AFMG Network Forum
AFMG Network Forum Index -> AFMG SysTune -> Windowing in Systune using TFC
Post new topic  Reply to topic View previous topic :: View next topic 
Windowing in Systune using TFC
PostPosted: Mon May 24, 2010 5:27 pm Reply with quote
soundarchitect
Member
 
Joined: 27 Jan 2010
Posts: 6




Hello,

I was wondering if anybody could explain the following to me:

I'm trying out the different windowing and smoothing options in Systune and I noticed that if I use the TFC window I loose a lot of resolution in the low frequency part. It also seems that there is a roll off of the frequency plot around 30 Hz.

It seems to me that if I use TFC the resolution through the spectrum should remain reasonibly constant. Perhaps am I missing the point of TFC?

Below is a screenshot of the Systune window with the windowing enabled.

Hope anybody can help or explain.




Regards,
Dennis Slot

_________________
Dennis Slot
Sound designer / consultant
Netherlands
View user's profile Send private message
PostPosted: Tue May 25, 2010 12:00 am Reply with quote
Doug Fowler
Forum Moderator
 
Joined: 29 Jun 2009
Posts: 11
Location: Saint Louis, MO USA




Dennis -

A couple of things, first :-)

When you set the 'window size' for the TFC window, keep in mind that value is the length of the window at 1KHz. 2KHz will be half that value, 4K half again, etc.

Going the other direction, 500 Hz will be twice as long, 250 four times as long, etc.

You have set the 1KHz window to be 200 msec long. That would make the 125Hz window eight times that, or 1.6 seconds.

So, a more reasonable value for 1KHz would be the first thing to look at.

Try -1 (from) and 41.5 (to) to simulate the window sizes for 24 PPO. Realize this is not an FPPO measurement, but merely simulates the window sizes used to implement that scheme. Until just recently, another FFT product used these window sizes.

You can then adjust the size of the 1KHz window (which also adjusts all the others) to fit the reflectivity of the room you are measuring in.

Your magnitude response is dead flat, looking like an electronic measurement.

Is that what this is?

The resolution will remain constant. If this is an electronic measurement, can you take a measurement in a reverberant space and post the result?

-doug
View user's profile Send private message
PostPosted: Tue May 25, 2010 1:38 pm Reply with quote
Charlie
Member
 
Joined: 28 Apr 2005
Posts: 51
Location: Charlotte, NC USA




Hi Dennis,
As Doug pointed out the TFC window you applied is 1.6 s at 125 Hz. At lower frequencies it is even longer. Given your FFT size of 1.37 s, this means that at these low frequencies you don't have a window other than the FFT block size. I don't think the TFC window would be further limiting the frequency resolution in this LF region.

Is the 30 Hz roll-off perhaps something in the device you were measuring? Do you have the band-pass filter engaged (Tools tab)? This can also result in the type of disply you are seeing in the LF region.

_________________
Charlie Hughes
View user's profile Send private message Visit poster's website
PostPosted: Tue May 25, 2010 11:34 pm Reply with quote
soundarchitect
Member
 
Joined: 27 Jan 2010
Posts: 6




Hello Charlie and Doug,

Thank you both very much for your reply.

To answer Dougs' questions: Yes, I've used an electronic measurement to try these things. And I expected the measurement to be flat all the way, and it does when I change the window to anything other than TFC. I will do another measurement (again) with a microphone to see what happens, but I've seen before that the resolution drops which is strange. Especially if think of it of as a simulation of the 24 or 48 FPPO systems. The programs that use this technique have pretty much the same resolution throughout the spectrum.

And to Charlie: there are no band-pass filters engaged. I will try using a different input or use the RME mixer to really match the inputs to see if there is anyting going on in there. Or does Systune only apply the window to one of the two inputs??????

I'll keep you posted and hope you can say anything more with this information.

_________________
Dennis Slot
Sound designer / consultant
Netherlands
View user's profile Send private message
PostPosted: Wed May 26, 2010 6:20 pm Reply with quote
Doug Fowler
Forum Moderator
 
Joined: 29 Jun 2009
Posts: 11
Location: Saint Louis, MO USA




Dennis -

The only thing here, using the window start and stop numbers I gave you, that has anything to do with 24 PPO is the window sizes used for each frequency.

TFC window is applied after deconvolution, not on the inputs (I think...., Charlie?).

Anyway, TFC uses a different window size for each frequency. Some of the other products use some percentage of octave banding to do this. Charlie, again? :-)

So TFC is different in that fashion.

If you zoom in on some area of the display and move your cursor along, it should change in equal increments. Do that, and let us know what you see.

I already know the answer.... :-)

-doug
View user's profile Send private message
PostPosted: Sat Apr 23, 2011 8:20 pm Reply with quote
Torben
Member
 
Joined: 19 Apr 2011
Posts: 3
Location: Germany




Hey Doug,

you wrote, that you set the TFC time at about 40ms for 1kHz to simulate a 24PPO. But I think this is not correct, because 24 points per octave at 1kHz has a frequency resolution of 40Hz and so in my opinion you need a window time at 1kHz of about 20ms to simulate a 24PPO and 50ms to simulate a 48PPO.

Please correct me if I'm wrong.

Torben
View user's profile Send private message
PostPosted: Mon Apr 25, 2011 5:09 pm Reply with quote
Bruce
Member
 
Joined: 19 Apr 2005
Posts: 459
Location: Minneapolis, MN, USA




You no doubt used a 1KHz BW to get 40Hz resolution. However, the Octave band at 1kHz has about a 630Hz BW, giving about a 26Hz resolution and roughly 38 ms length of each bin.

_________________
Best Regards,
Bruce C. Olson
View user's profile Send private message Send e-mail Visit poster's website
PostPosted: Wed Apr 27, 2011 6:32 pm Reply with quote
Torben
Member
 
Joined: 19 Apr 2011
Posts: 3
Location: Germany




Hey Bruce,
thanks for your answer. Yes you are right. I forgot that 1kHz is the Centerfrequency and not the starting point of an octave.

-Torben
View user's profile Send private message
PostPosted: Fri Apr 29, 2011 4:10 pm Reply with quote
Greg L
Member
 
Joined: 01 Dec 2009
Posts: 7
Location: Minneapolis, MN




First of all, applying a variable time window does not change the frequency resolution. If the analysis software calculates the resultant transfer function by changing the equivalent FFT block size, then the resolution will change.

A better way to look at variable time windowing is that, by doing so, we are creating a new and different system, with an IR that is frequency dependent. The manner in which it is created allows us to infer things about the original system.

Most importantly, making assumptions about TFC based on methods used in other analysis software, past or present, may not be correct.

Greg
View user's profile Send private message
AFMG Network Forum Index -> AFMG SysTune
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
All times are GMT  
Page 1 of 1  

  
  
 Post new topic  Reply to topic  


Powered by phpBB © 2001-2003 phpBB Group
Theme created by Vjacheslav Trushkin
Variation by CodeWeavers and AFMG