LT4294 Power delivery issue

If problem appears we disconnect board from power, it cools, we reconnect it back, it classifies and works till the next failure. 

Hello,

I have investigated meanwhile the issue and here are some more info. In attached snapshots RED(Ch3) is a hot swap N-CH DRAIN, BLUE(Ch4)-SOURSE, YELLOW(Ch)-GATE: 
Snapshots presents classification against the PSE and failing conditions.

I also measured temperature on a package. It somehow raised up to 70 degrees, probably from the power supply that located no far from it. The GATE disassertion appears when the temperature reaches about 70 degree on the package(measured with a thermopcouple and it is not very precise). Once again, we have no problem to operate with the board if AUX input(of LT4294) is used. The mode of operation not changing during the test. We expecting to consume about 30W at this mode.

Please suggest what can be wrong here.

It would be nice to get at least a hand wave here.

Apologies for the delayed response.

Have you tried checking the voltage at the AUX pin? You mention the board ambient around the IC quite high, and I see from your schematic you are using schottky diodes for the AUX pin. That diode looks like it could have 40uA of leakage current at 50V and 60C junction temperature. If this flows into the auxiliary divider, it might be enough to look like an auxiliary power source is connected. This would cause the hotswap FET to turn off, then leakage current would go away so the FET would turn back on... and rinse and repeat. Try re-running this test with R111 removed and let me know how ti goes. I would also recommend replacing these diodes with silicon or lower leakage schottkys.

Best Regards,

Eric

Hi Eric,

I understand you theory, but there is no situation when both, AUX and POE are connected at the same time. I tested when only AUX(12V, lower board temperature and works fine) is connected or when only POE is connected(reach ~70 degree on package and fails).

At time I waited for response I have made another test: I cooled the board with the fan, so the maximum temperature on LT4294 on a package surface reached 55 degrees. In such situation no fail occurs.

I also changed the hot swap transistor(Q5) to the TO220 package suited for 110A continues and soldered it outside the board, just to be convinced that there is no repeated transistor issue.

Do you have another idea? Do you think the test you suggested still worth to implement?

I can order the LT4294HMS#PBF (125 degree qualified) do you think it worth it? Can be a batch issue?

Thanks,

Andrey

I still think this might be a leakage issue. You would not see this issue with the auxiliary supply connected, only when just PoE is connected and the board is hot. During PoE operation, the AUX pin should be at 0V, but if there is approximately 14uA leaking through D6 that might be enough to make pull the AUX pin above its threshold voltage. This would turn off Q5, which would cause the voltage across D8 to droop which would decrease the leakage current. Less leakage current would cause the AUX pin voltage to decrease, so Q5 will turn back on again. If D8 is still leaking this process could repeat. 

You could test this by either disconnected D8 or R111, it'd be nice to rule out this as the cause of this behavior. I would also suggest changing D6, D7, and D8 to silicon diodes as these are just for signaling an auxiliary supply is present.

Best Regards,

Eric

Hi Eric,

Great idea!! Thanks for sharing your suggestions. I definitely will test it and let you know the results.

Andrey

Hi Eric,

I have removed the D8 from the assembly and it seems that problem was solved. I will manage to replace this diode.

Thanks again,

Andrey