Author Topic: How this linear PSU circuit works, and why it doesn't (Read 29500 times)

bogdan2014 · « **on:** March 12, 2020, 06:36:25 am »

I've built this linear lab PSU circuit based on a project found on the internet. I admit it was pretty much copying and adapting sections of it to fit my needs, without actually understanding how it's supposed to work, so consequently I'm stuck with troubleshooting it. I suck at linear circuits so I'm asking for help.
The schematic is posted below. Ignore the tracking pre-regulator part as it's bypassed for now. What's basically hapenning is that I only get about 5V out regardless of the action of the two potentiometers (current and voltage). I suspect there's something wrong with Q3/Q4 circuit, since that should work on its own, maybe R9/ R10 values are too high?

xavier60 · « **Reply #1 on:** March 12, 2020, 07:54:24 am »

Assuming that the output is unloaded, for any non-zero setting of RV2, pin 7 of U2b should be high. Q2 should be turned on hard with next to no Collector voltage and providing a solid ground path for Q1's Emitter. Confirm this.

With VR1 set half way, what voltages are on pins 1,2 and 3 of U2 as well as its supply pins? And check the output voltage again.

The voltage at each end of R10 might be useful.

mikerj · « **Reply #2 on:** March 12, 2020, 08:32:23 am »

TL072 is not a suitable device for this circuit, it suffers from phase reversal when it's limited common mode input range is exceeded which is almost certainly the case in this circuit.

Measure the voltages on the input and output pins of both op-amps and report back.

ggchab · « **Reply #3 on:** March 12, 2020, 08:42:49 am »

To always sink some current ? But useless when the output is close to 0V
Some years ago, mjlorton has made some videos on a power supply using a similar circuit. But I don't he finished the project.

bogdan2014 · « **Reply #4 on:** March 12, 2020, 08:46:19 am »

Quote

Assuming that the output is unloaded, for any non-zero setting of RV2, pin 7 of U2b should be high. Q2 should be turned on hard with next to no Collector voltage and providing a solid ground path for Q1's Emitter. Confirm this.

Confirmed.

Quote

With VR1 set half way, what voltages are on pins 1,2 and 3 of U2 as well as its supply pins? And check the output voltage again.

With RV1 half way, 2.5V at pin 3, 0.5V at pin 2, 10.7V at pin 1. Supply voltage is around 12V. And pin 1 of Q1 is at almost 0. Output voltage is 3.1V (with 12V input).

Quote

TL072 is not a suitable device for this circuit, it suffers from phase reversal when it's limited common mode input range is exceeded which is almost certainly the case in this circuit.

It's actually an LM358 there, I've used the TL072 for the footprint. I also have some other op-amps I can try.

Quote

WTF is that LM334Z (U6) doing from V+ out to GND?

That's a constant current sink, it gives a minimum load to the circuit. It's currently not populated on the PCB.

xavier60 · « **Reply #5 on:** March 12, 2020, 09:01:08 am »

Check the voltage across R10 and all other voltages in the output stage. Measure the voltage across Base and emitter of Q3.

bogdan2014 · « **Reply #6 on:** March 12, 2020, 11:15:13 am »

I think I found the culprit, Q3. Replaced it and it seems to be working, I get the full output and the voltage adjustment works. I'll will test the rest of the circuit too.

xavier60 · « **Reply #7 on:** March 12, 2020, 11:24:45 am »

I'm curious about the stability. Check for oscillations at the op-amp's outputs if possible.

mikerj · « **Reply #8 on:** March 12, 2020, 11:25:44 am »

The ground side of the voltage feedback divider appears to be on the wrong side of the current sense resistor, so it won't compensate for the voltage drop across it.

bogdan2014 · « **Reply #9 on:** March 12, 2020, 11:31:10 am »

Yes, that is true. Will fix that.

xavier60 · « **Reply #10 on:** March 12, 2020, 11:49:25 am »

Actually just about everything should be grounded to the top of the CS resistor except for the negative of the unregulated supply and the op-amp's negative supply pin. There will then be the problem of ripple current through R1 will also flow through the CS resistor.
Is there a link to the original design?

Edit: Forget all of that for now, it will mess up the current regulation. Test it as is for now.

xavier60 · « **Reply #11 on:** March 12, 2020, 12:27:45 pm »

This design is similar to yours, https://www.eevblog.com/forum/projects/simple-diy-linear-psu/msg2951262/#msg2951262
It can be used as a guide to how everything should be grounded with respect to the CS resistor. The ripple in the reference bias current can be solved with a constant current source.
The design I have linked wont work properly with a pre-regulator.

bogdan2014 · « **Reply #12 on:** March 12, 2020, 05:58:57 pm »

https://www.mjlorton.com/wp-content/uploads/2015/08/Bench-PSU-v25-Schematic.pdf
This is the original design

xavier60 · « **Reply #13 on:** March 12, 2020, 10:44:48 pm »

Quote from: bogdan2014 on March 12, 2020, 05:58:57 pm

https://www.mjlorton.com/wp-content/uploads/2015/08/Bench-PSU-v25-Schematic.pdf
This is the original design

I see a current source for the reference zener, another LM334 could be used in your design.
The way the grounding has been done, the way I see it, the load regulation will change with the CV Pot setting. The output voltage would increase with load for some part of the setting range.
Test your design for stability first. For C5 to be effective, put a 10K resistor in series with the signal path from the CS resistor.

Miti · « **Reply #14 on:** March 13, 2020, 02:33:40 am »

Shouldn’t there be a diode at the output of U3?

bogdan2014 · « **Reply #15 on:** March 13, 2020, 11:07:04 am »

Quote

Shouldn’t there be a diode at the output of U3?

Damn, good catch

Well, good thing I don't plan to use the pre-regulator anyway.

Quote

Test your design for stability first. For C5 to be effective, put a 10K resistor in series with the signal path from the CS resistor.

Yes, there's place on the PCB for that resistor, but I had a jumper in place.
I poked around with the oscilloscope and found some oscillations from pin 1 of the op-amp all the way through to the output. It's actually audible.

xavier60 · « **Reply #16 on:** March 13, 2020, 11:39:17 am »

Quote from: bogdan2014 on March 13, 2020, 11:07:04 am

Quote
Shouldn’t there be a diode at the output of U3?
Damn, good catch
Well, good thing I don't plan to use the pre-regulator anyway.
Quote
Test your design for stability first. For C5 to be effective, put a 10K resistor in series with the signal path from the CS resistor.
Yes, there's place on the PCB for that resistor, but I had a jumper in place.
I poked around with the oscilloscope and found some oscillations from pin 1 of the op-amp all the way through to the output. It's actually audible.

Q1 and Q2 add too much gain to the CV and CC loops. I have not heard of anyone being able to properly stabilize that type of design.
It sounds like you have already made a PCB?
It might be possible to adapt the CV loop from the design I have been using but not the CC loop.

bogdan2014 · « **Reply #17 on:** March 16, 2020, 06:22:20 am »

Yes, I already have the PCB manufactured. I don't think it's worth adapting it, so I either find a way to reduce the oscillations or re-do the whole project.
Maybe lowering the gain somehow?

bogdan2014 · « **Reply #18 on:** March 16, 2020, 08:18:24 am »

I re-did the project based on your PSU. Looks better and simpler. I'll go ahead and do a PCB.

xavier60 · « **Reply #19 on:** March 23, 2020, 01:47:23 am »

If a TLC072 op-amp is used, the addition of R7A will solve the power down voltage spike problem.

bogdan2014 · « **Reply #20 on:** July 27, 2020, 08:57:20 am »

I'm back at this project and I'd like to post the schematic before sending the project to the PCB manufacturer. If there's anyone still watching this topic, please check and report if anything is wrong. Thank you

xavier60 · « **Reply #21 on:** July 27, 2020, 09:37:26 am »

Quote from: bogdan2014 on July 27, 2020, 08:57:20 am

I'm back at this project and I'd like to post the schematic before sending the project to the PCB manufacturer. If there's anyone still watching this topic, please check and report if anything is wrong. Thank you

There are some serious mistakes.
Make the current sensing the same as this,
https://www.eevblog.com/forum/beginners/lm324-power-supply-with-variable-voltage-and-current/msg3089254/#msg3089254
Notice that Im using a white LED in Q1's Emitter circuit. But stick with your Zener because of the 12V control rail.

Copy the TL431 circuitry from this,
https://www.eevblog.com/forum/projects/linear-lab-power-supply/?action=dlattach;attach=751023

It's good that you are using an independent supply for the control rail.

eblc1388 · « **Reply #22 on:** July 27, 2020, 09:53:56 am »

As someone has already commented.

Edit: I've overlook a mistake. Even with the changed current sensing connection, it will not work as the voltage drop across the sensing resistor will be negative w.r.t. ground. This requires the Opamp comparator be wired as inverted amp for the circuit to work.

bogdan2014 · « **Reply #23 on:** July 27, 2020, 10:21:03 am »

Can I ask why there's a voltage divider following the adjustment pots and why the CC pot divider is reference to the return, not the 0V? The values seem quite high.

xavier60 · « **Reply #24 on:** July 27, 2020, 10:53:49 am »

Quote from: bogdan2014 on July 27, 2020, 10:21:03 am

Can I ask why there's a voltage divider following the adjustment pots and why the CC pot divider is reference to the return, not the 0V? The values seem quite high.

If you mean the divider after the CV pot in the simplified version? It's because the 8V control rail is being used as the reference.
I don't quite understand the 2nd question. That isn't really a divider after the CC pot, they are more so summing resistors.
Make R6, R7, R2, R3 all 10K. C6 and R17 aren't needed. 2mA is enough for the TL431.
U4 should return to GNDA.
The CV compensation should come from the ORing node.

EDIT:
R4 actually is too high because you are using a higher value CS resistor. What max current are you aiming for?

xavier60 · « **Reply #25 on:** July 27, 2020, 11:15:11 pm »

The opamp's GND pin 4 needs to go to GNDA.
With R8 to 5V, Q1 wont turn on, it needs to go to 12V but this might cause too much maximum drive for Q1.
Include a resistor from the ORing node to GND, 33K for now

bogdan2014 · « **Reply #26 on:** July 28, 2020, 04:48:24 am »

I'm aiming for 5A so I chose a 0.22 ohm resistor. 1V across it at maximum current and 5W power dissipation.
What is the ORing?

xavier60 · « **Reply #27 on:** July 28, 2020, 05:11:47 am »

Quote from: bogdan2014 on July 28, 2020, 04:48:24 am

I'm aiming for 5A so I chose a 0.22 ohm resistor. 1V across it at maximum current and 5W power dissipation.
What is the ORing?

For 5A, consider using 100mΩ or less. I use s 50mΩ CS resistor.
The CV or the CC opamp can pull down the Base voltage of Q1 via D5 or D6. I call the junction of D5 and C6 the ORing node.
Also, depending on what your unregulated voltage is, 5A might not be possible to do safely with one 2N3055.

bogdan2014 · « **Reply #28 on:** July 28, 2020, 07:02:28 am »

I can fit two of them on the heatsink, so that's not a problem. I chose 5A since it was easier on the calculations, but I doubt the transformer will be able to deliver it, and that I'll ever need that much current.
I added all the mods and an extra transistor footprint so I have everything on the PCB. I'm thinking, does the auxiliary power supply have to be referenced to GND or GNDA like the op-amp?

xavier60 · « **Reply #29 on:** July 28, 2020, 08:00:14 am »

Quote from: bogdan2014 on July 28, 2020, 07:02:28 am

I added all the mods and an extra transistor footprint so I have everything on the PCB. I'm thinking, does the auxiliary power supply have to be referenced to GND or GNDA like the op-amp?

Either way, some unwanted current will flow through the CS resistor.
With my bench supply based on this design, I derive the 8V control rail from the unregulated input so its current has to flow back to the unreg negative. This was fine until I added a digital panel meter, powered from the 8V which caused the complication of its current draw flowing through the CS.
So if you don't intend adding a digital panel meter, reference the AUX supply to GNDA. You might even make it selectable.

EDIT: You still need to take the CV compensation from the ORing node. It will greatly reduce voltage overshoot when recovering from an overload.

bogdan2014 · « **Reply #30 on:** July 28, 2020, 08:15:19 am »

Quote

You might even make it selectable.

That sounds good. I think I can accomodate a jumper.

ChenBH · « **Reply #31 on:** July 28, 2020, 11:29:36 am »

Newbie here:
Why is there a diode from E to B of Q4? guarding from returning current from a non-resistive load? Why use a combination of PNP and NPN in the darlington configuration?

bogdan2014 · « **Reply #32 on:** July 28, 2020, 12:06:57 pm »

The diode serves as a protection to return the current to the bulk capacitor when the transistor is closed.
The pnp/npn combination is used because it's a series pass topology, and it works by pulling the base of the pnp transistor to ground instead of applying a voltage to it.

xavier60 · « **Reply #33 on:** July 28, 2020, 12:08:23 pm »

Quote from: ChenBH on July 28, 2020, 11:29:36 am

Newbie here:
Why is there a diode from E to B of Q4? guarding from returning current from a non-resistive load? Why use a combination of PNP and NPN in the darlington configuration?

D9?, it's to prevent the B-E junctions of the 2N3055's being reversed biased if an external voltage is connected while the PSU is powered down and damaging the transistors.
I don't see why a PNP Darlington wouldn't work as well. I think that NPN silicon power transistors are still perceived to be better and cheaper than the PNP type. Google it.

bogdan2014 · « **Reply #34 on:** July 29, 2020, 05:16:51 am »

Updated schematic and PCB

xavier60 · « **Reply #35 on:** July 29, 2020, 10:42:06 am »

Quote from: bogdan2014 on July 29, 2020, 05:16:51 am

Updated schematic and PCB

The CV compensation needs to be taken from the ORing node. It will greatly reduce voltage overshoot when recovering from an overload.
Will Q3 and 4 be mount under the PCB on a heat sink?
5 amps through a bridge rectifier without a heat sink would be pushing the limits. Even if it doesn't fail, the high temperature could damage the PCB and nearby components as well as the solder joints.
I use a 15A bridge on the heat sink.

Some header pins for 12V out could be useful later.

xavier60 · « **Reply #36 on:** July 29, 2020, 07:35:00 pm »

R7 and R8 need to connect to 12V,

bogdan2014 · « **Reply #37 on:** July 30, 2020, 04:54:05 am »

Transistors are TO-3 style (metallic case) so I'll mount them on a heatsink placed outside of the enclosure and run wires to them. The rectifier has a hole for mounting on the heatsink, so I'll do that if necessary.
EDIT: fixed the negative pin of the LV bridge rectifier. Also added a copper plane to that return node.

xavier60 · « **Reply #38 on:** July 30, 2020, 10:05:32 am »

That looks like a very simple design. One more thing. If the CV pot develops intermittent wiper contact, the output voltage might go high. A simple precaution is the put a bleed resistor from the wiper pin to GND, a 470K. A bypass capacitor wouldn't hurt either.

Beware of knock off Bourns 10 turn pots. After some use, the resistance wire can break. This will cause the PSUs output to jump suddenly from zero to full voltage as the pot is rotated. There is no simple precaution for this.
Bochen 5 and 10 turn pots are proving to be more reliable for me.

bogdan2014 · « **Reply #39 on:** July 30, 2020, 10:10:52 am »

That's the disadvantage of this topology. I've had lab power supplies from the 80s that gave out 40V instead of 5V because of bad pots. For sensitive circuits I have a programmable power supply, but there are times where a good old analogue PSU is better since it allows you to vary the voltage gradually (not stepped). Plus I have loads of unused 2N3055 and three identical transformers that would be a shame not to put to good use.
If there aren't any other suggestions, I'd like to send the files and get the PCB done.

xavier60 · « **Reply #40 on:** July 30, 2020, 10:20:53 am »

Only one last possible problem. If you are really using the TL072, it wont work properly with it's inputs close to ground voltage.
I use the TLC072. Even an LM358 will work ok. The current limiting response will be a bit slower.

bogdan2014 · « **Reply #41 on:** September 08, 2020, 10:26:08 am »

Boards have arrived. On first sight, everything seems ok. I had to lower R4 to 100k to get the full current range.

xavier60 · « **Reply #42 on:** September 08, 2020, 10:32:11 am »

Will you be testing it?
A load transient recovery test will show how well the voltage regulation responds to sudden changes in load current, also how stable it is.

bogdan2014 · « **Reply #43 on:** September 08, 2020, 12:09:57 pm »

I only did a quick test with a 12V/20W halogen light bulb. I'll be doing some more tests with different loads.

bogdan2014 · « **Reply #44 on:** January 07, 2021, 08:40:50 am »

I finally managed to assemble the entire thing in an enclosure and did some tests. The main issue is the heatsink, which gets crazy hot. I might have to add a fan or find a bigger one.
Some turn-on turn-off measurements below.

xavier60 · « **Reply #45 on:** January 07, 2021, 09:41:58 am »

The turn on looks unusually slow.
Depending on how your transformer's secondary windings are, it might be possible to reduce the unregulated voltage for when the output is set to low voltages.
With mine, the two 15Vac secondary windings are relay switched to parallel or series to best suit the set output voltage.
I also have a large fan cooled heatsink. It originally belonged to an aircon inverter, and cut in half.
https://www.eevblog.com/forum/projects/linear-lab-power-supply/?action=dlattach;attach=726186;image

bogdan2014 · « **Reply #46 on:** January 17, 2021, 04:54:50 pm »

The project is available on gitlab for those interested: https://gitlab.com/bogdan-kicad/eevblog-linear-psu

Quote

Depending on how your transformer's secondary windings are, it might be possible to reduce the unregulated voltage for when the output is set to low voltages.

I could do that, yes. The transformer has a 12V tap off of the 24V secondary winding.

bogdan2014 · « **Reply #47 on:** October 06, 2021, 05:13:35 am »

I'm in the process of building the second PSU of this kind. I'm happy to say that the first one has proven to be very reliable. However there's one thing that made me curious. In this topology if the control circuit fails, the output goes to the highest possible voltage. I've seen this in most linear power supplies, but is there a way to build it so that the op-amp biases the output transistor instead of pulling the base down?

xavier60 · « **Reply #48 on:** October 06, 2021, 07:06:14 am »

Nice job!
I don't think it's possible to design a control circuit that can be trusted to inherently fail to safe.
Assuming that the power transistors are kept well within their SOA and safe temperature, the greatest risk I found was the resistance wire in the 10 turn pots breaking, suddenly sending the PSU to its maximum voltage setting.
This is the design being currently used in my own PSU.
https://www.eevblog.com/forum/beginners/lm324-power-supply-with-variable-voltage-and-current/msg3582664/#msg3582664
Later in the thread there is a tap changer circuit.
With this topology, a load response test needs to be done to check the stability.
The compensation can change with different parts used.

bogdan2014 · « **Reply #49 on:** October 06, 2021, 05:44:25 pm »

The first one has a 10-turn pot for voltage and a standard pot for current. This time I'll use standard pots with fine/coarse adjustment. The 10 turn pots are chinese "bourns" and not once I had them fail.
I'm not happy about the panel meter, these are the only digital ones that I could find and that display both voltage and current, but don't like them. I'm thinking of building one myself with an MCU.

xavier60 · « **Reply #50 on:** October 06, 2021, 08:23:41 pm »

Quote from: bogdan2014 on October 06, 2021, 05:44:25 pm

The first one has a 10-turn pot for voltage and a standard pot for current. This time I'll use standard pots with fine/coarse adjustment. The 10 turn pots are chinese "bourns" and not once I had them fail.
I'm not happy about the panel meter, these are the only digital ones that I could find and that display both voltage and current, but don't like them. I'm thinking of building one myself with an MCU.

The few 4 digit meters I have bought lately are much better, the current metering is auto-zeroing.
I make some comment here, https://www.eevblog.com/forum/projects/volts-amps-digital-meters/msg3685030/#msg3685030.
An MCU could be used for other tricks also. Mine toggles the PSU On/Off with the press of a momentary button and controls the fan.
When the 10 turn pots started failing, I added protection that watches for sudden jumps in wiper voltage.
Since switching to Bochen wxd3-13-2w 10k pots, I have had no more open wire type failures. They can go scratchy. Some PAO oil fixes that.
The wxd3-12-2w series pots are 5 turn and shorter.
My MCU also does the transformer tap changing.

bogdan2014 · « **Reply #51 on:** November 25, 2021, 10:35:20 am »

I ordered some Bochen pots and tried them. When the voltage pot reaches one end (the 0V one), the BD140 transistor goes shortcircuit between E-C. No idea why, it didn't happen with the normal pots. I'm also investigating using digital pots, and I can also use an LCD with the MCU to measure the voltage and current.

xavier60 · « **Reply #52 on:** November 25, 2021, 10:42:31 am »

Quote from: bogdan2014 on November 25, 2021, 10:35:20 am

I ordered some Bochen pots and tried them. When the voltage pot reaches one end (the 0V one), the BD140 transistor goes shortcircuit between E-C. No idea why, it didn't happen with the normal pots. I'm also investigating using digital pots, and I can also use an LCD with the MCU to measure the voltage and current.

That's very strange. Was the PSU loaded? What is the unregulated voltage into the regulator?

xavier60 · « **Reply #53 on:** November 25, 2021, 11:13:45 am »

Is the BD140 on a heat sink? How many times has it failed in the same way?

bogdan2014 · « **Reply #54 on:** November 26, 2021, 05:56:11 am »

It failed each time I tested it with the new pot (twice until now). It was on a heatsink but it never got hot, and it failed very quickly, I just heard a small pop and that was it, no smoke no heat. No other components were damaged. I teste both with the PSU loaded and unloaded, same thing. I'll investigate further today.

bogdan2014 · « **Reply #55 on:** March 19, 2022, 04:45:57 pm »

I have two identical transformers (30VAC and about 3-5A) and I'm thinking of making a dual PSU out of them. I will be building two of these boards and run them in series, and I'll have dual pots to adjust voltage and current at the same time. However, I can't fit two voltmeters and two ammeters so I'm thinking of only using one, but measuring the current at the center tap. Is that possible?
The panel meter is one of those chinese digital ones with seven segment displays, with both volts and amps reading in the same unit. It measures current at the return terminal.

xavier60 · « **Reply #56 on:** March 19, 2022, 11:34:05 pm »

Just curious about how the earlier PSU project is working out?

bogdan2014 · « **Reply #57 on:** March 20, 2022, 07:38:07 am »

Quote from: xavier60 on March 19, 2022, 11:34:05 pm

Just curious about how the earlier PSU project is working out?

It's actually that one that I'm trying to modify into a dual tracking PSU. I gave up on multiturn pots and just used plain old single turn ones and it works as it should.
I even played around with digital pots but that's another story.


EEVblog Main Site	EEVblog on Youtube	EEVblog on Twitter	EEVblog on Facebook	EEVblog on Odysee

Author Topic: How this linear PSU circuit works, and why it doesn't (Read 29500 times)

Share me