Other Parts Discussed in Thread: EV2400, BQEVSW
We build low quantities (3-5 at a time) of power supply modules that have a small (1100 mAh) 4 cell LIPO for emergency battery backup. Our latest build is operating in a way that is different from previous builds. The LCD display flashes periodically, and somewhat erratically. After much sleuthing I discovered that it was behaving as though every time it came out of sleep to measure the current, at (SLEEP CURRENT TIME) intervals, the LCD would flash (not blink) for (CHG RELAX TIME). Initial values were 45 for (SLEEP CURRENT TIME) and 60 for (CHG RELAX TIME) which seems to make the actual behavior somewhat erratic. I played with both (SLEEP VOLTAGE TIME) and (SLEEP CURRENT TIME) and (SLEEP CURRENT TIME) was the setting that altered the behavior, which makes sense
However when I change CHG RELAX TIME to 6 and SLEEP CURRENT TIME to 15 the behavior becomes quite regular in time and easily observable and about what one would expect. Changing SLEEP CURRENT TIME to 150 changed the behavior as one would expect.
Next I looked at why the current builds manifest this behavior. There is a difference between current and earlier boards. The layout had to change, many parts moved and a calibration register, (BOARD OFFSET) had to change from -34.7 to -104.8.
The tech manual explains that the chip enters charge mode if the current > (CHARGE CURRENT THRESHOLD), which is 100 in all the builds. What I did was by successive approximation determine the threshold that this behavior manifests itself at (on 1 unit that flashes) that threshold was 151 . That is if I set (CHARGE CURRENT THRESHOLD) to 152 the chip does not enter CHG mode upon waking from (SLEEP CURRENT TIME), at 151 it does. (sample of 1 board)
Important points. This happens on a system where the chip has been calibrated, and the battery has been idle for hours. It must come out of sleep. With the timers set to the longer values (CHG RELAX TIME) = 60 and (SLEEP CURRENT TIME) = 45 I could connect the EV2400 (running the bq20z655-R1 v0.03) software (while the LCD was flashing), and read the status (SBS screen - Refresh) while flashing the DSG bit would be GREEN (two places) and go to RED once the flashing stopped. I verified this on the PRO screen by reading SMB WORD 54. (8447 or 8445 when not flashing, 8405 when flashing) The EV2400 seems to stop the chip from going to sleep, so normally to observe this behavior I had to leave it disconnected. Also with the board calibrated and the (BOARD OFFSET) as above the EV2400 typically reports 0,0 for current and average current.
SO... it seems as though somehow upon waking from (SLEEP CURRENT TIME) our latest builds (at least) make some sort of transient, glitchy current reading that can trip the chip into CHG mode for (CHG RELAX TIME). I don't rule out that earlier builds with the lesser board offset operate similarly, with a lower then 100 (but still rather high) (CHARGE CURRENT THRSHOLD). I am not the original designer, having inherited it, so I don't know what early issues might have been been found.
Question is, has any one seen something like this? Any fixes, or suggestions? Could it be related to the what seems (to me) the rather large (BOARD OFFSET) value that we have to use?
