I am benchmarking an app using the batteryhistorian.
The result is a chart where every percentile is a "tick".
After I've measured the battery consumption
and let the battery drain to half, I would like to be able to save time
by making a change and restarting a measurment without having charged
the deviced back to full.
Is it a fair approximation to assume that a drop from 85% to 84% is the same
as a drop from say 64% to 63% when benchmarking an application?
Assuming your app is not a heavy power hog, it is reasonable to extraploate discharge. But, since the discharge isn't strictly linear, it would give better results if
Battery drop interval is more , say 5 to 10% and not 1% as proposed
Battery drop is not measured below 30% of battery ( to avoid greying into non linear portion of curve)
Measuring at 1% drop inherently introduces rounding off errors- if your phone has two digit accuracy, anything between 75.90 to 75.99 would show 76% . In the worst case, the error could be doubled (it won't fall to 75% till it goes below 75.90). Which means a 20% error, and that can be reduced to 2% over a 10% drop
Avoid measuring when the battery is fully charged to reduce errors. Measure at about 85% for the first reading and around 50% (or higher ) for second reading after tweaking your app
Q & A