following a decrease in gain due to an increased noise floor.
This is controlled by --adaptive-range-scan-delay and defaults
to 5 minutes (vs. the default rescan delay of 1 hour)
This makes the adaptive gain code only process some fraction of all samples,
rather than every sample. This helps a lot with CPU on slower machines.
We default to a 50% duty cycle to save some CPU without really affecting
the behavior much.
We split the one-second adaptive gain blocks into 20* 50ms "subblocks"
Each subblock is either entirely processed or not processed at all,
based on the configured duty cycle. (This means that the duty cycle
can only be specified in approx 5% increments, but that should be
fine)
For adaptive dynamic range, there's no further complications - we
just look at fewer samples to compute the noise floor.
For adaptive burst, we need to scale up the measured burst rate by
1/duty cycle to account for the samples that we didn't process. It's
also more likely to cause functional changes as it's quite possible
for a relatively small number of loud messages being missed to
affect the computed rate substantially. Not many options there other
than increasing the duty cycle when adaptive burst handling is wanted.
Update adaptive gain control loop to handle interactions between burst detection and dynamic range control.
Internal doc updates & naming cleanups.
Stats tweaks.
This PR adds basic adaptive gain support, which adjusts SDR gain on the fly based on the noise & signal levels seen.
There are two control mechanisms:
Dynamic range control is enabled by the --adaptive-range option. This adjusts SDR gain to try to achieve a minimum dynamic range, regardless of the exact hardware in the RF path.
Burst (loud message) control is enabled by the --adaptive-burst option. This decreases SDR gain when undecodable loud messages are heard, allowing for better reception of nearby aircraft at the expense of range.
This is only the basic implementation - see the PR for remaining work to do.
Oliver Jowett