REW: export smoothed impulse responses (.wav), but "measured" and not "minimum phase"...

[Stewix]

REW always puts the ref time at the peak, but you may change it afterwards.

Oh, ok...

So I was right in saying that one way to export impulses so that they are aligned is to apply to them the same left width, while keeping their respective ref time unchanged...

Good to know!...

You have been awesome as always!...

Many many thanks!...

 

[John Mulcahy]

So I was right in saying that one way to export impulses so that they are aligned is to apply to them the same left width, while keeping their respective ref time unchanged...

That would align their peaks, but peak width and position varies with system bandwidth so the impulses would not necessarily be time aligned.

 

[Stewix]

That would align their peaks, but peak width and position varies with system bandwidth so the impulses would not necessarily be time aligned.

Ok, I think you are referring to the fact that, for example, a subwoofer impulse, having only low frequency, would have a "larger peak"...

But when two IRs can be considered "time aligned" then, when their starting points (where they start to raise from noise floor) are aligned?...

 

[Stewix]

That would align their peaks, but peak width and position varies with system bandwidth so the impulses would not necessarily be time aligned.

PS: moreover, impulses "derived" from the same system (like in my specific case: 4 impulses of the same stereo gear piece), would have the same system bandwidth and, therefore, peaks aligned would also mean impulses time aligned, right?...

 

[John Mulcahy]

two IRs can be considered "time aligned" then, when their starting points (where they start to raise from noise floor) are aligned?

Yes.

impulses "derived" from the same system (like in my specific case: 4 impulses of the same stereo gear piece), would have the same system bandwidth and, therefore, peaks aligned would also mean impulses time aligned, right?

Barring any substantial difference between the channels, yes.

 

[Stewix]

Yes.

Ok, thanks...

Barring any substantial difference between the channels, yes.

I'm not worried about the differences between the channels, but the difference between the "normal" channels (L-R), and the "crosstalk" channels (R-L), is that the "crosstalk" measurements have a considerably poorer sound to noise ratio (it was inevitable); but, also in that case, the bandwidth should be the same; so I think I'm good to go...

I would have simply kept their relative timings, but the time differences where so small that the "cumulative shift" end up putting, in some cases, the "normal" channels (very very slightly) after the "crosstalk"channels (which, if I'm not wrong, should be physically impossible); so I opted, in those cases, to align the peaks instead...

 

[Stewix]

Hard to say, probably easiest to just try it.

I've done some more tests regarding using REW for downsampling external impulses, and I've seen that REW seems to do a better job if I apply a window when exporting/converting...

So, let's say I import a church impulse, for example, that has a tail that lasts 3 seconds (3000 ms)...

What is the "rationale" behind the window size? Maybe I can "imitate" the window size that usually REW applies by default, so that, in the example of the church impulse, one should apply a right width of 3500 ms (while letting the left width as REW calculates it)?...

 

[John Mulcahy]

You can override the default window settings in the Analysis preferences. REW's defaults are for domestic spaces, larger spaces would need longer right windows to capture their decay.

 

[Stewix]

You can override the default window settings in the Analysis preferences. REW's defaults are for domestic spaces, larger spaces would need longer right windows to capture their decay.

Yes, thanks, but I was not asking this...

I was asking if there is some reason, regarding the audio quality, to let some more "space" on the right...

Let's say that a "reverb tail" stops completely at 3000 ms; is there a "resolution benefit" in letting 500 ms more, setting the right width to be 3500 ms (and not simply 3000 ms)?...

I'm also not talking about the fact that, if one does not want the window "form" to change anything, one can chose the window to be "rectangular" (though this can be a further consideration maybe, because I think that those "forms" can act like some sort of "fade in" and "fade out")...

 

[John Mulcahy]

One purpose of the window is to smoothly truncate the ends of the response to avoid response artefacts. Rectangular windows do not do that and can result in frequency domain ringing depending on the level at the point of truncation.

Using a longer window, extending beyond the point the response drops into the noise floor, mainly adds noise.

 

[Stewix]

One purpose of the window is to smoothly truncate the ends of the response to avoid response artefacts. Rectangular windows do not do that and can result in frequency domain ringing depending on the level at the point of truncation.

Using a longer window, extending beyond the point the response drops into the noise floor, mainly adds noise.

Ok, thanks...

I think then that, regarding the example of the reverb tail that stops at 3000 ms, having a right window of 3000 ms is the right thing to do...

But if I use the "Tukey 0.25" form, shouldn't I let just a bit more ms, so that that "form" does not "alter" the end of the tail?...

 

[John Mulcahy]

Could do, sure, though the level should already be very low by then if the IR has good signal-to-noise.

 

[Stewix]

Could do, sure, though the level should already be very low by then if the IR has good signal-to-noise.

Of course, I agree regarding the level...

But what should I imply from this, that the "Tukey form" would then be somewhat "useless", because with a very low level also the artefacts would be practically non-existent, or that the reverb tail should be so low that the "Tukey form" wouldn't "harm" its sound, while also avoiding the artefacts?...

I suspect it is the second, but I'm not sure...

 

[John Mulcahy]

The second.

 

[Stewix]

The second.

Thanks for the umpteenth time!...