Self oscillating flyback converter (SMPS)

[faxmodem]

Hi there,

I've been trying to repair a switching power supply for ages, and I thought it wise to step back and study the theory.
After some research, I identified said device (an SR98 Atari 520ST) as being a self-oscillating flyback converter.
Now I grabbed a copy of "Switchmode Power Supply Handbook, Keith Billings, 1st edition, 1989, McGraw-Hill" and there's a similar circuit (Fig. 2.6.4)
and a detailed explanation which I don't understand. So I fired up circuitjs and tried to simulate it, in the hopes it would give me some insight
to the inner workings of the circuit. Unfortunately, there are no values in the book, and so I tried to guess them and now have a somewhat functioning oscillating primary, but regulation doesn't work at all.

I would be very grateful to get some help on finding the right values, and understanding the circuit. You can find the link to the circuit at falstad.com.

Thanks,

Fabien

[faxmodem]

I just realized the book scan at archive.org required to log in to be viewed. There's a copy of the diagram on e-magnetica.pl with a simplified explanation and for your convenience here's the diagram.

[Konkedout]

I used to design commercial AC-DC flyback power supplies which were self oscillating.  This was > 30 years ago.

Self oscillators may not simulate properly because noise may be needed for the oscillation to begin.  Noise is always present in real circuits, but simulators....not so much.

I would send a schematic if I had one to send....but I think I would need to re-design.  The schematic you show may be OK (yeah it lacks values) but I do better to design than to reverse engineer an existing schematic.

The earlier AC-DC designs used bipolar transistors such as MJE13005 which may no longer be available.  Now it is all MOSFET main switches, and boundary mode also is sort of obsolete but I think it is worth understanding.

If you mainly want to learn, I recommend that you start off trying to make a critical conduction boost converter, as that avoids the difficulty of having a good flyback transformer.  The transformer itself is probably the biggest challenge.  If you DO want to do a flyback converter, make your job easier by using a larger sized PQ core.

What is your desired input voltage, output voltage, output current?

I recommend: Search for key words: Boundary mode, critical conduction, flyback converter.

[Xena E]

Hello Fabien.
The circuit is basically a blocking oscillator based on the positive feedback between L1 and L3, without the regulation the output voltage would rise until circuit losses or more likely component breakdown limited it.

A proportion of the output is applied to the comparator A1 non inverting input, the inverting input of which has a stable voltage provided by the regulator diode Ref. When the non inverting input exceeds the inverting, the output switches high, turning on Q2 shunting away the input to the switching transistor.

Q2 is also used to provide current limit as it also senses the voltage across the switching transistor emitter resistor, R4.

As for some of the specific component values, that would be a hard task as we would need a lot more information about the original circuit.

If you have one to repair then can the original not be reverse engineered?

If, it is a repair, then the most likely faiures are the capacitors, switching transistor, or transformer.

Hope that helps.

X

[xavier60]

A common failure point is the start-up resistor, R4 R1 in the example.
Careful, if the start-up resistor is open circuited, the large HVDC capacitor or reservoir capacitor can stay charged and be a shock hazzard.
I have had this kind of SMPS spring to life while bring probed with multimeter probes when the start-up resistor is open.
Also, avoid powering it with any components removed that disable output voltage regulation.

[Konkedout]

A common failure point is the start-up resistor, R4 in the example.
Careful, if the start-up resistor is open circuited, the large HVDC capacitor or reservoir capacitor can stay charged and be a shock hazzard.
I have had this kind of SMPS spring to life while bring probed with multimeter probes when the start-up resistor is open.
Also, avoid powering it with any components removed that disable output voltage regulation.

Maybe we are looking at different schematics, but I see the startup resistor as R1.  And yes the startup resistors do fail.   In addition to the power rating of R1, watch the voltage rating.  Be conservative about the voltage rating of that resistor...and do NOT ignore it!  It is OK to put a few resistors in series to get the needed voltage rating.

[xavier60]

I should have said "R1"

[faxmodem]

Thanks for your answers.
They helped me experiment in a more educated way with the values, and I think I found a stable setup !
I still have to figure out how to avoid overvoltage on the output just after switching on :



It's going up to 18V and then stabilizes around 15V with some ripple.
One probably stupid question I have is about the Op-Amp. It's an ideal one in the simulation with -15/+15V, but if I were to try this circuit out on a breadboard, how would I feed it the -15/+15V ?

Here's an updated version of the simulation

[schmitt trigger]

You don’t require a negative supply in this particular application.

[faxmodem]

And where can I get the positive supply from ?

[Xena E]

And where can I get the positive supply from ?

The output!

[Konkedout]

You need to observe the specifications of the....

I do not see whether it is an op amp or a comparator.  Usually an op amp will be used with negative feedback, and many comparators (such as LM339) need output pullup resistors.  So identifying the device is one issue.

The (op amp or comparator) will have a common mode input voltage range.  In a uA741 op amp or LM311 comparator, the input voltage range does not reach either the negative or positive supply rail.
In an LM324 type op amp or LM339 comparator, the input voltage range extends down to the negative rail but not up to the positive rail.
Some (mostly lower voltage devices) have rail-rail input.  And some are RRIO which is rail-rail input and output.

But from the looks of your circuit, I think your input voltage is far enough above the (amplifier) negative supply rail so that I do not think you have a problem.

[faxmodem]

When I try to feed the op-amp the output voltage, the output voltage drops a lot and isn't stable (1-2V)

[xavier60]

The snubber should go to GND and 10 ohms is way too low for the start-up resistor. It makes better sense swapped with the 100K.
I spent some time on the simulation with no success.
Time would be better spent tracing out the actual SMPS.

[Konkedout]

Given that this is a non-isolated flyback (mosquitofront?)  converter, I would replace the op amp (is it that?) and zener with a TL431 and (PNP transistor w emitter resistor).  The TL431 is cheap and versatile.   It can be used as (reference + error amplifier.)  Millions of them (maybe billions?) have been sold and they were almost universally used in AC/DC power adapters.

I drew the attached with KiCad.  Values for R1 and C1 are just an educated guess.  R2 is a gain set resistor which value depends upon the circuitry being driven.

[Xena E]

Looking closely I'd say the circuit as it is, has to have component values calculated correctly to work, these are just random.

The output sensing and voltage reference for instance should sit the values centrally within the bounds of the comparator (?) supply rails.



I agree with Konkedout, to do this now the go to component would be the 431.

X

[faxmodem]

First of all thanks so much to all of you for taking the time to try to help me, I really appreciate it !
I have to admit many of your answers are a bit over my head and I should give it some time to understand them properly.

Now, due to popular demand I'll post the schematic of the original PSU I'm trying to repair, and also the simulation.. Unsurprisingly it uses an TL431 :-)

I've got a working device, and a faulty one. I replaced all the electrolytic caps on the faulty one, and swapped most of the components with the working one to no avail. The symptom is a very low, unstable voltage on the output, and my scope doesn't seem to be able to sync to Vbe of Q1 and Q2 (using an isolation transformer).

[xavier60]

There are a lot of interactions happening in that primary control circuitry.
There is usually a resistor across the opto's LED in all SMPS designs that use the TL431. Makes it difficult to determine if the TL431 is involved with the problem or not.

[faxmodem]

Correct me if I'm wrong, but I think I ruled out the secondary by feeding it 5V through the output (disconnecting mains of course) and then measuring the resistance across the optocoupler's transistor while varying the voltage. I don't remember the direction, but the resistance varied nicely with input voltage.

[xavier60]

Correct me if I'm wrong, but I think I ruled out the secondary by feeding it 5V through the output (disconnecting mains of course) and then measuring the resistance across the optocoupler's transistor while varying the voltage. I don't remember the direction, but the resistance varied nicely with input voltage.

That's a useful test. The resistance across the opto's transistor should abruptly start dropping at very close to 5V, corresponding to the TL431's REF pin approaching 2.5V. If you see a change in the transistor's resistance well before 5V is reached, put a 1K resistor across the LED.

[xavier60]

BTW, dont increase the voltage any more than necessary. From what's shown, there is nothing preventing the TL431 from drawing enough current to damage the LED.

[Konkedout]

There are a lot of interactions happening in that primary control circuitry.
There is usually a resistor across the opto's LED in all SMPS designs that use the TL431. Makes it difficult to determine if the TL431 is involved with the problem or not.

Yes the TL431 has an idle current which could get up to 1 mA.  So you want bypass the opto LED using a resistor which will pass at least 1 mA before the LED turns on.