0-30V, 0-3A PSU - Audioguru's version

[rstofer]

AFAIK the board from the kit does not include the 10 V zener to reduce the supply to the not so critical OPs. So it would need better OP for all 3. The difficulty here is getting DIP8 OPs that can stand a high voltage (e.g. 40..44 V), without paying a premium price (like for OP27). Plain old 741 could work - though a little slow (especially slew rate for current limit -  so quite some delay before CC mode sets in).

So, assuming a 24 volt transformer, so about 35 volt max input voltage, what are the options for common, cheap, through-hole op-amps?

Given that you will be designing a new board, why limit things to through-hole?  Assuming that SMD components can meet the output current and dissipation requirements.

As to high dollar op amps, what difference does it make for a one-off power supply?  $1 versus $10 is pretty insignificant when compared to the total project cost.  OTOH, if the goal is to commercialize the design, BOM costs matter.

[Kleinstein]

For a one off unit cheap OPs are not that important, but even if not looking for the price the choice of easily available DIP8 OPs with at least a 40-44 V supply is not that large. Ordering from Mouser or similar could be a hassle from some countries.
Many audio OPs like NE5534, OPA604 would need more current from the negative supply. So possible but not not really attractive.
The right version of the 741 would be an option - though slow. The OP27 could also work - though high in supply current. The mentioned MC34071 can be tricky to get in DIP. So an SO8 - DIP Adapter may be needed.
The TLE2021 might also be an option - though also difficult to get in dip.

Starting with 24 V AC, the positive supply may reach some 35 V worst case.  In addition one needs a negative supply for many OPs - especially the TL081 in the kit really needs some -3 V headroom on the negative side for the input common mode. Even with an extra diode to limit it, this could need some -4 V.  Single supply OPs could work with less negative supply. So if single supply one could get away with a slightly  lower supply rating.

With 24 V AC the circuit would likely not give up to 30 V out - worst case it's more like 26 V depending on the OP. So even with a 44 V rated OP one would be near the edge with 28 V AC that may be needed to really get 30 V out.

With a slightly modified circuit one could a a few small improvements, already discussed in other threads:
1) limit the worst case voltage with something like a Zener. So less headroom needed for high mains
2) Add extra filtering for the OP, so that it would see less drop due to ripple. This can reduce the dropout voltage and thus alow a lower transformer voltage. It could also safe a little on the filter cap - which is still too small for 3 A.
3) reduce the voltage for the other 2 OPs (they only need some 15 V)
   this can also speed up the onset of CC mode
4) add back the protection against a peak on turn on / off
5) maybe add some fixed fast current limit as a fallback for transients
6) fix the shunt layout
7) Limit on how far negative the set voltage can be drawn done by the CC mode regulator: just a diode to ground as a hard clamp.

Even the modified circuit would like an OP that can stand some 35 or maybe 40 V depending on the common mode and output range.

[soldar]

Given that you will be designing a new board, why limit things to through-hole?  Assuming that SMD components can meet the output current and dissipation requirements.

As to high dollar op amps, what difference does it make for a one-off power supply?  $1 versus $10 is pretty insignificant when compared to the total project cost.  OTOH, if the goal is to commercialize the design, BOM costs matter.

Through-hole is much easier for beginners, repairs, etc.

You are right that for a one-of build the cost of the op-amp is trivial but I am just thinking of making it a good design in every way and that is one aspect.

I am looking at a commercially-bought bench PSU which I reverse engineered when it failed and it has a single LM723 which includes voltage reference, current limit, etc. This unit switches transformer windings to limit transistor dissipation but I am thinking I could copy the schematic removing the transformer switching part.

Another thought is that secondary voltages which require very little current can be obtained using transformers from old wall warts. I have several crates of these. 

[floobydust]

These 0-30V 0-3A PSU threads always seem to fizzle out because the requirements are never nailed down.
People want it all - low cost, reliable, high performance, easy build.

But this is unrealistic starting with a $5 kit that is based at the absolute max. ratings of many parts, as a chinese design prone to smoke.
So right off the bat, people have to add parts and cost just to get the PSU reliable and meeting its original specs, capable of running more than a few seconds at rated output.

Audioguru did fix things by adding an additional pass transistor and lowering Vcc for the op-amps.
I would add power-up/down overshoot & backfeed protection and limit Q2's dissipation to make it dependable, like the odd Mastech and Manson models.

For high performance I use an auxiliary transformer and regulated supply for the op-amps.
To keep things cheap, change to rail-rail op-amps or one's with PNP diff amp. and consider a Sziklai NPN power transistor configuration.

[not1xor1]

This a followup to https://www.eevblog.com/forum/beginners/diy-0-30v-0-3a-again/

It seems that project was not completed successfully.

I am attaching from that thread what I believe is the final design.

Is that considered to be the correct and final design and confirmed to work correctly?

If so, is it possible to buy the PCB?

In this thread I would like to come to a final, confirmed working, design, including specs, schematic, list of parts and PCB design.

I am analyzing it. I see the op-amp on the right does the voltage regulation and it looks quite simple and straightforward.

The op-amp and 5.6 volt zener in the center provide the voltage reference. I think it would be quite possible to simplify that and get rid of the op-amp and provide a simpler voltage reference. It might not be so exactly precise but I think it would be more than enough for most applications.

The op-amp on the left does the current limitation control and Q3 lights the LED when current limit is active.

I do not see why the 10V zener that lowers the supply voltage to U3 is needed. Maybe it can be omitted?


Maybe I would add an "ON" LED, just to show voltage at the main capacitor.

I think that the only possible comment for this PSU design is: 
I wonder how many ms can it survive Bob Pease's file test 

[soldar]

These 0-30V 0-3A PSU threads always seem to fizzle out because the requirements are never nailed down.
People want it all - low cost, reliable, high performance, easy build.

But this is unrealistic starting with a $5 kit that is based at the absolute max. ratings of many parts, as a chinese design prone to smoke.
So right off the bat, people have to add parts and cost just to get the PSU reliable and meeting its original specs, capable of running more than a few seconds at rated output.

Audioguru did fix things by adding an additional pass transistor and lowering Vcc for the op-amps.
I would add power-up/down overshoot & backfeed protection and limit Q2's dissipation to make it dependable, like the odd Mastech and Manson models.

For high performance I use an auxiliary transformer and regulated supply for the op-amps.
To keep things cheap, change to rail-rail op-amps or one's with PNP diff amp. and consider a Sziklai NPN power transistor configuration.

Yes, I have come to realize it's the "design by committee" phenomenon.  In the end each one has to take the initial design and tweak it for his own needs and desires.

I am attaching the schematic as I was redoing it. No great changes, just re-organizing it to make it simpler to analyze and understand.  The more I think about it the more I think I won't go ahead with the project. The center op-amp which provides the reference voltage can easily be eliminated and replaced with a few discrete components. The right hand op-amp which provides the voltage control can also very easily be replaced with a few discrete components.  The current sensing resistor dissipates a bit too much heat when it can be replaced by a much lower value and sensing with a current mirror as I have done in the past. Still, it is fun analyzing how it works.

Many people seem to build a circuit they do not understand, just to save some money, and then go crazy when it does not work, sometimes by bad design but most often by mistakes made when assembling it. For me most of the fun is analyzing and improving circuits.

[soldar]

I am looking at a bench adjustable power supply which I built some twenty years ago or more. It is very similar in concept but only uses discrete transistors. Three transistors and few other components do the voltage regulation and reference voltage which takes two op-amps in the circuit we are considering. Now, maybe some characteristics are better with op-amps but you cannot compete with the simplicity of the discrete transistor schematic.

[floobydust]

It's a good building block for voltage control. The pass transistor config (double or triple Darlington/Sziklai) is what I think is best.

But... discrete transistor solutions fizzle out doing the adjustable current-limit feature. It's always difficult to measure mV without drift and linearity problems, more so on the high side (if you use the sense (emitter) resistors). I haven't seen any good variable current-limit solutions with discretes or even the LM723 PSU's.

[soldar]

It's a good building block for voltage control. The pass transistor config (double or triple Darlington/Sziklai) is what I think is best.

But... discrete transistor solutions fizzle out doing the adjustable current-limit feature. It's always difficult to measure mV without drift and linearity problems, more so on the high side (if you use the sense (emitter) resistors). I haven't seen any good variable current-limit solutions with discretes or even the LM723 PSU's.

hmmm... I have a commercially built 30V - 5 A variable power supply which uses LM723 and the current limiting works fine.  I also have several self-designed and built units with discrete components and they also work fine. Where do you find the problem?

[floobydust]

LM723 has a current limiter for short-circuit protection, it's not adjustable with a potentiometer because it's a simple transistor E-B junction. Changing sense-resistors with a rotary switch is something my 1978 Radio Shack lab dual bench supply does lol.
This is where the antique MC1466L does better, but it needs a dedicated floating rail. It's almost discrete - two diff amps and some glue.

If you have seen a way for PSU adjustable current-limiting without a negative rail or dedicated floating rail, let me know.

On the chinese kit, I hate the -ve rail and the need for the V ctrl op-amp to run at the top rail while the I ctrl op-amp runs on the low-side. I can find away around it all, but haven't built it to prove it works.

So far, the only low cost idea is a low-side sense resistor and op-amp with PNP diff amp inputs, like LM358 (too slow) or MC34071 looks ok, for inputs that can go to ground. This is staying with cheap op-amps, not the ones worth twice the entire kit price.

[soldar]

LM723 has a current limiter for short-circuit protection, it's not adjustable with a potentiometer because it's a simple transistor E-B junction. Changing sense-resistors with a rotary switch is something my 1978 Radio Shack lab dual bench supply does lol.

I see what you mean.  Well, yes, it is a limitation and I have used different work-arounds. As you say, a separate floating rail is one way. That is what commercial Chinese PS does.

What I usually do is have a fixed resistor and a pot in parallel so it allows a certain range of adjustment but, yes, it is not the best. I have never thought of switching resistors but it is not a bad idea. The pot in parallel allows continuous adjustment.

What I have used with good results is a current mirror configuration. That requires a very low resistor value to produce a significant signal so that the heat dissipation in the resistor is kept low.

[soldar]

LM723 has a current limiter for short-circuit protection, it's not adjustable with a potentiometer because it's a simple transistor E-B junction. Changing sense-resistors with a rotary switch is something my 1978 Radio Shack lab dual bench supply does lol.

Come to think about it, switching resistors seems like a very bad idea because you are left with no resistor while switching.

Now that I think about it, many years ago I did something like this but it only had two ranges, high and low, so the switch only connected and disconnected a second resistor in parallel. The current indicating galvanometer instrument was connected to the same resistor and so the power supply had two ranges, 2A and 0.2A. In the low 200 mA range the instrument always indicated just under 10 mA, even with no load connected, because of some internal current in the PSU.

So I am coming to the conclusion that the basic voltage regulation can be done with discrete components or with an LM723 or other op-amp and the current regulation can be done with discrete components but it means using several transistors and so this part may be done with an op-amp to make it with fewer parts. The op-amp for the reference voltage I think can be eliminated.

The advantage of using an LM723 is that it includes the voltage reference.

Regarding using secondary floating voltages, I understand the desire to avoid them in commercial products where cost is a consideration but for one-of home-brews I have several crates filled with wall warts that I would be happy to put to some use.

[not1xor1]

LM723 has a current limiter for short-circuit protection, it's not adjustable with a potentiometer because it's a simple transistor E-B junction. Changing sense-resistors with a rotary switch is something my 1978 Radio Shack lab dual bench supply does lol.

Come to think about it, switching resistors seems like a very bad idea because you are left with no resistor while switching.

is it so hard to conceive a resistor permanently connected and just switch other resistor in parallel to that?

One might even use low Rdson MOSFETs for switching a binary network of resistors.
In any case that kind of current limit depends both on Vbe drift, temperature and the output voltage so a few milliohms of drift would be the least of the problems.

[iMo]

This is something which works nice at least in the simulation. Experienced users may try.
The R5 is actually an 1k pot/trimmer where you can set the current limit from 20mA to full in pretty linear way.
The zener and 1k resistor is not related to the current limit circuitry (you may remove it).
The Q3 feeds a constant current (around 800uA) into the base of the internal CL transistor. That current is being "balanced" against the current set by the R5, sensing the 0.56ohm shunt resistor.
The 1n4148 with its series resistor have been selected such the scale of the R5 pot is somehow "linear".
EDIT: this is from my simulation in a different thread:

Code: [Select]

R5 (w/ 100ohm in series)[ohm]   I_shunt_CL (0.22ohm)
1000     0.82mA
900     22mA
800     157mA
700     315mA
600     503mA
500     732mA
400     1.02A
300     1.37A
200     1.75A
100     2.14A
0       2.53A

[Kleinstein]

Switching the current limiting shunt is not that simple, but can be done.  Just a 2 nd resistor in parallel would add the switch / MOSFET resistance to the lower shunt - this can be a problem with a 0.1 Ohms or similar shunt.  Given that in a 723 type current limit the accuracy is poor anyway, a MOSFET would be acceptable.

The better way would be to have the 2 shunts in series and have a fixed path to read the voltage. The switch would change the path of the current, going one way or the other.  Ideally one would need a make before brake type switch. If not, a diode could help.

In the original circuit the OP for the reference can be eliminated, but a cheap OP (LM358 class)  is not a bad solution to keep the current through the reference constant. This can help as the current through the reference would also end up to go through the shunt. The OPs for the reference and current regulation don't need a high supply and can be relatively simple types - down to the LM358 level, though the current control may like a more precise and maybe slightly faster one. Its already way to much text about possibly saving a 20 cent OP.

[rstofer]

is it so hard to conceive a resistor permanently connected and just switch other resistor in parallel to that?

It might be easier to use two or more very small resistors in series and then just short them out to change ranges.  No make-before-break required.  Of course, the switch will need an adequate current rating.

[Audioguru]

I think if you are going to make the old Greek kit or a Chinese copy then the proper opamp that is still available should be used and its input offset adjustment should be correct like this:

I also got it wrong, this correction is as per the datasheet:

[jaycee]

Having recently built a 0-18V 0-1A bench supply, and plans to build a 0-30V 0-3A, here's my five cents on the matter.

First off, powering the opamps from the same supply as the raw supply voltage is fraught with issues. The best way to alleviate these problems is dont do that - most linear bench supplies have a separate supply voltages for the opamp, with the midpoint tap connected to the main regulator's output. This is a bit of a headf*** at first, but it works extremely well - i believe it was HP/Harrison Labs who came up with this "Regulator within a regulator" topology.

This may sound complicated because you need the extra transformer tappings, but you can always use a second small transformer, it doesnt need to be powerful, so some tiny 5VA thing from an old digital bedside clock works very well. In fact that is exactly what I used when I was prototyping my supply... and I could have built the end unit like that too, I just happened to find a suitable transformer core that I could re-wind with custom secondary taps.

Tap switching is well worthwhile doing to vastly cut regulator dissipation. Use a 2x18V transformer for 0-30V (remember it sags vastly under load so you need quite a bit of overhead), connect the windings so you have 0-18-36, and you can easily arrange a circuit which can switch between the 18V and 36V AC taps.

I've attached the main schematic of my power supply. It works well, doesnt oscillate or do stupid things under load, and uses only bog standard TL082 opamps. I haven't shown the front panel board, but it is only metering and display. The controls are 10K linear pots wired to CN2 with their top wired to VREF, the bottom wired to AGND, and the wipers to V_FINE, V_COARSE, and I_CONT respectively.

[jaycee]

Here's a few pictures I took during prototyping it. As you can see, there's a tiny transformer with 9-0-9 secondaries. This provides enough for +/- 5V rails, and the raw rectified voltage is good enough to switch a relay with.

[floobydust]

Using separate (floating) rails for the op-amps is best, it gets rid of almost all of the problems but adds cost and parts: additional small power transformer, fuse, bridge, filter caps, voltage regulators etc. The extra $10 might not be desirable, on a $5 kit.
It also makes new issues:
When you switch power on/off, the op-amps' rails and reference may or may not collapse/rise before the main DC voltage. Power-off under a heavy load, the main DC falls before the opamps & reference, but under a light load the main DC can stay up after the op-amps' rails first dropped off. Power on, you have to scope the rails rise and make sure the op-amps' rails come up ahead of main DC.
Either way, there is an unpredictable output overcurrent/overvoltage spike to look after.

TL0xx series JFET input op-amps exhibit phase-reversal at low common-mode voltages, below 3-4V above the negative supply.
These $0.10 op-amps do a "surprise" which can smoke your load with overvoltage or overcurrent when you weren't looking.
So I'm not sure about the mod lowering the -ve rail from -5.1V to -1.4V and deleting the UVLO, not introducing a new problem. edit: not relevant with mod's different op-amp TLE2141

[jaycee]

I dont have any issues with undervoltage on my build, but... I do have the OUTPUT_ON connected to a microcontroller on the front panel, which will drive the output off when power sinks too low such that it's brownout detector puts the processor into reset. It would be trivial to make OUTPUT_ON controlled by an "AC loss" detect circuit too.

Phase reversal also shouldnt be a problem here, as it's unlikely the opamps will ever see anything near their rails.. however I used TL082 because I had them... I suspect any opamp would work just fine, though the compensation might need adjustment. I think I actually prototyped it with a 4558 first as I had it to hand.

The cost of the additional small power transformer (not a cost at all if you can wind your own transformer) + associated bridge diodes, filter caps and regulators are small when the end result is a much better power supply.

[Audioguru]

TL0xx series JFET input op-amps exhibit phase-reversal at low common-mode voltages, below 3-4V above the negative supply.
These $0.10 op-amps do a "surprise" which can smoke your load with overvoltage or overcurrent when you weren't looking.
So I'm not sure about the mod lowering the -ve rail from -5.1V to -1.4V and deleting the UVLO, not introducing a new problem.

The datasheet for the TL081 with phase reversal shows its input common mode range is 5V above its negative supply, the newer TLE2141 without any phase reversal has its range all the way down to and including its negative supply.
One TL081CP at Digikey costs $0.53US today but in China a copy or reject might cost $0.10US. The TLE2141CP must be ordered at a minimum of 400 at Digikey but they sell one TLE2141IP at $2.09US today.
One TLE2141 opamp in the 0V-30V circuit needs a -1V negative supply only because its output must go a -0.7V doped voltage drop less than the circuit's 0V when current limiting a shorted circuit output. Then its positive supply is safe to +42.7V.

[floobydust]

fixed - I missed the change to TLE2141 in OP for your mods.

Trying to improve this $5 chinese kit- without doubling its costs, is a good challenge. There is no other kit out there for beginners, aside from LM317 kits but no adjustable current-limiting feature in them.

This PSU kit uses no op-amp IC's, just four TL431's https://www.eevblog.com/forum/beginners/cheap-chinese-adjustable-lab-power-supply-kit-on-ebay-(-3ag1-xiaolin-)/msg1412511/#msg1412511
Try to do adjustable current-limiting with a TL431, for fun as a design puzzle. Voltlog #155 found it has stability problems, CC oscillates.

[Audioguru]

"Cheap" usually means "poor quality".

[floobydust]

I didn't realize this chinese kit has been out there for 10 years!
1000's of postings across the Internet trying to improve it, or suggest throwing it in the garbage. It's some form of turd that keeps on smelling.
I find it sad one weak design can proliferate and the rest of the world cannot be organized well enough to fix it. Communism++

For OP, perhaps check CC mode. It looks like the -1.4V rail adds hum to CC mode which brought about a change to LM337T: Paul's DIY electronics blog is at revision 6 but 5 looked more complete, sch attached.

I found a higher rev (original mods) schematic with the shutdown transistor left in.

late edit: I found the original 0-30V 2mA-2A PSU circuit, Practical Electronics Nov. 1978.
Project on pg. 41 but big 12MB pdf:   https://americanradiohistory.com/Archive-Practical-Electronics/70s/Practical-Electronics-1978-10.pdf