NiCd batteries should be fully discharged before charging in order for them to deliver the expected performance. This experimental circuit automatically discharges a battery to an end-point of 2.0V (ie. 1.0V per cell) then shuts down. This is only a "back of an envelope" idea - I have not built it.
When running, the battery discharges through LP1 and R4. R4 should be a 10W wire-wound part so that it is suitably under-run. The total discharge current will be about 1.5A which is what is recommended for a size F cell. IC1 monitors the supply voltage as follows. When the unit is in operation pin 2 is high, and since it is a pull-up output (as the little symbol shows) it pulls pin 8 up to the supply voltage. R2 and R1 (which should be 1% parts) drive the threshold input at pin 3. When this drops below 1.15V (the internal ref.) then pin 4 goes high. Since this is a pull-down (open-collector) output this turns the transistor off. At the same time pin 2 goes low (open-circuit) thus adding R3 to the potential divider chain. This adds hysteresis and means the circuit will not switch on again until the supply rise to about 3.9V, well out of range. For this reason SW1 has to be incorporated to get the circuit going.
Q1 is any old pnp darlington transistor. It is driven to saturation by IC1. LP1 should be a torch bulb, around 300mA, and R4 is 1.8ê. If you use a high current halogen bulb, or an led then you need to adjust R4 accordingly to maintain the total current at around 1.5A. Its a good idea (but I forgot when I drew the diagram) to decouple pins 3 and 8 with small capacitors, say 100nF polyester.
IC1 is a MAXIM part - the MAX8212. Note that the 8211, which is similar, cannot be used instead. Also, other manufacturers' parts, eg. ICL, may not work. The 1.15V ref. is not quite as accurate as required, but should be sufficient if 1% resistors are used, and the circuit operates at 25°C. Before constructing this you ought to check the values of R1, R2; the end-point tolerance; and the performance of IC1 against a datasheet.
Contact David Gibson for further information.