0002647: LEQ(m) calculation has bugs

ID	Project	Category	View Status	Date Submitted	Last Update

0002647	DCP-o-matic	Bugs	public	2023-11-07 15:25	2024-01-02 02:23

Reporter	Carsten	Assigned To	carl
Priority	normal	Severity	major	Reproducibility	always
Status	resolved	Resolution	fixed
Platform	Mac	OS	OS X	OS Version	10.12
Product Version	2.16.x
Target Version	2.16.67

Summary	0002647: LEQ(m) calculation has bugs
Description	Looks as if LEQ(m) calculation has a bug in that the results vary with the number of audio channels. E.g. the same audio gives different results when e.g. 6 channels have been assigned in the DCP, vs, 8 or 16. As a result, you may get correct LEQ(m) values when you use a 5.1 source file in a 6ch DCP, but when you create a SMPTE DCP with 8 or 16channels in the mxf, the LEQ(m) will not calculate to the same number. That doesn't seem to happen with LUFS. I guess both LUFS and LEQ(m) should only use active channels, and there should also be a limitation to useful audio channels, that is, exclude HI/VI-N, ATMOS TC, dbox tracks, etc. That would mean, look only at channels 1-6, 11/12, and also take into account audio matrix. See https://dcpomatic.com/forum/viewtopic.php?t=2290
Tags	No tags attached.

Branch	2647-leqm
Estimated weeks required
Estimated work required	Undecided

carl 2023-11-11 00:24 administrator ~0006062	Carsten mentioned some problems with this.

Carsten 2023-11-11 16:23 manager ~0006063	How is the channel assignment done for LUFS/R128 vs. LEQ(m)?

carl 2023-11-11 18:57 administrator ~0006065	Good question, it looks like LUFS/R128 only takes 5.1 and ignores anything else.

carl 2023-11-15 13:31 administrator ~0006071	Need to work out why Carsten gets failure to analyse audio with this branch (mentioned on the forum).

Carsten 2023-11-15 13:40 manager ~0006072	Occasionally, I get that error message when trying to analyse audio. Usually it's fixed after deleting the analysis folder. I just set up a project to reproduce this new issue with 2.16.66devel: Create new project. Per default, number of DCP audio channels is set to 6. Add an audio MXF with 16 channels from a DCP previously generated with DCP-o-matic. In the matrix, only channel 1-6 are assigned, 7-16 remain unassigned. Choose 'Show Graph of audio levels'. The window opens, but it says: 'Could not analyse audio'. Set number of DCP channels to 16 under DCP->Audio. After that, analysis is performed successfully, and it looks as if the LUFS/LEQ relationship makes sense.

carl 2023-11-15 16:50 administrator ~0006073	Thanks - this should be fixed now, I'll make another build.

carl 2023-11-20 06:35 administrator ~0006086	Hopefully all fixed in 4ae0193c125cc70176eb1660919d507fbf5378e4

mhm 2024-01-02 02:23 reporter ~0006180	Maybe related to 0001107

Date Modified	Username	Field	Change
2023-11-07 15:25	Carsten	New Bug
2023-11-07 15:25	Carsten	Status	new => assigned
2023-11-07 15:25	Carsten	Assigned To	=> carl
2023-11-07 15:27	Carsten	Description Updated
2023-11-07 15:27	Carsten	Estimated work required	=> Undecided
2023-11-07 21:23	carl	Target Version	2.16.x => 2.16.67
2023-11-07 21:26	carl	Relationship added	related to 0001991
2023-11-07 23:17	carl	Status	assigned => confirmed
2023-11-08 00:00	carl	Branch	=> 2647-leqm
2023-11-08 00:00	carl	Status	confirmed => tests running
2023-11-11 00:24	carl	Status	tests running => feedback
2023-11-11 00:24	carl	Note Added: 0006062
2023-11-11 16:23	Carsten	Note Added: 0006063
2023-11-11 16:23	Carsten	Status	feedback => assigned
2023-11-11 18:57	carl	Note Added: 0006065
2023-11-15 13:31	carl	Status	assigned => feedback
2023-11-15 13:31	carl	Note Added: 0006071
2023-11-15 13:40	Carsten	Note Added: 0006072
2023-11-15 13:40	Carsten	Status	feedback => assigned
2023-11-15 16:50	carl	Note Added: 0006073
2023-11-19 23:03	carl	Status	assigned => tests running
2023-11-20 06:35	carl	Status	tests running => resolved
2023-11-20 06:35	carl	Resolution	open => fixed
2023-11-20 06:35	carl	Note Added: 0006086
2024-01-02 02:23	mhm	Note Added: 0006180