Technical Notes on the Ramping Algorithm
It's analyzing the thumbnail from each image to determine how much it's under or over exposed relative to the starting image, and then feeds that info to the ramping algorithm
The ramping algorithm then filters (removes any outliers) and integrates the exposure readings as they are passed to a tuned PID function for tracking
- The PID tuning is split into two parts -- one set of tuning parameters for rising light levels, and a separate set for falling light levels. This allows the system to "lag" a little bit for sunset so that it feels like it's getting darker, and also lets it aggressively "over-respond" to sunrise preventing it from blowing out.
- Additionally, after the PID step, an offset is applied based on the overall exposure value, compensating by -1 stops (configurable) during the night and 0 stops during the day (and interpolating in between). This helps night feel like night.
- All of the parameters above will be user accessible, but the defaults should always work so unless a person is curious to experiment, it should not be necessary to adjust anything. Right now there's a total of 16 parameters it uses for tuning, but I'm going to try to abstract that to a "more responsive/less responsive" slider setting. Still, the current defaults are working wonderfully.