We have ignition. TP5100 fiery fail

nz_lifer

Member
Joined
Aug 26, 2020
Messages
87

A frame from the full view of the camera.
vlcsnap-2021-04-11-08h46m00s071.png

Out of the 80 TP5100 modules I have, this would be the 4th one that has failed.
2 just didn't charge. 1 almost went out like this but I caught it in time during initial commissioning.

Late on a rainy Saturday night I was adding another battery to my test bank.
The charging system was on the other side of the covered deck. About 5m (16ft) away.
First visual sign of trouble can be seen at 33:17.
Last flash at 34:01.
I don't notice anything till 34:30 when the smoke get to me.
I don't find the source till 47:00, as the system was just ticking along.

It seems the only damage so far has been limited to that slot and some melted cell holders either side.
I have not removed the failed TP5100 to check if the mosfets underneath are still ok.
Won't do that until spares have arrived and I can bring the slot back online.
The system otherwise is still going as normal.


--Analysis.
Could this have been prevented?
Yes, if I had used premium modules? Is there such a thing unless you build it all yourself.

Duration of the fire could have been reduced if the 12v power was removed as soon as signs of trouble were detected.
Either a temperature sensor or current sensor on each module.
Or one on the 12v rail with some maths to calculate how much should be going in from the known draw of each slot.
ie there should be 10w due to 2 slots charging. Greater than 12w there may be a problem.
That is assuming it needed much power to flame like that.

Generous component separation prevented a cascade failure.
It did slide down the posts it was soldered to and one end was touching the main pcb.
I think that is what allowed it to damage the cell holder next to it.
Flames coming out at an angle.
Might put a kink in the post to stop it doing that.

The system sitting on 12mm thick tiles, on a metal shelf.
It should withstand a complete meltdown.
My only concern is if cell are ejected beyond they shelving.
Hopefully the existing sensor prevent it getting to that stage.



--Going forward
I will add a smoke detector to the system.
It will trigger an input in the monitoring system.
Activate an extraction fan.

Will investigate the power monitoring.
Can not add it to current system as power drawn monitoring is not precise enough.

Safe processing everyone ;)

The system pictured
 

Attachments

  • 20210410_204731.jpg
    20210410_204731.jpg
    1.2 MB · Views: 287
  • 20210410_204738.jpg
    20210410_204738.jpg
    919.9 KB · Views: 259
  • 20210410_204742.jpg
    20210410_204742.jpg
    999.3 KB · Views: 287
  • 20210410_204747.jpg
    20210410_204747.jpg
    1.1 MB · Views: 328
Glad nothing really exciting happened. Thanks for sharing the video, the situation, and the future resolution. 👍
 
Wow, scary. I've used these modules briefly before but wasn't impressed. I blew up 2 of them trying to use them in 2s mode before discovering that they're unstable at 2s until you add your own decoupling capacitors; maybe this would make them more stable in 1s mode also? I believe if you measure the capacitors used on this board, they're missing a few that are listed as needed in the datasheet.

If using a TP5100 module for charging 2s, you need to at minimum add 2 caps (0.1uF 50V) on VS and VIN, since those are the critical nodes for the switching PSU. This should be sufficient to make it stable, otherwise the TP5100 IC will most likely blow up.
tp5100_fix.png
 
Difficult to say from the pictures, but...
If I had to guess, I'd say that the power supply to the modules is inadequate/too noisy.

At peak charging, the board uses 4.2v * 2A * 16cells = 134.4W, so 11.2Amps on the 12V supply side. Looks like you have 2 boards daisy chained, so 22.4Amps?
The 12V cables don't look particularly thick.
The 12V traces are 4mm wide, just enough thanks of the looping splitting the current into 2 separate paths. But likely inadequate if supplying a second board. Unless you have 2oz PCBs.

But I think the biggest issue is that the 11.2Amps (22.4Amps?) is only the average draw. The TP5100 is essentially a buck converter, which draws current in peaks. It's impossible to say how high the peak current of 16 (32?) TP5100s is without an oscilloscope, but I'd guess it's ginormous.
I doubt the 10uF capacitors on the TP5100 modules are sufficient to smooth out the voltage in this system, and the caps of the power supply is probably too far away.
On the other hand, the TP5100 should work fine even if the voltage should drop to 5V... :unsure: So maybe just a shitty module after all.


My suggestions:
Look at the 12V supply on some of the modules with an oscilloscope. If noisy, then
Beef up 12V cables, no daisy chaining
Add generously sized capacitor to each TP5100 module input
Reduce current of TP5100 module


On extra safety features:
Perhaps a simple axial glass fuse for each TP5100 module? Or perhaps a thermal fuse? A resettable PTC fuse might also work well, since they kinda activate based on temperature.
 
Hi! Sorry to resurrect this thread, but I'm also having problems with my TP5100. This isn't the prebuilt module, but my own fabricated PCB.
I set it to single cell, 500mA. Vin 12V (power supply, 115kHz non-linear, capable of 6A, also tried with 12V battery pack). It won't charge. I tried 3 boards now, one worked for one full charge (~2000mAh), then gave up the next day.
10uH inductor (should perhaps be 22uH as I'm powering from 12V). Followed the datasheet with caps, diode, etc. (Only difference I made is is 1M pullups on the LEDs, as I'm reading the signals with an MCU, and TP5100 CHRG and STDBY pins are open-drain.)
It puts out 2.5mA or so (battery voltage stays the same) when I power it with 12V, nothing when off with battery still connected (almost nothing in reverse).
But powered without battery, the BAT pin jumps to 4.1-4.2V (as it should), but drops to 3.8V-ish as soon as I connect the battery.
I also noticed both LEDs light at the same time. When touching the CS pin with a unconnecting (floating) test lead, "charging" LED stops. But no change in voltage out/current out.
If I pull the CS up to VREG, it outputs ~8.4V. So the chip is responding somewhat.

Most datasheets say 18-20V max, but the one I bought (from LCSC) says 10V (https://datasheet.lcsc.com/lcsc/221...Extension-Microelectronics-TP5100_C379389.pdf)
Maybe Top Power ASIC revised the chip and upped the specs, or it's the same chip, but limited the VIN due to failures. I emailed the manufacturer, but maybe someone here knows?
 

Attachments

  • tp5100-my.png
    tp5100-my.png
    187.9 KB · Views: 28
Back
Top