LM324 Power Supply with variable voltage and current

[mike_mike]

Yes, the probe is connected to the PSU's output terminals.
I checked second time, using another PCB. Please have a look at the screenshots. I think that it is better now.
I had a problem with the PSU before the screenshots from my previous reply, because I accidentally connected a 12V/1A PSU to the output GND of my PSU and I think that I damaged some components.

EDIT: It looks like previous screenshots, I compared it with the wrong screenshot.

[xavier60]

Ok, all good then. It was looking like the problem that was caused by R7 when it was 10K.

[mike_mike]

The damaged component from one of the PCB's was the C2 capacitor (47uF), the output capacitor. I replaced it with a new one and the response looks like the attached screenshot.

[ejd.pol]

Hello,

I stumbled on this thread by accident, and it fits perfectly with some vague plans I have to build a micro-controlled PSU.
This means that (eventually) i would like to make some changes, but to prepare for that, I want to understand the design.
So, I started from the Kicad files posted recently by Mike_mike (many thanks for that!), and first rearranged the schematics
with the purpose to make it simpler to follow the flow of how it operates, and so give a better overview.
In particular, the two opamps now point from left to right, reducing the number of wires going the wrong way.
This makes the schematics more tidy. See for yourself in the attached screenshot. (I have preserved the component designators.)

I like the simplicity of the CV/CC controls, and I will certainly keep the shutdown connections.
Admittedly, I have made a first change: replacing the TL431 reference by a REF02.
This change is not because of the accuracy: in a later step, the two pots will be replaced by DACs,
such that a micro can control the voltage and current settings. The PSU can then be calibrated in software.
The change is intended to increase stability of the output voltage over temperature and age.
It may be necessary to raise the 8V supply by a few volts to accommodate the new reference.

At this point I have the following questions regarding the design, I hope you can help to answer them.
(I have learned a lot from scanning the previous 47 pages of posts, but may have missed an answer or two.)

1) In the local feedback loop of U1b (CV output), R5 (4k7) is connected so at to include D1 in the loop.
   Why is D1 included in the local feedback loop? (Alternatively, R5 could be connected to the output of U1b.)
2) What is the purpose of R7A1 (10k)? (And why does it have an unusual designator?)
3) What is the purpose of R10? Especially when Q1 starts drawing current (depending on the power stage),
   it causes the reference for Q1 (emitter voltage) to rise, making life more difficult for the opamps.
4) What is the purpose of R41 (10M)?

In the new schematics, I have drawn boxes around subsections of the circuit that are interesting for playing around with.
The box around the reference and the two pots is clear: this is where the microcontroller should hook into.
(I don't trust a micro to control the pass stage directly; too many uncertainties.
But it is good for controlling the setpoints, sensing user controls (except shutdown), and displaying information.)

There is also a box around the resistor divider providing the feedback signal for CV control.
It may be advantageous to use special resistors to match the stability of the reference and pots (or DACs).
Also here, accuracy is less of a concern.

Finally, I have drawn a box around the pass-transistor stage and its driver.
The pass stage should behave like a PNP transistor. In case I understand the CV/CC control loop well enough,
connecting a power PNP transistor (2955 or so) to the three terminals should produce a working PSU. (Is that correct?)
To increase output power, it is then possible to hook up instead a Sziklai pair consisting of a smaller PNP
that drives one (or multiple) larger NPN power transistors to do the heavy lifting.
But the behavior of the attached circuit should resemble the behavior of a PNP transistor
(The emitter sinks current, the base and collector source current. This is why I have relabeled
the connection terminals to indicate their function according to the single PNP transistor configuration.)

So far, so good (I hope), and all feedback is highly appreciated: comments, corrections, and suggestions.
In case there is any interest in the new schematic and/or planned modifications, I can post the Kicad files as well.

Kind regards, Evert-Jan

[xavier60]

1) In the local feedback loop of U1b (CV output), R5 (4k7) is connected so at to include D1 in the loop.
   Why is D1 included in the local feedback loop? (Alternatively, R5 could be connected to the output of U1b.)

That greatly reduces voltage overshoot when the PSU transitions from CC to CV mode. Short discussion here. 
https://www.eevblog.com/forum/projects/oscillation-in-psu-simulation/msg3023338/#msg3023338

2) What is the purpose of R7A1 (10k)? (And why does it have an unusual designator?)

Depending on the type of opamp used and at what supply voltage it loses control of its output, R7A1 is there to help prevent Q1 from being turned on during power down due to loss of mains power and causing a voltage spike at the PSU's output. It may not be needed.
The purpose of the LED is also to prevent power down spikes as well as allowing the opamps to completely turn off Q1. Although I haven't tested it, the LED could be replaced by a TL431.
The unusual designator, because it was an afterthought that wasn't tidied up.

3) What is the purpose of R10? Especially when Q1 starts drawing current (depending on the power stage),
   it causes the reference for Q1 (emitter voltage) to rise, making life more difficult for the opamps.

The path from the Base of Q1 to the PSU's output is a transconductance amplifier, Voltage signal in, current out. R10 sets the gm or gain of the output stage. The gain needs to be high enough to keep Q1's current low. Too much gain could cause instability, especially with having a sluggish Sziklai pair in the signal path.

4) What is the purpose of R41 (10M)?

It applies some positive bias to the inverting input of the CC opamp, ensuring the CC can be set to zero.

Keep in mind that loads placed on the 8V that return current to ground, the right-hand side of the CS resistor R18, will be measured.
 This is ok if the current is low and not fluctuating.
Otherwise, a separate control supply will be needed.
I used the TLC072 in my build of the design, now superseded.
The pass element needs to be high gain PNP such as a Sziklai pair or Darlington. I have not tested a Darlington.
With my build, the output is turned on and off via the Shutdown input. Using this method causes such an abrupt turn on, the charge up of the output capacitor causes brief current limiting, followed by a small voltage overshoot.
A better controlled turn on could be achieved by enabling and disabling the reference.

I'll go over your post again.

[ejd.pol]

Many thanks for your answers! They greatly help to improve my understanding of the design.
Also the remark about the current being included is definitely something to keep in mind.
Apart from the change of the TL431 to the REF02, the designs should be the same.
Do you see any disadvantage in increasing the 8V supply to 10V?

[xavier60]

Many thanks for your answers! They greatly help to improve my understanding of the design.
Also the remark about the current being included is definitely something to keep in mind.
Apart from the change of the TL431 to the REF02, the designs should be the same.
Do you see any disadvantage in increasing the 8V supply to 10V?

Increasing to 10V will increase the current limiting response time to sudden overloads. It will need to be tested but may not be a problem.
It can be dealt with in 2 ways. A faster opamp. The TLV9162IDGKR replaces the TLC072.
Also, by raising the operating voltage of Q1's Base. Presently, it is being set by the convenient ~2.7V drop of a blue or white LED.

[mike_mike]

@xavier60 , I would like to ask a question: I need to use the tap changer circuit for another power supply. For example, a LM317 power supply or a LM723 power supply (0-30V/ 0-3A - schematic attached).
What are the modifications that I need to do on the tap changer schematic (schematic attached) ?
I will also remove the DC off switch, I only need the tap changing part of the schematic.
Should I first connect J1 to J2 ? (in1 and in2 are connected to the PSU output)

[xavier60]

@xavier60 , I would like to ask a question: I need to use the tap changer circuit for another power supply. For example, a LM317 power supply or a LM723 power supply (0-30V/ 0-3A - schematic attached).
What are the modifications that I need to do on the tap changer schematic (schematic attached) ?
I will also remove the DC off switch, I only need the tap changing part of the schematic.
Should I first connect J1 to J2 ? (in1 and in2 are connected to the PSU output)

Yes, connect J1 to J2. It should work normally.
BTW, the design that has the LM723 is a Harrison type.

[mike_mike]

Regarding the schematic at reply #1182, I mounted 4 x TIP35C instead of 2XBD911. I also wanted to use a higher power driver transistor instead of BD139. I chose TIP41C, and I got the attached screenshot on the oscilloscope:


Also, if I replaced BD139 by BD241C, I got the attached screenshot:


The load was 11R at 34.11V output voltage, and the current potentiometer was at maximum.
Do I need a higher speed driver transistor ?

I mean that in the first screenshot the recovery is very slow while in the second screenshot (the one with BD241C) the recovery has some damped oscillation but it recover faster.

[xavier60]

Regarding the schematic at reply #1182, I mounted 4 x TIP35C instead of 2XBD911. I also wanted to use a higher power driver transistor instead of BD139. I chose TIP41C, and I got the attached screenshot on the oscilloscope:
(Attachment Link)

Also, if I replaced BD139 by BD241C, I got the attached screenshot:
(Attachment Link)

The load was 11R at 34.11V output voltage, and the current potentiometer was at maximum.
Do I need a higher speed driver transistor ?

I dont understand what's happening in the first test so put it back to what gives the second result where it seems to need compensation adjustment. First try replacing C10 with a 10K and 1nF in series.
EDIT: Start with the condition that has the ringing.

[mike_mike]

EDIT: Start with the condition that has the ringing.

Ok, the condition that has the ringing is the one that uses BD241C as driver transistor.
I tested in the same condition as above, but with 10k in series with 1n, instead of C10. The result is attached.

I forgot to mention, my schematic is modified.
The modifications are:
1. U2 is TL082
2. R23 = 1 ohm
3. R5 =10k
4. instead of P2 (rectifier), I have a power supply which has 7812 and 1912.
5. RV1 is 50k
6. I use 4xTIP35C instead of 2xBD911
7. I use BD241C as driver instead of BD139.

The schematic is from an old magazine from Romania. I made the modifications, because the members of a romanian forum recommended them.

... and before replacing LM358 by TL082, the power supply was oscillating.

[xavier60]

That's good enough to leave as is or you can progressively reduce the capacitor to improve the settling time.

[mike_mike]

I reduced the capacitor to 220p (0056.png) and to 100p (0059.png).
I think that the one with 220p is better.
Is it necessary to reduce the capacitor lower than 100p, to check if the circuit is stable ?

[xavier60]

That's a very nice response with the 220pF. As the capacitor size is reduced further, ringing will be expected, so no point.

[mike_mike]

I made a test, using the old PCB of the LM723 power supply and a new group of 4 x TIP35C and BD241C.
This time the results were a little bit different, I found that on the screenshot is a damped oscillation, please have a look at the screenshot.
 

[xavier60]

Likely because Type 1 compensation using just a capacitor is not suitable for current sourcing designs like the Harrison.
It should be Type 2, series RC.   

[mike_mike]

Should I use some random values for R and C ?
I saw that there are a lot of calculations around this Type 2 compensation, probably a simplified method is to check using different values for R and C.
https://developerhelp.microchip.com/xwiki/bin/view/applications/power/switching-regulators-/digital-compensator-design-tool/type-II-analog/

[Jwillis]

Something concerns me with the last schematic you posted. J1 33Vac, C61 and  C62. Is that an error or is that the way it's supposed to be. Just that there's two polarized capacitors on an 33V AC line. Or is that supposed to be rectified 33V.

[mike_mike]

@Jwillis It is a mistake. There are about 40-41 Vdc. The transformer is a 2x15Vac. I am using the tap switcher with this supply, otherwise the voltage drop on the power transistors is high and the power dissipation is making them very hot.
The rectified voltage is about 40-41 Vdc.

[xavier60]

Should I use some random values for R and C ?
I saw that there are a lot of calculations around this Type 2 compensation, probably a simplified method is to check using different values for R and C.
https://developerhelp.microchip.com/xwiki/bin/view/applications/power/switching-regulators-/digital-compensator-design-tool/type-II-analog/

I dont know enough to take this any further.  That response shape indicates good stability along with a fast response. The only thing likely to upset it is to put a very low ESR capacitor directly across the output at the PCB.

[mike_mike]

I am trying to mount all the components into a case, and make a new power supply with LM723 and TL084 schematic using the 2x15Vac transformer.
I also have the tap changer transformer (12Vac) and the 2x12Vac transformer for the LM723.
I wanted to use only 2 transformers, the 2x15Vac and the 2x12Vac for the LM723. It is possible to power the tap changer, the cooling fans and the digital voltmeter from the same 2x15Vac transformer as the power supply ? Is that possible ? Or from the 2x12Vac transformer ?

[xavier60]

I am trying to mount all the components into a case, and make a new power supply with LM723 and TL084 schematic using the 2x15Vac transformer.
I also have the tap changer transformer (12Vac) and the 2x12Vac transformer for the LM723.
I wanted to use only 2 transformers, the 2x15Vac and the 2x12Vac for the LM723. It is possible to power the tap changer, the cooling fans and the digital voltmeter from the same 2x15Vac transformer as the power supply ? Is that possible ? Or from the 2x12Vac transformer ?

Initial thoughts are, the fans only can be powered from the 2x15Vac transformer.
The tap changer might be able to be powered from the 2x12Vac transformer with modifications so that it can accept negative sensing input voltage.  Because the DVM can only measure positive voltage, it needs to be powered WRT the PSU's negative output. So the unmodified tap changer might as well be powered from the same supply.
 Keep in mind that ground on a Harrison PSU is the positive output.

[mike_mike]

I could use a transformer for this power supply which has a secondary voltage of 2x15Vac, and another secondary of 28Vac. I could use the 28Vac secondary for the tap changer circuit. First I could bring down the voltage to about 15-20Vdc using a LM317T and then I could power the tap switcher (the input of L7812) from the output of the LM317T.

[xavier60]

I could use a transformer for this power supply which has a secondary voltage of 2x15Vac, and another secondary of 28Vac. I could use the 28Vac secondary for the tap changer circuit. First I could bring down the voltage to about 15-20Vdc using a LM317T and then I could power the tap switcher (the input of L7812) from the output of the LM317T.

That should be fine. The voltage would be too close to the limit of just an LM7812.