compensation capacitor is used.

[David Hess]

That should be better than my design. It should also be more stable. Others have said that my op-amp based design might oscillate but I've not managed to make it do that on the bench, even though no compensation capacitor is used.

The simplified tracking design does not track with line and load changes and does not track in current limit.  I think its major advantage is simplicity.  Adding a single operational amplifier to balance the output as shown in the EDN example solves that.

I have never had a problem controlling an integrated regulator with an operational amplifier.  If the operational amplifier is slow enough, then frequency compensation is not a problem.  If frequency compensation is needed, then it is not difficult to add.

None of these designs are suitable for a variable output bench supply unless extreme simplicity is desired and an integrated design is unlikely to need universal tracking.

[Neomys Sapiens]

There are also some specialized ICs for building symmetric power supplies:
the RC/XR4194 (Raytheon/Exar) and the SG1501 (Silicon General/Unitrode), albeit the latter does not go below +/-10V. Both have Current limiting and use external transistors for higher currents.
Also see LM125 (NS, see AN82). The circuits in the datasheet of the CA3094 (RCA/Intersil) and
 the applications for the MC1461/1463 also have some circuits.

I just went though my index of appnotes and articles: all build-your-own ones with one knob controlling two outputs use a double pot or related, so they are not tracking each other. NS AN127 gives a tracking regulator using OPA's (LM143).

[b_force]

What power do you need?

If i am right I read between the lines you're just starting?
So why not making one yourself? It's not that hard.
In general, why don't people make stuff themselves anymore?

[Shock]

Tenma = Farnell, Element 14
Korad = Shenzhen Korad Technology

[Awesome14]

I use a Sorensen quad 15-4. It's cheap, digital, bullet-proof, does parallel and serial, so pos/neg voltage no problem. I have 2 of them, $300 ea. off eBay.

[David Hess]

I just went though my index of appnotes and articles: all build-your-own ones with one knob controlling two outputs use a double pot or related, so they are not tracking each other. NS AN127 gives a tracking regulator using OPA's (LM143).

Most tracking supplies do not track each other including the example in National Semiconductor application note 127.  It like most has a second output configured as a gain of -1 inverting power amplifier.  Maybe call that half tracking?

In practice however the distinction is almost always irrelevant and even less so with bipolar supplies.  With multiple output supplies, it can be very important; some circuits will destructively fail if the higher supply voltage fall below the lower ones.  Sometimes protection diodes are included in the design to prevent this from happening.

[Neomys Sapiens]

Most tracking supplies do not track each other including the example in National Semiconductor application note 127.  It like most has a second output configured as a gain of -1 inverting power amplifier.  Maybe call that half tracking?

Concur on this. I did not take a look at the circuit. I also think that I saw another, more interesting circuit somewhere (which was, inexplicably, not in my index). A possibility that was not mentioned was using a power OPA for creating a virtual ground.

Also, it becomes obvious why all those circuits are not made up of 'regular' LPS, that is, CV/CC ones with a rectangular characteristic, as the control rule would become tricky.
The function described for the Oltronix B202 (and probably our PS503s) is one possibility (and maybe the one that makes the most sense). If we assume two PS tracking each other in CV mode with the same CC limit setting and the load drawing different currents out of the two, nothing happens as long as both stay below the CC limit. When one reaches the limit, forcing the other supply into CC mode too, it would try to force the same higher current into its load, maybe damaging it. Or should the function be implemented in such way, that the lower current is tracked? But that would be two tracking current sources only, or did I get this wrong?

[David Hess]

If we assume two PS tracking each other in CV mode with the same CC limit setting and the load drawing different currents out of the two, nothing happens as long as both stay below the CC limit. When one reaches the limit, forcing the other supply into CC mode too, it would try to force the same higher current into its load, maybe damaging it. Or should the function be implemented in such way, that the lower current is tracked? But that would be two tracking current sources only, or did I get this wrong?

I have never seen tracking applied to the output currents of a bipolar output.

With the Tektronix PS503A, each current limit only applies to its side and the output voltage of both sides track a reference voltage.  (1) So the two sides do not interact at all; it is just two power supplies permanently wired in series with one combined tracking voltage control, two individual voltage controls which set the maximum voltage of each side and the ratio between them, and two current controls.  When tracking mode is not enabled, it is the same as if the tracking voltage control was set to maximum.  (2) This way of operating may seem weird but it closely duplicates how some tracking regulators work in circuits.

In that EDN example where the outputs are balanced against each other, each sides tracks the other and current limiting on one side causes the voltage on that side to drop.  The other side now tracks the *voltage* of the side which is being current limited so it also drops but it still uses its own current limit.  The advantage of this over how the PS503A operates is that when the PS503A is used as a single 0 to 40 volt power supply, the lower current limit causes both outputs to drop simultaneously all the way down to 0 volts.  With a PS503A, one side drops down to 0 volts (actually -0.6 from the protection diode across that output) and then the other side only drops if its current limit is exceeded.  The disadvantage is that you might prefer only one output to lower its output voltage when current limiting occurs.

Which way is better?  The balanced method will be better for maximizing power supply rejection in some circuits making it better as part of a built in power supply.  The tracking method used by the PS503A is probably better for a bench power supply used for development and other work except in the case where it is being used as a single 0 to 40 volt output power supply.

(1) This allows the PS503A to be remote programmed.

(2) The way this works is that with tracking off, each output of the bipolar supply is set independently.  Now when tracking is turned on, both outputs may be adjusted from 0 to the previous output voltage with the single tracking control and they always maintain the same ratio.  This allows low voltage circuits to be powered without risk of applying a higher voltage during adjustment if only the tracking control is used.