BQ24780S: Mosfet drivers not working properly

Hi Anastasios, 

I don't see any obvious issues with your schematic. To begin diagnosing the issue I would start with probing to check if the FETs themselves could be shorted to ground. 

Best Regards,

Aaron 

Hello Aaron,

Thank you for the quick response!

I assume you are referring to the gates of Q5, Q6? The gate voltage was a few millivolts above the "AC" input, 0.5V at most.
I will double check on Μonday, but I didn't notice any shorts to GND. 
The FETs were also replaced on one PCB, without any improvement.

In fact even Q5 seems to conduct somehow, albeit with a very high Rds (around 600mOhms). 
Without load the Vsys output is around 17.5V, I assume that means that Q6 does not conduct at all, so the output current passes through Q6 body diode.
When load is applied sometimes you can get the full output current (of course with a very high voltage drop), sometimes Q5 stops conducting entirely. 
Another weird thing that I can't explain: When I tried removing R20 (AC sense shunt), the BQ24780S IC got fried.

Could these issues be somehow related due to the fact that I haven't connected a regular battery for the initial tests or due to firmware configuration?
Or perhaps there is some latching output protection on the FET driver that gets triggered due to the relatively high gate capacitance of the FETs?

Hi Anastasios, 

Anastasios said:

Another weird thing that I can't explain: When I tried removing R20 (AC sense shunt), the BQ24780S IC got fried.

Could these issues be somehow related due to the fact that I haven't connected a regular battery for the initial tests or due to firmware configuration?
Or perhaps there is some latching output protection on the FET driver that gets triggered due to the relatively high gate capacitance of the FETs?

Without the sense resistor, the IC can no longer limit adapter current so the IC getting damaged in this case does not seem abnormal. I don't think you using a regular battery vs a supply or any firmware configuration would cause this issue either. I would suggest trying to use the same FETs that we use on the EVM and see if that helps.

Let me know your results are from testing the FETs shorts and I can give further suggestions. 

Best Regards,

Aaron 

Hello Aaron,

The (Q5) FETs are indeed damaged in all 5 prototypes, as you suspected. D-S is shorted and G-S measures ~100 Ohms!
Q6 seems ok in all instances.

I tried replacing Q5 once more and reducing the gate resistor (R24) to 4R7, but it got damaged again after a couple basic tests.
Unfortunately I can't replace the FETs with CSD17308Q3 because these are 3x3mm package, whereas BSC052N03 is 5x6mm. But at first glance the dynamic/gate charge specifications seem very similar and the Rds on is about half.
What could be causing this problem?

Looking forward to hearing your insight/suggestions

Hi Anastasios,

When the gate resistance on the FET was still 10ohms, did it take repeated attempts to fail (like with the 4.7ohm) or was it the first time? A quick suggestion would be to see what happens when you limit the input current to lets say 1 amp or so and see if ACDRV can be driven properly then.

Hi,

I am not sure if there was a significant difference with the smaller gate resistor.
The load current in today's test was pretty low, up to 2A, while the battery terminal was unconnected. There is a fairly large copper polygon connected to the drain of Q5 as you can see below, so I don't think it's a thermal issue.




I did some further testing, under static conditions and a load of 2A the circuit seems to work properly.
Apparently the FET is damaged when the "AC" input is hot plugged (with some load being present on the output). As you understand, this is quite likely to happen under use, so it should not happen on the final design.
Maybe it would help if I decreased the turn-on delay of the FET by reducing the values of C34-C29?

Hi Anastasios, 

I think that would be worth giving a try. I attached the recommendations below. 

I would also recommend a snubber circuit on the input even if the input is coming from a buck converter. Is there a particular design reason for the absence of one? 

Best Regards,

Aaron 

Hello,

The buck converter is on the same PCB, only a few cm away from Q5, so we considered the snubber unnecessary. 
I will do some tests tomorrow with a snubber and a mosfet with higher Vds rating, to eliminate the possibility for damage due to voltage spikes.

I was also considering that, since we do not need the Learn function, perhaps we could replace Q5 with a short permanently. Q6 will not allow current to flow back to the 18V converter, so it should work ok. 
Will the BQ24780 have an issue if CMSRC is shorted to the +18V input?

Hi Anastasios, 

I think could work, there should not be an issue with CMSRC being connected to 18V input. Let me know if you have success using the snubber. 

Best Regards,

Aaron 

Hello, 
I tried adding a snubber with 2R2 & a few cap values, from 100nF up to 4.7uF. There was no significant improvement, Q5 still gets damaged.
I also re-checked the input voltage on an oscilloscope, couldn't measure any overvoltage events with or without the snubber. Just minor dropout (200mV max on the drain of Q5), when connecting a heavy load. Nothing that I would expect to cause unstable response from the charger or otherwise damage components.

Is there any other disadvantage to omitting Q5?
Of course the charger won't be able to disconnect the 18V input from the load due to the internal diode of Q6, but we generally don't mind about this. The 18V converter can be disabled from it's own enable pin if necessary.

To give you a better understanding, the circuit is meant to be used as an uninterruptible power supply.
It will be always connected and supply power from the 18V rail to the system under normal conditions, the battery is only meant to provide power to the system when there is a blackout or brownout from the mains supply.

Hi Anastasios,

The only issues I can see is the inability to regulate input current as well as what you have described below. But as you stated the input is coming from a buck so I do not see any obvious issues with removing the FET. 

Anastasios said:

Of course the charger won't be able to disconnect the 18V input from the load due to the internal diode of Q6, but we generally don't mind about this. The 18V converter can be disabled from it's own enable pin if necessary.

Best Regards,

Aaron