Perspective 1: State-based-empty. Empty is defined as the state/moment when no more capacity can be taken from the battery, even at light-load. A looser definition is to allow for <1% consumable at light-load.
Perspective 2: Voltage-based-empty. Empty (0%) is defined as the moment when empty voltage is first crossed
These 2 definitions are incompatible. More specifically, the battery has a state, and loaded voltage can never correlate to a state, so the “Voltage-empty” perspective is very different from the battery state, and suffers “moving empty” troubles if the loading changes.
However, almost all applications are stuck with “Perspective 2: Voltage-based-empty”, because the bottom-line is that other chips on the system will fail below some specific voltage. Therefore the system’s UVLO defines the battery’s empty-voltage, and the application requires “voltage-empty” perspective. Therefore, that’s what we always support in our modelling.
The resulting troubles of supporting voltage-empty all come from the moving empty problem. Here is a summary of the associate troubles:
- Lower deliverable capacity for heavier loads (or, more undeliverable capacity below Vempty).
- Lower deliverable capacity for colder temperatures (or more undeliverable capacity below Vempty).
- Lower deliverable capacity for higher resistance cells (“”)
- Much lower deliverable capacity for higher empty voltages
- Result is that 0% must be declared at a variable-actual-battery-state. Example:
- At heavy-cold-3.4V 0% should be declared although battery is actually 20% state. The remaining 20% is still available >3.4V if the loading is reduced.
- At light-load-room-3.4V, 0% should be declared while the battery is actually 1% state.
- Therefore empty moved 20%. In practice we see empty move between 3% and >80%, depending on the application.
In extreme cases, when moving-empty is a performance challenge, customers need to consider the following:
- Battery becomes like “not a battery”. If the challenging heavy-cold case results in huge capacity lost below Vempty, then in addition to gauging eventually failing, the battery is also beginning to behave like “not a battery”. When the battery stops behaving like an energy source, the problems are bigger than “gauging accuracy problems”, and the customer really should evaluate whether the condition is realistic, or if Vempty is realistic (can they operate below?), etc.
- Can Vempty be reduced? It’s always wise to relate to the real UVLO constraint of the system. If the system can operate to 3.1V, or 3.2V, then it’s wiser to target the gauge accordingly.
- Is loading realistic? Engineers are often very careless to consider focusing the gauge to 1C or C/2. Many modern electronics cannot be profitable unless they deliver run-times longer than 10 hours. Only the realistic “heavy-persistant-load” should be considered. Any load pulses briefer than a few seconds should not be the target focus for the gauge.
If there remains some highly challenging use case, then the application really should consider ModelGauge m5 (MAX17055 family) instead of basic ModelGauge (MAX17048 family). ModelGauge m5 directly addresses the moving empty and performs much better at much more aggressive loading conditions and empty targets.