BQ27421-G1: BQ27421-G1:

Are you able to control the load current (e.g. turn off high power parts of your system when you configure the gauge)? If so, you can minimize the initial error. What's the minimum load (C rate) that can't be turned off when configuring the gauge? A C/2 or C/5 load would be too high. If you go down to C/50 or lower then the error will become negligible (depending on state of charge).

>When the host is running and therefore able to configure the fuel gauge

The gauge should only be configured when the gauge was reset (ITPOR = 1), otherwise it should not be configured. Please make sure that you don't configure the gauge unless ITPOR = 1.

ITPOR gets cleared after configuration and SOFT_RESET. So this is a great indicator that the gauge was reset and needs to be configured again.

>In our application, the battery is always connected. 

This is perfect. That means you only have to configure the gauge once or whenever the cell voltage drops below 2.5V (which should never happen anyways).

>The host could stop charging long enough to configure the fuel gauge, and then re-enable the battery charger. Is that a workable solution?  If not, what is the proper sequence that should be followed?

The goal is to get a true (or as close to true) OCV measurement. This method seems complicated - also the cell won't have a true OCV after a fairly long time (as the voltage takes some time to converge to its final open circuit value) so this is probably not feasible. If this is not well controlled, then the error may not be any better than if you just take a C/5 loaded voltage measurement after SOFT_RESET and let the gauge estimate the OCV (it tries to subtract the voltage drop with an estimated cell resistance but this is not a fully defined equation so there's always an error that is more or less dependent on current and cell state (depth of discharge)).

>Does "relaxed" mean no charging or discharging current?  

Relaxed means that the current is low enough for the cell capacity to not have a significant impact on voltage measurement vs. true open circuit *and* that the voltage has stabilized (it will spring back after a significant load so it takes a while (like 30 minutes or so) until it is stable - that's when it is relaxed enough for a high accuracy estimate of depth of discharge, which is the reference for all gauging).

>How long must the battery be relaxed before issuing the soft_reset?

Ideally, the voltage should not change by more than 1uV/s. But that is ideal and often only possible when the system is turned off (e.g. it's not in use over night - the gauge tracks this automatically so it will eventually self-correct, if the initial OCV measurement was not great).

Because your battery is always connected, the problem with a SOFT_RESET after the gauge first (! --- it will only have to be configured once) powers up, will self-correct over time (it will already diminish after the first charge as the gauge re-adjusts when it detects a valid charge termination).

--> The most important rule is, please do not configure the gauge unless ITPOR = 1. That will ensure that the gauge can fully adjust to the state of the battery. Every time you configure the gauge and re-start the gauging algorithm, the gauge will start off with more or less error due to the OCV measurement problem.