BQ40Z50: Cell balancing when the battery is rarely in RELAX state

Hello Vasily,

I have assigned this to an engineer more familiar with the BQ40Z50 to answer your question.

Sincerely,

Wyatt Keller

Balancing requires that the gauge knows how unbalanced the individual cells are. The way the gauge determines this is by taking voltage readings for each cell when they are relaxed. Without this information, the gauge cannot calculate the cell balancing times. If an application never allows the cells to relax then cell balancing (and also gauging in general) will not work well (or at all).

Thank you for the clarification.

Is there any way to manually enable/disable the balancing FETs inside the gas gauge commanding from an external host?

I also wonder if it's possible to configure the relaxation period externally and reduce it from 5 hours to a lower value?

Hello Vasily,

I don't believe that the BQ40Z50 offers manual control over the cell-balancing FETs.

Yes, you could configure the Relax Balance Interval to a value lower than 5 hours in the Cell Balancing Config. However, I am not sure if it is advisable to lower it below 5 hours.

You can read about the different configurations that you can modify in Cell Balancing Config in the Technical Reference Manual(Section 13.5.12). 

Best Regards,

Luis Hernandez Salomon

Hello Luis,

I appreciate the update, although I'm not sure we're talking about the same relaxation period intervals.

I believe the Relax Balance Interval you refer to is the interval between recalculations for required balancing time specifically for balancing at rest (if it is even enabled). The ability to recalculate the balancing time still depends on the availability of the OCV measurement, which is taken also after 5 hours upon entering the RELAX state (Figure 7-2 on p.60 of the doc you linked), or when dV/dt drops below 4 μV/s. What I was interested in is whether the latter 5-hour OCV calculation delay can be manually configured, because both balancing modes (at rest and during charging) depend on it.

What I realized is that as long as the rate of change of voltage drops below 4 μV/s, an OCV measurement gets acquired (Figure 7-2 on p.60) -> delta Q calculation can be performed -> cell balance timers get updated -> balancing can start. Two questions: can you confirm balance times can be recalculated in less than 5 hours as long as dV/dt is below 4 μV/s ? Can the 4 μV/s threshold for measuring OCV be configured?

The second and most important question refers to this paragraph on p.62:

"The cell balancing time for each cell to be balanced is calculated by: dQCelln × Bal Time/mAh Cell 1 for Cell 1 or and dQCelln × Bal Time/mAh Cell 2–4 for Cell 2–4. The cell balancing time is stored in the 16- bit RAM register CellnBalanceTimer, providing a maximum calculated time of 65535 s (or 18.2 hrs). This update only occurs if a valid QMax update has been made; otherwise, they are all set to 0."

From what I see, being able to manually write to CellnBalanceTimer would solve my problem even faster, allowing almost direct control over balancing FETs. The host can make dumb but good enough decisions on which cells to balance (at least during charging) based only on cell voltage measurements, bypassing the necessity to wait for a good OCV measurement. My problem is, I can't find the mentioned registers in the RAM map. They are only mentioned once in the TRM. Question: can you help me locate these cell balance timer registers and confirm that they are writable?

I sincerely appreciate your feedback,

Vasily

Hello Vasily,

1.Can you confirm balance times can be recalculated in less than 5 hours as long as dV/dt is below 4 μV/s ? Can the 4 μV/s threshold for measuring OCV be configured?

Yes, if it's below 4uV/s it should take a measurement before the 5 hours. I reached out to others in the team and there's no way to configure the 4 μV/s nor the 5-hour time delay for the OCV. So these parameters will always be the same.

2.Can you help me locate these cell balance timer registers and confirm that they are writable?

It is not possible to directly set the cell timers as these are done automatically.

Could you please explain using a block-diagram what your system is trying to do? We may be able to assist you better if we have a better idea on your application.

Best Regards, 

Luis Hernandez Salomon

Luis, thank you for the clarification. Below is the overview of our application.

We want to reliably and accurately track a SoC of a battery used to power a BLDC motor. We may cycle the battery continuously (no rest) for 48 hours at a time, and cycling for 16 straight hours is common. When cycling continuously, we will start discharging as soon as charge terminates, or as early as at 80% of SoC. We discharge the battery at 5C to 10C, and charge at 1C to 2C. We would like to be able to use passive balancing during these charge cycles to avoid having the cell voltages diverge.

I would genuinely appreciate if you could help me figure a few more things out.
1. Given our application and the previous posts in this thread, it seems to me passive balancing can't be done with BQ40Z50 alone because of rare relaxation periods. Do you agree?
2. If we add external passive balancing, is it going to interfere with SoC estimation accuracy of BQ40Z50 in any way?
3. Are there any other TI solutions (for gauging, balancing, etc) that you think would better fit our application?

Kind regards,

Vasily

Hello Vasily,

I've been discussing with people in the team for a solution.

I will be giving an answer before the day ends.

Best Regards,

Luis Hernandez Salomon

Hello Vasily,

There is a configuration that we had overlooked that might be able to help you use the BQ40Z50 in your application. This would enable you to cell-balance and accurately keep track of SoC without the need to wait long times to relax.

Cell-balancing can occur during charging under normal operation, however, in your application this usually will not work as relaxation time is usually required to re-calculate the balancing times coefficients needed for cell balancing. However, the relaxation time that is usually required for this can be by-passed by enabling Fast Qmax Updates (See Section 6.4.3 of Technical Reference Manual) which would remove the need to wait for OCV readings. This setting is actually in place in case of infrequent relaxation such as your application.

When you have this enabled, as long as you charge your battery and trigger a Valid Charge Termination(when the battery is fully charged) and you discharge to less than 10% SoC, you will get the Qmax updates needed to re-calculate the cell-balancing timers without the need of the longer relaxation times. 

Regarding your external cell-balancing question, SoC accuracy should not be affected by external balancing. You can read more about external cell-balancing in this Application Report.

Best Regards,

Luis Hernandez Salomon

Hello Luis,

Thank you very much for pointing out this configuration, it might be just what we need. 

The last bit I'm missing is intuition for whether BQ40Z50 is the best component in TI's gauge lineup for our application. Navigating the balancing issue, I discovered the BQ4050 gauge, which uses CEDV vs Impedance Track for estimation. Page 3 of a reference BMS design by TI suggests BQ4050 is very well suited for high current applications:

"Many gauging algorithms do not work well when the motor drive current goes more than 3 to 5 C more than the rated 1 C rate of the battery. The bq4050 use CEDV for the gauging algorithm and works excellent in designs well above the 1C rate up to 25-50C rates."

Based on this, I have two questions.

1. Which gauge would you suggest for the application I described above (discharge at 5-10C)?

2. Does the accuracy of SoC/capacity estimation of BQ40Z50 at all suffer at such high discharge currents? We may or may not configure Load Select to option 7 (Max Avg P Last Run). This question is critical for me.

Luis,

Quick follow-up question on a point you made:

"However, the relaxation time that is usually required for this can be by-passed by enabling Fast Qmax Updates (See Section 6.4.3 of Technical Reference Manual) which would remove the need to wait for OCV readings. "

However, the very same section 6.4.3 suggests that a single OCV reading is still required:

"Instead of taking two OCV readings for QMax update, a Fast Qmax update requires only one OCV reading ... The differences in requirements allow the Fast Qmax feature to have a QMax update at the end of discharge (given one OCV reading is already available and discharge below 10% RSOC) without a longer relax time after a discharge event."

What am I missing?