Bugzilla – Bug 1362
Measured selection rounded to time controls in Analyze Contrast
Last modified: 2018-08-20 11:51:43 UTC
The time controls in "Analyze Contrast" are always in the form hh mm ss + hundredths. If the actual selection start or end time is not an exact hundredth of a second, the effect rounds the selection times before measuring. In some cases this can have a very significant effect on the measured value. As the selection cannot be defined within the Analyze Contrast GUI (see bug 1361), the measured selection 'should' be the actual selection (to the nearest sample) rather than the rounded value.
Fixed in https://github.com/audacity/audacity/commit/03915b443 The measured duration is now taken from the selection rather than from the (read-only) time controls.
This bug seems to have been fixed last year.
(In reply to Steve Daulton from comment #2) > This bug seems to have been fixed last year. It would probably help to use the Steps to Reproduce box to make it easier for a wider range of people to test. Guessing at some reasonable steps, in 2.1.2 on Windows 10 I created a DTMF tone at 1.0 amplitude and mouse-selected from 1.085 seconds to 1.135 seconds as displayed in Selection Toolbar. Most of the selection was in silence, but the last 12 milliseconds selected were the start of a tone. The TimeText controls in Contrast see the selection as 1.08 seconds to 1.13 seconds and the RMS as "-17.2 dB". ACX Check says the RMS (not RMS (A) ) is -13.5 dB. If I move the right edge of the selection back to 1.134 seconds and remeasure, Contrast sees no change in the dB value. Saving this as a project (selection of 1.085 to 1.135s) and reopening in HEAD, Contrast sees the RMS of the selection as -13.52 dB and ACX as -13.5 dB (same as before). If I move the right edge of the selection back to 1.134 seconds, Contrast now sees a lower RMS as expected (-14.19 dB). As in 2.1.2, the RMS result in Contrast correctly does not change when changing the Selection Format in Selection Toolbar. If I move the project to Mac I get the same result. Given Steve tested on Linux this is good enough for me to pass it.
