This project is inspired by the TritiLED project. The key idea of the project is that every LED has its optimal current which is typically much higher than a few microamps but much lesser than the rated current of given LED. Its author Ted Yapo uses a PIC MCU and some passives to generate current pulses. He thinks using a MCU for blinking a LED can be considered an overkill and I agree. I would like to make my own version of TritiLED without a MCU. I will rely heavily on Ted Yapo's work and when I do something differently I plan to compare my motives and results with his.
Requirements for the project:
- Use common, loosely specced parts
- Simple (and cheap) circuit
- Predictable, repeatable and reasonably stable parameters
- Settable current draw with reasonable minimum 1-2 uA or less
Many circuits I find on the web use some components that are not easy to get. It may be some ancient part (such as unijunction transistor), some IC with strict requirements (i.e. very low supply current op amp), or a discrete part with some non-trivial requirements - very low threshold voltage or specific R_DS(ON) of a MOSFET, exceptionally low or high gain of a BJT. This also limits resistor value - 10M is the highest common value of a resistor. Avoiding parts unavailable in THT is also preferred.
Paying premium for extending a coin cell battery lifetime by 10% is nonsense. Cost and complexity must be balanced with possible gains. This also limits resistor usage - pair of 10M resistors in series may be acceptable to "create" a 20M resistor. Using 100 of 10M resistors to get 10 of 100M resistors is bad.
The parameters (LED current, duty, frequency) should depend mainly on easily measurable parameters - preferably values of passive components. "Nearly constant" voltages as V_BE of a transistor, forward voltage of a diode or a LED may be used too - it is possible to make a guess, measure result and change passives slightly to get the desired result. Characteristics which are highly variable (either part-to-part or temperature dependent) such as transistor gain or diode reverse leakage current should not be used.
The quiescent current should be so low that the circuit will keep some reasonable efficiency even at average current drain of 1-2 uA. Since 1 uA is about 10 mAh per year it means run time more than 10 years for one CR2032 cell should be possible.