The current battery warning/shut-down constants for the GoPiGo are inappropriate for the new Li-Ion batteries

[jharris1993]

Background:
Modular Robotics/Dexter Industries decided to depreciate the older 8-cell NiMH batteries in favor of a monolithic Li-Ion battery module and dedicated Li-Ion charger module.

• NiMH battery performance varies over wide limits based on the manufacturer and quality of the cells.  This made it difficult to use the robot for any period of time and/or predict the useful runtime for the robot.
• The exposed contacts on the 8-cell battery holder created the risk of short-circuits causing potential damage to the battery holder and/or the robot.

As a consequence of all this, (along with possibly other issues), the decision was made to depreciate the 8-cell battery package in favor of a better quality, and safer, Li-Ion battery pack and dedicated charger.

Issue:
The originally designed firmware for the GoPiGo robots, including the GoPiGo3, has built-in battery voltage points that trigger:

• The power LED to turn yellow warning the user that the battery power is getting low
• The power LED to turn red to indicate that the battery power is critically low, additionally disabling the motors.
• The GoPiGo board commands a formal operating system shutdown if the battery voltage drops below a certain critical threshold.

Unfortunately, these battery voltage points were calibrated based on the discharge curves for a "typical" collection of eight NiMH batteries.  Since the Li-Ion battery pack automatically cuts power at a voltage considerably higher that even the NiMH warning threshold, the GoPiGo3 does not have an opportunity to trigger a safety-shutdown.  In fact, it does not even reach the "low battery - yellow power LED" warning stage before automatically cutting power.

It is an established fact that un-commanded and spontaneous shutdowns of a system like Raspbian/Linux can seriously corrupt the media and it is firmly established, (and documented on the MR/DI GoPiGo site), that a controlled shutdown is necessary to avoid corruption of the media.

Request/Proposed Solution:
The calibration constants for the battery thresholds should be changed to accommodate the new discharge curve of the new Li-Ion batteries.

• This could be hard-coded into a firmware update
  -- and/or --
• Both sets of constants could be supported in firmware with a corresponding setting in the calibration routine to select which set to use - defaulting to the Li-Ion set.

[CleoQc]

unfortunately it's impossible to find a voltage level for the new battery that will be reliable, from one battery to the next.
So won't fix, because can't fix unless we go to a different battery chemistry.

[jharris1993]

unfortunately it's impossible to find a voltage level for the new battery that will be reliable, from one battery to the next.

I disagree as this "issue" is true for any battery chemistry. Alan has researched this, (and I have independently confirmed his results), that suggest voltage levels that are appropriate. I was able to use this data to modify the circuitry in the TalentCell batteries to make the "fuel gauge" more accurate.

Of course, these set-points are always only approximations regardless of chemistry.