BQ25798: IC catching fire

Hello Dave,

I haven't seen this scenario before. This is really strange.

I have a few questions to help debug this:

• Do you have the BATP pin connected to the battery pack? If so, how is it connected?
• What's the charge current from the BQ25798?
• If you have boards to spare, is the BQ25798 damaged reliably when the BMS disconnects the battery?

Dave Besson said:

Is it possible that the BQ25798 while charging in boost mode from the 5V USB-C Input overshoots the max FET voltage internally when the battery is suddenly disconnected and whatever the protection I have on all power paths, the surge comes from inside the chip and thus, destroys the FETs way before the TVS have a chance to do their job?

I'll think about this and get back to you.

Best Regards,
Ethan Galloway

Hi Ethan

The BATP pin is decoupled to ground through a 47nF Cap and goes through a 100ohm resistor to the battery (and pins 22 - 23 of the IC) and the VBAT has a 18V TVS (as well as VSYS and VBUS) to avoid transients. The schematics is pretty much what is in the datasheet as "typical application" but without the input FETs, without the battery FET. The 5V ( 9V / 12V with a USBCPD compatible adapter) from my USB-C connector is fed straight to VBUS, VAC1 and VAC2.

I'm setting the REG03 with 0x64 (100) which gives me 1 Amp of charge current and ILIM is also set to get close to 1A.

I don't understand what you mean by "damaged reliably", do you mean every time the battery is disconnected the IC burns? That I'm not sure, it seems to be a very specific set of parameters that when met, the chip burns. At first I thought it was when I powered it on after a full charge, but this last time the board was sitting peacefully on my desk while charging, the only thing that happened that is consistent with the 3 fires I had is that the battery had just finished charging. And I have a dozen more boards that have been charged fully and they never failed.

I don't have alot of spare boards to test with, but I'll try another one and see if I get similar results when manually pulling the battery plug when it's charging and near fully charged.

One option I thought was to lower the max charging voltage but if the battery is disconnected then reconnected without powering the device on, the default 16.8V will be selected by the 17.4K resistor on the PROG pin. Unless I use a 10.5K to set the BQ for 12.6V and then change it back to 16V in software? I'm not sure if it's the right way to go.

Hi Dave,

The charger is expecting the pack protector OVP to be higher than the charger's VREG setting.  If the protector opens while the charger is in CV/taper mode before termination, the charger's CV loop could overshoot with only the min recommended capacitance at BAT and/or SYS. Adding more (>100uF) capacitance on SYS and/or BAT prevents overshoot.

I would have expected the TVS diode to clamp the output to prevent this.  Other customers have used such a diode successfully.  Is the diode close to IC SYS and GND pins?  Can you remind me if you have the shipFET?  If no shipFET then the 47nF on BATP is not really doing anything.

Regards,

Jeff

Hi Jeff

I have 20uF (2 x 10uF) on VBAT and 50.1uF (5 x 10uF + 1 x 0.1uF) on VSYS, which is what's recommended in the datasheet.

The TVS I'm using for these tests is the Bourns P6SMB18CA with a breakdown voltage between 17.1 - 18.9V. Why this one? That's what I had on hand. I use those to protect the 20V gates of IGBTs on some other designs. The diodes are located as close as can be to the BQ25798, next to the caps at the output of VSYS and VBAT (and I also have one on VBUS to clamp the input transients when connecting the USB cable for charging). Should I push the caps farther out and put the TVS between the IC and the caps?

That's why I thought this could come from residual energy in the inductor that would induce a high spike if the battery disconnects just at the right time (or wrong time) and since the surge comes from within the IC itself, the TVS couldn't clamp it fast enough and some internal FET would short and toast the whole IC.

I'm not using the ship FET. But I still put the 47nF on BATP because the datasheet says: "Recommended if hot plugging 4S battery packs with long leads or PCB traces." While I'm not supposed to hot plug the battery, the low / high voltage cutoff will do so. They also don't specify what length is "long" leads or traces, so I put it there because my battery packs have about 8 inches of wire.

I included a screenshot of my PCB where the BQ is located, D7 / D8 are my TVS. You can only see 3 caps on VSYS here, the 2 others are closer to the VREGs.

I have reached out to my battery pack manufacturer to ask them if it's possible to change the charging cut-off voltage and I will try that, push it to like 18V.

I will try to add more capacitance to the output of VBAT and VSYS. Does it need to be all 10uF ceramics? Or can I just stick a good old 330uF electrolytic in there and place it anywhere? The datasheet specifies: "Low ESR ceramic capacitor such as X7R or X5R is preferred for the output decoupling capacitor and should be placed close to the SYS and GND pins of the IC". But is it for the minimum capacitance only?

This is the aftermath:

Every time the damage is similar. The chip cracks from the bottom in the middle and you can see flames shoot out mostly from the bottom and from the right side.

Hi Dave,

If the ground planes are good, the TVS diodes should be okay.  The 47nF on BATP slows down the CV loop response so I suggest removing it.  Also, if there is a way to slow down the package protector FET turn off time with a series resistor and/or more gate to drain capacitance, that might help prove our theory at least.

Regards,

Jeff

Finally got a reply from the battery manufacturer who won't help me at all so I need to find a plan B. I thought about lowering the charging voltage so it never reaches the BMS cutoff voltage, but the problem is, if the battery gets too low and the low voltage protection disconnects the battery, when plugging the charger the BQ25798 will POR and revert to 16.8V until the device is turned on and the MCU can reprogram the BQ25798 so it leaves the system vulnerable.

What if I "hardwire" the BQ25798 to use a 3S battery instead (using a 10.5k instead of a 17.4k)? If I understand the datasheet correctly, I could just reprogram the CELL bits (REG0A) to a 4S battery and change VREG to 16.5V? There's still a small chance that the watchdog in the BQ25798 will revert VREG to 16.8V in which case I could set the STOP_WD_CHG bit to stop the charging on a watchdog timeout, and in the case of a POR, VREG would revert to 12.6V so no problem for the battery pack.

Hi Dave,

Setting 3S default sounds like a safer option.  You could even slowly increase VREG in small steps.  Keep in mind each cell setting (1S,2S,3S,4S) has a restricted range for VREG, i.e. if you slowly increase VREG in 3S there is a point where you will have to change to CELLS register to 4S to continue increasing VREG

Regards,
Jeff