• State
  • Locked Locked
  • Replies 14 replies
  • Subscribers 63 subscribers
  • Views 2583 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

BQ34Z100 R1 fuel gauging and misleading remaining capacity.

martin gibbs
Intellectual 520 points

I am fuel gauging a 3200mAhr pack and discharging at a 50mA rate as a test. The remaining capacity and time is adjusted slightly several times until we get to the voltage drop off knee: The remaining capacity is suddenly adjusted downwards in the last 15 minutes. This is a problem (medical applications) when expecting say10 mins but end up with 2 and being informed at the last moment. Any ideas what could be happening and what to adjust please?

It happens when the battery is left for a while ( say +2 days) before commencing discharge. I'm guessing its something to do with quiescent discharge but that should have been captured by the fuel guage. 

  • 0
    TI__Guru 73830 points
    Martin
    You should not see the RM drop that much quicker near the end of discharge. There are several things that can affect this. ChemID match, optimization cycle, parameters, etc. Can you provide a log file with your observations and an srec from the pack that you are using? I will review them and see what I can find.

    Tom
  • 0 in reply to martin gibbs
    Intellectual 520 points

    Toms folder.zipThanks Tom
    Here's the log file and the graphed data. This one is a 750mA DSG but you can see that at the end of the first cycle the RC drops 11 mins too early but on future cycles its OK. The cut off is 22000mV but the RC drops dramatically just before hand

    It was left PRES enabled (no discharge) 2 days prior to the first discharge cycle.

    Thanks for having a look Tom, much appreciated.

  • 0 in reply to martin gibbs
    Intellectual 520 points

    zipped.zipIn addition this the graphed data but with a 50mA dsg. The RC again drops quite dramatically, which we don't understand. Thanks,.

  • 0 in reply to martin gibbs
    TI__Guru 73830 points
    Martin,
    The chemistry selection tool shows that ChemID 2080 is not the best fit for your cell, but it should work. You have the Cell Termination Voltage set to 2500mV (2000mV stack voltage), but you do not discharge to this level. Do you want the SOC to reach 0% at 20000mV?

    I also noticed that the Load Select = 1 and Load Mode = 3. Load Mode = 3 is not a valid setting. Are these parameters switched? These will affect gauging as well.

    The first cycle of a new pack may performa differently from other cycles, because the gauge updates Qmax and the Ra table to fir the new set of cells.

    Tom
  • 0 in reply to ThomasCosby
    Intellectual 520 points

    Hi Tom,

    Thanks for that information. I have included into the fuel gauge and the situation has improved. However I am still experiencing the Remcap and RMT2 MT falling to zero after just reporting say 10 or 15 minutes to go at the end of discharge.

    The situation gets worse the longer the battery is left (say a few days) after a charge.

    It appears to me that the fuel gauge is not accurately accounting for quiescent leakage from the cells when the battery is enabled in circuit.

    I calculate the battery's quiescent (FETS enabled)discharge to be 2.7mA.

    How can I best set the fuel gauge to account for this please?

  • 0 in reply to martin gibbs
    Intellectual 520 points
    Also I forgot to ask.
    You mentioned that the chem ID of 2080 is not best fit. Would it be possible from the info you have to say which one would be.
    I am using the one provided by E one Moli for the cell we're using. Thankyou.
  • 0 in reply to martin gibbs
    TI__Guru 73830 points
    Martin
    The gauge adjusted the capacities as it was designed to do. The RM and FCC adjustment took place at the first Ra grid point simulation and it improved the accuracy. The accelerated adjustments near the end of discharge were due to fast convergence. These adjustments should be reduced as learning cycles occur and the Ra table and Qmax updates. There is one odd thing that I see. The gauge tried to set SOC = 0% at about 2800mV cell voltage, where the cell voltage was set to 2500mV in the previous gg file. Was the Term V changed in the device?
    Tom
  • 0 in reply to ThomasCosby
    Intellectual 520 points
    Hi Tom,
    Thanks for getting back so quickly.
    You are correct; I have changed the Term V, as advised, as our EOD V is 22000mV. It is now set at 2750mV (8S).

    The collapse of the fuel gauges reported capacity close to the EOD, as I stated earlier, seems to get increasingly worse when the battery is left for more days after a full charge.

    May I infer the following from your response.

    i. That the Chem ID = 2080 for the Molicel (INR18650A) is correct or at least very close enough to work.
    ii. That our 2.7mA quiescent discharge (which is measured across our sense resistor) is accounted for by the impedance track algorithm. Or does the IT algorithm manage to take account of this without seeing it across the sense resistor?

    Many thanks for looking at this Tom, we urgently need to get a measure whether our battery is fit for medical purposes. I'm sure you can appreciate for eg, a ventilator battery reporting 10 mins left but after 1 minute saying now its 0, is going to be a problem for our customers.

    Best regards

    Martin
  • 0 in reply to martin gibbs
    TI__Guru 73830 points

    Martin,

    ChemID 2080 was in the acceptable list, but it was not the best option. I do not have the list and would need to rerun the analysis. Increasing the TermV will reduce FCC and RM, so that may be why we see the adjustments. They should get better after a couple of learning cycles. The CCDeadband is a hidden parameter that sets the limit for the current to count towards passed charge. This is set to about 1mA, so the 2.7mA will count even though the Deadband mat prevent it from being report as a current.

    Tom

  • 0 in reply to ThomasCosby
    Intellectual 520 points
    Thanks for the response Tom, we really appreciate your input on this,
    I'll be trying your suggestion and I think we'll revisit the chem ID supplied by Moli too.

    But on a more general note....
    The battery settings, hardware and sense resistors are configured for a 20A max discharge: Do you think the fuel gauge to be able to accurately account for such low discharges (~3mA) accurately over a few days before its "main" big discharge? Could this be why it is out by 10 to 15 minutes, finishing earlier than forecast?

    Best regards
    Martin
  • 0 in reply to martin gibbs
    TI__Guru 73830 points
    The gauge will accumulate any charge that is above the CCDeadband threshold, so it should accumulate the 3mA current while at rest. If this current is always present, then the gauge may never update Qmax. The current may not allow the cell voltages to stabilize.
  • 0 in reply to ThomasCosby
    Intellectual 520 points
    Is there a hidden setting that would help the situation, such that the voltage would be then seen to be stable?
  • 0 in reply to martin gibbs
    TI__Guru 73830 points
    No, the voltage has to stabilize to support Impedance Track. The gauge cannot set an accurate Qmax estimate, if the voltages are dropping. Will you always have a 2 to 3mA current present when at rest?