Calibrating Reduced Order Particle Observers for Accurate State Estimation
Calibrating reduced order particle observers optimizes usable cell capacity and fast charging rates while preventing lithium plating through precise surface state tracking.
Fixed Point Math
Meaning
Numeric representation formats substitute floating-point arithmetic with scaled integer operations to execute deterministic calculations on low-cost microcontrollers. Embedded battery management system firmware uses fixed point math to perform state estimation and filtering without requiring hardware floating-point units. The scope covers scaled integer representations where fractional precision is determined by a fixed radix point location.
It stops where dynamic exponent scaling or native floating-point hardware processing is used.
Radix Scaling
Integer bit allocation determines both the dynamic range and fractional resolution of represented physical variables. Applying fixed point math requires selecting fractional bit lengths, such as Q15 or Q31 formats, to balance precision against register width constraints. Scaling factors convert physical battery values, including millivolts and microamps, into standard integer words for fast bitwise shift operations.
Overflow Handling
Saturation logic prevents wrap-around errors during accumulator overflow events in state of charge integration loops. Truncation and rounding choices introduce quantization noise that accumulates over extended operating cycles.
Execution Velocity
Deterministic cycle counts lower interrupt latency in safety-critical cell monitoring microcontrollers lacking floating-point hardware. Executing integer math routines reduces clock cycle requirements per control loop execution. Reduced execution time lowers microcontroller power consumption while freeing processing budget for safety diagnostic functions.