DIY bench power supply PSL-3604

[inteljoe]

I like the design a lot. I may borrow a few ideas for my own projects.
That said, I hope someone out there will put this (or something similar) together in kit form for purchase. I'd buy one.

[serge_m]

hope someone out there will put this (or something similar) together in kit form

Hi Inteljoe,

I am working on this. PCBs and some components are already available, but I am not ready to send full kit yet. Will keep you informed.

[Liv]

Firmware update - version 2.04.
Fixed setting of V, I values and OVP, OCP, OPP thresholds for TOP_V more than 36 V or TOP_I more than 4 A.

[VEGETA]

Hi there Liv, let me tell you: this is a great project!

I have started one for myself [ https://www.eevblog.com/forum/beginners/mosfet-linear-regulator-circuit/ ] which I would like you to take a look at it and help if you can. I hope I can get enough knowledge to complete it! that thread contains the main mosfet regulator circuit only, other stuff will be added later as my problem is that it is not stable and I can't get it to behave well in low currents below 80mA as well as it doesn't accept input voltages higher than say 3v of output voltage which is just funny.

anyway, I have some questions:

1- where is the main regulator transistor in your design? I didn't see it.

2- please tell me how did you make it stable? I mean the diode+caps+resistors in every feedback loop and op-amp circuit. This is a nightmare for me right now as I don't know how to choose them.

3- how did you make it behave well and stable in low currents like 1mA range?

4- what is the drop voltage of your linear regulator (I aim for 1v only to make it efficient, may not be good for dynamic loads).

5- your voltage sensing is interesting. is "remote sensing" the default way or is it an option? I see resistors from out+ and out- to the diff amp but I don't know their job.

6- You used TL082 opamp for critical stuff like current diff amp, is it sufficient? in my thread they suggested LT1678 which is a lot pricey. as for comparator op-amp they are not critical thus something like lm358 or lm324 is good enough in my opinion. As for TL072 they suggested not using it because it will need a negative voltage rail unlike single supply ones like LT1678.

I have some ltspice circuits in my thread but none of them is stable over the whole range. I appreciate your help and looking forward to learn skills like yours.

your friend,

VEGETA

[Liv]

1- where is the main regulator transistor in your design? I didn't see it.

I am using the stacking transistors in order to reduce power dissipation. This is equivalent to switching voltage ranges, but it happens automatically and quickly.

2- please tell me how did you make it stable?

Sleepless nights in the PSpice.

3- how did you make it behave well and stable in low currents like 1mA range?

I do not see any problem with small load currents. Sometimes, problems may arise from loads that have a complex impedance. In such cases, my PS is not stable, due to the absence large capacitor at the output. But this capacitor can be added directly to the output terminals, while PS becomes stable.

4- what is the drop voltage of your linear regulator (I aim for 1v only to make it efficient, may not be good for dynamic loads).

I have a multi-level power, if we take the intermediate levels, the minimum drop is very small. But at the upper level it reaches of several volts.

5- your voltage sensing is interesting. is "remote sensing" the default way or is it an option? I see resistors from out+ and out- to the diff amp but I don't know their job.

Remote sense is optional. My copy of PS without separate sense terminals. The sense signal is internally connected directly to the output terminals. But you can make separate terminals. The resistors are needed to ensure that when broken sense wires output voltage does not greatly increased. This is a standard solution that can see the circuit Agilent PS.

6- You used TL082 opamp for critical stuff like current diff amp, is it sufficient?

This choice has been made to reduce the cost and simplify the search component.

in my thread they suggested LT1678 which is a lot pricey. as for comparator op-amp they are not critical thus something like lm358 or lm324 is good enough in my opinion. As for TL072 they suggested not using it because it will need a negative voltage rail unlike single supply ones like LT1678.

My scheme will not work with a single supply with any opams. Such a scheme can be done, but I do not aspire to it. PS quality will be worse. Get the negative power supply is very simple with the main winding, I just connected a rectifier bridge through capacitors.

I have some ltspice circuits in my thread but none of them is stable over the whole range.

That's a very difficult question. When calculating the stability of the circuit need to use load parameters. For a power supply they are not known, because the load can be any. Therefore, to make stable power supply is impossible in principle. For example, Agilent draws graphics with PSU stability area for different load inductances and capacities.

[VEGETA]

I am using the stacking transistors in order to reduce power dissipation. This is equivalent to switching voltage ranges, but it happens automatically and quickly.

You mean VT11 to VT14? while VT7 is the gate driving transistor? if so, then they are in series with each other.
Then it is fed from "out" to "out+" and "out-" which are the true output terminals while the current sense resistor is connected between out- and the true ground, right?

Sleepless nights in the PSpice.

On that we agree xD.

I meant for example, why picking 10nf+10kR as compensation (with that diode like VD19), instead of 1nF+10KR or something else? is it trial and error or is there a guided method for doing so? both will yield sleepless nights as well. I don't have PSpice thus I will rely only on LTSpice.

I do not see any problem with small load currents. Sometimes, problems may arise from loads that have a complex impedance. In such cases, my PS is not stable, due to the absence large capacitor at the output. But this capacitor can be added directly to the output terminals, while PS becomes stable.

it is common to have this low currents issue, I mean when you put a current limit of 10mA or so... not when it is in CV and the load draws 10mA. By large capacitor do you mean something like 100uF aluminum (electrolytic) capacitor? or do you mean you have some circuit to add it when necessary?

What is the minimum load current that your supply can set in CC mode? does it suffer from high capacitive or inductive loads?

I have a multi-level power, if we take the intermediate levels, the minimum drop is very small. But at the upper level it reaches of several volts.

You mean V1,V2,...etc rectified from the transformer taps? But you still got to pick an output voltage and this is where the drop is determined.

Remote sense is optional. My copy of PS without separate sense terminals. The sense signal is internally connected directly to the output terminals. But you can make separate terminals. The resistors are needed to ensure that when broken sense wires output voltage does not greatly increased. This is a standard solution that can see the circuit Agilent PS.

As far as I can tell, out+ is connected to R83 and R84 in series... after that it is "sense+" which is fed into R70 and R61 in series which form the positive input of the op-amp (and the same for the negative one) while R57 is for the gain. R75+C33 are compensation for stability.

Now what I meant is why not connecting "sense+" directly to the out+ since it gives /10 gain via the op-amp? why there are R83+R84? without this you won't be able to sense the voltage thus what is the remote sensing here which is optional? you mean the "sense+" and "sense-" external connection via XP3 (to be able to sense the voltage outside of the supply)?

My scheme will not work with a single supply with any opams. Such a scheme can be done, but I do not aspire to it. PS quality will be worse. Get the negative power supply is very simple with the main winding, I just connected a rectifier bridge through capacitors.

And why is that? what is the gained benefit of using negative voltage rail? and why it will be worse quality supply? someone mentioned that it will be slower CC response with single supply op-amps like mine but I don't know why while your design seems fast which gave you the freedom to pick lower quality op-amps. Is there a reason for this?

That's a very difficult question. When calculating the stability of the circuit need to use load parameters. For a power supply they are not known, because the load can be any. Therefore, to make stable power supply is impossible in principle. For example, Agilent draws graphics with PSU stability area for different load inductances and capacities.

And how stable is yours? people mentioned the trials of putting different types of loads especially low current ones and high capacitive ones as well as pure inductive loads. So you mean I should finish the design then try these loads one by one?

As for stability, I just learned stuff about ac analysis with bode plots and it does seem the key to make power supplies stable. what points do you test to determine that?

___

thanks!

[Liv]

You mean VT11 to VT14? while VT7 is the gate driving transistor? if so, then they are in series with each other.
Then it is fed from "out" to "out+" and "out-" which are the true output terminals while the current sense resistor is connected between out- and the true ground, right?

Yes.

is it trial and error or is there a guided method for doing so?

I watched the phase margin for different loads. It was necessary to solve complex optimization problem that just happened by trial and error.

By large capacitor do you mean something like 100uF aluminum (electrolytic) capacitor?

Yes. Most PSUs are 470 uF output capacitor. I would do without it.

What is the minimum load current that your supply can set in CC mode?

1 mA, it is limited by the resolution of DAC. Analog circuit can provide arbitrarily small current (within the OPAmps errors, of course).

You mean V1,V2,...etc rectified from the transformer taps? But you still got to pick an output voltage and this is where the drop is determined.

I do not understand what the question is.

Now what I meant is why not connecting "sense+" directly to the out+ since it gives /10 gain via the op-amp? why there are R83+R84? without this you won't be able to sense the voltage thus what is the remote sensing here which is optional? you mean the "sense+" and "sense-" external connection via XP3 (to be able to sense the voltage outside of the supply)?

My copy of the PSU has no external sense terminals. XP3 internally connected to the Out + and Out-. In this case, the R83-R86 are not needed, as VT11 and VT12. But if you add external sense terminals, these elements are desirable.

what is the gained benefit of using negative voltage rail? and why it will be worse quality supply?

In single supply PSU is difficult to provide a precise work near zero voltage or current. Since even the Rail-to-Rail op amp output voltage can not reach the ground. It will be necessary to shift all the signals. I also shifted the signals, but for a different reason - unipolar DAC and ADC can not be calibrated at zero.

And how stable is yours?

I use the PSU for almost 3 years, during which time the two times met the oscillation. And it was not inductive or capacitive loads. For example, when connected as a load current source using MOSFET. It was necessary to add a capacitor on the output. I can not do absolutely stable PSU.

[Crumble]

Hi Liv,

This is a nice power supply, well done! I have no insights in the software (yet!), but the design and execution of the hardware is impressive! I have two (small) questions here:

The toroid is custom manufactured I guess?

Toroid based on 105W transformer for halogen lamps. Secondary winding (12V) was removed and replaced by 4 windings 2 x 8.5V and  2 x 10V.

Why are the voltages different for the two sets of windings? Why not make them all ~9V? Just wondering. If one has a transformer custom made it might be easier if all windings are the same number of turns (AFAIK they are usually wound together, thats why I'm asking).

This is how it seems they wind toroids:

If you do a precision power supply, it must be purely linear. Or use low-noise preregulator working at mains frequency. An example of such amateur PSU with 1 mV step here.

Is this yours too? I am seeing the same style of case here... ;-) It looks like they sell this case in in NL too, they seem to be dirt cheap (<€2,00 for 5+). Don't know the difference between Z-2A and Z-2AW, but I might just buy one and check what I'll get.  I have had issues finding decent affordable cases here, so this is good news.

Really respect the work you did!

Regards,
Crumble

[Liv]

Why are the voltages different for the two sets of windings? Why not make them all ~9V? Just wondering. If one has a transformer custom made it might be easier if all windings are the same number of turns (AFAIK they are usually wound together, thats why I'm asking).

I wind the transformers myself, so it does not matter what the windings are. The design of the PS allows the use of identical windings.

Is this yours too?

No. This project was made by one developer from Ukraine.

they seem to be dirt cheap (<€2,00 for 5+). Don't know the difference between Z-2A and Z-2AW

The quality of these cases is not very good, but it is adequate to such a price. Typically, the letter W denotes the presence of a ventilation.

[Crumble]

I wind the transformers myself, so it does not matter what the windings are. The design of the PS allows the use of identical windings.

Is the handwinding the reason the windings are different, or did it have some technical relevance? I can imagine the dropout of 4 mosfets (when outputting a high voltage) is more than for only 1 (at low voltages), and that you would use the higher voltage windings for the higher ranges to compensate. I could not find the way the windings were connected in the schematic though. It is good to hear the supply is not critical about some variation in voltage. I think serge_m must have noticed this too preparing the kit he proposed making.

The quality of these cases is not very good, but it is adequate to such a price. Typically, the letter W denotes the presence of a ventilation.

Ok, that is usefull to know, at least I won't have to worry about the size. Think I'll order one and check how I like it. I bought a case from Banggood recently, but the quality of the plastic was only mediocre, it had minimal ventilation slots and it had some play in it. It also took €10 rather than €2, so even if this case is the same quality I'd be quite a lot better off.

Regards,
Crumble

[Liv]

Is the handwinding the reason the windings are different, or did it have some technical relevance?

In theory, this PSU structure allows you to arbitrarily select all 4 voltage levels. In practice, the lowest stage is used to obtain a voltage of 5 V, it must provide this taking into account the drop on the LM317. Another limitation is the simplification of rectifiers - pairs of windings should be the same. If completely separate rectifiers are used, this requirement is eliminated. The voltage levels of the individual stages can be selected based on the practical conditions for using the PSU. When the output voltage is close to the voltage of one of the stages, the power dissipation is minimal.

I can imagine the dropout of 4 mosfets (when outputting a high voltage) is more than for only 1 (at low voltages)

In a linear PSU, the voltage drop across the pass transistor is much higher than on an fully open transistor. Therefore, the additional fully open transistors in the chain have almost no effect.

[zii]

Hi everyone, and thanks Liv for sharing this awesome project. I have wanted to build the PSL-3604 for quite some time.
I have already gathered all parts, soldered PCBs, flashed STM32 and I am left with winding my transformer, wiring it all together and screwing it to heatsink (I also have the same Z-2AW enclosure).

So I have a question regarding the transformer windings.
Do I understand correctly, that those recommended "2x 8.5V" and "2x 10V" windings as You have mentioned should be 8.5V and 10V in amplitude when the maximum load is applied? So that after synchronous rectifier and filters, the highest tap would be (close to) 37V with respect to GND and that there would be 1V headroom for 36V regulation. Is this high enough for losses in the path and operate normally ?
Or should I actually wind ~8.5V and ~10V windings (effective value), such that after filter caps the highest tap would be around 2 * (8.5V + 10V) * sqrt(2) = 52.3V ?

And if my transformer allows such power, could I potentially go (by rewinding transformer) from 36V to for example 40V (as long as the output protection diode allows it and I am able to modify firmware accordingly) ?

Cheers,
Zigmars

[Liv]

My transformer: 2 x 8.6 V + 2 x 10.2 V effective without load. Rectifiers out: VN = -13 V, V1 = 11.1 V, V2 = 22.4 V, V3 = 34.8 V, V4 = 47.2 V, VG = 58.6 V (all boards connected, no output load).

You can make any voltage stages. The first stage V1 should provide sufficient voltage for +5 V regulators. The rest can be chosen arbitrarily. The maximum output voltage, which is supported by program, is 99.9 V. But you need to change the gain of U9: 1, so that the full scale corresponds to 3 V at U9:1 output. The same is true for the current scale, which can be up to 9.99 A.

[Branquinho]

Olá, não sou muito experiente na eletrônica to apenas começando e too precisando de unidade de fonte de alimentação gostei muito do seu projeto tá de parabéns tou pensando em fazer passo a passo. essas pci você fez manualmente ou mandou fazer numa fabrica especializada em pcis é possível fazer em casa.

[Liv]

The boards for the PSL-3604 are made at the factory. But I made boards for similar PSL-2401 and PSL-2402 manually using a laser printer and an iron. For beginners these PSUs can be too complicated.

[Branquinho]

Olá amigo esse display de 7 seg  onde você comprou eu não acho ele é catodo ou anodo o buzzer eu não acho nem um deles da lista que você deixo em loja nem uma

[Liv]

7-seg LEDs - common cathode.
Buzzer either of the two shown in the schematic diagram.

[Branquinho]

Olá amigo vou manda fabrica as minha placa pela  seeed studio ou easyEDA mais o aquivo gerber não funciona não aparece as medidas ta dando erro por que o arquivo gerber

[Liv]

I dont know. These gerber files were given to the PCB manufacturers in Ukraine, everything was fine. You can try to send the PCAD-2006 files to the PCB manufacturer.

[Branquinho]

Hello friend, how are you? these leds REC-S3461CG-G / REC-S3461CSR-G has no where to find it does not have the easiest model to find that I can use psl-3604 this leds is a special model thanks...

[Liv]

Any leds with a common cathode and suitable dimensions will work. For example, Toyo E40361-L.

[Branquinho]

Thank you very much for sharing this great project. Congratulations.

[Chris56000]

Hi!

Afternoon Liv, how's the LD8020 Electronic Load Project getting along?

I'd love to make your completed design rather than buy a Chinese one - in my opinion, Chinese gear needs to be taken apart, examined for safety issues and corrected where necessary, a set of schematic diagrams drawn out and a B.O.M. written up (and if possible!) the firmware extracted before it can occupy a permanent lab place in my opinion, whereas all your excellent projects have all the documentation on hand before you begin!

Chris Williams

[Liv]

Unfortunately, the LD-8020 electronic load project has not yet progressed. Now I have difficult times, there is no way to do hobbi projects.

[pradipk]

Hi,

I really like the power supply design, specs and the simplicity. Thank you for sharing the all the details to us and its really very helpful to understand working of it. Please keep such a good work up .

May I know that is that product still active and where can I buy that (online)? Is it available in India anywhere please?

Cheers,
- Pradip