BQ34Z100-G1: Sudden Jump in Remaining Capacity

Hi Matt,
Did you set your LFP relax flag? Do you have smoothing enabled? What is the status of the relax jump ok and relax smooth ok flags?

You will need to have gain access to one of your packs and try to get log data showing this behavior and a gg file for analysis.

thanks
Onyx

Hi Onyx,

The LFP relax flag is set.  Smoothing is enabled.  Both RELAX_JUMP_OK and RELAX_SMOOTH_OK are enabled.

I will potentially have an easier time getting gg files before and after, but getting logged data of the jump happening has been very difficult due to the inconsistency of whether or not it occurs and the exact timing.

For most packs, Qmax did not seem to be updating during the two complete cycles.  Could this be an indication of a bad golden image learning?  Could that lead to or at the very least play a role in these types of jumps?

I've created a new golden image with a heavily learned fuel gauge and have begun to bring through 5 packs with this new golden image to see if the issue persists.

Thanks,

Matt

Hi Matt,
It is possible temperature changes occur and rem cap simulation occured. Try successfully reproducing the issue and then disable both the relax jump ok and relax smooth_ok... This will hold the SOC at 30%

thanks
Onyx

Hi Onyx,

I successfully logged the jump occurring.  It appears that the Remaining Capacity jumps when OCV measurement takes place and again with each subsequent update to DOD0.  Not sure why it is taking 48hrs plus to do an OCV measurement once in RELAX mode.  I have confirmed on separate batteries that disabling both the RELAX_JUMP_OK and RELAX_SMOOTH_OK prevents Remaining Capacity from jumping, although True Remaining Capacity still jumps in the same manner.

I've attached the data for your review. Would this behavior indicate that our stored DOD(OCV,T) table is way off for our battery chemistry?

Thanks,

MattFG0060 No PIC 30 Pct Relax 3rd Night.log

Hi Matt,
These jumps are inevitable because the gauge performs rem cap simulations in relax mode. Because you are using LFP cells, it takes a while for the cells to relax especially after charge hence the gauge waiting for a long time (48hours) before the OCV is taken. The LFP relax flag controlls this.

so because the cells voltages relaxes higher, when the gauge performs simulations at the different votlages you will get different rem cap results.

thanks
Onyx

Hi Onyx,

I understand the LFP cells can take a little longer to relax, but not to this degree and the voltage and temperature were not still moving.  It seems like there is something else going on.  I picked out some data points that help demonstrate this.

Time (minutes)	RemCap	FullChgCap	Voltage	Temperature	TrueRC	TrueFCC	DOD0Time
20 Hrs Earlier	2090	7290	13208	24.8	2307	7507	442
0	2104	7290	13208	24.3	2322	7507	765
100	2105	7290	13204	24.9	2324	7507	792
101	2820	7343	13204	24.9	2820	7343	0
161	3185	7341	13204	24.6	3185	7341	0
221	3411	7334	13204	24.5	3411	7334	0
281	3577	7332	13204	24.3	3577	7332	0
341	3692	7328	13204	24.2	3692	7328	0
642	4157	7313	13204	23.4	4157	7313	0
942	4180	7303	13204	22.8	4638	7297	0
1544	4652	7302	13200	23	4652	7302	0
2145	4665	7306	13200	23.1	4665	7306	0
2746	4669	7300	13196	22.9	4669	7300	0
2 weeks later	4696	7215	13184	22.1	4696	7215	0

You can see that when I collected data the night before the voltage and temperature were virtually identical and remains essentially the same throughout the collection. 

And here is a graph of the Remaining Capacity over Time showing the nature of the pattern of the jumps:

It seems like the gauge has no idea what the remaining capacity should be at a fully relaxed voltage and temperature.  Would this indicate bad learning or an incorrect Data Flash constant somewhere?

Is there an up to date Application Report on how the gauge handles the Fuel Gauging Algorithm and any documentation on how this is adjusted when configured for LFP cells?  A year or two ago it seemed like the best document that described how the Impedance Track algorithm worked was SLUA450 from 2008.

Thanks,

Matt

hi Matt,
The flatness of the voltage profile of LFP cells make it difficult for for rem cap simulation during relax. can you confirm if you have the DODWT flag set or clear? That document you pointed out is the IT detailed description we have.

thanks
Onyx

Hi Onyx,

I can confirm the DoDWT flag is set.

I guess my major hangup is I understand that the voltage profile of LFP cells is very flat which would make it difficult to run remaining capacity simulations in relax based purely on Voltage and Temperature and stored resistance data. 

What does not seem right is that the gauge consistently thinks that a 27-28% SOC battery is actually 55-66% SOC.  If our raw pack is charged to that level (which is consistently 13.20V +/-0.01V) and then paired with the gauge it will think the pack is 65% SOC.  It would seem that the gauge should be able to make a good guess, albeit with a large error due to the flatness of the voltage curve, but its guess/simulation just seems consistently and significantly wrong in a specific direction.

Maybe none of this really matters if the LFPRelax, DoDWT,  and SMOOTH flags are set and the RELAX_JUMP_OK and RELAX_SMOOTH_OK flags are cleared since the RMC and FCC will just base and update themselves on passed charge, but the degree/magnitude to which these other values are off makes me concerned there is something else that is not set or learned correctly.

Thanks,

Matt

Hi Matt,
Did you check what the flat region of your chem id is and if those voltages are within the flat region?

thank
Onyx

Hi Onyx,

I can confirm that when fully relaxed the battery voltages in this region are extremely flat.  For our 4s pack there is a 0.03V difference between 30% charge and 60% charge.

Could you please confirm that for an LFP pack that we want the LFPRelax, DoDWT,  and SMOOTH flags set and the RELAX_JUMP_OK and RELAX_SMOOTH_OK flags cleared?

Are there any other relevant flags we should be aware of for our LiFePO4 application.  Would it be helpful to attach a gg file of our current configuration?

Thanks,

Matt

Our chem ids have a flat region which is listed in the excel sheet when you download the chem updater for bqstudio or bqEVSW. what is the flat region for your chem id?
thanks
Onyx

Hi Onyx,

I believe this is what you are referring to:

Description	Chemical ID	Vqdis_min	Vqdis_max	SOC_min, %	SOC_max, %	Rhf,mOhm	Rhf_lo	Rhf_hi	Qmax	Ra0_charge
LiFePO4/carbon	457	3277	3330	34	96	8	-31	31	80523	50

So for our 4s pack, that would mean that OCV measurements should not be occurring between 13.108V and 13.320V?  The jump I collected data had occurred when the OCVTAKEN flag became set (indicating OCV measurements occurring) while the pack was at 13.204V at 24.9 degrees C. 

Please let me know if any other information would be helpful.

Thanks,

Matt

Hi Onyx,

Just wanted to make sure you saw my last reply.  Looking forward to your response.

Thanks,

Matt