Improve safety with temperature measurement - Li-ion 18650

[putz182]

Hi there, It's been I while that I wanna build a battery pack for my skate board. It is originally based on PbA battery. However, another thing that keeps bothering me is battery recycling. Besides, I have access of a bunch of cells from notebook batteries. So, I wanna reuse it. Please, understand that this is my goal!
Thus, I will use electronics to improve it and keep it safe...

I actually know the problems and safety issues regarding parallel and series arrangement. BMS can solve most of the problems for series applications, however, parallel doesn't. To be more clear, I wanna mix batteries with different capacities. Ex: 2500mAh (4.2V) and 1500mAh (4.2V), in order to reach 4000mAh.

1. My first approach is reduce current drain, for instance 3Amax (2500 + 1500). It seems reasonable for stationary applications, but not for transiente like the skate board.
2. I also wanna add a second layer protection to work with BMS, and here is my question.

I will produce a pcb board as Patrick did:



However, I wanna place a SMD termistor near each cell connection. If temperature increase, a SMD mosfet, driven by a pull-up resistor will conduct and change the logic level to zero (which is the voltage that BMS is monitoring).

I need some tips and opinions. I also hope that, if BMS stops reading 3.7V, but go to 0V, it will open the circuit.

[amyk]

See this previous discussion: https://www.eevblog.com/forum/beginners/connecting-different-sized-lithium-batteries-in-parallel/

[putz182]

Thanks amyk, I did already.
From my point of view, everything is inconclusive. That's why my question is about the circuit, not about the battery itself.
I will perform my own tests, but before I start, I want to develop some protections.

I want to fake a 0V for BMS and force it to shutdown. 0V is emulated with a NTC.

[sleemanj]

A thermistor for each cell?  That's crazy IMO, nobody does that.  And on top of that PCB, the thermal transmission is I doubt going to be that precise, or fast, or accurate in any way.  I would embed some thermistors in the pack (if ant all), you could do one thermistor for 4 cells (in the middle). As for your circuit idea in general, it's going to depend on the BMS/protection you are using if it will be happy with that.

Otherwise, that PCB... looks like a massive PITA way to build a battery.  It looks like you need to feed through a contact wire from each individual cell through the PCB, and solder it to the PCB, I don't know how it's connected to the battery, is it a spring or directly welded? 

If it's a spring then that's a lot of pressure being applied to the PCB, and if it's welded/soldered then you have a lot of work to remove a cell!  This is not a good design if you ask me, not that you did.


Also: Consider LiFePO4 cells, that way you don't have to worry very much about it going thermonuclear if things go wrong.

[szan]

I see another problems with this pack design from mechanical point of view - how those 18650 Li-on battery cells are connected to those two PCBs on the top and the bottom?
It looks like vibrations while ride skyte board might simply brake soldered/spot welded connections and create dangerous bad connections where sparks can apear and generate additional heat...
I think, It should be more versatile design  to avoid mechanical stress.

 

[Daixiwen]

This will create a safety issue rather than solve it. First the MOSfet will not suddenly start to conduct and take the voltage to 0. This will happen gradually, and as the temperature increases the voltage reported to the BMS will slowly decrease.
But what is worse is that when you are charging, as the cells heat up, your battery will report a voltage that is lower than the actual cell voltage to the BMS. As a result the BMS will end up overcharging one of the cells, causing a thermal runaway and burning your battery down (as well as your house, if you charge it inside).

[Siwastaja]

If you really want to monitor individual temperatures, just build a tree of analog muxes, or use digital sensors in a shared bus, or whatever. It will be very expensive and complicated anyway, so I don't see the point in that strange solution which, if I understood correctly, tries to fake the measured cell voltage in case of bad temperature?

Thermal coupling between the sensor and cell is important if you want to react quickly to weird occurrences with bad cells. Maybe use multiple thermistors per cell, coupled to different areas so you can catch localized heating? Needless to say, it completely defeats the cost savings obtained by using crappy second-life cells.