Home Theater Forum and Systems banner

1 - 10 of 10 Posts

·
Registered
Joined
·
7 Posts
Discussion Starter · #1 ·
Hello

I use REW for teaching and this issue constantly arises. When analysing the RT60 of an impulse response a certain frequency range will show zero reverb time (see attachment). I am very familiar with REW and am quite sure I am doing everything correctly, though maybe not ;) Has anyone issue come across this and have a solution? Or is it a bug?

thanks
Alan
 

Attachments

·
Premium Member
Joined
·
6,711 Posts
It is a failure of part of the lundeby routine when faced with an impulse whose initial decay in a frequency band is very fast and very deep. I can fix it by tweaking a threshold in that part of the code, but that may cause issues with other measurements, I'll need to check the effect on some other data.
 

·
Premium Member
Joined
·
6,711 Posts
I have made a modification in V5.17 beta 4 to accommodate the initial decay issue that response highlighted. Let me know if you see odd results with any other responses.
 

·
Registered
Joined
·
7 Posts
Discussion Starter · #4 ·
Hi John

Thanks for acting on this :) When I open the IR that was showing unusual results in the new Beta version the results are much better. But if I open another IR in the Beta version I get strange results again. When comparing the RT60 results of this new IR in V5.16 with the Beta version everything is the same up to around 300Hz then they start to differ, but both giving strange results. I have attached images from both versions and the mdat file from the beta version.

Thanks
Alan
 

Attachments

·
Premium Member
Joined
·
6,711 Posts
That looks like an export of a windowed response. One of the things the RT60 code needs to establish is where the noise floor lies, on that response there's no way of telling where the floor is because of the effect of the truncation and the window shape that was applied to it. For RT60 calcs really need to use an unwindowed response.
 

·
Registered
Joined
·
7 Posts
Discussion Starter · #6 ·
You lost me a bit there ;) Is it something I am doing to cause these "errors" or still something within the software. I don't see any windowing options in relation to RT60.
 

·
Premium Member
Joined
·
6,711 Posts
I mean the response, which is an imported IR, looks like it was windowed before it was exported, so the noise floor has been lost. That's not to say there shouldn't be a way of dealing with that, but Lundeby uses the noise floor to work out where to start the backward integration for the Schroeder curve that is used to derive the RT60 values, and without that noise floor that gets difficult so it fails. The noise floor appears as a flat region on the log plot of the IR. Here's the latest response (top), compared to the ORTF response to illustrate the difference.

ir1.png

irortf.png
 

·
Registered
Joined
·
7 Posts
Discussion Starter · #8 ·
Ok got it. The IRs I am using were created using a plugin within Ableton Live, I guess I should have mentioned that earlier. I also use REW to create IRs but not these ones. I guess that plugin must have performed some sort of windowing function.
 
1 - 10 of 10 Posts
Top