Confused about the difference between FDW and smoothing

[JoachimStrobel]

I recorded responses that I have trouble understanding:

These are two large speakers in a 20 sqm room, placed near corners – there is no other options. And I prepare myself for a DRC measurements with REW.

Now, the first display shows a 20 sec 10-20 Khz sweep with time alignment and default windows with a 1/24 smoothing. I checked the imulse and both speakers are "plus". It is as it is, awaiting DRC work…

And then, I use a FDW of 1/11 octaves and see a severe dip at 9 kHz. Of course, there are other dips, but these are not my concern as they are nudges and could be seen before and relate to the room (?mostly). But the one at 9 kHz is suspect. Why does the FDW filter see it, and the raw data or smoothed data das not see it? And is does not matter if I use 1/11 or 1/40, it is still there.

Meanwhile I changed the amp and play via recorded wave through the whole system, but it is the same. So, this must be the loudspeaker? But why does it only show in an FDW processed sweep? Is FDW processing then a must? Or does FDW emphasize things that are not important.

 

[John Mulcahy]

The FDW gets narrower as frequency increases, so things that may be filled in by later reflections become visible.

If you measured both speakers playing then the response will have comb filtering effects from even very slight differences in the distance from the mic to each speaker, that makes the responses hard to interpret. About the only time it makes sense to measure more than one speaker at a time is when looking at the low frequency behaviour that would result from mono bass content fed to the mains.

 

[JoachimStrobel]

The FDW gets narrower as frequency increases, so things that may be filled in by later reflections become visible.

If you measured both speakers playing then the response will have comb filtering effects from even very slight differences in the distance from the mic to each speaker, that makes the responses hard to interpret. About the only time it makes sense to measure more than one speaker at a time is when looking at the low frequency behaviour that would result from mono bass content fed to the mains.

....... I measured the speakers separately.

 

[JoachimStrobel]

I guess the topic that I have touched has been discussed extensively here already

Feature Request: Frequency Dependent Windowing

Dear John: Many thanks for your continued advancements in performance in REW! If I understand correctly reading the latest beta notes, you have conquered the OSX measurement issues with Java that plagued us before and led me to use ASIO on PC? -------------- I'd like to make my "annual"...

www.hometheatershack.com

It seems to me, that the FDW „sees“ the loudspeaker‘s direct sound, and that loudspeaker has a „dent“ in the tweater. The room‘s reflection average that and it is hence hardly visible in the slightly smoothed data. That speaks a bit against the common wisdom that FDW processes will lead to a loss of information while spatial averaging will enhance information. I will do some pink noise stuff and some 0 deg mic measurements to check.

 

[kmajjo]

Hello,
What are the options to obtain the FDW options window ?
Br
Jaky

 

[kmajjo]

IR windows

 

[JoachimStrobel]

Hello,
What are the options to obtain the FDW options window ?
Br
Jaky

I am not quite sure what you mean. REW uses a fixed cycle or octave window for FDW. As an half educated amateur I understood, that using two such settings for low and high frequency could be beneficial. For the low frequency the octave setting should be longer as to resolve lower frequencies and capture some room modes, while for high frequency one would use narrow settings to focus on the direct sound. I guess there is now way to splice those together in REW.

 

[Spennis]

I recorded responses that I have trouble understanding:

These are two large speakers in a 20 sqm room, placed near corners – there is no other options. And I prepare myself for a DRC measurements with REW.

Now, the first display shows a 20 sec 10-20 Khz sweep with time alignment and default windows with a 1/24 smoothing. I checked the imulse and both speakers are "plus". It is as it is, awaiting DRC work…
View attachment 37566

And then, I use a FDW of 1/11 octaves and see a severe dip at 9 kHz. Of course, there are other dips, but these are not my concern as they are nudges and could be seen before and relate to the room (?mostly). But the one at 9 kHz is suspect. Why does the FDW filter see it, and the raw data or smoothed data das not see it? And is does not matter if I use 1/11 or 1/40, it is still there.
View attachment 37567

Meanwhile I changed the amp and play via recorded wave through the whole system, but it is the same. So, this must be the loudspeaker? But why does it only show in an FDW processed sweep? Is FDW processing then a must? Or does FDW emphasize things that are not important.

Old thread, but I saw the exact same thing when trying FDW for the first time last night. Did you figure this out?

Also, I'm wondering, can FDW be applied at different stages? I thought maybe it had to be applied as a measurement setting before taking a measurement, but apparently I was able to apply it afterwards to an old measurement last night.

 

[John Mulcahy]

If a measurement is made with more than one speaker playing there will be severe comb filtering causing deep notches in the combined response. Windowing is a post-measurement process and can be changed as desired.