- Generated by
1.15.0
Knobs for measure_delayed_passthrough. More...
#include <latency_evidence.hpp>
Public Attributes | |
| std::int64_t | max_delay_samples = 8192 |
| Largest delay considered, in samples. | |
| double | null_floor_db = -60.0 |
| The residual at the best delay must be at least this far below the input level for the output to count as "a delayed copy of the input". | |
| double | ambiguity_margin_db = 12.0 |
| The best delay must beat every non-adjacent competing delay by at least this margin. | |
| std::int64_t | tolerance_samples = 0 |
| |measured - reported| may not exceed this and still be a match. | |
Knobs for measure_delayed_passthrough.
The defaults are deliberately conservative: they would rather refuse to measure than report a delay they are not sure of.
| double pulp::test::audio::DelayedNullOptions::ambiguity_margin_db = 12.0 |
The best delay must beat every non-adjacent competing delay by at least this margin.
This is what stops a periodic stimulus (a pure sine nulls just as well one period late) from producing a confident wrong answer.
| std::int64_t pulp::test::audio::DelayedNullOptions::max_delay_samples = 8192 |
Largest delay considered, in samples.
The search window is [0, max_delay_samples]; a true delay beyond it reads as not_measurable.
| double pulp::test::audio::DelayedNullOptions::null_floor_db = -60.0 |
The residual at the best delay must be at least this far below the input level for the output to count as "a delayed copy of the input".
If it is not, the processor is doing something other than passing audio through and the policy reports not_measurable rather than guessing.
| std::int64_t pulp::test::audio::DelayedNullOptions::tolerance_samples = 0 |
|measured - reported| may not exceed this and still be a match.
1.15.0