Really newbie question about the estimated time delay

[Bubo]

Hello everyone ! It's my first post on AvNirvana, as I am still pretty new to RoomEQWizard. I'm trying to let go of Smaart, and to use REW for tuning sound systems in live settings. Also, I'm french, so I might write english weirdly.

Today, I tried to align in time just a suboofer and a fullrange speaker, and I came across something I didn't understood.

I did first two mesures, one for the subwoofer, and one for the fullrange. The t for the sub was 34.9ms, and 17.2ms for the FR one.
Now, If I use "Estimate time delay" on both of these (I didn't change the timing offset when taking these measures), I get 17.2ms for the FR, and 24.3ms for the subwoofer, which get me a difference of 7.1ms of delay. So I put in my FR speaker 7.1ms of delay to match it with the sub.
I listened to it with a colleague from work and it seemed ok.

But then I got to the the Impulse category of the Overlays window, and I got confused.
When filtering my measures with a 100Hz 1/3 filter (my cross-over frequency), I saw that puting 7.1ms really made the two impulse NOT align, and moreso, it was better aligned without any delay. So we tried to measure form the overlay window what was the most convergent delay, and it was 1.1ms.

And when taking measures with both of the speakers active, the 1.1ms cleary got me the best addition in my cross-over area (the two speakers were phase-compatible so it don't seems like a """phase""" problem). But with our ears, it seemed like the 7.1ms sounded better (not by much, but still, with 4 blind comparaisons we preferred each time the 7.1ms).

So my question is this : How does the Estimated time delay works, and how can it suggest a delay that doesn't converge the IR on the frequencies of my cross-over ? I understand that the time delay is calculated based on the enery peak of each measure, but i wonder how the energy peak and the cross-over at IRs can be at differents delays. Also, are we completly deaf for preffering the 7.1ms delay ??? Maybe ???

I'll attach the .mdat file, if anyone want to look at the curves, and i also have a photo of the setup if you need (but it's really just a mic on a stand, 2m apart from the loudpeakers)

Thanks !

 

[John Mulcahy]

Estimate IR delay uses cross-correlation with a minimum phase version of the response, it isn't suitable for alignment though. For that you should measure with a timing reference (as you appear to have done) and use the Alignment tool on the All SPL graph.

 

[sm52]

Neither one nor that delay gives an ideal result. And they can't give. The sub has a roll-off slope of approximately 18 dB/oct. The crossover frequency is approximately 90 hertz. Try a roll-off slope of about 24 dB/oct. So that the crossover frequency becomes about 80 hertz. Or lower the sub level by 5 dB, and make the roll-off slope about 12 dB/oct.

 

[John Mulcahy]

 

[Bubo]

Wow, thanks for the quick answers !

I forgot about the alignment tool, it really is a powerful tool. And to answer to sm52 answer, the two speaker are have a DSP inside with no way of tweeking the parameters. So we're kinda stuck with these filters.

I will try again, using the alignment tool this time. And one more question : why is the alignment tool the only (?) place I can filter my IR at any frequency ? And why can't I choose between 1oct or 1/3oct ? Is the algorythm filtering the IRs in the alignment tool a different one than in the filterd IR window ? It feels easier to align with the IRs than with the Phase plot...

 

[John Mulcahy]

The alignment tool impulse alignment mode uses zero phase 1/3 octave filters. The Filtered IR graph is primarily aimed at examination of individual bands of RT60 analysis, it has options for 1 octave or 1/3 octave and forward, zero phase or time reversed filtering.

 

[Bubo]

Ok that makes sense.

Thank you for your quick answers !