Home Theater Forum and Systems banner

REW V5.17 Beta 10 smoothing bug

3K views 16 replies 3 participants last post by  EarlK 
#1 ·
Hi John,

The smoothing controls are no longer working properly in REW V5.17 Beta 10.
- "Remove Smoothing" doesn't work, while "1/48 & 1/24" give the same "look" ( & that "look" seems to be some variation of smoothing values derived from between 1/12 > 1/24 ).

I can't say how long this problem has been present ( though back in May a transducer I measured was unexpectedly smoother looking compared to my expectations )..

They ( smoothing controls ) are working properly in REW V5.10 build 3618 ( which I have recently reinstalled to check out my suspicions ).

I'm running Windows 8.1 ( btw ).

:)
 
#2 ·
I'm on Win 7 x64 and can't verify this. However, I do see one quirk. I have an MDAT file with two nominally identical measurements that were saved with all smoothing disabled. The frequency range of the measurements is 40 Hz - 200 Hz. If I enable 1 oct smoothing, it works okay, with a start frequency for the smoothed trace of about 56 Hz. Then I disable smoothing and the trace returns to the unsmoothed state with a start frequency of 40 Hz once again, just fine.

Now I set smoothing on one of the measurements to 1 oct, then save all measurements to MDAT. Then I close REW, open it back up, and open up the saved MDAT. If I take the measurement that was saved with 1 oct smoothing, and subsequently disable smoothing on it, the result is unsmoothed, but the start frequency of the unsmoothed data is now 56 Hz instead of the original 40 Hz. So it appears to have lost some of the data or bookkeeping information required to regenerate the full data set of the original unsmoothed data.

I've attached the project below.
 

Attachments

#3 ·
I don't see that behaviour Earl, can you post an mdat that displays it and how you expect it to appear? Bear in mind that for measurements which are log spaced (96PPO) the minimum smoothing is 1/48, can see that in the Info panel. Measurements can be converted back to linear spacing by disabling the log spaced option in the Analysis preferences and re-applying the IR windows. 5.13 had some changes to smoothing, small improvements to the Gaussian approximation in the smoothing filters.
 
#6 ·
Hi John,

Here's the mdat file as requested.

Here are 2 pics that demonstrate the smoothing discrepancies ( as displayed on this laptop ).

The first is a single trace from the Decay tab ( set to 1/48th smoothing ).



The second pic is from the SPL & Phase tab ( same trace ).



I'm running the latest graphics driver for this laptop ( maybe thats' the problem :scratch: ).

FYI, this is an ancient ( 8yr old Dell ) machine . I am running Java 8 with an up-to-date Win 8.1 OS .

:)
 

Attachments

#8 ·
Hi Earl, the difference you are seeing is because the measurement has a frequency dependent window applied, which has its own smoothing effect. There isn't an option to use a frequency dependent window for the decay or waterfall plots (that's on my todo list), but the spectrogram has a wavelet mode which is produced using a frequency dependent window.
 
#9 ·
Hi John,

I agree that the 2nd pic I posted looks to have had a FDW smoothing window applied to it ( but I'm not making that choice ).

Specifically ( since I find your response a bit ambiguous ) , how do I engage the smoothing option ( assuming it exists ) to make my second picture look like my first picture?

The thing is, I have already tried to get rid of any over-riding global smoothing within the "Analysis" tab ( plus I'm not aware that I ever choose to apply something like FDW ).
- I have unchecked every box within the Preference's Analysis tab ( to try to disengage any & all global smoothing choices , but to no avail ) closed that window, shut-down REW, rebooted & re-opened REW / all to no avail.
- FYI, I can select "Remove Smoothing" from the graph menu ( for any file ) and smoothing is definitely not removed.


I have chosen "Delete Preferences & ShutDown", then Uninstalled REW & ReInstalled REW 5.17 / but this doesn't clear up this situation ( I no longer have 5.10 installed / which was a temporary, exploratory measure ).

:)
 
#10 ·
All the measurements in that file had a frequency dependent window, to remove it use the IR Windows dialog for each measurement and de-select it. Also check that frequency dependent window is not selected in the Impulse Response window defaults in the Analysis preferences.
 
#11 ·
Thanks John! That is indeed where the smoothing was being applied.

Let me ask, Was this global ( FDW smoothing ) a choice that I made in the past ?

Or is FDW now a default choice for every newly installed program ( my cleared Preferences not-withstanding ) ?

:)
 
#13 ·
Thanks for that!

So why doesn't deleting preferences & shutting down the program, then uninstalling the program & rebooting the computer ( before re-installing the program ) not give me the expected behavior ( of no FDW being applied to any newly acquired file ) ?

Or put another way, why does an older local file preference get carried forward ( after everything above is executed ) to be effectively made into a ( an unseen ) global preference ( for all new captures ) ?

Or in other words; Why doesn't the "Preferences" window ( & all choices made within it ) reign supreme ?

:)
 
#14 ·
Delete preferences and shut down works for me, a previous selection of applying FDW in Analysis preferences gets cleared (along with everything else being set back to defaults) and new measurements after REW is restarted don't have FDW applied. On Windows preferences are stored in the registry, delete preferences and shut down removes that registry key and all its contents. There is no other repository for the preferences.
 
#15 ·
Thanks John for guiding me through the wilderness on this matter ( I've sent a donation your way ).

For those of us ( in our 60's and who are more likely ) to forget where we've been within the program you might consider adding a pop-up dialogue box ( when a user is in the Smoothing Menu ) challenging that "Remove Smoothing" is impossible due to FDW being in place.
- You could always offer the user the option ( then & there ) to disable FDW before granting the request to remove all smoothing.

Additionally, when FDW is enacted ( and a user chooses 1/48 smoothing / or other ) it would be handy to actually see FDW-1/48, FDW-1/24, FDW-1/12, etc. indicated within the lower legend ( there's obviously quite a visual difference of any FDW-1/xx smoothing versus regular 1/xx smoothing ) .
- That simple indicator would have most likely prevented this thread.

Thanks Again! <> EarlK
 
#16 ·
Although the FDW has an effect on the response that looks like smoothing, it is a form of variable windowing of the impulse response rather than a choice of smoothing - its effect is due to progressively narrowing the region of the impulse being analysed to steadily remove all but the direct response. The combination of removing the room's influence at higher frequencies (and so removing comb filtering) and the reduced frequency resolution of an ever narrower window combine to produce a response that looks smoothed, particularly at high frequencies. It doesn't really qualify as smoothing though. Having some indication when an FDW has been applied is a good idea nonetheless, I'll add that to the list and think about how it might best be done.
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top