LM5022: Boost converter fails.

Hi Timothy,

Thank you for using the LM5022 in your design.

I could hazard a guess that upon startup, you are running into some current surges, which draws a lot of current through the FET and diode. The current surge may be much higher than what you have sized the components for. You can probably solve this from making a longer soft-start time.

It is possible that there is a bit of ringing at the switch node, which may exceed the diode's and FET's ratings.

It is also possible that you have low FET drive from adding resistance at the gate, which can cause the FET to heat up since it will be conducting for a longer time than needed.  

Finally, although I do not think it is likely, your duty cycle is not that high, but at higher duty cycles, your FET is carrying the current for a larger amount of time. It is possible that the FET will short out first, which after a while will melt the connections and become open. Therefore, current will flow through the diode, damaging that as well.

If you replace the damaged components with those with higher ratings, do you see the same issue? Additionally, can you provide a schematic with any available scope shots such as switch node, VCC, and COMP? 

Thanks,

Richard

Hi Richard,

     Thanks very much for your response. I tried to add a schematic to my original post, but apparently that didn't work - so I've pasted a schematic into a Word document - I hope that works.

Since the circuit fails after about a minute, I don't think it is likely a start up issue. Other than the Schottky diode, the rest of the components seem to be well above the needs, and exactly what the TI tool had recommended. The FET feels quite cool until the point of failure. There is no resistance between the LM5022 and the FET gate. There could be some ringing at the switch node, though when I looked at it, it seemed OK. I do think that as the load increased, the failure happened more quickly - with no load, I didn't see a failure.

If you can see the schematic, I added an LC filter at the output of the converter - could that be causing an instability?

Thanks again for your help.

Tim 

Hi Tim,

I'm unable to see the word document, so if you attach an image that would be good. However, adding the system level stability is altered by the extra LC filter, which the TI tool does not take into account. If you remove the extra LC filter, does the converter operate properly?

Thanks,

Richard

Hi Richard,

      Thanks for your feedback. I will try removing the filter, but I just wanted to check - can you see the schematic now?
Thanks,
Tim

Hi Tim,

Yes I can see the schematic now, thank you for uploading the image! Hopefully the modification will lead to some fruitful results. 

Thanks,

Richard 

Hi Richard,

       With a higher power diode, things seem quite a bit better. I want to run a thermal simulation, but I can't find it anymore in the new Webench Power Designer. Can you help me find it?
Tim

I found out that I had to go back to the old webench to run a thermal simulation. We have not had a failure after switching to the heftier diode; however, we do  notice that Q1 is getting hot - we measure 100C on its package after 10 minutes. Using the thermal simulator, Q1 doesn't even rise to 40C. Our theory is that since we had to replace Q1 from being damaged in previous tests, we are not able to manually solder it well enough into the board so it can get proper heat sinking. 

Q1 is BSZ520N15NS3 G

Do you think we should expect this part to get hot, or do you think the problem is poor soldering when we had to replace it?

Thanks,
Tim

Hi Tim,

According to some calculations I did, it looks like you should see about a 10 degree C rise in Q1 assuming the worse case rds-on, 52 milliohms. Give some leeway regarding the temperature rise (+/- 5 degree C), and assuming an ambient temperature of 25 degrees C, it would look like you should see about 40 degree C based at the case

That being said, with a temperature rise of 75C at an ambient of 25C, that would imply you are dissipating more than 1W of power through the FET or that theta ac is higher than expected. This would back calculate to about 400 milliOhm resistance, so you would have about 350 milliohms from somewhere else. This could possibly be due to a incomplete soldering job so perhaps it cannot dissipate heat as much, dirty traces, or part variations. 

Can you try to de-solder the FET and clean the board properly and see if this will clear the issue? 

Thanks,

Richard  

Hi Richard,

      We've tried resoldering the part, and we really feel like it is well soldered. We've tried other parts, and still get the 75C rise. Since it is a plastic package,  we are measuring on the top side. I'm assuming that you ran the calculations with Pd=0.44W and theta junction to ambient of 60 K/C. But I did notice that for this package they state 60 K/C assuming 6 cm square of 2 oz. copper. We have a smaller area (maybe 1 cm square) and only 0.5 oz copper.

Do you think this is the issue?

Thanks,

Tim

Hi Tim,

Yes- not having enough area and thickness of copper will increase your temperature rise. I would recommend increasing the area size and thickness of copper.

Thanks,

Richard

Hi Richard,

     Since we already have a PCB, I can't change the copper on this revision. So I chose the same transistor in PG263 package and reworked it into the board. However, I find that if we draw current above 1A, then the output voltage starts to drop - it still seems like the circuit is regulating, but the output voltage falls to 95V. What could be happening? The transistor is spec'd at 21A, 150V, and 53 milliohms -very similar to the original transistor, but a different package with much better thermal capacity.

Thanks,
Tim

Hi Richard,

      We've encountered another problem. I tried a different transistor (IPB530N15N3) that has similar specs except that it was in a T0-263 package so could handle more heat. Rds is 53milliohm, ID 21A, Vds 150V.   Here's what I see: the output is at 100V if the load is < 0.15A. When it gets above that, the output starts to drop. It still looks well regulated per the scope, but the output is 95V at .0.2A.Can you give me some insights as to what might be going on?

Thanks,
Tim

Hi Tim,

Are you examining the voltage at the load? Or are you examining at the output of the converter across the capacitor? Since you are increasing the load, you will see resistive drops. With a drop of 5V at 0.2A, is there 10-ohm's in between the output of the converter and the load?

You may also be hitting current limit.

To get a clearer picture, can you monitor the COMP pin voltage when the drop from 100V to 95V occurs?

Thanks,

Richard

Hi Richard,

       Thanks for your response. I am measuring the voltage at various spots, including across the capacitor, so I'm quite confident that I'm getting an accurate reading. The oddity to me is that the output voltage stays steady at 100V up to 0.15A, but then starts to decrease

         I don't think I'm hitting the current limit - with a 40 milliohm shunt, I would need 12.5 Amps to hit the 0.5V threshold if I understand how that works correctly.If the current limit was kicking in, what behavior should I expect? 

I will try to get you a picture of the COMP pin.

Thanks,

Tim

Hi Richard,

       I'm attaching a picture of the COMP pin waveform. Let me know if you cannot see it. I see a similar waveform whether or not a load is attached to the circuit.

Thanks again for your help.

Tim

Hi Tim,

Thanks for attaching the waveform- I can see it clearly.

If current limit was being hit, you would see the MOSFET gate signal terminated until the condition is removed. 

It seems like COMP is dropping below 0V. This seems out of range for the pin of the part. Can you replace the part and take a scope shot of COMP?

Thanks,

Richard