DIY Stackable Single Cell LFP Protection Board - Feedback Wanted

[BoomStick0117]

Hi All,

I've been working on another project in which feedback would be much appreciated!

I got some raw 105Ah EVE LFP prismatic cells a year (or two) ago. (time flies...)
I decided I wanted to build my own cell protection circuit for over/undercharging.

This idea came up since I work with large SLA battery strings (lighting inverters) at work and find that the batteries are never fully balanced before installing.
As a result batteries would overcharge and fail over time as weaker batteries had a significant capacity decline... Shunting that current though, could keep the rest of the string healthier!

I ended up adopting this idea for Large LFP cells. I originally wanted to put them on the 280Ah+ cells making this idea more practical, but events at the time made that... unlikely in the states. My budget opted for 105Ah cells instead!

I am fully aware that BMS boards are readily available on the market and are likely better designed (and safer) than my basement etched monstrosities, but this project was meant more for learning than retail/commercialization.

Background aside, heres the circuit!

Single cell LFP Battery Protection Board

The circuit has 3 separate parts/functions:
-Undervoltage lockout
-Voltage regulation via Shunting
-Over voltage Alarm/"Shutdown"

UVL is pretty simple, circuit needs 3.0V to engage Q1 allowing the rest of the circuit to function. (I ended up replaced Q1 with a P-MOS)
Once engaged, Q2 is deactivated, lowering the de-activation threshold to around 2.7V
When Q1 is active, the Optocoupler is powered and will send an "OK" signal to the charger.
These signals are meant to be series to an "OK" input on a monitor board and high powered MOSFET Bi-directional switch in the future.

Voltage regulation is done via IC2 and Q3, as voltage rises above 3.4V, Q3 starts to shunt current up to 2.8A
(for space I ended up going with 10W 1.2 Ohm ceramic resistors)
Once 2.8A is reached, the voltage will continue to rise as Q3 should be fully engaged and can no longer sink more current (limited b Rshunt)
*Q3 is not a BS170 - forgot to update the FET PN, its actually a STP80NF03L-04 (30V 80A, low Vgs)

Over Voltage is similar to UVL. When Q1 is active, the Optocoupler is powered and will send an "OK" signal to the charger.
However, when the battery voltage raises above 3.5V, it will disable the opto and un-short R17 making the re-enable voltage threshold around 3.45V

Yes, I probably should've put the V-ref for both Shunt Reg and OVP before Q1...
This circuit was designed for ultra low power consumption, I believe quiescent current in UVL is only a couple uA, and only around 0.5-0.8mA when active!
This means that I can keep these boards attached to the batteries for months without fear of over discharge less than 1Ah per month!

Issues I ran into while designing and building:
-Q3 Gate voltage... I originally used IRFP1010s which I had plenty lying around, but the VGS was too high to operate properly on a >3.5V supply.
I ended up switching to STP80NF03s I found in Jameco's clearance section online. (I bought more than half of what they had left!)
These ended up working much better since the Vgs range is much much smaller.

-I originally wanted to use ribbon cable connectors to allow a neat feedthrough of all the battery voltages on one cable, but space constraints said otherwise, remembering this idea was meant for batteries 3x the size.

-During testing on the bench, I found the circuits to randomly start drawing 1-2Amps when transitioning between UVL and active. I ended up tracing this to my pull up for REF2 (R11) was way too high, causing a very tiny window where the Q1 was not fully on and causing a >1.5V to leak through. Vref ended up being lower than the sensed Vbat (via R12/R13) causing the Opamp to engage the current shunt. This was fixed by lowering R11 from 100K to 33K.

**Yes, I know these glass fuses probably aren't the best choice but I have hundreds in a bag from who knows when... My thoughts are if something was to short enough to cause over 5A, the copper traces will vaporize long before we reach the 10kA the cell is capable of delivering.. Right?

Boards are etched in Muratic Acid + Hydrogen peroxide, traces made with toner transfer.
All parts are hand soldered, cleaned with 99% IPA and dried.
Boards are bolted across the cell with some large screw terminals with the thread drilled out.

I've attached the circuit, PCB layout and a photo of intended installation.
8 painstakingly assembled boards. I still have 8 more to go...

Let me know your thoughts!
Suggestions thoughts and comments appreciated!
Please let me know if my assembly is a giant fire hazard waiting to happen!

Thanks,
David