Feeding Independent Current-Measurements to Fuel-Gauge Algorithm

[Tyekuhn]

Hello,

In a system that incorporates an nPM1300 (PMIC), we draw large amounts of current from our battery for certain portions of our circuit, and we must bypass the PMIC for this portion of the circuit.  Specifically, we have a high-current DC/DC circuit which has its input directly connected to the battery (bypassing the PMIC, and the VSYS output), however, the rest of the system runs off of the PMIC output (VSYS).  Because of this bypass, we must measure/estimate the battery current draw of this DC/DC circuit independently of the PMIC battery-current measurements, and feed this independent measurement to the fuel-gauge algorithm.

The question is:  Is there any precedent for doing this?  Is it expected that this would succeed or fail, with regards to getting accurate fuel-gauge information?

Thank you,

Ted

[runsiv]

Hi

As this is an unite case we need a bit more information

What kind of battery you are planning to use (chemistry /capacity for ex.)?

Operating temperature range? 

How often such bypassing are made / For how long/ and what's the peak current during such bypass points?

Any estimate of the current consumption/ loading behavior in normal operating mode and device operating time?

Regards

Runar

[Tyekuhn]

Hello,

We are using a LiPo pouch-type battery, 1000mAh, 3.7V nominal...standard LiPo chemistry.

Operating temperature range is ~20C to 40C; this is a wearable application.

This bypassing occurs in the active-mode of our device, so nearly all of the time.  The peak current can get up to about 1200mA (1.2A); and this can last for several minutes as battery voltage gets lower.  Our device is roughly a constant power device, so as battery voltage gets lower, current increases, and vice-versa.  However, nominal current is closer to 800mA; this can last up to 30 minutes.  The normal operational mode of our device has the battery being completely discharged in approximately 60 minutes.  So a 1000mAh battery is nearly discharged in 60 minutes, and then fully charged again before used again.

When our device is not in active-mode, the current draw is far less, <10mA.

Our charge current is set to 500mA right now, if that's of any consequence.

Thanks!

Ted

[runsiv]

Hi

This is not something that we can support out of the box with our solution today, so we can't promise anything. You can still try with our fuel gauge solution and evaluate its performance for your application.

The nRF Fuel Gauge Fuel Gauge algorithm relies on instantaneous battery current and voltage measurements and uses a pre-profiled battery model to estimate the state of charge. Good thing is you don't really need to accurately count every coulombs going in and out of the battery for a stable fuel gauge prediction. 

For the initial evaluation, you could use nPM1300 EK and nPM-PowerUP PC application for the evaluation. I believe the evaluation would be straightforward and fast since the run time of the device application is 1 hour.

Here are some steps. 

• First they need to profile their battery at room temperature using nPM-FG Board for better accuracy. 
  ( There is also a pre-profiled LiPol (LP602760) 1000mAh battery model available for testing)
• Export that model to the host SoC to initialize the fuel gauge.
• In the nPM1300 EK, you have the access to VSYS pins where they could connect the rest of the system which runs off of the PMIC.
• There is also an extra VBAT Pin they could use to connect high-current DC/DC circuit independently out of the PMIC.
• Use the graph tab to evaluate fuel gauge predictions.

Regards

Runar

[Tyekuhn]

Thank you for the response.  It looks like this is unexplored territory!  We will run some tests, and I will report some results here, for anyone else interested.

Thanks!

Ted

[runsiv]

No problem, that's why we are here. Sounds good to me, don't hesitate to ask if there are any questions or issues along the way

Regards

Runar

[Tyekuhn]

1st update on test results.  For this test, I use the nPM1300 EK loaded with a battery model that was created by profiling our exact battery.  At the battery terminal of the EK, I also connected an external electronic load (thus bypassing the PMIC).  With the EK running, and communicating with the nPM-PowerUP PC application, I discharged the battery using the external electronic load (at a rate of about 0.6C).  On the graphs in the application, the battery drop was clearly visible, and the application also reported decreased remaining battery capacity.  The battery was completely discharge, and then I allowed the EK to charge the battery, figuring that might improve the fuel-gauge estimate.  I then discharged the battery a 2nd time, hoping the fuel-gauge estimate would give better results.  However, the capacity estimate was very inaccurate (off by ~70%), i.e., when the battery was actually completely discharged, the fuel-gauge said there was still 70% remaining.

So it seems using only battery voltage, the fuel-gauge isn't very accurate.  Not an unexpected result.  The next test will also feed the fuel gauge algorithm estimated current on the path the bypasses the PMIC.  However, I don't think I'll be able to use the EK and PowerUp application for this, I'll have to use a custom application on an nRF device so that I can input the estimated current into the function calls of the fuel-gauge.

[runsiv]

Thanks for the update. I don't have to much to say myself but I have passed on the information to the PMIC team in case they have some observations 

Regards

Runar

[runsiv]

Update: The PMIC team asked the following: If it was possible to repeat the test with logging enabled? If you could share the logged files so that we can have a look what's going on there. Also some screenshots of the graph window will help.

Regards

Runar