TPSM63604: TPSM63604 catastrophic failure

Hi Wilson,

Sorry for the inconvenience, can you send the schematic and layout for me to review? Also, how often is this type of failure did you see? Does it happen every time on every board?

Thanks,

Andrew

We've had no issues powering off a benchtop power supply, 12V battery packs, or other systems at the integration level. This only happened on a particular 24V system that we plugged into, likely driven by a bank of lead acid-like batteries. It's likely very low output impedance and able to source a ton of current.

We ran some tests on benchtop where we apply power in hot plugged fashion (off a 24V/5A PS) to get a sharp dV/dt on the input. On occasion, we're able to measure a spike > 40V at the node where TPSM63604 VIN is.

I'm trying to attach a pic of the circuit but it's difficult with this tool.

Hi Wilson,

The way you described it, it does seem likely that the 42V abs max may be violated. You may be able to solve this the line transients damping high ESR capacitor at the the input of the device.

The EVM uses a 100uF Aluminum electrolytic capacitor (p/n UUD1H101MNL1GS) at the input to add some VIN dampening.

Thanks,

Andrew

Andrew,

Thank you for your reply. I've got some questions.

1. Can an overvoltage violating the 42V abs max on the datasheet cause a permanent short of VOUT to GND? And in our extreme case, a violent thermal event that completely destroyed the VIN section of the device (melted PCB with black soot particles)?

2. Is there any possibility that there's a initial inrush that the OVC protection (ie. via hiccupping) of the device is unable to guard against? To my knowledge, hundreds of uF of bypass caps on the output should be no problem for this device to handle.

3. You're saying I should put the high ESR cap at the input CLC or as part of the bank of input caps? I'm not sure how a ESR cap would reduce the voltage transient at VIN of the device given that we're seeing the high amplitude ringing occur at the output of the CLC.

Thanks,

Wilson

Hi Wilson,

1. Yes, I've seen a similar outcome for VIN abs max violation.

2. TPSM63604 cannot protect itself from overvoltage.

3. Dampening the LC filter you have implemented will reduce the effect of a 40 Line transient at effect the input. It slows the response, and hopefully by the time it sees the full voltage of the lead-acid battery, it's the 24V nominal voltage.

Thanks,

Andrew

Adding something like a 30V clamp at VIN can also protect our device.

Sorry, on (2), I was trying to get at whether I can reliably eliminate inrush current as the culprit. It was the leading root cause before because the failure occurred when we connected to a system capable of high amperage delivery but we hadn't had the same problem anywhere else. What do you think?

On (3), I cannot dampen the LC resonance by increasing the ESR of the caps but I can cause a much quicker decay (initial amplitude remains the same) if I add DCR to the L.

We will consider that. Since we are hoping to operate at the full 36V input range, I will need to clamp it similar to the voltage input into the CLC.

Yeah you could try adding DCR to the L to see if that helps with the VIN overshoot. I'm interested to see how high VIN terminal of our device gets for this test.

Thanks,

Andrew

Hi Wilson,

50V (even for short period of time) will kill the part. I think you may want to consider using a 65V rated part for this application.

Thanks,
Andrew

Andrew,

Actually, our max system input is 36V and might get reduced. The 50V I'm talking about is the ringing and hold caused by the CLC and cap bank. My intent is to redesign that entire circuit with a tail-to-tail load switch soft start at the input, followed by the CLC. The effect should be very little spiking (or none) going to VIN of the part above the max system input. ie. the part's VIN should be held below absolute max of 42V at all times. Does that make sense?

Thanks, Wilson

Hello Wilson, 

Andrew is currently out of office in observance of the US Holiday, he will respond soon as he returns to office. 

Thank you for your patience. 

Regards, 

Oscar Ambriz