Questions about modifications to AliExpress SMPS units

[intabits]

My questions relate to modifying these cheap SMPS units available from AliExpress:-
https://www.aliexpress.com/item/1005005626231303.html?spm=a2g0o.order_list.order_list_main.53.30481802SpMQDT


But for some background: The SMPS will be used to power two of these OPA541 power opamp modules that I have already:
https://www.aliexpress.com/item/1005002532091168.html


These modules can run from up to +/-40V and deliver 5A continuously. They reportedly work quite well, as described by HWCave in this video:-
https://www.youtube.com/watch?v=925X0RDUy9w&t=241s&pp=ygUNSFdDYXZlIE9QQTU0MQ%3D%3D

But at higher output currents and lower output voltages, thermal considerations and keeping the OPA541 in its SOA can be difficult:-
https://www.youtube.com/watch?v=ERJ9Om-eBqo&pp=ygUNSFdDYXZlIE9QQTU0MQ%3D%3D
(Fans and some heatsink improvements are required)


PSU questions
I have two of the 36V 15A supplies on order. So I can put them in series to get a +/-36v supply, but then the two OPA541s would not be fully isolated and independent. Four smaller supplies could do that, but that's getting a bit ridiculous. So I'm wondering if these supplies can be modified to provide + and - outputs. From the work done by "imajeenyus" linked below, all these supplies seem to have circuitry similar to this (with appropriate components to match the required specs):-


Question 1: Would it be feasible to duplicate the shaded area at the top right to provide a negative output (with reversed component polarities)?
Although the regulation is only done from the positive side, would an asymmetric load on the -ve side be reflected at least in part on the +ve side?
With some minimum load on the +ve, it would quickly match the -ve, no?
Is there a practical way to bring the -ve side into the regulation loop?   

Question 2: One solution for the thermal and SOA issues is to reduce to voltage from the PSU as required, and Imajeenyus has come up with a way to modify the adjustment range of the PSU to cover a 40%-120%, as described here:-
http://imajeenyus.com/electronics/20151028_smps_variable_voltage/index.shtml
He achieved this by replacing the adjustment section (shaded area at left of schematic above) with this:-


This allows manual voltage control to mitigate SOA problems, but it would be nice to adjust this automatically by monitoring the amplifier output voltage.
This is beyond my analog design skills, but I'd be interested in any ideas or pointers that I could investigate towards doing this.

But maybe it could all be digital, as there will be a micro-controller floating around in whatever I end up doing, and it would also be nice to make the power supply programmable. So I'm thinking of digital control of the PSU voltage. I envisage using a digital potentiometer in the circuit above, which will need to be driven by a digital isolator, with a local +5v supply. And some sort of peak detection circuit for feedback to the controller. (the MCP41HVX1 digital pot can handle the 36v environment)

Your thoughts?

[moffy]

Such a question depends upon the detail of the power supplies you are modifying, being 'similar' is not sufficient to provide an answer. If you can find an exact schematic for the supply you are purchasing then some advice might be possible.

[Phil1977]

I once tried to use a digital potentiometer in the feedback loop of a step-up-converter - unfortunately it locked up due to EMI or any disturbance and the DC-DC went havoc, destroying its diode, the digital potentiometer and the arduino.

But what worked was a NPN-transistor parallel to the adjustment potentiometer. The regulation loop of the converter was filtered so well that I could directly feed the NPN with a 10kHz-PWM signal from the microcontroller. If you want to I can sketch up the schematics.

I think you are aware that an analogous regulation loop (the OpAmp) needs its supply voltage before its too late. A digital pre-regulation can only work if the signal gradients are either quite slow or predictable.

[Whales]

If the transformer does provide a split (3-terminal) output then yes I could see you modifying the highlighted bit of circuit to give +/- outputs.  The two half rails would be weaker however (half the current limit?)

I do not expect all power supplies to be wired that way or even have split-output transformers.

Question 1: Would it be feasible to duplicate the shaded area at the top right to provide a negative output (with reversed component polarities)?
Although the regulation is only done from the positive side, would an asymmetric load on the -ve side be reflected at least in part on the +ve side?

I believe this is called "group regulation", lots of ATX computer power supplies do it.

An audio amp shouldn't present asymmetric loads on the rails during normal operation, so I suspect it would be fine.

Question 2: One solution for the thermal and SOA issues is to reduce to voltage from the PSU as required,

Controller ICs need feedback with certain parameters.  Sometimes using too large or too small resistors can upset the feedback loops (as per Phil1977) causing things to misbehave and/or blow up.

Small tweaks are OK, but installing pots may or may not work.  Also if the pot is dirty then the PSU might blow up whilst you are turning it.  It would be safer to used fixed voltage dividers, but again be wary of switching between them with a switch (depending how you implement it).

Digital control of voltage dividers for PSU feedback probably has gotchas too  EDIT: Do digital pots have capacitance/bandwidth issues, or do they work OK?

[Hiemal]

When you're switching/messing around with the feedback using digital control the best solution is to usually feed the midpoint of the voltage divider with a resistor and DAC output. There's a calculator built into Ti's Power Stage designer tool that lets you figure out what values to use for whatever resistances you have in place in circuit. Bonus is too is this method also doesn't mess with your circuits loop/gain characteristics.

[thm_w]

Sounds incredibly complicated.

- What voltage do you need, is +/-36V and 5A necessary?
- Why do the two opamps need to be isolated?

[pienari]

Is the psu quality any better then few years ago?.
What i measured 12v 30A psu:s ab. 5yrs ago psu:s make lots of noise.
12v 40A black ones was so bad i cant plug it to mains when im listening HF radio.
S9 +30db noise all over the bands.

[Roehrenonkel]

Hi intabits,
 
...why so complicated?
You could connect + to + and - to - and create a "virtual ground" with a resistive divider
(maybe with an OPAmp to buffer).
 
Good luck

[intabits]

Thanks for all the helpful and thought-provoking responses!

Such a question depends upon the detail of the power supplies you are modifying, being 'similar' is not sufficient to provide an answer. If you can find an exact schematic for the supply you are purchasing then some advice might be possible.

By "similar" I mean almost "exactly", with only component values changed to meet the specs for each model. That's what I read somewhere while digging into this. In any case, I'd like us to assume the schematic is correct for the purpose of my questions. Obviously I'll be checking that when I get my specific units. But in the areas of interest indicated, any variations will hopefully be minor, and can be allowed for. I think my questions are general enough to proceed on that basis.

I once tried to use a digital potentiometer in the feedback loop of a step-up-converter - unfortunately it locked up

due to EMI or any disturbance and the DC-DC went havoc, destroying its diode, the digital potentiometer and the

arduino.

Sounds not too good. I'll still get some DP's and play around with this. But I'll use an isolator so that Arduino should be safe.

But what worked was a NPN-transistor parallel to the adjustment potentiometer. The regulation loop of the converter was filtered so well that I could directly feed the NPN with a 10kHz-PWM signal from the microcontroller. If you want to I can sketch up the schematics.

Your description is clear enough. This is a neat solution, the NPN could even be an optocoupler's transistor. The replacement adjustment schematic shown was to get a more linear response to potentionmeter angle, but that's not so important when a micro is controlling it. I'll definately play around with that.

I think you are aware that an analogous regulation loop (the OpAmp) needs its supply voltage before its too late. A digital pre-regulation can only work if the signal gradients are either quite slow or predictable.

That was a concern, so maybe automatic adjustment is not such a good idea. With fans and a modified heatsink on the opamp, one could have a larger headroom between PSU output voltage and the opamp's needs. Anything less than the full 36v output would be a help.
And I do envisage that mostly it would be a slow and predictable signal. And in that case, no automation is needed, as the PSU voltage can just be programmed accordingly.
One example usage would be as a programmable "variac", by driving the opamp into the secondary of a step down transformer.

If the transformer does provide a split (3-terminal) output then yes I could see you modifying the highlighted bit of circuit to give +/- outputs.  The two half rails would be weaker however (half the current limit?)
I do not expect all power supplies to be wired that way or even have split-output transformers.

The PSUs I'm getting are 15A units, so current won't be a problem. I'm hoping that they are wired like that. I should know in a few days...

I believe this is called "group regulation", lots of ATX computer power supplies do it.

Duh! Of course, I hadn't thought of that. Many SMPSs only use one output for feedback, so that's a good sign.

An audio amp shouldn't present  loads on the rails during normal operation, so I suspect it would be fine.

The amp is DC-coupled, and will be used to extend the capabilities of a function generator. I imagine there will be cases where the signal will be asymmetrical, or have a DC offset. So something to watch out for.

Controller ICs need feedback with certain parameters.  Sometimes using too large or too small resistors can upset the feedback loops (as per Phil1977) causing things to misbehave and/or blow up.

Small tweaks are OK, but installing pots may or may not work.  Also if the pot is dirty then the PSU might blow up whilst you are turning it.  It would be safer to used fixed voltage dividers, but again be wary of switching between them with a switch (depending how you implement it).

Yes, I'll need to experiment carefully. Some of the reading I've done suggested that there should be resistors across the pot to set a safe output should the wiper get "scratchy".

Digital control of voltage dividers for PSU feedback probably has gotchas too  EDIT: Do digital pots have capacitance/bandwidth issues, or do they work OK?

I shall find out...

When you're switching/messing around with the feedback using digital control the best solution is to usually feed the midpoint of the voltage divider with a resistor and DAC output. There's a calculator built into Ti's Power Stage designer tool that lets you figure out what values to use for whatever resistances you have in place in circuit. Bonus is too is this method also doesn't mess with your circuits loop/gain characteristics.

Minimum destabilizing influence on standard operation seems like a worthwhile goal. I'll look into experimenting with what you've described as well.

Sounds incredibly complicated.
- What voltage do you need, is +/-36V and 5A necessary?
- Why do the two opamps need to be isolated?

Doesn't bother me if it gets complicated, I'm quite happy for this to become a substantial project if I can justify it.
It will be nice to have a general purpose box with two 36V/15A programmable PSU's, as well as the power opamps to suit.
The "variac" example function mentioned above can make use of all the power it can get.
In the general purpose piece of equipment that I'm thinking about, two isolated halves would always be a good thing as it's less restricted and more flexible.

Is the psu quality any better then few years ago?.
What i measured 12v 30A psu:s ab. 5yrs ago psu:s make lots of noise.
12v 40A black ones was so bad i cant plug it to mains when im listening HF radio.
S9 +30db noise all over the bands.

I don't know, I hope not, I'll find out!

...why so complicated?
You could connect + to + and - to - and create a "virtual ground" with a resistive divider
(maybe with an OPAmp to buffer).
Good luck

Thanks. But creating a "virtual ground" for a high power supply? Wouldn't that require another power opamp, just to hold the virtual ground still? (and connecting the two PSUs together would only be required if I can't get a negative output from each)

[Phil1977]

I think you are aware that an analogous regulation loop (the OpAmp) needs its supply voltage before its too late. A digital pre-regulation can only work if the signal gradients are either quite slow or predictable.

That was a concern, so maybe automatic adjustment is not such a good idea. With fans and a modified heatsink on the opamp, one could have a larger headroom between PSU output voltage and the opamp's needs. Anything less than the full 36v output would be a help.
And I do envisage that mostly it would be a slow and predictable signal. And in that case, no automation is needed, as the PSU voltage can just be programmed accordingly.
One example usage would be as a programmable "variac", by driving the opamp into the secondary of a step down transformer.

That sounds like a nice idea for an universal AC source/sink. Combine a digitally controlled PSU with a high-power-opamp and make it configurable to be either in "Low-noise-low-power"-mode or in "what-if-we-use-more-power"-mode.

Maybe you could even connect a clipping detector. E.g. a second low power opamp parallel to the inputs of the high-power-opamp to rectify the voltage difference between its inputs. You could build a peak-detector and connect it to an ADC of the microcontroller. If clipping is detected, the controller could raise the supply voltage. If no clipping is detected for e.g. 1min you could marginally reduce the voltage again.

[moffy]

There are some subtleties in the schematic shown, the capacitor C10/3.3uF in series with the main transformer suggests some sort of resonant/semi resonant topology especially with the low value filter inductor on the secondary which is placed in the ground leg.

[Roehrenonkel]

Hi intabits,

Thanks for all the helpful and thought-provoking responses!

...why so complicated?
You could connect + to + and - to - and create a "virtual ground" with a resistive divider
(maybe with an OPAmp to buffer).
Good luck

Thanks. But creating a "virtual ground" for a high power supply? Wouldn't that require another power opamp, just to hold the virtual ground still? (and connecting the two PSUs together would only be required if I can't get a negative output from each)

Yes, you're right, how stupid of me.... B-(
I forgot about the load! It would work for a symmetric (Bridge)-Amp (load not grounded).
With the + to + i meant +PSU to +Amp-supply...... ;-)

Good luck

[intabits]

That sounds like a nice idea for an universal AC source/sink. Combine a digitally controlled PSU with a high-power-opamp and make it configurable to be either in "Low-noise-low-power"-mode or in "what-if-we-use-more-power"-mode.

Maybe you could even connect a clipping detector. E.g. a second low power opamp parallel to the inputs of the high-power-opamp to rectify the voltage difference between its inputs. You could build a peak-detector and connect it to an ADC of the microcontroller. If clipping is detected, the controller could raise the supply voltage. If no clipping is detected for e.g. 1min you could marginally reduce the voltage again.

I'm not sure what you mean by "Low-noise-low-power" and "what-if-we-use-more-power" modes. But I assume you mean:-
a) Avoid clipping by giving it all the voltage it could possibly need, or
b) Minimize dissipation, but increase voltage if it starts clipping.

The peak detector that I mentioned earlier was along those lines of your clipping detector, but my concern was response time. (Which I suppose is why you proposed those two modes.)
I was thinking of comparators watching for the output voltage getting too close to the supply rails.

That could always be done, and makes its use optional. And simpler (for me at least) than trying to combine some automatic analog circuitry with the voltage control stuff.
Added to the list...

There are some subtleties in the schematic shown, the capacitor C10/3.3uF in series with the main transformer suggests some sort of resonant/semi resonant topology especially with the low value filter inductor on the secondary which is placed in the ground leg.

I'm certainly no SMPS expert (which is why I'm asking these questions), so that may well be the case.
Are you saying that these components may influence the modifications that I'm proposing?

But my understanding is that:-
The cap is there simply to block DC and stop everything frying if the transistors are not switching. Which could be due either to a failure, or  because some fault or other condition requires the PSU to shutdown. 
The inductor is just the normal filtering used in SMPSs to help supress high frequency switching noise on the output.
Both components seem to be almost universal in the ATX PSU schematics found at https://danyk.cz/s_atx_en.html

I'm also curious to know why the other side of TR2's primary is not connected directly to the switching point of the transistors (Q1C to Q4E), but instead via a winding on TR1?
This also appears to be common practice, but sometimes its a separate inductor. So being part of TR1 must not be important? And it doesn't interfere with TR1s transistor driving function?

[moffy]

There are some subtleties in the schematic shown, the capacitor C10/3.3uF in series with the main transformer suggests some sort of resonant/semi resonant topology especially with the low value filter inductor on the secondary which is placed in the ground leg.

I'm certainly no SMPS expert (which is why I'm asking these questions), so that may well be the case.
Are you saying that these components may influence the modifications that I'm proposing?

But my understanding is that:-
The cap is there simply to block DC and stop everything frying if the transistors are not switching. Which could be due either to a failure, or  because some fault or other condition requires the PSU to shutdown. 
The inductor is just the normal filtering used in SMPSs to help supress high frequency switching noise on the output.
Both components seem to be almost universal in the ATX PSU schematics found at https://danyk.cz/s_atx_en.html

I'm also curious to know why the other side of TR2's primary is not connected directly to the switching point of the transistors (Q1C to Q4E), but instead via a winding on TR1?
This also appears to be common practice, but sometimes its a separate inductor. So being part of TR1 must not be important? And it doesn't interfere with TR1s transistor driving function?

This is a quote from this site: http://imajeenyus.com/electronics/20151028_smps_variable_voltage/

"Bridge transistors + base drive + main transformer
(TR1 is the base drive transformer, I've also called it "gate" transformer sometimes. TR2 is the main transformer.)

The two bridge transistors (Q4 & Q1) switch one end of the transformer's primary between 0V and the full DC supply voltage. There is some very clever trickery going on here which I only barely understand. First, the extra resistors such as R14, R13, R8, R4 bias the main transistors on slightly during startup (keep in mind that the auxiliary supply isn't available during startup, so the TL494 isn't running). One transistor will turn on slightly quicker than the other. If you look closely, note that the bottom end of the main transformer's primary isn't connected directly to the midpoint of the two transistors - rather, it goes through a winding on the base drive transformer. As current starts flowing in the main transformer primary, it induces a current in the base transformer windings, one of which will assist the already-on transistor, switching it fully on. Thanks to some resonant and saturation trickery (probably involving C10, in series with the transformer primary), this whole process repeats for the other transistor, and the whole bridge self-oscillates. This will provide enough power to get the auxiliary supply up (it reaches about 10V, but that can vary), and start the TL494 running, at which point it takes over and controls the switching of the bridge transistors.

Another extremely neat feature of this configuration, in addition to the self-starting capability, is that the TL494 doesn't have to provide the full base drive current to the bridge transistors - the base drive current actually comes from the primary current, coupled through the base drive transformer. The drive transistors on the primary of the base transformer simply manage to control which of the main transistors is held on by the primary current.

All that is an extremely loose and incomplete explanation. Fortunately, there is a fantastic page which describes exactly how it does work - Manfred Mornhinweg has a page on building a 13.8V 40A supply and his design uses nearly exactly the same circuit (or rather, the Chinese supply uses the same circuit as he, since his one probably came first!).

Fortunately (2), understanding the actual operation of this part isn't essential to understanding the rest of the supply, so I wouldn't worry about it too much. It Just Works™."

The base switching transformer configuration, is similar to that used in self resonant induction heaters, which use LC at a resonant frequency, hence my mentioning of C10, resonant switching allows for near zero volt switching ZVS which helps reduce EMI and switching losses, it is also a tuned system which can be disturbed by adding extra devices without proper thought. It's unusual to see the filter inductor in the ground leg of the secondary, normally it's, in this case, on the positive output leg. As the 'Transformers' would say there is "more than meets the eye."

[Phil1977]

I'm not sure what you mean by "Low-noise-low-power" and "what-if-we-use-more-power" modes. But I assume you mean:-
a) Avoid clipping by giving it all the voltage it could possibly need, or
b) Minimize dissipation, but increase voltage if it starts clipping.

The peak detector that I mentioned earlier was along those lines of your clipping detector, but my concern was response time. (Which I suppose is why you proposed those two modes.)
I was thinking of comparators watching for the output voltage getting too close to the supply rails.

That could always be done, and makes its use optional. And simpler (for me at least) than trying to combine some automatic analog circuitry with the voltage control stuff.
Added to the list...

I was just thinking loud and wrote it down here... I just like playing with opamps and analogue signals, that´s why I would monitor the bias voltage of the high-power-opamp and feed it to a peak detector.

Monitoring the output voltage and comparing it to the PSU-voltage with a comparator should fulfil the same purpose though it may be a little less sensitive in high-load-situations. The bias-voltage of the opamp is a quite clear indicator if the amplification and feedback loop is healthy or if anything is wrong - but anyhow, innovation happens if you follow a variety of ideas!

And yes, my proposed two modes are exactly what you describe. One low-(fan-)noise-low-dissipation mode and one full power mode for maximal signal integrity.

By the way: Astonishing how clever this ZVS-like control circuit is...

[intabits]

This is a quote from this site: http://imajeenyus.com/electronics/20151028_smps_variable_voltage/

All that is an extremely loose and incomplete explanation. Fortunately, there is a fantastic page which describes exactly how it does work - Manfred Mornhinweg has a page on building a 13.8V 40A supply and his design uses nearly exactly the same circuit (or rather, the Chinese supply uses the same circuit as he, since his one probably came first!).

Fortunately (2), understanding the actual operation of this part isn't essential to understanding the rest of the supply, so I wouldn't worry about it too much. It Just Works™."

I overlooked the theory part of the imajeenyus article. At the time I was only interested in the "Designing a new feedback divider" section. I should have read all of it (and will). And I'll google for the Mornhinweg stuff as well.
After my last post, I did some further reading and did see mention of resonances being an integral part of SMPS operation, so thanks for making me aware (and wary) of possible issues there. Similarly for the output inductor.
But I'm also hoping the "It Just Works™" approach will get me by for now...

...I just like playing with opamps and analogue signals...

Me not so much. I'm more comfortable with digital things, so comparators are attractive to me.
Also, I'm just not clear on what your method is. When you say "The bias-voltage of the opamp is a quite clear indicator if the amplification and feedback loop is healthy or if anything is wrong", sounds like something I shouldn't ignore. 
By "a second low power opamp parallel to the inputs of the high-power-opamp to rectify the voltage difference between its inputs" and a peak detector, are you saying that a significant difference across the high-power-opamp's inputs indicates clipping? Because the difference means the feedback loop is no longer able to keep the inputs at the same voltage, because it's hit the supply rail?

That makes sense to me if I've got it correct, and seems worth exploring. But a bit more tricky to interface with a micro (versus a comparator driving an opto-coupler).
I want to make the two SMPS+OpAmp systems fully independent and floating, so that they can even be put in series. 

Anyway, the PSU's have arrived, so time to verify them against the schematic above, and build a test-bed to explore the ideas described above!

[Phil1977]

Frankly speaking I do not know if the "bias-voltage-peak-detector" as proposed by me is the better solution for your application. I´ve seen this design in a very old instrumentation amplifier, and it gave a really good visual feedback by an attached panelmeter if clipping is detected or not.

But if you just want to know if the supply voltage is sufficient, your proposal (voltage comparator of output voltage against supply voltage*0.9 e.g.) may be the better way. There is more than one reason for a too high input voltage difference of an opamp, and insufficient supply voltage is only one of them. But your idea focuses on checking the voltages, so it may be better suited for this feedback circuit.

In conclusion: I think this is a design that definitely needs a prototype or simulation. It´s probably less efficient to discuss theoretical advantages /disadvantages than making a test with a scope and breadboard. You can like or dislike Elon Musk, but "try and fail early" really is one of the best ways to learn...

Anyhow, its great you can discuss these ideas in this forum... please always give feedback about the result!

[intabits]

The power supplies arrived a couple of days ago.
First up: test output voltage and current on a dummy load.

Wired up...



Test setup. Dummy load of 9x12V50W halogen bulbs. (Wired in 3P3S)



Voltage & Current. Volts measured at load, after the ammeter. 36Volts@13Amps. (463Watts)
(Imgur decide to flip this upside down)



Internals - Top



Internals - Bottom. Flipped horizontally to match top image.



Above two images merged. After buggarizing around with graphics libraries for 2 days, I finally have a vaguely usable top & bottom overlay image.


There is provision for a second pair of output diodes.
But not for +/- output, looks to be in parallel with the other, for higher output current models.

Now for some experimenting with modifications...

[moffy]

Congrats, at least a cursory look seems to have quite a similarity between the schematic and your purchased PSU. The dual output diode is fed from both sides of the output winding instead of one side which makes sense, because with only one dual output diode that provides balance for the transformer current, and C8 looks like C30 in the schematic and the inductor is in the 0V leg of the output.

[intabits]

Yes, turns out that it's very similar.

Somehow I'd got it into my head that the TL494 is a 14 pin chip. And seeing that the my 540W version had a 16 pin chip, a closer look showed it to be a  KA7500B. So figuring its a different chip (and circuitry) altogether, I set about reverse engineering the circuitry around that chip. Then I saw a video stating that the TL494 and KA7500B are in fact, identical chips.
Checking the TL494 datasheet, I saw that it was indeed a 16 pin job after all. And furthermore, that my reverse engineering had produced the same circuit as given above by Imajeenyus for the 400W model. So I completed the reverse engineering for the whole circuit. Here is the original circuit with changes in the 540W version shown in the shaded areas.


This is the circuit for my 540W model, with part designators for the PCB shown above. (rearranged and minor errors fixed)


My speculation that the extra winding on the drive transformer could be a separate inductor appears to be incorrect.
According to DiodeGoneWild @17:45 in https://www.youtube.com/watch?v=wNnOfF1NkxI, that extra winding is for feedback to make the main switching circuit self oscillating at startup, until the auxiliary winding of the main transformer produces supply power for the chip, which then takes over control of the switching.
And this winding also provides extra current to operate the switching transistors, reducing the current requirements on the smaller transistors connected between the chip and the drive transformer. (not sure how that black magic works)

[Phil1977]

If you look closely, note that the bottom end of the main transformer's primary isn't connected directly to the midpoint of the two transistors - rather, it goes through a winding on the base drive transformer. As current starts flowing in the main transformer primary, it induces a current in the base transformer windings, one of which will assist the already-on transistor, switching it fully on. Thanks to some resonant and saturation trickery (probably involving C10, in series with the transformer primary), this whole process repeats for the other transistor, and the whole bridge self-oscillates. This will provide enough power to get the auxiliary supply up (it reaches about 10V, but that can vary), and start the TL494 running, at which point it takes over and controls the switching of the bridge transistors.

Another extremely neat feature of this configuration, in addition to the self-starting capability, is that the TL494 doesn't have to provide the full base drive current to the bridge transistors - the base drive current actually comes from the primary current, coupled through the base drive transformer. The drive transistors on the primary of the base transformer simply manage to control which of the main transistors is held on by the primary current.

All that is an extremely loose and incomplete explanation. Fortunately, there is a fantastic page which describes exactly how it does work - Manfred Mornhinweg has a page on building a 13.8V 40A supply and his design uses nearly exactly the same circuit (or rather, the Chinese supply uses the same circuit as he, since his one probably came first!).

Fortunately (2), understanding the actual operation of this part isn't essential to understanding the rest of the supply, so I wouldn't worry about it too much. It Just Works™."

The base switching transformer configuration, is similar to that used in self resonant induction heaters, which use LC at a resonant frequency, hence my mentioning of C10, resonant switching allows for near zero volt switching ZVS which helps reduce EMI and switching losses, it is also a tuned system which can be disturbed by adding extra devices without proper thought. It's unusual to see the filter inductor in the ground leg of the secondary, normally it's, in this case, on the positive output leg. As the 'Transformers' would say there is "more than meets the eye."

Isn't this extra winding exactly what moffy has explained?

[moffy]

Just a small comment, on your schematic ZD1/ZD2 are inverted.

[intabits]

Isn't this extra winding exactly what moffy has explained?

Yes. I did not give that post the attention it deserved. My Bad!
It mostly went over my head, as I had not yet wanted nor attempted to get a deeper understanding of these things. And being mostly concerned with the original questions of making a negative output and controlling the output voltage remotely, I unfortunately only skimmed that post.
At least I did take the "...I wouldn't worry about it too much. It Just Works™." advice to heart!

Just a small comment, on your schematic ZD1/ZD2 are inverted.

Oops! I'll fix that... Thanks!

[moffy]

Just an FYI, C8 is just an antisaturation cap to prevent DC through the transformer core.

[intabits]

So I've got external voltage control via PWM working pretty nicely. And via an optocoupler enables the supply to be fully floating.
A short video of the operation:-


Here is the schematic:-


It's very simple, and requires minimal change to the PSU - one resistor removed, and two wires added.

The internal modifications (also for negative output voltage):-


But it's all pretty fiddly, and next time I would implement this completely differently.


Once again I find my self wondering what I was thinking. My very first question in this thread about negative output and whether + and - voltages would track together, given that regulation only watches the + output. I must have been thinking of two secondary windings on the transformer, but there is only one. In fact I'm really just connecting a bridge rectifier to a center-tapped transformer, so of course the voltages will track, subject only to load imbalances that could affect the drop across the diodes and ripple voltage on the filter capacitors.

So I added that, but obviously stuffed something up. When I applied power with my negative output filter and terminal board connected, something in the PSU cracked - so I immediately switched it off. You can see that one of the caps is about to pop:-


I haven't investigated this yet.