TOP267 35W DESIGN FAILED EMI

Posted by: bbmilo on 07/11/2012

Hi,

This 13V 2.7A TOP267 design recently failed the radiated and conducted emission but passed all the other EMC tests (immunity, harmonics and flicker, transient burst etc.) The circuit had a RCZD clamp and the drain voltage waveform was squarish. Circuit and waveform are attached.

Hence I changed to use the 'lossless snubber circuit' according to Abraham Pressman's book. It's a CLD circuit. We haven't gone back for another EMC test yet but the waveform now looks much rounder. Circuit and waveform are attached. We also have to add a series resistor on the HV side of the optocoupler due to safety standard requirement. With these changes, we made another board with different layout so to fit everything in. And one obvious problem we have now is that the PS has problem starting up sometimes and it has to try for a few seconds and eventually recover. How can I fix this problem?

Please also check my layout and suggest improvement as needed for emc purpose.

Many thanks.

Comments

PLS upload your new schematic. I can found the old schematic, but the problem happens at the new schematic with lossless snubber. I guess the issue is caused by the capacitive charge to the snubber. Let's discuss it when the new schematic is available.

Thanks

Hi, Thanks for your reply. The diagrams were submitted in the pdf format and apparently the web didn't like it. Re-attaching the files below.

Hi Cochrane,

I've just captured more waveforms - The output rectifier voltage, drain voltage and bias voltage.

The noisiest one is the bias voltage. I forgot to mention before, the flyback transformer has double earth screens in between the primary and secondary windings as required by the safety standards.

bbmilo -

Can you please attach your PI Expert design files, your PCB layout as well as an EMI scan so I can see where the problem areas are? It would also be helpful if you could include any transformer measurements you've made.

From the waveforms you submitted, it looks like your clamp circuitry isn't put together correctly.

Also, how did you determine the components for your output rectifier snubber circuitry? Did you determine them experimentally or from measurements?

-The Traveler

Your layout is not bad.

PLS check which frequency caused the emi issue, and then check the frequency of DCM ring, switching node voltage of topswitch and secondary output diode. Hopefuly you can find the root cause of the EMI, and then fix it.

Use a bigger filter and Y cap can help too. Maybe you can use bigger Y cap( the one from secondary return to DC plus) to see what may happen. I suggest staying with the board with RCD snubber as long as the efficiecny is not a problem.

Hi Cochrane,

Thanks for your reply. The voltage waveforms of Vdrain and output were already attached in my previous post. Should I look at the current waveforms as well?

Using a bigger filter and Y cap will be my last resort, I'll try to kill the root cause of the EMI first, as how you put it.

Efficiency is not a problem for us. However, the RCD snubber gives a squarish waveform which I think would have generated more noise, don't you agree? Don't you think the new CLD snubber works?

Hi Traveler,

Thanks for you comments.

A few changes have been made from the attached PI Expert design files:

1) RCZD clamp nextwork, as suggested by a local PI support person for Asia Pacific,

2) The control pin capacitor, from 47uF to 10uF as we need shorter re-start interval,

3) Transformer construction, has two internal earth screens as required by the safety standards,

4) OVP circuit removed as there will be other voltage protection downstream,

5) The optocoupler series resistor has been finetuned to stablize the system,

The snubber for the output rectifier was straight from PI Expert and no measurement has been made. I will now try to look at this area as that seems to be the noisiest area from the waveforms shown in my previous post.

In my later board, I have removed the clamp circuit and only have the snubber to slow down the ringing and not to limit the peak as I thought this gives a squarish waveform which contains high harmonics frequency.

I would like to send you more files, but can I send it to you via an email?

Cheers.

Hi again,

I've been playing with the snubber circuit a bit more and came up with the circuit and Vdrain as attached. The spectrum analyser shows much quieter scan above 20MHz. So it's looking promising for EMI.

However, the PS sometimes struggles to start-up. It doesn't matter whether it is full load or no load, with high Vin or low Vin. Once it's up and running it'll continue to run.

Any suggestion how can this be fixed?

Thanks.

Test start up without R6,I guess soft finish circuit not work for quickly turn on turn off test. The soft finish circuit is important for the IC to start up sucessfully, because it provides some current through the opto at the first start up.

Hi Cochrane,

I've tried that deleting R6 but didn't make any difference.

In my previous post I said that the circuit won't start up in both full load or no load, actually it's only in full load that it can't start up. It starts up okay with the load disconnected.

I have also tried reducing the size of the output capacitor to 1000uF. But that didn't make any difference either.

How else can I improve the feedback circuit?

Thanks.

I checked your schematic again and again, it looks ok. I know this issue happens after you change the snubber circuit. So i guess you test with a brand new board. I wonder if some part could be wrong during the placement.

Another thing, would you like to double check the load current during start up.Maybe you can try a ressitor load first. The start up maybe caused by the E load.

I donot think the issue is caused by the loop.

Reduce R8 to make sure opto provides enough current for C pin. The loop gain will be higher with lower value R8.

Hi Cochrane,

I found some slight difference in my soft-finish circuit compares to the one in the TOP-HX datasheet. Should the 10k resistor be connected to before or after the output filter inductor?

Also, I have a couple of fuses inserted in the output circuitry. One thermal fuse between the output diode and output capacitor and a normal fuse between the output filter inductor and the feedback trace to the programmable zener. Would these fuses affect the feedback loop?

Cheers!

Hi, I found that the 10K resistor is placed before the post stage filter, and it should be ok .

The thermal fuse between diode and output cap really affects the performance of the power supply. I do not know how it woks, but definitely it blocks energy to output cap. Feedback loop should be affected by the thermal fuse since there is some resistor to block the energy transfer and results in some diference of transfer function of secondary side.

For the fuse between output incutor and feedback resistor, I donot think loop will be affected by it. The resistance is very low and there is a inductor in series with it.

PLS remnove the thermal resisitor after the output diode first, and then check if the start up issue can be solved. If the issue still there, check the output waveform and the output of the bias winding voltage.Output voltage will tell us if the issue is caused by overshoot or not reaching setpoint. The bias voltage must be good enough to provide enough current to control pin. Short the 1k resistor between C8 and R13.