Chemistry selection tool problem

Thank you, Tom.

I will record a new log, this time with the rest period and will see if that works.

Hello again Tom

I recorded a new log, this time with the relaxation period. I got 400 as the most suitable chemid but the program shows an error of 25%. What might be the problem?

I will attach the log. I didn't record the charging cycle, so I artificially inserted some current values in the log so that the program would detect the end of charging and end of relaxation. I also used a single cell this time instead of six to make the measurement more accurate.

Kaarel,

I checked the data with a special program and the shape of the discharge curve looks good as compared to the OCV curve for ChemID 400, but there is a large gap between them. Is the 2A discharge rate between a C/5 and C/10 rate for your cell? If it is higher than C/5, then that may help to explain the gap. We can characterize the cell and create a new ChemID, if you would like to submit cells. Please provide an email address and I can send you the information on submitting cells for characterization.

Tom

Hello Tom

Thank you for the quick reply.

The cells are 11Ah so 2A should be suitable. But when setting up the test I did notice around 100mV voltage drop on the power cables coming from the cell - the voltage measurement wasn't done on the tabs of the cell but on the power cables. Can this measurement error be the cause of this gap?

I added my e-mail address to my profile. Can you see it?

Does the characterization explicitly requires submitting the cells or could I just do the cycling of the cells myself and send you the data? Our application is meant to work with many different batteries and it might happen that some of those also don't have a correct chemid in your database, if I would have to send you the cells every time it might take a long time and cost much.

Kaarel

Kaarel,

The voltage drop will contribute to the gap. The cell voltage sense contacts should be kelvin connections to the cells. An accurate ChemID is critical for good performance with IT gauges. You may need to consider a CEDV gauge, if it is not going to be practical to verify the ChemID for each cell type. You can check for a match using the Mathcad program. If you are going to change frequently, then you should consider purchasing a Mathcad license. A full characterization needs to be done in our lab, because we have a special characterization process.

Tom

Hi Tom

Verifying the chemID for each cell type is not a problem. I was just thinking about the cases when there isn't a suitable ID in your database. If there are many cases like these then it would take up a lot of resources to send the cells each time. I don't have any specific reason to think that there would be a lot of such cases, I was just playing with the thought. And as I understand you constantly update the database with new ChemIDs?

The current cell under test is a no-name chinese battery, and the MathCad program could still find a match for it. I will try the test again with the kelvin connection and see if the error percent is better this time.

Our application is dealing with light electric vehicles and we are using this fuel gauge. Is the fuel gauge suitable for use in electric vehicles? I can see from the documentation that there are two main profiles for SOC estimation - constant current and constant power. Electric vehicle batteries aren't discharged according to neither of these. Will the gauge get good accuracy under these conditions?

Also can you please give some examples of TI's CEDV gauges?

Hello Tom

I did a new measurement cycle, this time the error was about 7mV as opposed to the previous 100mV. Couldn't get it any lower, because the wire coming from the battery's negative terminal also carries all the current which goes through the sense resistor as in the reference design. The tutorial for finding the ChemID (chemselect_cont) says the error should be under 1mV - should the measurement then be done with a separate data logger? It seems that such a low error value is hard to achieve when using the fuel gauge itself for doing the logging?

This time MathCad gave me bestchem = 414, with a max error of ~12.7% and ChemID 400 was at the 4th place.

Is the new chemID the correct one as compared to the old one which was 400? Can you check with your special program. I have attached the log I used. Link for the log. Also is 12.7% accurate enough or should I send the cells to you for characterization?

Any feedback on the suitability of the fuel gauge for light electric vehicle usage?

Best regards

Kaarel

Kaarel,

I checked the ChemID candidates with my program and I would recommend ChemID 435 over the others., becuase its OCV table has a better fit to your discharge data. We recommend that the %error be less than 10% for LiFePO4 cells, so you may want to have a cell characterized if you encounter gauge issues. The bq34z100 can be used with light electric vehicles and we will release a new firmware update early next year that will improve performance. We also have a new family of device that support high cell count applications and they use the CEDV type gauging algorithm. Reference bq78350 on the TI website.

Tom