LNK304 failure at startup

 What would be most helpful would be to see the actual drain current and drain voltages at start-up.  I would do this by using a current probe between Vboost and the LNK304 Drain and then use a high-impedance differential probe to capture the Drain to Source voltage of LNK304.  This would give us a better idea of what's happening across the MOSFET.  Although you're saying that the drain voltage isn't exceeding 600V, without actual waveforms, there could be some ringing effects during start-up that are exceeding the 700V MOSFET rating.

If the problem is an issue with exceeding the MOSFET breakdown voltage, moving to a larger device won't solve your problem.  I would double check the drain current and voltage waveforms before proceeding further.

-The Traveler

I put a loop of wire between the drain pin of the LNK304 and the pad on the PCB and cliped the Lecory CP030 current probe onto the wire. I also connected a Lecroy ADP305 high voltage differential probe across Drain to Source of the LNK304 (+ on drain, - on source).  The input voltage is 300vac (424 peak). The scope traces are attached. "fri" is the startup at 1ms per division. "fri1" is an individual ontime of the lnk304. These are just traces at "normal operation". The drain current is 532mA peak (over the datasheet limit of 400mA @ 400v). I have tried a slow start cap across the 13k resistor and values from 1uF to 10uF limit the drain current swell to less than 400mA. Reducing the output cap from 220uF to 100uF shortened the length of the swell but the amplitude remained about the same. Are there any downsides to adding the slow start cap? (maybe dynamic consideratons?).

Thanks,

Kevin.

adlkelly - 

On the WF capture called fri.jpg, could you please repeat the capture but do the following:

-Use single trigger mode

-Increase the sample rate to the absolute maximum (the effective 500kS rate of the capture you provided is too slow to see the high-speed drain voltage/current waveforms accurately).

-Set the Lecroy to trigger off of the drain current waveform so it starts capturing as soon as the device starts.

-Change your time base to something a little smaller, maybe 500uS/div (for better resolution)

Perform the same WF capture, but locate the points of peak drain voltage and peak drain current (may not occur together) and zoom in on these areas.  Provide the 2-3 images in your reply.  

Another question, what are the output specs fo your PFC stage?  Is it a 380V PFC output?

-The Traveler

fri2.jpg shows the drain current waveform at startup with a timebase of 2us/div

fri3.jpg is the  peak current waveform at 1us/div

fri4.jpg shows the slow prgression of boost. At first it's at the peak of acin 300vac->424v peak. At around 200ms the pfc comes on and boosts to 525v then settles back to around 475vdc. The product is designed to ramp to 525 at turn on and then settle to a maximum of 500vdc. When I looked at the waveform using faster timebases I didn't see anything above what the values shown on the slow time base.

(Note: input voltage is 300vac during all scope traces)

Attached is a word doc showing the layout and schematic. Top copper is red, bottom copper is green. The "T" ref des (like T38) are test points. D3 is provided so that, at startup, the boost caps can be charged directly from rectified AC as opposed to taking the parallel path - thru the boost inductor. C10 is a .01uF high voltage ceramic cap which bypasses the drain pin. DB1 is the diode bridge. C6 and the parts to the left are the EMI filter. There are 2 outlines for C7 so that we can use multiple case sizes.

Attached are 3 shots of LNK304 during failure. When you power up a product before conducting a surge test, our surge generator puts out a short 600v pulse at the same time as AC is applied (some sort of capacitive coupling). We have see that the LNK304 has problems with this. Today we added a 10uF slow down cap and were able to do 100 powerups witout any failures. When we took the slow down cap off of the product it failed on the second power up. Usually the product will fail within 15 power cycles. A unit with a 1uF slow down cap also failed quickly.

The jpg labeled "LNK304 Failure CH1 Boost CH2 LNK Switch Node CH3 12V Startup on Surge Generator1.jpg" is

CH1 – Yellow – Boost voltage at LNK drain

CH2 – Pink – LNK switch node at LNK source

CH3 – Blue – 12V output

The jpg labeled " LNK304 Startup CH1 Boost CH2 LNK Switch Node CH3 Inductor Current CH4 Rect AC - Surge Generator Startup13.jpg" is

 CH1 – Yellow – Boost voltage at LNK drain

CH2 – Pink – LNK switch node at LNK source

CH3 – Blue – Inductor current

CH4 – Green – AC input

This one had a 1uF slow down cap.

The jpg labeled " LNK304 Startup Failure with No Startup Cap CH1 Boost CH2 LNK Switch Node CH3 Inductor Current CH4 Rect AC - Surge Generator Startup1.jpg" is

CH1 – Yellow – Boost voltage at LNK drain

CH2 – Pink – LNK switch node at LNK source

CH3 – Blue – Inductor current

CH4 – Green – AC input

Best regards,

Kevin.

Looking at your scope plots as well as your design parameters, I think the issues you're experiencing are due to a couple of factors:

-With such a high input voltage, your inductance value is too small.  LinkSwitch-TN has a minimum on-time associated with it (as do most of our products).  This means that regardless of the current limit feature, the MOSFET will be on 

 approximately 0.5us.  If the di/dt value is so large (due to the high input voltage) that the drain current is exceeding the device limits in 0.5us, this could easily explain the product failures you're experiencing.

I see a couple different solutions being easily implemented:

-Increase your soft-start time so that the PFC voltage has finished settling down before the LNK304 starts switching.

-Increase your inductance value to slow down the drain current di/dt so that the controller can respond in a reasonable amount of time to the device current limit.

-Switch to a larger device that is is more in line with the drain current values you're seeing in your real product performance.

My preference would be to use one or both of the first two options.  The third would probably work but with your di/dt values being so large, you're at the edge of how quickly the controller can respond and it's still possible you might still experience product failures due to the MOSFET not being able to turn-off fast enough.

-The Traveler

Do you have a recommended inductor value and current?

We got good results with the 10uF soft start cap. Are there any effects other than causing the +12 to come up more slowly. (I'm thinking of dynamic effects?)

The soft-start capacitor works by changing the feedback resister ratios at frequencies greater than DC.  This allows for a slightly longer start-up time than would normally be the case.

The one thing this doesn't do however, is change how long the device takes to start switching from application of power to the drain pin.  I would definitely reccomend checking the drain current and voltage waveforms during startup (especially when your input voltage is at its maximum) and see what the peak values look like.  You may still need to go to a higher inductance or a larger LinkSwitch-TN device.

-The Traveler

I would start simply and just try taking two of the inductors you're already using and add an extra one in series.  See if you can find an inductance that will slow down the di/dt of the drain current enough so that the MOSFET on-time is decently above the MOSFET minimum on-time.  You want to make sure that there is enough time for the controller to react to the internal current limit circuitry.  If the MOSFET current limit is being reached *before* the end of the MOSFET minimum on-time the MOSFET and controller can be damaged by excessive drain current.

-The Traveler