This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

LMG5200: Failure with Vin to SW, SW to GND failed short in low power state

Part Number: LMG5200

NOTE: I would like to move this thread to a private thread so that I can post schematic images. Once that is done I will add an additional schematic image. How can I do that?

Hello, 

In our design, two LMG5200's are used to form an H-bridge based buck-boost converter. Recently, the buck side (U9 below) of the converter failed short, with Vin to SW, SW to GND.

Below is the pin-to-pin resistance measurement I took for the failed IC (U9 in the above schematic):

    Black Probe
    1 - Vin 2 - HB 3 - HS 4 - HI 5 - LI 6 - VCC 7 - AGND 8 - SW 9 - PGND
Red Probe 1 - Vin  X 0.1 157k 170.3k 169k 0.1 105.8k 0.2 0.4
2 - HB 0.5  X 157k 170k 168k 0.5 105k 0.3 0.1
3 - HS 157k 157k  X 328k 325k 157k 262k 157k OL
4 - HI 170k 171k 328k  X 339k 171k 277k 157k 157k
5 - LI 168k 168k 326k 339k  X 169k 274k 168k 168k
6 - VCC 0.2 0.1 157k 170k 168k  X 106k 0.1 0.1
7 - AGND 106k 106k 263k 276k 276k 106k  X 106k 106k
8 - SW 0.2 12 157k 170k 168k 0.8 106k  X 106k
9 - PGND 24 0.7 157k 170k 170k 0.1 106k 0.4  X

When the failure occurred, the converter was outputting about one amp. The input voltage and output voltage were almost equal (32V) with the buck-boost converter switching in buck-boost mode. 

Investigation of the failure so far

1. The initial thought would be shoot through or something of that nature, however we have done analysis on the deadtime of the converter and everything appears to be in order. Additionally, the supply feeding the converter is current limited to 9A, and this is a 10A part. There is some capacitance at the input (under 10uF) where additional shoot through current could have come from, however after analyzing the system I think this not as likely of a cause of the failure. 

2. Another thought has been ringing at Vin or the switch node. There is an input LC filter that could certainly be causing severe ringing. I probed the input node (green), and the switch node (blue). The math function (orange) subtracts Vin and SW, for the voltage across the top FET.

It looks like the SW node is ringing low, below the -5V limit called out in the datasheet. It looks like it rings as low as -10V in some cases. Is this a potential failure mechanism? If so, what exactly does this negative voltage ring kill in the IC? I guess I would have expected ringing high (over the 100V limit in the datasheet) to have been the cause if ringing was the issue. 

Lastly, is failing short typically the result of too much current/a heat based failure, or a voltage based failure resulting in some sort of dielectric breakdown?

Thanks

Jesse

  • Hi, Jesse,

    Sorry to hear of your difficulty. My colleague will send you a friend request, and then you can send them the schematic via DM.

    Best regards, 

    Don

  • Hi Jesse, 

    From the provided oscilloscope screenshots, it seems like the main indicator of device failure is the excessive ringing in the switch node voltage. You are correct - as specified in the data sheet, the switch node voltage must be kept above -5 V to avoid causing permanent damage to the device. 

    This high switch node ringing is likely caused by high parasitic inductance in the high di/dt buck loop. This loop encompasses the input capacitors and the LMG5200. For best results, this loop should be implemented vertically and kept as small as possible. Please see Section 11 of the data sheet for layout guidelines and examples. 

    The short between Vin to SW, and SW to GND can be attributed to a high current, which can cause potential migration of metallized layers. 

     Best Regards,

    John

  • Hi John,

    Thank you for the reply, it is much appreciated. 

    Is there any more information you can give on what exactly fails when the node rings low (below -5V) and fails? If I were to send this failed chip to a lab for investigation, could they find tell-tail signs of a switch node based ringing failure to corroborate that this negative voltage killed the IC?

    Secondly, the ringing to me would imply a voltage based failure, however it sounds like the fail-short is the result of current. Can ringing cause a fail-short, or would the fail-short imply something else killed the IC?

    Lastly, I will send you the schematic block shortly in your messages.

    Thank you!

    Jesse

  • Hi Jesse,

    There is spacing on the die between the drain and source of and if the abs. max is exceeded, the voltage will be too great for the spacing and essentially “arc” across, causing a short between these two potentials. When this happens, there is an inrush of current that flows through the device, causing further damage. So in a way, the device failure is caused by voltage and current simultaneously. Many times, this can also damage the gate structure since it is in close proximity as well. If the current is unlimited, it may result in the bond wires “fusing” which means they act like an open fuse - i.e, there will be no connection from the pin to the die.

    Best Regards,

    John

  • Ah, that makes sense. So will this same arcing happen if I exceed the -5V 'negative' limit of the switch node as well?

  • Yes, this is an abs. max rating that must not be exceeded for the reasons listed above.

    Best Regards,

    John