This is a complex issue. If you measure it with the current on, resistive effects in the path come into play. IF you turn the current off, the voltage starts to drop, so you must measure at EXACTLY the same delay every time.
I once built a charger that used a 555 as the timing and voltage detector. The voltage dropped for the entire off time, but since the off time was constant, the thing worked great. I tried to duplicate the operation using the PC parallel port for control. This FAILED miserably because the sample time latency was not constant. I leave the current on.
This is another tradeoff area. First consider how much current you want. If the current is too low, the zero slope may not happen. Too much and temperature effects cloud the issue. I've been using C as the charging current with reasonable success.
I monitor the current every minute with the current left on for the reason above. I terminate after five consecutive readings of the same value Or at the first lower value.
Here are some of the reasons why.
Noise is an issue if you try to measure small changes.
My voltage measurement is just the voltage readback from a programmable supply. It has only 10mV resolution. It really should be better, but since I'm charging at least 4 cells most of the time, it works ok.
For series cells, the zero slope happens at different times for each cell. In aggregate, it smears out. The problem gets much worse for NiCd cause the pack may never go negative. Most of my charging is for NiCd. The 5 minute compromise seems to work for me.
You're gonna find this varies all over the map depending on the age and abuse of the cells. I use 1.6V/cell as an overvoltage termination, but sometimes, even that is not high enough for OLD computer packs. Don't forget to compensate for the resistance in the test setup. My wire and circuit breaker are 0.3 ohm total. Charging one cell at
4A, the error term is almost as large as the cell voltage.Depends on how far apart the readings, see above.
Don't forget to terminate on time. Once you've put 1.2X C into the pack, you might as well quit. Don't forget to terminate on temperature. If you start with a full pack or if something goes wrong, you want to shut it down before it explodes or catches fire.
Cell life will be better if you're conservative.
mike