L9963E in 12S setup, middle cell shows too low voltage

[TSuik.2]

Hello!

I have a L9963E in 192 cell EV battery setup, there's 16 modules daisy chained, each handling one 12S pack. The bottom module has a L9963T and stm32 chip that handles the whole pack and communicates with car through CAN bus. Everything is working!

However I noticed that every module reports slightly lower voltage for cell 7 (Vcell 9 in L9963E). This is the cell that sits right next to the "middle gap" where I have two unconnected cell inputs. VCELLx_EN is "0" for those two cells.

RLEC 0 (=0):
  Volts: 2971 2969 2971 2969 2969 2967 >2925< 2967 2971 2969 2964 2940
  Temps:   21   21   21   21   21   21   34
  faults: 00   timeout: 0
RLEC 1 (=1):
  Volts: 2979 2979 2979 2979 2976 2979 >2945< 2979 2979 2981 2979 2971
  Temps:   21   21   21   21   21   21   30
  faults: 00   timeout: 0

Volts are in mV. Left side is lowest cell (Vcell1), right side is highest (Vcell14). As you can see the cells marked with > < have about 35-40mV lower voltage than the rest. Also the rightmost cell is slightly lower but not that much. In reality when measured with multimeter, all the cells in module are within +-2mV.

What is causing this? I think I have the unused pins connected correctly in the schematic (see below). Unused pins are shorted together and connected to the lower cell through 1k resistor, as instructed in the datasheet.

I haven't yet charged the battery to see what the difference is fully charged, but I feel 40mV error in the top end can be catastrophic. I can "fix" it in software but that's of course not ideal.

[TSuik.2]

This is the relevant part of the schematic. Datasheet says:

◦ Unused pins must be first shorted together to eliminate differential noise
◦ Shorted traces must be connected to the PCB connector of the lower mounted cell, through a 1 kΩ resistor

However, there seems to be some leakage current involved. I measured 46mV over the 1k resistor. VCell9 voltage is measured between C9 and C8  pins, which now shows 46mV too little.

I removed the 1k resistor and shorted the unused pins directly to C6. I'm now getting exactly correct results and haven't noticed any ill effects. Is there any particular reason for the advice to use 1k resistor? I assume it would work for every other pin, but in my opinion C8 and C6 must be tied directly together for the voltage measurement to work.

[TSuik.2]

Continuing my monologue. :)

I have nearly all diagnostic features enabled. I noticed this bit on cell open wire diagnostic:

"Diagnostic is present just on enabled cells (VCELLx_EN = 1). The mechanism used for this detection is based on a diagnostic pull down current (IOPEN_DIAG_CX), which allows to measure the voltage drop generated on the external RLPF resistance connected in series to Cx pin."

And IOPEN_DIAG_CX is defined as 40-60uA, 50uA typical. Now if C6 pin has 50uA pulldown but C8 doesn't (because VCELL8_EN=0), doesn't this create exactly the ~50uA current draw over the 1k resistor I'm seeing?