As you well know batteries are chemical devices and don't recover instantaneously due to the rates of diffusion to recreate the charge separation needed for normal cell voltage after a sizable (for the type of battery) load. It takes under 100 microseconds to switch modes and do a full ADC measurement on a slow controller and much less than that for the switch and sample-hold function.

http://batteryuniversity.com/learn/article/testing_lithium_based_batteries

I do know that but, because I know that very well, I also know that there are actually three voltages in this experiment.

1. Loaded I*R voltage drop
2. Immediately after disconnecting "unrecovered" battery "unloaded" IR drop (your measurement)
3. Recovered battery unloaded

The alignment of the voltage drop for 1 and 2 depend 9n the load, chemistry of the battery and age (discharge) of the battery. They may align well or they may be off by a considerable amount.

 

It is moot. I redesigned the board so I have a pin that turns on the load as well as a pin to read the voltage.

Cool. You can still spend a little time doing battery voltage recover analysis to get a good idea of SOC.

 

Is it safe to assume I should trip my low battery alarm somewhere just above 2V which is the minimal operating voltage of the pic?

 

I do know that but, because I know that very well, I also know that there are actually three voltages in this experiment.

1. Loaded I*R voltage drop
2. Immediately after disconnecting "unrecovered" battery "unloaded" IR drop (your measurement)
3. Recovered battery unloaded

The alignment of the voltage drop for 1 and 2 depend 9n the load, chemistry of the battery and age (discharge) of the battery. They may align well or they may be off by a considerable amount.

So it is a question of recovery after all. Tks...

 

So it is a question of recovery after all. Tks...

I hope the sand is dry in Oregon, wet sand must be annoying when it sticks in your beard.

Anyhow, recovery is correct but, because an error of 2 volts or more is possible on a 9V lithium battery between measurement 1 and 2 makes that option useless unless some tests are done to prove otherwise.

 

Is it safe to assume I should trip my low battery alarm somewhere just above 2V which is the minimal operating voltage of the pic?

Don't forget the diode voltage drop. Use schottky diodes to reduce that.

 

Don't forget the diode voltage drop. Use schottky diodes to reduce that.

Thanks I know that. And it didn't answer my question. What would be the safest operation voltage for those little button batteries?

 

Thanks I know that. And it didn't answer my question. What would be the safest operation voltage for those little button batteries?

The answer to your question depends on the exact battery used, the current draw and voltage drops between the battery and load. Most have datasheets with Continuous Runtimes to 2.0 Volts but that might be below that you need to run the circuit.
CR2032 data.
http://data.energizer.com/pdfs/lithiumcoin_appman.pdf

 

Sure you won't see the voltage under actual load but the battery voltage won't recover instantly when the load is off due to the time it takes for the chemistry to work. That should be plenty of time to make a relative voltage measurement of good and bad cell using the same pin as load with serial resistor and measurement ADC.

I think that's right but it would be good to test. There probably would be some recovery before the sample so figuring out a fudge-factor through experimentation would be helpful. The other one-pin approach is to build a sample and hold circuit but that seems like overkill.

But, I think I saw there are other pins available - a 2 pin solution is the "Occam's Razor" here.

 

I think that's right but it would be good to test. There probably would be some recovery before the sample so figuring out a fudge-factor through experimentation would be helpful. The other one-pin approach is to build a sample and hold circuit but that seems like overkill.

But, I think I saw there are other pins available - a 2 pin solution is the "Occam's Razor" here.

It was a now moot possible solution to the one pin problem that would need an offset constant experimentally found for the missing voltage under load measurement data point. If the variation was small (like a voltage delta under 10%) then the recovery to open circuit voltage measurement would align well with the missing data point of voltage under load.

Some of my battery testing methods use the two dual pulse method with a voltage slope calculation as part of a Quick and Dirty SOC calculation.
http://data.energizer.com/pdfs/batteryir.pdf

 

Can't only have one pin left. Unless I can figure out how to do everything with the same pin.

Some people shift a bit stream into a shift register when they run low on pins.

Don't forget you can free up other pins to leave at least one analogue and one digital to sample the battery.