diyBMS problem when charging [SOLVED]

[CrAzY_DrIveR]

Hi,
I have been building my car battery pack using the diyBMS.
This project was developed for diy powerwalls but i thought it would be great for my car.

Now that it's all working i have a few problems.

When i drive the voltages are very slow to show any variation, i thint it's every 6sec.
When i charge the modules go crazy and i lose all communication.
The voltages go from 3.8 to 4.2 but it's not possible as the batteries are 240Ah and i am charging at 3Ah or 14Ah.

so i have a 32s3p battery (https://www.facebook.com/106.Electric)
Each bms is conected in a way that all 3p have the same lenght of wire to it.
Also there are temperature probes screwed in the terminals but don't worry about that.







All the cables are twisted.
All the cables until the last 10cm (when they go to the modules) are shielded.
The shield is grounded to the car chassis (negative) close to the controller and at each of the 3 packs.



Any idea of what i can do next?
I can't charge the car but i can drive it
Should i give up and get the 4-channel BMS from johu?

[arber333]

Been there...

1. Try routing comm cables away from HV cables or even 12V power cables. Even 30cm will be better as emi influence goes by Lsquared.
2. If you use shield for your comm cables it is very important how you connect shield to chassis. You can connect on one end or on both ends. Experiment as sometimes connecting both ends causes loops that attenuate voltage not sink it.
3. Inspect your comm optocoupler rc circuit with a scope. Sometimes RC filter is off and you can improve signal by changing the C value.

Wait... You use different code in modules? Try sorting that out first. You must use the same code for all your modules.

[CrAzY_DrIveR]

i will try that thanks.
i flashed all the modules at the same time before closing the battery packs so i was surprised to get that error.
inside and 10mm close to the busbars i have the short twisted cables (5-20cm) could this also be the problem?

EDIT:
The front pack has a different code But individually it can take 20Ah and only 10 crc errors (outo of 350packets). No crazy cell overvoltage

[arber333]

i will try that thanks.
i flashed all the modules at the same time before closing the battery packs so i was surprised to get that error.
inside and 10mm close to the busbars i have the short twisted cables (5-20cm) could this also be the problem?

Make your twisted cables go 90deg to busbars.
I think your problem might be poor crc protection in code or poor filtering. If BMS was made for domestic use designer might not have noticed any anomalies from relatively small current. Or if the system was designed for low S count there might be a problem with string response time.

[CrAzY_DrIveR]

I only had time to test all the packs individually and they work perfectly, only the front pack has 10 crc errors out of 350 packets,
the rear and the middle pack show no errors while charging at 20Ah.
Next i will test the cables from pack to pack and see if i find the one trat cuts communication.
The strange thing is that only if i connect all packs to the bms controller they will go crazy while chanrging and start balancing the first cell of each pack because it shows more than 4.2v, but connecting one pack do the bms no cell or bms module goes crazy...

[CrAzY_DrIveR]

New tests today i reached a conclusion.

Working only on the front pack rx0 to tx7 i have 8 modules with the same latest code,
the good big controller also updated.

The worse the comunication is the worse is the voltage fluctuation on the 1st cell of the pack.
This should not be possible but:

-When the controller is conected directly to the pack plug the first cell fluctuates from 3.849v to 3.858v when charging at 20Ah.
-When the controller is conected with 5mt+2.5mt cables on rx and tx the first cell fluctuates from 3.849v to 3.929v when charging at 20Ah.

If i concect to the other packs at the same time there is one cell that jumps to 4.2v or 4.5v when charging and that is just impossible.

How can the communication affect the voltage reading?

[celeron55]

Maybe taking measurements also with a multimeter in each situation would help. I have no clue - but when I have no clue, I take more measurements.

[CrAzY_DrIveR]

Hi, so i finally finished updating all the modules to the same code and i have the same results.
But new testing reveals that if i connect the shield (from pack 1 to pack 2) to car ground i lose all communication and the first cell of pack 2 starts balancing(i can see the red balancing bar and module temperature goes up).
If i remove all shields and just let the wire loose the first cell of pack 2 fluctuates from 3.8 to 4.08 but does not go to balancing (4.15v).
Today i will conect a wire to that cell and see what the real voltage is, but i suspect it's around 4.8 like all the others and like when i conect the bms to only one pack instead of 3.

[CrAzY_DrIveR]

I pulled wires from the first 3p of the rear pack that fluctuates when charging to get the real voltage reading as you suggested.

[CrAzY_DrIveR]

Now i understand why i can't shield to the car.
I still get a wrong voltage reading when charging with long cables even if i lower the noise on the line.
i forgot i had a small osciloscope:

ALL charging at 20Ah

controller directly to rear pack 12s


controller to 32s with long cables (osci on the cell 8 to 9, 5 meters +-)


if i ground the 8-9 cable to the car





if i ground the 8-9 cable to cell 9 ground

[arber333]

Do you think a bad isolation is possible?
What do you use for Rx on each module?
My BMS uses Pic12f directly on the Tx line (4V) and on the next cell i have optocoupler on Rx. That opto needs input resistor determined by opto CTR and its properties.

[CrAzY_DrIveR]

still why noise would change the voltage value of a cell?

https://github.com/stuartpittaway/diyBM ... ircuit.pdf

bms.JPG (17.21 KiB) Viewed 8100 times

[arber333]

still why noise would change the voltage value of a cell?

If the code is setup to look for the highest value than if the AD value would overflow (FF) BMS would signal cell at highest value.
This is why i only go until FE value. Any FF value would signal overflow, trigger alarm condition(watchdog), drop comms and reset voltage poll.
If you look carefully in my BMS github master code BMS AD values are compressed into a single byte from 2.50V to 4.35V in 254 steps .

If you can go on the opto pin and scope signal when you connect a single cell vs when you connect complete bank. When you see your culprit try to bypass hat module and see the rest of the pack. If all works it is just the bad module. I had my share of them. But strangely they would only manifest if i connected cells out of propper order. Since i use only single opto on positive input negative is left connected and voltage accumulates. If i would connect cells by bank i would probably burn couple of pics on the positive side of the bank.

[CrAzY_DrIveR]

I am shortening the cables i rerouted the problematic cable and it helps a lot, grounding will help even more.
Still the bms should be able to work with that noise maybe some day it will get another redesign.

edit: Grounded all long comunication cables to the RX of the module that recieves the signal and now it works no voltage fluctuation on any cell while charging.