Relying on a full negative-delta-V reversal assumes a dip large enough to measure - an assumption that fails for warm cells, gentle currents and high-capacity designs whose voltage merely flattens. Robust NiMH chargers therefore add two gentler detectors: zero-delta-V, which terminates when voltage stops rising and holds a plateau for a configured time, and inflection-point detection, which stops at the change in curvature just before the peak, limiting overcharge heating. Underneath all of them sits the unglamorous but indispensable maximum-time safety timer, the one termination that cannot be fooled by a missing signal. This paper specifies all three and defines their place in a defence-in-depth charge controller.
STMicroelectronics' NiMH charger documentation describes a zero-delta-voltage variant for exactly this situation: with NiMH the post-peak drop is very slight, on the order of 5 to 10 millivolts, and may not cross a conventional -delta-V threshold; instead the detector declares full charge when the filtered voltage has not increased by more than a small band over a set dwell time - a plateau rather than a reversal.
Plateau detection is inherently a time-in-band test: the controller asks whether dV/dt has stayed within plus-or-minus a small window for long enough that the curve has genuinely levelled, distinguishing the end-of-charge shelf from the slower mid-charge rise by dwell duration and by arming it only late in charge.
ST's AN417 frames the inflection-point method as especially adapted to NiMH because it does not wait for any voltage fall at all: it detects the change in the voltage curve's curvature (the second derivative) that precedes the peak, terminating before the cell enters the strongly exothermic recombination plateau. Stopping earlier in exchange for a small top-off phase measurably reduces heating and extends lifetime.
The method trades algorithmic complexity for cell health. Estimating a second derivative noisily demands heavier filtering and careful arming, and it can stop marginally before 100 percent, which is why ICs that use it pair it with a brief reduced-current top-off to complete the charge - the same optional phase the bq2002 provides after a fast-charge termination.
Three numbers define zero-delta-V: the voltage band defining 'flat' (tight enough to distinguish the shelf, loose enough to tolerate ripple and ADC quantisation), the dwell time required inside that band (long enough to reject a momentary pause in rise), and the arming point in time or delivered charge (late enough that the mid-charge slope is already small). Together they convert a smooth curve into a deterministic stop.
Because the plateau is shallow, the detector is usually OR-combined with dT/dt and the absolute limits rather than used alone; it is most valuable in warm conditions where no dip appears but the curve still visibly shelves, giving the firmware a voltage-domain stop that does not depend solely on the thermistor.
Every signal-based method presumes a signal exists. A thermistor can detach, a voltage channel can be miscalibrated, and -delta-V can simply never form at low current; without an independent bound the charger would continue indefinitely. The maximum-charge-time timer supplies that bound: started with fast charge, it forces termination or switchover at a conservative multiple of the nominal charge time for the rated capacity and current, regardless of every other input.
Charge ICs list maximum time among the standard fast-charge terminators alongside -delta-V, maximum voltage and maximum temperature precisely because it is the backstop that makes the others safe to rely on. The timer is sized from capacity divided by current divided by an assumed efficiency, with margin for a deeply discharged cell but not enough to permit prolonged overcharge.
At 0.1C overnight charge there is no -delta-V and little thermal signature, so the timer is the primary and appropriate terminator: a 0.1C charge is sized to about 14 to 16 hours to account for efficiency and then stopped. Panasonic's charging guidance treats long timer-based trickle as the least preferred method and bounds any maintenance current tightly - the subject of the next paper - distinguishing a one-time slow-charge timer from indefinite topping.
The first figure sequences plateau and inflection detection against a conventional -delta-V stop, and the second stacks the termination layers from earliest inflection through plateau, -delta-V, dT/dt to the hard timer, illustrating defence in depth: the controller stops on the first trustworthy layer and the timer guarantees that some layer always trips.
Specification should state, for each operating condition, which terminator is expected to act first and which guarantees a stop: inflection or -delta-V in cool fast charge, plateau or dT/dt when warm, the timer in slow charge or fault, and absolute voltage/temperature as immediate fault limits. Verification injects failures - detached thermistor, suppressed dip, stalled cooling - and confirms the timer and absolute limits catch each, measuring the resulting overcharge to bound worst-case heating.
Layering detectors this way is what allows a charger to be both aggressive enough to finish in two hours and conservative enough never to rely on a millivolt signal that may not exist. The remaining steady-state question, after fast charge ends, is how - and how much - to keep a full cell topped up, which the maintenance-charging paper addresses.
Weijiang Power designs and manufactures nickel-metal hydride cells, matched packs and charging-ready configurations for consumer, industrial, medical and mobility customers, and supports partners with charge-protocol guidance, IEC 61951-2 performance files, IEC 62133-1 safety evidence and charger co-validation. Share your cell format, charge rate, thermal envelope and cycle target and our engineers will specify a cell-and-charge combination that protects both runtime and service life. Review the range on the products page.
