Forward converter stability

Hi to all.
I'm trying to build a forward convertor (220Vac to 12Vdc 25A) , but am having a problem with stability.
It's a single switch forward converter with a reset winding in the XFMR. I'm using 800V mosfet for switching.
The converter is using the old uc3842 controller chip. It is working fine (Nice and stable) at high output power ,
but at low output currents (below about 1 Amp) its starting to sing nicely.
Anyone got any useful ideas on how to get this thing stable from say 100mA to 25A.
Or am I on a hiding to nothing here?
Cheers
Neddie

- - - Updated - - -

Just found something to add to the above post. I've noticed that under low load the transformer current( Which I'm measuring across the current sense resistor) is no longer a linear ramp , but starts to ramp up very quickly. A bit like a diode curve when going forward biased. This tells me the transformer is going into saturation. Why would the transformer be going into saturation under low load , and not full load?
It's difficult to get exact readings as I'm trying not to blow it all up again :0)
When things go "pear-shaped" , it tends to be a complete rebuild of all the switching electronics , which gets tiring after a while!!

A circuit would be helpful! Have you a current sense resistor in series with the source? it looks like the current sense circuit of the uc3842 does not work properly at low input voltages. Put a little DC forward bias on it and reconfigure the transformer? :-(
Frank

Not sure what you mean by low input voltages?
Relevant part of cct attached.

- - - Updated - - -

Not sure what you mean by low input voltages?
Relevant part of cct attached.
Sorry quality is a bit rubbish , I'll see if I can improve it.

- - - Updated - - -

zoomed in area of ic

Could you be more specific on the kind of instability? Is it chaotic, or is it actually oscillating at a coherent tone?

It certainly would be odd to have saturation issues at light load. Can you verify that the transformer is actually reset before the FET turns on (drain voltage should be equal to input DC voltage)?

Also I notice you're using an aux winding for control power bootstrapping. This can be problematic with the forward topology, so you may want to try operating it without the aux winding (just use a DC bench supply).

Instability definitely chaotic :0)
Difficult to see if the transformer is reset properly due to the fact that the switching is all over the place and difficult to trigger on
all the different waveforms. What I think could be happening is due to the erratic behavior of the circuit under light/no load the switch is being turned on BEFORE the core reset is finished. One or two cycles like this and bang :0(
With this type of convertor , do I need to use a different compensation scheme. I don't know much about this stuff unfortunately.
There is type 2 , type 3 , and various other schemes. Am I using the wrong one here?

I'll try the separate supply idea , thanks.

Chaotic instability at very light load isn't a rare problem, the main problem is that at light load the converter goes discontinuous and thus its voltage gain becomes much higher than in continuous mode. In extreme cases, the duty cycle may need to be extremely small, and in such a case the PWM modulator is more susceptible to noise and interference, which is why the instabilities usually look chaotic.

First I would try opening the feedback loop and directly control the COMP pin voltage. Put your desired light load resistance on the output and then manually adjust the control until you get the desired output voltage. Look at your waveforms and see if the circuit is chaotic in this open loop state.

Many modern controllers need to use some sort of burst mode or pulse skipping method to operate properly at light loads, though such methods generally increases output ripple voltage.

Interesting thoughts!!
I've never tried opening the loop and manually controlling the COMP voltage.
Could be a little scary with the high powers available :0) I normally put a light bulb (60 or 100W) in
my mains supply to limit the current when testing at light (no pun intended) loads. Still enough energy in the bulk
storage caps to do quite a bit of damage though.
The 3842 is a pretty old device , pulse skipping is not available AFAIK.
Cheers
Neddie

When manually controlling the supply in DCM, your biggest risk is overvoltage on the output. Either put a big voltage clamp on the output, or start with a heavier load and decrease it gradually.

It's not terribly difficult to implement some sort of burst mode with external circuitry. Basically you just have the converter shut off when the error voltage is below some threshold, and put some hysteresis on that threshold. This will likely still generate some audible noise, but at least it won't be chaotic.

You may find this useful:
https://www.ti.com/lit/an/snva555/snva555.pdf

Here is a commonly used compensation scheme


- - - Updated - - -

in your schematic i don`t see the compensation network because of bad quality of first picture

The above circuit is not so much for compensation as it is used for the voltage regulation.Rf and Cf can have a role in the overall compensation. Compensation is generally achieved using the amplifiers (feedback and comp pins on the 3842) and their various surrounding components.
I've tried mtwieg's suggestion of opening the control loop and manually controlling the COMP voltage. The pwm seems to go from a very narrow pulse and jumps up to about 25% duty cycle with narrow pulses inbetween , very unstable.
What is this indicating?
Cheers
Neddie

At very light loads / no load, you may not get enough o/p out of your aux winding to power the IC properly and it may be re-starting. Also the gain of the power circuit is different at no load and a slight amount of PWM will cause the o/p to go overvolts, stopping the PWM (and the aux output) until the Vo drifts back to just below the regulation point at which time the PWM comes back and the cycle repeats.... perhaps adding a min load to the output (2W) will be helpful, check the voltage level on the aux output, if you see serious dips you will know what is happening.

KA431 and TL431 is the same. and it includes amplifier.
as i understand, you trying to transfer the output voltage information to primary side and make compensation there. where you get this shem? or you invented it?

Compensation is done on the error amplifier of the uc3842. I'm certainly not inventing it , it's common practice.

neddie said:

I've tried mtwieg's suggestion of opening the control loop and manually controlling the COMP voltage. The pwm seems to go from a very narrow pulse and jumps up to about 25% duty cycle with narrow pulses inbetween , very unstable.
What is this indicating?
Cheers
Neddie

Click to expand...

Might be the typical problem with current sensing on the FET source, which is that as you get to narrow pulses the initial current spike due to Cgs charging is tripping your PWM latch early some of the time. Did you look at your error voltage and control supply voltage at the same time?

I have got a filter for this exact thing , but I'll check that it is actually working :0) I've also found another article that
shows noise being getting into the oscillator circuit. This results in a spike on the oscillator ramp (cap charging up ) signal. This will cause a premature trip of the output. They recommend using a bigger cap than I'm using , so must check this out as well :0)
Thanks
Neddie