Opinions and suggestions about this mV - mA accuracy power supply (10V - 1A)?

[Verdefluox]

Hi folks,
from some time, as title suggests, I'd like to build a precision power supply. I don't know if I'm aiming too high, but here are the specs I'd like to achieve:
- Range: 0 - 10V, 0 - 1 A
- 1 mV, 1 mA accuracy over their respective range
- CV/CC capability
- possibly built only with THT parts (I'm not that great yet in handling SMD)
- I'm doing this project also to learn more in depth about
- I'd like it programmable (Arduino, maybe)? This shouldn't regard this post, it's just to make the context clear

For the main board, the current design is based upon the Elektor precision PSU from 1982 block scheme (https://worldradiohistory.com/UK/Elektor/80s/Elektor-1982-12.pdf page 16 of the PDF). I'd like to ask about a general opinion about it so far (if the general architecture so far is okay etc.), and more in specific some of my choices and thoughts are the followings (not an exhaustive list):

- I went with this architecture as I've built the Elektor one and found it quite solid, also by doing some tests on LTSpice I found that it gave me less problems (also in terms of stability) in simulations with respect to a classical architecture (current control followed by a voltage control). Note that this may be strictly related to simulations issues.

- the discrete instrumentation amplifier can be substituted with a proper one, even though I don't know if its necessary given the accuracy needed. For this, I found the AD620, but never having used one I don't know if it's proper for the job.

- since I started building this from scratch (and by this I mean starting by simulating a simple op-amp voltage regulator), I encountered for the first time problems about stability, but I don't know if I'm correcting those problems in the right way (talking about the bypass capacitors and resistors near the op-amp).

- I also don't know if I'm following the right procedure about the buffer in the voltage control.
Initially, I went with a simple divider with two 10k, and then I added the buffer in order to ensure that in case of CC all the current actually goes to the load. Then, I substituted the divider with a direct resistor to the inv. input of U2. I based my choices on the simulation results, but I'm sure that after a lot of iterations and changes I'm missing some point and I don't know if this is right at all. Again, consider that this is the first time I'm actually evaluating these things, I know I could go with a simple voltage regulator but the purpose for me is also to learn something more than an off-the-shelf solution.

- the pic for now shows only the main board, didn't look almost at all about a voltage reference proper for the job, or the way to generate the reference voltages for the CC/CV controls (an ADC IC or maybe some discrete architecture, like, for example, an R-2R ladder). Some suggestions are welcomed, even off-the-shelf modules.

This is mostly a general overview for now. Obviously I may have missed something, and every suggestion is welcomed.
Thanks for all the help in case someone replies

COMPONENTS USED
- 2N2222 https://www.onsemi.com/pdf/datasheet/p2n2222a-d.pdf
- BD241C https://www.onsemi.com/pdf/datasheet/bd241c-d.pdf
- OP07 https://www.analog.com/media/en/technical-documentation/data-sheets/op07.pdf
- AD620 https://www.analog.com/media/en/technical-documentation/data-sheets/ad620.pdf

[blackdog]

Hi Verdefluox,
Due to the structure of your schematic this Power Supply will not work properly.
It is very slow, and by that I mean that both the current and also the voltage loop are very slow.

One of the rules of thumb that I always follow with linear Power Supply's is this, first of all the Power Supply has to be very gentle on itself.
This means that you will have to have a fast stable current loop and a additional peak current limiter.
These two functions keep the power transistor in one piece and thus your connected load as well.

Because of the amount of opamps you use, it becomes very difficult to keep this Power Supply stable.
The result is a much too slow reacting system.

You usually get the best results by using only one opamp per loop with a reasonably fast opamp, with low offset, bias, noise and generous Phase Margin, like the ADA4625-1.

Kind regard,
Bram

[Verdefluox]

Hi blackdog,
initially I didn't pose myself the problem about the op-amp performance, as the original Elektor project used simple 741 and worked without any problems. But as you say so, I suppose that you mean mainly referring to slew rate, settling time and phase margin. I actually happen to have around some fast op-amp, the AD826, which (after a quick check) should fit the purpouse (again, asking for a confirm, I don't know much about in-depth details of op-amps).

About the reduction of the number of op-amps:

- I was already thinking to substitute the instrumentation amplifier block with the AD620, that should already replace 3 OP07 with one job-optimized IC.

- I redrew a possible schematic (attached below), which uses only two AD826 (plus the instrumentation one). I also removed the buffer in the voltage control, as I thought that (since it's a DC power supply), once at regime I'd have a current only into the voltage control op-amp, which should be negligible anyway.
I also removed it because I thought that the input on the non inv. input of the CV has to be generated, and the previous configuration needed half of the output voltage (due to the divider). This can be problematic, as, for example, if I wanted 1.234V I would actually need an input of half that value, and didn't sound great for me, especially at very low values (close to 0V - I would have to enter the uV range).

Honestly, I don't know if I should reduce it more. I'm also wondering if this overall unusual configuration has disadvantages I don't see over a classical CC followed by a CV, or viceversa.
What do you think?

COMPONENTS
- AD826 https://www.analog.com/media/en/technical-documentation/data-sheets/ad826.pdf

[Kleinstein]

The more common form has the shunt to measure the current on the high side. This simplifies the current measurement quite a bit, but complicated the voltage loop. The problem with the shunt at the high side is that the amplifier needs quite good CMRR also for the higher frequencies that are relevant for the loop stability. This is not easy. Also the slew rate for the amplifiers there could be a limit. So the supply can have a problem with a sudden short and depending on which amplifier gives up first show a large signal oscillation even if stable for a small signal.

Multiple amplifiers are not ideal, but also not a clear no go. The 2 OPs left in the circuit are actually the ones that are not critical, as they are slowed down with the capacitor anyway. There the OP07 may be just fast enough, especially if the current signal already is amplified. The tricky part is the amplifier for the current signal, as here additional gain is wanted and this needs more GBW.
The version with an emitter follower output stage  has it a bit easier with voltage control, but is more tricky with current control.

[David Hess]

U3 is not doing anything useful.

The instrumentation amplifier made up of U4, U5, and U6 is not needed.  Reference current error amplifier U1 and its control voltage directly to the output current shunt.  The control voltage can be referenced to the current shunt using a voltage controlled current sink, which only requires one operational amplifier, and a pair of precision resistors.

[Verdefluox]

The instrumentation amplifier made up of U4, U5, and U6 is not needed.  Reference current error amplifier U1 and its control voltage directly to the output current shunt.

Sorry, I didn't understand if you mean that I don't need that discrete configuration (this was going to be an instrumentation amplifier in any case), or if you mean that I don't need it AT ALL (in which case I don't know which architecture you suggest).

The 2 OPs left in the circuit are actually the ones that are not critical, as they are slowed down with the capacitor anyway. There the OP07 may be just fast enough, especially if the current signal already is amplified. The tricky part is the amplifier for the current signal, as here additional gain is wanted and this needs more GBW.

Regarding the simulations, OP07 didn't give me any problem either, even though it's not that fast. Mind also that this project is not an ultra precision source, but still needs to be accurate. I guess I chould try, OP07 seems to be still sufficient for the job (again, just according to various simulations), plus I can correct its offset (not that its maximum one should interfere much - it's 75 uV according to the datasheet).

Some thoughts:
In the while I re-thought a bit about a general overview, and I assuming the OP07 is a go for controlling the transistors, the problem shifts towards how to compare the output parameters and also how to generate them on the + pins (and made me also reconsider various things):

READING ON - PINS
I'm considering two alternatives:
- go with this idea of using various kind of amplifiers and control it "classical"
- using an ADC

First idea:
FOR VOLTAGE: Modify the op-amp configuration in order to add a gain on that too (x10, x100)
FOR CURRENT: Go with the AD620 and work with different gains for different ranges (basically the same idea for voltage)
At least in the lower ranges (mV, mA) this should make output parameters easier to be generated with sufficient accuracy (since ex. 1 mV becomes 100 mV, more easily generated)
I'm wondering if it's better going with:
- a relay IC/various relays (or a switch in any case, so that it's controlled by an Arduino, for example) and some standard 1% resistor, and leave the rest to a calibration. Maybe I can add a trimmer in between too
- go with 0.1% resistors directly, but I don't know if at this level calibration is still needed (I'm not down to uV levels, but you know)

Second idea:
As stated before, using an ADC and going back to the controller.

GENERATING ON + PINS
For the higher ranges, (since I'd like to be able to set, for example, 5.000 and 5.001 differently -same idea for current-), I think the problem lies more on how to generate them on the + pins accurately. Also two ideas here:
- go with DACs and make it full digital
- use multi-turn pots from the buffered reference

First idea:
Go with simple DACs, and based on the fact that I think its a good idea be able to generate more or less one decade below my target accuracy (in other words: I want 1 mV -> being able to generate something with steps of around 100 uV), I have the following:
FOR VOLTAGE: Use a 16 bit DAC, as on 10 V range I would have every step on around 150 uV. I don't think I have some easily available choices (especially in through hole packages), so I think a module or an SMD part will have to fill the gap.
FOR CURRENT: Since here the value is more relaxed (1 mA in 1 A range), a lower bit DAC can still do, so I can also save some money.

Second idea:
Self-explained above, use multi-turn pots. Maybe also couple together a 1 turn and a 10 turn one (sort of coarse - fine).

[Benta]

Elektor Article: "404 Not Found".

[David Hess]

The instrumentation amplifier made up of U4, U5, and U6 is not needed.  Reference current error amplifier U1 and its control voltage directly to the output current shunt.

Sorry, I didn't understand if you mean that I don't need that discrete configuration (this was going to be an instrumentation amplifier in any case), or if you mean that I don't need it AT ALL (in which case I don't know which architecture you suggest).

The instrumentation or difference amplifier is not required.  The error amplifier can make the comparison between the current and the current setpoint *at the output voltage* which just becomes the common mode voltage.  Besides simplifying the frequency compensation, which replaces the error from the common mode rejection of the instrumentation amplifier, which needs to be trimmed, with the inherently high common mode rejection of the operational amplifier.

An example directly relevant to your design is shown below in the Tektronix PS501 schematic.  U70 is the current error amplifier which directly compares the voltage across the current shunt on the output to a fraction of the voltage across R65 in parallel with R66 which is also referenced to the output.

For an electronically adjustable design, R65 in parallel with R66 is fixed and an adjustable current is used to control the voltage across the resistance.

With a MOSFET output stage, the same circuit works on the drain side of the MOSFET since the gate current is negligible, however placing the current shunt in series with the output can have advantages for frequency compensation, although the PS501 does not take advantage of it by taking AC feedback before the current shunt.  This would be more important with a common emitter/source output stage like with a low dropout regulator.

One other thing to watch out for in your design is that the maximum differential input voltage of the OP07, or most precision operational amplifiers, is only a couple volts or less.  This should not be a problem if the series resistance at the inputs is sufficiently high, but it is something to watch out for.  It may be an issue because when one error amplifier is controlling the output, the other error amplifier is open loop.

[Verdefluox]

Elektor Article: "404 Not Found".

Should be fixed, but it's really easy to find it with a search.

An example directly relevant to your design is shown below in the Tektronix PS501 schematic.

Interesting design. I took some days to look at it (sorry, it takes me time plus I was a bit busy, hence the late response) and to some alternatives too. I'll try to simulate it and look it more in depth, maybe redrawing it in a clearer way.
Thing I cannot still solve with this method, however, is how to add a gain for the lower ranges (which is why an instrum. amp was a direct and easier idea for me) other than using different shunts for different ranges, but doesn't seem a very elegant solution to me. At a first glance I could say that I could change
Anyway, I found this built power supply from which, at least, I could take some parts scheme as a reference: the H24005 (block scheme attached). Even though it's an SMU I thought I could reuse some of the underlying blocks
https://www.eevblog.com/forum/projects/diy-smu-project/

One other thing to watch out for in your design is that the maximum differential input voltage of the OP07, or most precision operational amplifiers, is only a couple volts or less.

Hm, according to the datasheet (attached), I see that the differential input voltage goes up to +-30 V, but I never looked this detail in other op-amps, so I'll just take it as a thing to check for the future anyway.

[David Hess]

One other thing to watch out for in your design is that the maximum differential input voltage of the OP07, or most precision operational amplifiers, is only a couple volts or less.

Hm, according to the datasheet (attached), I see that the differential input voltage goes up to +-30 V, but I never looked this detail in other op-amps, so I'll just take it as a thing to check for the future anyway.

Ah, the situation is not quite as bad as I remembered for the OP-07.  The input differential pair is protected by a series connected pair of back to back diodes limiting the differential input voltage to about 1.2 volts.  There is a resistance in series with each input to limit the current.  What this means is that when the differential input voltage exceeds 1.2 volts, the differential input resistance drops and the input current rises beyond the input bias current.  This may cause problems in some circuits.