Another Noob PSU Design Needing Review, Please

[t1d]

This is a draft of a circuit that I propose to build. It is based on an old National Instruments data sheet design. I found it on the web. The author has made some changes to it. Specifically, adding a negative rail, so that the output can reach a true zero volts.

My design goal is up to 30VDC and up to 10A (I am using a more powerful BJT) for the output. Of course, the typical 30VDC/3A criteria are fine, if we make adjustments that end up with that output.

I am sure that there are better more modern designs, but this will do for my purposes. I would very much appreciate your thoughts. I do no want to go with a switch-mode solution.

Here is a specific question:
The Power BJT is rather expensive. I have lots of MOSFETs on hand. Might I use a MOSFET, instead? If so, can it be dropped directly in place of the BJT? If it would need additional circuitry, what would that be?

I eagerly await your replies. Thank you for your help!

[t1d]

Here is the author's link:
http://circuit-zone.com/index.php?electronic_project=835

Here is the author's original design:

[Kleinstein]

There are quite a few weak points with this design:
the voltage setting uses a fracion of the negative of -5 V to compensate the 1.2 V min voltage of the LM317. So this part also has to be quite stable.
The pots are used in a way that a bad contact would cause a higher output voltage, which is not very good. Chances are the OP could see too high a supply voltage with the higher voltage version.
A tantalum cap at the output (especially in the 30 V version - as it would be hard to use a 3 fold higher voltage rating here) is prone to catastrophic failure.
 
It is unlikely the regulator would be stable against oscillation for a low ESR capacitove load.

The given transistor may be large enough for about 2-3 A not really more when stying inside the SOA.

[t1d]

All good stuff, Klienstein. Through the years, you have always proven to be most helpful and reliable, to me. Thank you!

Yes, I would need to verify that all of the components can handle the higher volts and amps created by my component substitutions.
- The -5V supply would be from a LM7905, which I would think would be quite stable.
- What would be the solution for the risk of a bad pot contact?
- Yes, the op amp voltage needs to be verified.
- Yes, the Tantalum cap voltage needs to be verified. If it can not be sourced at that high of a voltage, a typical work-around seems to be to use a common ceramic cap. Or, an e-cap at x10 rating, IIRC.
- What is the solution to harden the circuit against low ESR capacitive load oscillation? Terminate with a 50 ohm resistor on the output? That seems problematic.
- As said, if the transistor is only good up to 3 amps, that is fine. Final testing will sort that out. Would using a MOSFET instead be of any improvement? If so, what is needed to incorporate it?

[iMo]

Also your voltages are at max or even outside the TL071 datasheet values.
The absolute max power voltage is 36V (yours is 38V) and the abs max input voltage is Vcc- + 36V (yours is 35V).

[Kleinstein]

The tantalum cap problem is relatively easy to resolve, just use a low ESR electrolytic cap, maybe in combination with a 100 nF or so ceramic.

For the stability, one would have to test - the specs for the LM317 are not that detaild that one could calulate this upfront. Not all LM317 are the same too - so a Ti part can be different in the details from a ON part. Usually it help to have some capacitance with significant ESR at the output. It may be enough or not.

The way the circuit is made, by using the pot's as variable resistors makes it hard to avoid problems with an open viper. I see no easy fix for this.
A LM7905 for the -5 V is OK. That should be stable enough for the rest of the ciruit.

For the power handling capability there is no big difference between BJTs and MOSFETs. Both don't like that much power. Especially at more than some 20 V one also needs to look at the SOA curve, bit just the simple power rating. The nominal max power is for very good cooling and one is usually better off with about 1/2 of that, so the heat senk can be at some 60 C instead of 25 C case temperature as assumed for the power rating. With so much power, the usual way around is either using a SMPS or using multiple transformer taps. Both are not simple, but at more than 100W this may be worth the effort.
For higher power it usually needs multiple transistors in parallel with resistors at the emitter to ensure current sharing.

For the input voltage is also depends on the source. With a transformer and filter cap there will usually be some extra ripple on top. One will than need more than 33 V (at least as wrst case) at the input to get 30 V out. This adds to the needs for the OP and SOA.

[t1d]

Thanks, folks, for your wonderful help.

Most of the issues appear to be from trying to get too much power out of the unit. I will gladly downgrade the specs to 30VDC/3A. Otherwise, I think Kleinstein sees this problem...

The way the circuit is made, by using the pot's as variable resistors makes it hard to avoid problems with an open viper. I see no easy fix for this.

I am not sure what Klienstein means by an "open viper." My guess is that he means some unforeseeable major problem? Maybe someone knows how to harden the circuit against such things?

Thanks for all the good help. Keep it coming. You are teaching much!

[t1d]

It makes sense to try to use whatever components that I have on hand...

On my thread regarding terminating the null set pins on a TL741, Terry Bites suggested bypassing op amps that use a null set function and just go with a precision op amp. That makes great sense.

I have three precision op amps in stock... All Analog Devices brand: 1) ADA4075 2) ADA4004 3) AD8675. Numbers 2 and 3 list 36 volts in there brief descriptions. I will have to read the data sheet on all three, as I have not used them before and I am not familiar with them. EDIT: 4) AD8671

I have lots of MOSFETs. Are there any particular model numbers that are commonly used in bench PSUs?

Cheers!

[Electro Fan]

I am not sure what Klienstein means by an "open viper." My guess is that he means some unforeseeable major problem?

Might have just meant “open wiper”?

[xavier60]

During normal operation, the LM317's ADJ pin is 1.2V below the OUT pin.
With VR2 at max setting, half of the 1.2V will be applied across the opamp's input pins.
A precession opamp is not needed.

[t1d]

During normal operation, the LM317's ADJ pin is 1.2V below the OUT pin.
With VR2 at max setting, half of the 1.2V will be applied across the opamp's input pins.
A precession opamp is not needed.

Good point!

[floobydust]

TL071 exhibits phase-reversal and non-linear gain, you don't want those surprises in a power supply lol.

[xavier60]

Im not certain what a TL741 is. If equivalent to the LM741, it would be a reasonable choice. From my own observations, input common mode range is about 200mV short of its supply voltage.
Current regulation doesn't imply short-circuit protection. There will be a brief and large current spike when the output is shorted.
Difficult to know if it should be dealt with or not. 

Extra: I overlooked the fact that the -5V rail adds to the total opamp's supply voltage. This reduces the margin between supplying enough voltage and what's not safe. It would need to be close to 31V to the + supply pin.

[iMo]

In the potentiometer (or a trimmer) a "wiper" slides over the resistive surface.
It could happen the wiper looses its contact with the surface while sliding on the surface, or because oxidation, or mechanical failure, etc.
There are designs where the output voltage (or current limit) drops down to MIN voltage, or, there are designs where the output voltage jumps to MAX voltage in case the wiper's contact with the surface is lost ("open wiper").
The better is the solution where output voltage drops to MIN voltage for obvious reasons.

[iMo]

High voltage opamps from ADI TABLE..

[t1d]

TL071 exhibits phase-reversal and non-linear gain, you don't want those surprises in a power supply lol.

The history of this design is that it comes from an old National Instruments LM317 data sheet. The attached picture shows they originally used a LM301A op amp. IIRC, it is still included in the TI LM317 data sheet.

I listed some op amps that I have in stock that look like they can handle 36 volts. With reducing the goal to 30V output, maybe one will work. But, I need to research them fully. I am still working on that.

The first step is to learn how to determine their Safe Operation Area. I know what the term means, but not how it is determined for a specific component.

Notice that the original did not bother with the 0 volt goal/-5V rail.

[t1d]

Im not certain what a TL741 is. If equivalent to the LM741, it would be a reasonable choice. From my own observations, input common mode range is about 200mV short of its supply voltage.
Current regulation doesn't imply short-circuit protection. There will be a brief and large current spike when the output is shorted.
Difficult to know if it should be dealt with or not. 

Extra: I overlooked the fact that the -5V rail adds to the total opamp's supply voltage. This reduces the margin between supplying enough voltage and what's not safe. It would need to be close to 31V to the + supply pin.

As I recall, the TL is supposed to be "comparable" to the LM, but the devil is in the details. And, as said, I am looking at the other op amps I have in stock. With just a quick look, the AD8675 looks like it might do. Data sheet below.

[t1d]

In the potentiometer (or a trimmer) a "wiper" slides over the resistive surface.
It could happen the wiper looses its contact with the surface while sliding on the surface, or because oxidation, or mechanical failure, etc.
There are designs where the output voltage (or current limit) drops down to MIN voltage, or, there are designs where the output voltage jumps to MAX voltage in case the wiper's contact with the surface is lost ("open wiper").
The better is the solution where output voltage drops to MIN voltage for obvious reasons.

That is a really good explanation! Do you have a link that shows the two approaches, or do you have any terms to search for them?

[iMo]

What does the D2 do on the pin8 of the LM301 in the original schematics?

Where is the V+ of the LM301 in the original schematics wired to?

Still I think operating an opamp at "absolute max voltage" specs is not a good idea..

Open wiper: nope, I do not have a schematics comparison handy..

[Kleinstein]

TL071 exhibits phase-reversal and non-linear gain, you don't want those surprises in a power supply lol.

The phase reversal happens at the low end and is thus not a problem here, as there is a -5 V supply. So the OPs inputs should hardly go below some 3.8 V from the OPs neg. supply.

The AD8675 would not work well, because the input common mode range does not include the positive supply. So it would need a relatively high supply (can still work with somthing like seprate filtering and voltage limit for the OPs supply).  If ordering new, the TL071H or OPA991 could be an option (with 40 V maximum supply voltage).

As shown, at the max setting the voltage across the shunt would be 1.2 V and thus relatively high, but to limit the heat there it may be advisable to change to Pot more to 50 K or less, so that the max would be more like 0.6 V or maybe even 0.3 V.  The current limit is not high precision anyway. So from the precision the TL071 should be good enough, unless one uses a very low (< 100mV) drop on the shunt.

[perieanuo]

not sure if D3 will ever work properly, but i haven't put the pen on paper to check if he can light correctly with current draw from adj pin, maybe it's safer to put that missing resistor
anyway i don't like that configuration with led entering conduction linear-like, i like the on/off approach, either the current protection (limit) is ON and LED is ON, either it's off. with your schematic i'm not sure it works that way

[t1d]

Extra: I overlooked the fact that the -5V rail adds to the total opamp's supply voltage. This reduces the margin between supplying enough voltage and what's not safe. It would need to be close to 31V to the + supply pin.

That is a really good point that must be included in the supply consideration, too... Meaning the transformer, etc. I am sure that I have some transformer with two, or more, secondaries that would work to supply the 33V and -5V rails.

As the negative rail only needs to output ~-1.25V, maybe its input could be reduced to -3.3V, or -2.5V. That would save a little headroom. Hmm... I need to see if I have any LM79xxs in that low of a voltage. I rather think not.

[xavier60]

Im not certain what a TL741 is. If equivalent to the LM741, it would be a reasonable choice. From my own observations, input common mode range is about 200mV short of its supply voltage.
Current regulation doesn't imply short-circuit protection. There will be a brief and large current spike when the output is shorted.
Difficult to know if it should be dealt with or not. 

Extra: I overlooked the fact that the -5V rail adds to the total opamp's supply voltage. This reduces the margin between supplying enough voltage and what's not safe. It would need to be close to 31V to the + supply pin.

As I recall, the TL is supposed to be "comparable" to the LM, but the devil is in the details. And, as said, I am looking at the other op amps I have in stock. With just a quick look, the AD8675 looks like it might do. Data sheet below.

A suitable opamp needs to be able to operate with its inputs at close to + supply rail voltage.
The "Input Voltage Range" spec doesn't look quite right.

Edit: Already mentioned by Kleinstein.

[iMo]

I would recommend to put the new schematics into the LTspice.. All parts are available.
Soon or later you will start to deal with questions around dynamic parameters (ie stability under various loads/settings, gains, phase margins, overshoots) and the simulation works pretty good as we know from past.

[Kleinstein]

D2 at the LM301 should speed up the recovery from saturation and thus the reaction the current limit somewhat. The higher speed of the tl071 would about make up for this.


If powered from a transformer, and this with some variations and hum on the main supply, it would be a good idea to have a seprate limit for the OPs supply. This could be a simple resistor +33 (or 35) V zener + emitter follower type circuit. So the OP would not see the variations and hum on top the main supply.

A simulation is definitely a good idea. However it depends on the details of the LM317 model, so the higher speed details in the response may be a bit off. Sill it is better than the calculation one can do in the head.