Interesting battery failure mode

28.04.2013 13:29

Thanks to my previous posts about Amazon Kindle, I have another broken specimen on my desk now. This one seems to have experienced an interesting battery failure.

Amazon Kindle 3 batteries

Kindle's battery has 4 terminals: ground, a positive terminal for power and SDA and SCL pins for I2C communication with the integrated battery management circuit. On a normal battery, the positive terminal is around 3.7 V above ground, depending on the charge level of the Li-ion cell and the I2C lines are on ground level, because they need external pull-ups.

This broken battery however has the positive terminal at 0 V compared to ground terminal while the I2C pins are at -2.5 V. I can't imagine what kind of failure mode could cause pins to go lower than ground, unless the polarity of the cell got reversed somehow. I don't see any way how a failure in the battery management circuit or a loose connection somewhere could cause such readings. I'm pretty sure it's not an artifact of my multimeter either, because the battery can draw some milliamps of current from the ground to one of the I2C pins. For the record, this looks like an original 1830 mAh battery. Date of manufacture is April 2011 and type is 170-1032-01 Rev. A.

The master I2C interface on the Kindle wasn't damaged though, because it boots and reads out battery state just fine when attached to a different battery. There does seem to be a problem with bad a connection somewhere on the motherboard, because it crashes if I lightly knock on the CPU package. Possibly a hairline crack in some solder joint. But that's a topic for some other time.

Posted by Tomaž | Categories: Life


Thanks for posting this. I've just discovered my kindle keyboard's battery has come to a similar end. 0v between - and +, and -2.57v on SCL and SDA relative to -.

Mine was also an original battery, as I bought the kindle new. The type is the same as yours, but the manufacturer date is August 2011.

I've just bought a new battery. I wouldn't have really bothered spending money on that but from your post I'm more optimistic about it working!

This battery has GND pin disconnected from the cell (presumably due to a bug/failure in internal protection circuit). And there is a "resistor" from GND to VCC that pulls up disconnected (contact side) GND pin to VCC. I2C pins are lower then VCC. Quite likely SDA/SCL are at the potential of cell-side GND (remember - GND is disconnected between cell and contact), hence "negative" voltage on I2C pins (because you are actually measuring with respect to VCC).

Two not-very-intuitive things to understand here are:

- Low-side switching. Battery protection circuit would measure voltage (and rarely, current too) of the cell and isolate cell from contacts if those are not in safe range, but why disconnect GND? Well, this has to do with how MOSFETs work. N-channel MOSFET are inherently more efficient, but require gate voltage to be above the source/drain voltage to open. Meaning, it's easier to do low-side switching (disconnect GND). Contrast this to more rare high-side switching where you have to use either lower-efficiency (higher-resistance-channel) P-MOSFETS, or have a "bootstrap" circuit to generate gate voltage (above VCC) for the high-side N-MOSFET. Neither is desirable in battery protection circuitry. Hence, low-side switching (disconnecting GND) is most common.

- "Resistor" from GND to VCC. This could be a short in protection circuit itself (for example, ceramic capacitors are often fail short-circuit), or it could be leakage through "logic" part of battery protection circuit. Hard to say without taking out and actually measuring protection circuit.

Add a new comment

(No HTML tags allowed. Separate paragraphs with a blank line.)