I agree with your reasoning, I'm just asking what you're actually measuring where you indicate sub only in those graphs. I imagine that you simply make the mains not play and that the DLBC filter is always the same between sub only and sub + main measurements.
The measurements were not taken by me directly but were provided by another user upon request.
Sub only refers to the combined state of Sub1 and Sub2. After processing between the subs is completed, they are finally integrated with the main speakers.
As you mentioned, the filters for Sub-only measurements and Sub+Main in the DLBC state are identical. They should be, as this is not a filter issue. It seems to be caused by delay instead.
So, what I was suggesting is that I suspect there might be an alignment issue between the speaker and the subwoofer.
I agree - I cannot imagine how to justify a GD of 100ms, or even 50ms as shown in the above screenshot if you can significantly lower it by 'manual' integration
Either something is wrong with DLBC's algorithm or there is indeed a justification for such high GD values (that I cannot imagine but still)
In fact, most users in the Korean community where I am active do not use DLBC. Instead, they use DLBM and adjust it manually. They see this as a stepping stone toward eventually reaching Dirac ART...... =(
Of course, I believe that small errors in the millisecond range are possible.
As I mentioned earlier, in certain specific spaces, there might be unusually strong phase shifts that could cause DLBC to become confused and provide incorrect values. However, what I find difficult to understand is why it consistently provides incorrect values in such a manner.
It's not just a small ms error.... As you mentioned, I tried to reverse-engineer this, keeping Dirac's intentions in mind as much as possible. Even after virtualizing it with BRIR and listening to it countless times, I still couldn't understand the intended outcome.
