Hi,
This question is important partly because: bq27421 is promoted by TI as the smallest and easiest-to-use Impedance-Tracking gauge in the market. Naturally one might think there are simplications be made, and the important questions of whether it can handle GSM pulses comes into attention:
According to 27421 tech ref SLUUAC5 p.37, OCV readings are read when
|AverageCurrent( )| < |Quit Current| for Dsg Relax Time (60s)
This makes sense for constant load current applications.
What if the current is pulse-like? There are applications like motor-driver, GSM pulse and alike in which high-state current is dozens or hundreds of times larger than low-state, and it is possible for |AverageCurrent( )| < |Quit Current| to be true over entire period (say several millisecond, not limited to GSM cases). So if the OCV readings are attempted based on AverageCurrent readings, then what happens if the readings coincide with moments of the high-pulse (and high voltage drop)?
In SLUA450 p.3, an iterative algorithm is outlined
“If the current during the OCV reading is non-zero, then an IR correction is done. The first iteration of DOD is found from the uncorrected OCV reading; then the resistance value is found from the R(DOD) table and used to correct the OCV value as OCV`=OCV-I*R. Then, the corrected DOD is found from OCV`. This method achieves the best accuracy if the current during relaxation mode is below a C/20 rate. This is why it is recommended that the DF.Quit Current not exceed C/20.”
Clearly the designer are aware of the cases of such coincidence, however the recommendation here of keeping current below C/20 simply doesn’t apply to pulsing cases, because current cannot be controlled below C/20.
In SLUA450 one there is extensive discussion on bq2750x family gauges for GMS profile. However, the discussion doesn’t apply to bq27421 because:
1. for bq2750x, if OCV was read when DF.Dsg > Current Threshold, it is ignored. However for bq27421 SLUUAC5 ONLY give the |AverageCurrent( )| < |Quit Current| for Dsg Relax Time (60s) time averaging criteria. So is it possible that the inaccurate OCV reading at coincidence moment is still taken?
2. Aside from OCV reading, does the coulomb sample-integration method used by bq27421 accurate/dense enough for pulsing applications (include GSM)?
3. What exactly is the sampling method bq27421 uses? Is it Delta-Sigma Coulomb counting?
Hui