Audiophile Linear Power Supply Questions

[lunar]

Here is a linear power supply for audio. Someone asked me to try to repair this.
It was outputting 42V, and it was because the original toroidal transformer had failed, and as
far as I could tell that was it.

It is designed to output around 24V with 1A, as listed on the back of the supply.
There are four transistors used. Most are smd NPN/PNP, the 2GW (npn) and 558
(pnp). The 2 power transistors are D45H onsemi (pnp). There is also 1 to-92 k363 which I
think is a FET. There are two 40V 5600uf Kemet caps as input filtering, and some
tantalums, passive resistors, and an unidentified capacitor type. There's also one
diode on the second section.

2gw      - https://www.mouser.com/datasheet/2/916/BC850C-3196933.pdf
558      - https://www.diodes.com/assets/Datasheets/FMMT558.pdf
d45h11g  - https://www.onsemi.com/pdf/datasheet/d44h-d.pdf


I have some questions about the design for this.


- I am not familiar with what they are doing here. Is this feedback to keep
the voltage stable with the transistors? Is there a name for this topology?
Note that it is divided into two sections. The first section has a voltage divider
going into the base of the first smd transistor, then there is a pair (which
is also on the end of the second section as well). This feeds into the first power transistor
of which there are two. One at the end of each section.

What is the point of the voltage divider on the base at the start? What is the point of the npn and pnp
pair at the end of each section?

- What is the FET doing in the second half of the circuit? It seems to measure
as a voltage reference. I only tested it so far with the 42V but it was outputting around
26 volts. I assume with 24V it will output less. Everything else on the circuit was
around 42V, the FET was the only voltage that was different. I haven't tested it since
changing out the transformer.
- Is anyone interested in me reverse engineering the
second half of the circuit? I've looked at this long enough, I could probably make
my own audio power supply if I wanted to by now. Maybe someone just knows it by
looking at it. It is simple enough.
- Is there any problem with them not addressing the inrush on the toroidal transformer?
The original transformer had failed, it was outputting 42V but it was a 24V transformer.
Most switchers I see have some kind of inrush protection via the mov, thermistor, and coils.
I notice that they did nothing here. I assume as a design decision to limit noise from
extra components, but was that wise?
I replaced it with a Hammond, but the hammond website notes that inrush current is
a problem with toroidals.
- What type of capacitors are the shiny metal ones? I have seen the black tantalums before,
but I haven't seen anything that looks like these metal capacitors that are used
on the output.
- Is it even worthwhile to put a bleeder resistor across the output for the capacitors?
It seems like something else that would just add to the overall noise. Though it does look
like they used a metal film, not something like a carbon. The closest replacement for
looks I could find was a 2W 2.2K stackpole, and it looks very close, if not identical, though
the color bands on the original don't match up 1:1.

Attached is a few photos of the board, and a text file of the falstad schematic for the first section.
There is also a screenshot of what the falstad schematic looks like.

EDIT: I went ahead and reverse engineered the rest of the circuit. I don't understand the K363 / GR91 component. It doesn't appear to be wired correctly, and the simulation doesn't do much (it overloads falstad a bit), even if adding a load. Anyways, if anyone is interested, there is the 24V 1A circuit, hopefully without errors.

[MathWizard]

So far with a J210 JFET as the SK363, I can't get it to work. And I don't see where the base current of the lower DH45 is supposed to go. If it's meant to go through the lower right BC558, backwards as some protection, it's not doing it in LTSpice.

Does your model work ?? I haven't tried to calculate anything, but my J210 has a turn on of -1.5V and is on a bit /w ~400uA, but w/ 24Vcc I get 21.2V on the emitter of the lower DH45 which is the series pass output element. But I only get 21.1V on it's base so it's way off.  The bottom of the diode is at 20.8V

[magic]

This schematic likely has numerous errors, little of it makes much sense.

[CaptDon]

I agree, that schematic can't do anything useful as drawn.

[lunar]

That was my fault, there were errors on the schematic. I found that if you make an overlay in Gimp/Photoshop with both images, and then draw in the second layer tracks, it's a bit easier to reverse engineer. The first time I tried to just trace it out with a meter, which was not the right way to do it. I know this is discussed in videos online, but this is the first time I used Gimp exclusively. I have tried importing as an image into KiCad before, but I had trouble with that. Gimp is better as you can adjust the opacity of the bottom layer as you work around the board.

See attached for the updated files. Measuring on the board, it looks like the JFET and the two DH45 are both shorted. But I am waiting on a desoldering station, and will remove them to double check. Nothing else on the board measures wrong (although I don't have an ESR meter). I checked all junctions of the transistors, the diodes, and the resistances while I was at it.

Reverse engineering and making schematics is not an easy thing to do when you don't know what topologies or whatnot you are dealing with. I think I need to find some resources on the subject.

If anyone is interested in the Gimp xcf file, I can upload it somewhere online (it is 42MB).

[Calvin]

Hi,

since the PCB features a nice silkscreen printing, it might just be helpful to also know the names of the parts on the schematic sketch.
Then someone might be able to sketch the different buiding blocks for easier recogition and correctness.
So far the JFET comprises a ccs ... not uncommon in the regulatory part of the reg.

regards
Calvin

[magic]

It would still be helpful if you could draw it more conventionally - high voltages up, low voltages down, input left, output right. But the function is becoming mostly apparent and it's a pretty simple circuit.

First you have some voltage divider and I'm not sure what's the point of the NPN. Are you sure the capacitor is on its emitter and not its base? Anyway, this generates some voltage which is ~90% of the input. Then you have the second NPN emitter follower and a pair of PNPs set up to amplify its collector current - it's a Sziklai pair with a Darlington pair inside it if you want names. Anyway, the NPN emitter is the output of this section and it's a somewhat-regulated voltage about 90% of the input.

The JFET is a constant current source and the two transistors below is another Sziklai pair set up as a Vbe-multiplier - a voltage referece. Not sure about the red dot - did you mean an LED? There is a circuit symbol for this... Anyway, the reference voltage is lowpass-filtered by 100k and 2x10μF and applied to another Sziklai-Darlington emitter follower.

The end.

[5U4GB]

This schematic likely has numerous errors, little of it makes much sense.

That's what you get when you ask ChatGPT to design a power supply for you :-).