0-70V, 0-5A Lab Power Supply Design

[H713]

Here's a version with the Vbe current limit circuit. R1 and R12 represent a 5K pot. Yes, there would be limiting resistors so that the operator can't disable current limiting. Diode D2 serves two purposes. Firstly, it protects the series pass transistors from an external high voltage. You laugh, but I've blown up multiple power supplies this way. It is also there so that I can get the full adjustment range without the need for a 6 ohm 200W resistor. If desired, I could add a "low current mode" by switching an additional diode in series with D2. That said, my current bench supply has a minimum current limit of 250 mA and I rarely wish it had a lower current limit range.

This won't be as accurate as the version with the op-amp, but it's quite rare that accurate current limiting is needed. Fast current limiting, on the other hand, is pretty handy.

[H713]

Interestingly, the Vbe current limiter suffered from similar instability issues to the op-amp based one. This (combined with the fact that a series RC network to ground from the base of Q6 stabilized it) leads me to believe that it is a stability issue with the series pass transistors. I seem to have shut them up with a 470 pF cap from the base to collector of Q6.

In testing, using a BD139 for Q7 showed relatively minimal thermal drift and generally performed well. Thermal drift over 2 minutes was good enough, less than 15 mA. The main issue that showed up is very poor ripple rejection while current limiting, even at lower frequencies (120 Hz). This may not be an issue, however, as I intend to use a capacitance multiplier in the filtering circuit anyways to achieve a relatively quiet DC bus without the need for 30,000 uF of filter capacitance, which takes up valuable chassis space (remember, this is a dual power supply in a 2u enclosure). Huge filter caps also have a big inrush current, which would require efforts to mitigate in order to avoid welding relay contacts in the tap switching circuit.

I intend to do a bit more testing on the op-amp based current limiting circuit tomorrow, then make a final decision. Beyond that, things are looking good and I am planning to start my PCB layout in the next few days.

[magic]

You will probably get better PSRR by replacing the Darlington's base pullup resistor with an active source.
Also with higher loop gain, did you add that resistor as David told you?

[H713]

Yes, I added that quite a while ago, just forgot to include it in the Spice schematic.

[b_force]

@b_force, the problem is a lack of adjustable current limiting. When doing initial tests on a circuit that I don't necessarily trust, either because it is new or intermittently failing. Once I am confident that it is behaving well,  I may turn that maximum current up to something like 3A- just so I don't vaporize PCB traces if something stupid happens.

Just add a current limiter?
That's not so hard, quite a straight forward circuit actually.

[Pawelr98]

Also, what is not shown in the schematic yet is a diode-capacitor network to feed the control circuitry, which drastically improves the ripple rejection.

Such network (control circuitry capacitor) as shown on the drawn schematic ?

Fairly simple way to cut off the ripple from the main tank capacitor.
Using shottky diode may be better as long as there's a 30-40V zener diode in parallel to protect it from accidental reverse breakdown during shutdown or when taps get switched (main capacitor bank going much lower than the control circuitry capacitor). Most common diodes are rated for 40-45V reverse voltage.

However there still will be some 60Hz ripple as the diodes in the main bridge are not equal as they have slightly different voltage drops.
It's not going to be much but any RC filter for the control circuitry supply should be tuned for <60Hz rejection.
Another small reason to go with boosted voltage as there's headroom for filters on the control circuitry side.
If active current source is to be used the boosted voltage will be even more handy.

As for current limiting using NPN transistor.
Put the BE juction directly across the current shunt.
Remember about BE junction beeing low voltage zener which can conduct in reverse.
When applied across the diode it won't be protected in any way.

In this case the current limit will be dependent on not only the temperature but also the current.
Diode drop will change in roughly 0.5V to 1V range which is quite a lot.
Extra source of error is the collector-base leakage current (disconecting the base pretty much made the transistor conduct all the time, shutting down the supply for good).
For silicon this is fairly low but with germanium I had far much better results when adjusting the potential of the emitter instead of base.
Far less dependent on the voltage put across the current limiting transistor.

As for current limit adjustment.
Either have the main shunt made out of few smaller shunts in series and switch the emitter (base will generally work too, but switch resistance should be low to minimize the influence of collector-base leakage current) between different points using CD4051 (digital control) or some mechanical switch.
That or add schottky diodes in series with the emitter.
Either one works.
Resistive dividers over the main shunts are, even with silicone transistors, rather tricky.
My bench power supply uses 100ohm potentiometer over the main output shunt and I can easily tell it's nonlinear.
Classic 723 circuit with default internal current limiting transistor used.

[duak]

The question was asked if there was a simple solution to the temperature dependancy of a VBE current limiter.  Some older Lambda linear supplies use a custom Lambda designed chip with an on board heater to stabilize the zener reference (and the other bits too).  I thought that an LM35 temperature sensor and heater transistor thermally connected to the current limit transistor could hold themselves at, say 75 C.  If done properly, there should be at least a 10X improvmenet in the stability of the current limit.  If the VBE sense transistor & heater were near the pass transistors and received some of the heat, and their case temeratures went above 75 C, it would naturally reduce the current limit.

By the way, I built my first bench supply in 1974 and used a 723 regulator with a modified VBE limiter that lit an LED when in current limit.  The stability wasn't half bad at low currents, since things didn't get very hot.  At high temperatures, there was a natural reduction in the current limit that was a bit of protection.

The next thought I had was to make the limiter less thermally sensitive by using another transistor to balance the VBE variation.  The attached circuit shows the basic VBE limiter using a differential pair.  Note, I run linux and don't have any spiffy ECAD packages so I used Oregano which is somewhat limited.  Hopefully, the schematic is clear enough.  I could provide the .oregano file and perhaps a SPICE file but I can't promise they'd be useable.

Q8 is the basic VBE current limiter sense transisitor.  Its base is driven by the current sense resistance voltage.  Due to superposition the voltages on the ballast resistors to combine to give the total load current.  It works for any number of sense resistors as long as they and the combining resistors are the same values.  The base of Q3 is driven by a variable voltage corresponding to the current limit setting.  The negative end of the current limit reference is connected to the bottom end of the current sense resistance which also happens to be the output voltage Vout+.  When the voltage on the base of Q8 is more positive than that of Q3, ie. reached the current limit threshold, Q8 will conduct more and draw current from the base of the pass section driver Q5 and reduce the output voltage.

Q4 sources current to pull up the input of the pass section and R4 must be adjusted to provide enough drive current to reach full output voltage at full load current.

Q2 limits the voltage applied to Q3 to reduce its power dissipation so that its temperature (and thus VBE) and hFE are more closely matched to Q8.  D6 is optional and increases Q3's VCE by one diode drop.   Q3 and Q8 work with VCEs on the order of 1 V and hFE is somewhat important.  I haven't simulated or tested this part of the circuit but I think it will work.

The rest of the circuit deals with the currents to operating the diff pair and current limit reference.  The diff amp emitter currents are taken to V-- with a current sink.  This sink must be able to draw more than the current source pulling up the drive can provide.  The current limit reference sink must likewise draw more current than the corresponding source can provide to prevent the output voltage from drifing upward when unloaded.

It looks more complicated than it actually is as much of it is needed for any supply.  The current source and sink for the current limit reference could be eliminated by using a floating supply like a DC-DC converter or just another transformer, rectifier and caps.  Op-amps could also be used but this circuit offers fast response as it really doesn't clamp and wind up like an op-amp.  BTW, D1 is probably not needed but in the worst case, it allows simple VBE limiting albeit at a higher current level.

[David Hess]

The question was asked if there was a simple solution to the temperature dependancy of a VBE current limiter.

Use a differential pair as shown in your example.  In the past, Tektronix commonly did it that way in their linear regulators and implemented fold-back current limiting as well.  Another advantage is that the sense voltage can be lower than Vbe.

In cost sensitive designs, the differential pair might be a diode and transistor instead of two transistors but there is hardly a reason to do that today.  There might be a suitable fast operational amplifier but the only advantage would be precision.

[H713]

The question was asked if there was a simple solution to the temperature dependancy of a VBE current limiter.

Use a differential pair as shown in your example.  In the past, Tektronix commonly did it that way in their linear regulators and implemented fold-back current limiting as well.  Another advantage is that the sense voltage can be lower than Vbe.

In cost sensitive designs, the differential pair might be a diode and transistor instead of two transistors but there is hardly a reason to do that today.  There might be a suitable fast operational amplifier but the only advantage would be precision.

This is essentially (but not exactly) what I was trying to do with the op-amp based current limiter.

[David Hess]

The question was asked if there was a simple solution to the temperature dependancy of a VBE current limiter.

Use a differential pair as shown in your example.  In the past, Tektronix commonly did it that way in their linear regulators and implemented fold-back current limiting as well.  Another advantage is that the sense voltage can be lower than Vbe.

In cost sensitive designs, the differential pair might be a diode and transistor instead of two transistors but there is hardly a reason to do that today.  There might be a suitable fast operational amplifier but the only advantage would be precision.

This is essentially (but not exactly) what I was trying to do with the op-amp based current limiter.

The difference is that the two transistor differential amplifier is so much faster but lacks the precision of an operational amplifier although performance is relative.  Those old Tektronix regulators with two differential amplifiers, one for voltage control and one for current limiting, were plenty precise under load and line variation.

In a power supply, I might implement fast current limiting with Vbe or differential pair as part of the pass element protection and a slower but very precise adjustable current limit using an operational amplifier.  I like using integrated regulators as power pass elements which results in exactly that except the fast current limiting is built in although with thermal and SOA protection.

Check out the schematics on old Tektronix oscilloscopes for examples.