L6566B - getting it to work

[kapurk]

I am trying to get a prototype SMPS board for a 12V 10A output using the L6566B to work. The circuit design has been done using the edesignsuite on the ST website. I have got a few boards made professionally for prototyping after designing them with KiCAD. I have a fair bit of experience with power supplies using the 3842 PWM controller but getting this L6566B board to work is proving to be a challenge.The controller takes its own time starting up. In fact it starts up quickly if I touch some points on the pcb with a multimeter probe. But then after startup it seems to shutdown or remain in burst mode the moment I put some load on it (less than 1A). I have tried disabling some of the protection features by tying Vff (pin 15) to ground and AC_OK to Vcc as suggested by the datasheet.
The circuit is designed for QR mode operation. For test purposes I am using a 0R33 resistor for Rcs and 15k for Rosc. The transformer is a PQ3220 that is taken from a 3842 based SMPS. Would the L6566B shutdown with minor variations in turns ratio? I have tried changing the ratio of Rovp_h and Rovp_l but to no avail.
The opto is K817C with a TL431 voltage ref.
Vcc measurements show a cyclic variation between approx. 8V and 13V. Vref is 5.06 - 5.07V. What is confusing is that the Pin 13 shows only about 0.2V whereas the datasheet says it is an accurate 1V source. The scope shows no waveforms or pulses when on load but it does show an occasional single pulse at the MOSFET drain on no load. This pulse is sufficient to charge the output capacitors to light up an LED which slowly goes dim and then brightens up when the next pulse comes.

Any tips on what could be going wrong?

 

[KlausST]

Hi,

The first that comes into my mind is a floating input, maybe accidentally by a bad solder joint.

But without schematic, board and additional informations it is hard to help

Klaus

 

[kapurk]

Here is the schematic:

 

[treez]

your c11 doesn't look big enough....10n or 100n here
your zcd divider looks a bit off....are you sure you are not ending up making it think its in overvoltage by too much volts on zcd pin?
R16 is too big..should be 1k or 750r.
you got no snubber on your d4 schotttky..you could be ov'ing your D4, shotkys don't like Overvoltage spikes.
your cout is massive..you sure you not triggering overload protection from startup...rem you must get into reg b4 certain time elapses.
you got pot rv1...big nono...wiper noise...not good.
you got no cs RC filter....you sure l6566b has deadtime enough for not using rc filter here?
you sure your ac_ok vopltage is high enough...maybe you going into uv lockout.
c5 is bad...no use it...you don't want time delay here..you want to be in sync with the bus.

you using constant freq or boundary conduction mode?

what your max primary clamp voltage at your min vin?.....you sure you all rated for that?

what is min ctr of opto?....you sure you correctly sized r15 in accordance?
you got no slope compensation in there...you using constant freq mode?..what is your max duty cycle (at min vin)?
...oh no!...ou have taken your output feedback from after L2 inductor...what value is L2.....it bad do this...you connect feedback to other side of inductor!...side nearest the secondary coil.

..you sure you do VFF correct..maybe you disable this at first, then put in in later when you get it working

what is d1 for?

what you do "test purposes"?...if your rcs resistor too big value then you shut down on overload protection...if comp pin rails then l6566 think it is in overload or output short circuit, and it shut down.

 

[kapurk]

Thanks a lot for the detailed reply Treez. Now here's the catch. The circuit design is not my own but from ST Micro's eDesign suite. I have already tried out disabling the various protection features like cs, zcd, Vff and also the various points mentioned by you. Now here is the interesting part.

After my last post I ordered an evaluation board from ST (burnt a 140$ hole in my pocket) to get a jump start and see how it works under various conditions. I plugged it in and it started working. I put a rheostat to load it unto 3.4A and ti worked perfectly. Then I increased the load upto 3.5A and sure enough the over current protection kicked in. I tried going beyond 3.5A by reducing the rheostat value but the short circuit protection kicked in and the output went to zero. I brought the load back within range and the output came up again. I varied the ac input from 100 to 250V and everything kept working fine. Then I switched it off and guess what, it did not turn on again!! When I checked the board I found that pins 4(gate driver) and 5(Vcc) L6566B had shorted out to ground (pin 3) internally. My bad luck? Or is it that the L6566B is one of those quirky chips that ST would like to quietly bury and forget about?

 

[Orson Cart]

first rule of power electronics, always buffer the gate drive with an emitter follower pair, this protects the IC somewhat for prototype analysis..., lucky its easy to change the chip... 10 ohms to the buffer will help too....

 

[treez]

Then I switched it off and guess what, it did not turn on again!! When I checked the board I found that pins 4(gate driver) and 5(Vcc) L6566B had shorted out to ground (pin 3) internally. My bad luck? Or is it that the L6566B is one of those quirky chips that ST would like to quietly bury and forget about?

ok but you saw the ncp1217 app note concerning a possible reason for that right?
..I am sure you know you cannot disable CS pin.
There is an overcurrent pin, if layout on their demo board is poor it will be triggered.
If you give me your tsfmr spec, I will simulate it for ou (is it BCM, or const freq)?

I will simulate it with my ltspice model I have for both modes..then I send it here to you

first rule of power electronics, always buffer the gate drive with an emitter follower pair

ok I know what you mean, but it does have a gate driver on it so it shouldn't be needed with good layout...I would put resistor 22k from gate to ground, and a series resistor as said, and also the schottky from gate to ground.

- - - Updated - - -

also, try increase c17 to 100nf or 220nf......then reduce r17 to 10k then 1k

- - - Updated - - -

also, what is full part number of the opto...PC817A...PC817D....the A version has ctr down at 50%, so not much, what is the comp pin current for "full stop"?...remember that you must have ENOUGH IN OPTO TO PROVIDE THAT (sorry for caps)

as before, r16 is too high value...should be 1k or 750r...but remember to downsize r15 so it can carry that and not drop too much volts

 

[kapurk]

Treez, Thanks a lot again for the detailed reply. I am going to work on your suggestions and then post my results.