I have an application where I am trying to balance 4 series-connected lead-acid batteries in a backup power supply. I recently started cycling the batteries more frequently, and have started measuring the individual battery voltages, and have noticed they have become quite unbalanced.
Since this is not a volume application, I attempted to use a DC2043B demo board in conjunction with a supercapacitor to accomplish my goal as simply as possible.
Unfortunately, during an initial powerup test, it seems the LTC3305 chip has become damaged, and I am trying to understand what happened.
Here was the state of the connections when I tested:
Everything seemed to work fine for 6-8 seconds, when I was about to measure to ensure Vreg, Vboost, etc, were operating correctly. But then the power supply tripped, and now trips whenever I connect it.
I now measure a resistance of 17 ohms from the V4 to GND pins on the chip, 6.7 ohms from V1 to GND and 23.6 ohms from V4 to V1. Between the other battery pins, it is high-impedance. I can measure a capacitance from the decoupling caps between these pins.
The only low-impedance path I can measure between V4 and GND seems to be through the LTC3305. In addition, there also seems to be some additional low-impedance path that has been introduced between V1 and GND. Between V4 and V1, I measure the series connection of these paths.
All of the NMOS on the board are measuring high-impedance from drain to source, and of course none of the fuses have blown.
So, it seems that some failure has occurred in the chip due to V1, V2, V3 being left floating. Is this a possibility? Is the LTC3305 designed in such a way to not support this kind of connection?
Unfortunately, in my application, I need to be able to occasionally disconnect the batteries without shutting down the charger/supply so that the load can continue to operate. I would prefer to not have to think to disconnect the battery balancer in this case to prevent its destruction.
Thanks for any insight you can provide.
Here are some follow-up observations.
I have not run the DC2043B/LTC3305 with V1, V2, V3 open, so I have not experienced this problem, and I can only speculate as to the cause. Generally speaking, of course, it is not a good idea to apply power to an IC with inputs open unless the IC is specifically designed to accommodate an open circuit (e.g., an internal pullup or pulldown is present). Specific to this situation, note that in the DC2043B C1-C4 and C12-C15 comprise capacitive voltage dividers. When a voltage is applied between V4 and GND the voltages at V1, V2 and V3 initially charge to a voltage inversely proportional to the capacitance (V=Q/C). Thus, the voltages at V1, V2, V3 are in general not distributed equally. Following the initial power application unequal bias currents from the V1, V2, V3 inputs may further unbalance the voltages at these inputs. Conceivably, an overvoltage condition could arise between the V inputs leading to unwanted behaviour of the IC, possibly leading to destruction. Again, this analysis is speculative.
I suggest operating the LTC3305/DC2043B with all inputs connected to the battery stack. The external power supply between GND and V4 should not be used. If there is some reason you do not want to do this, I would suggest adding balancing resistance between the V inputs or installing the optional zener diodes (D135-D138) in the DC2043B.
Also, if there is any question about applying the LTC3305, note that you can run the LTC3305/DC2043B while the battery stack is charging, discharging or idling.
Thanks for the reply, and information.
Unfortunately, I had assumed that the V1-V4 inputs were high-impedance, and would float between the levels of V4-GND due to the capacitive voltage dividers, and that there would not be a means for current to be sourced or sunk through them to cause a significant imbalance. But what you mention makes sense. I suppose some sort of latch-up could have occurred.
Since I have no reference device to compare, I will assume that the resistance I am measuring between V4-GND and V1-GND is indicative of the part no longer being serviceable.
It looks like populating the diodes or adding resistors would be a good solution. It's a shame those 4 zener diodes were not populated -- I'm not sure I understand why they would not be desired. It should work around the problem that still remains if batteries are to be removed from the stack. The battery balancer would need to be removed from the system whenever any individual battery is removed, since it seems an unsafe condition can exist during even the short time until all the batteries are disconnected.
Unfortunately, without access to rework equipment, I am not in a position to test these suggestions but I will note them for the future.