Optimising quantisation noise in energy measurement

We give a model of parallel distributed genetic improvement. With modern low cost power monitors; high speed Ethernet LAN latency and network jitter have little effect. The model calculates a minimum usable mutation effect based on the analogue to digital converter (ADC)’s resolution and shows the optimal test duration is inversely proportional to smallest impact we wish to detect. Using the example of a 1 kHz 12 bit 0.4095 Amp ADC optimising software energy consumption we find: it will be difficult to detect mutations which an average effect less than 58 μA, and typically experiments should last well under a second