Several issues with REW 5.20 beta 5

[AustinJerry]

I am noticing a couple of issues/peculiarities with the latest REW beta.

First of all, when loading a MDAT file with multiple measurements, the loading of the file is considerably slower than with previous versions. Is there some additional background processing happening as the measurements are being loaded? When searching through multiple MDAT files trying to find something specific, the slow loading times make the search much more tedious.

Second of all, when the same MDAT is loaded with different REW versions, shouldn't the measurement curves look the same? I have attached two screenshots, one from REW 5.19, and the second from REW 5.20 beta 5. Something very different is happening below 30Hz, yet the MDAT's are the same. Why is this?

 

[John Mulcahy]

The mdat format changed in 5.20 so there is some format conversion when loading files from previous versions. Part of the change speeds up the preview generation when browsing files so with 5.20 there may be less need to load measurements to see what they contain.

For that centre channel response please attach the mdat or a link to it.

 

[John Mulcahy]

As an aside, your older measurements may load more quickly if you set the IR oversampler to None in the Analysis preferences.

 

[AustinJerry]

Thank you for your response, John. Here is a link to the MDAT file:

https://www.dropbox.com/s/covww75b7xjn3xm/AustinJerry.mdat?dl=0

I have changed IR Oversampler to None, and it seems to make a slight difference.

Regarding the file conversion, if I load an older measurement file, which undergoes a conversion as it is loading, and then save the file, opening the new version is still slow. Since the file has been converted, would loading it subject it to the conversion logic again? I had expected the file to load quickly after being converted.

 

[John Mulcahy]

No, the conversion wouldn't happen on a file saved with the newer version. I'll look into it.

 

[John Mulcahy]

The plot difference looks to be a bug in the fractional octave smoothing for log spaced data. I'll fix it for the next beta. In the meantime you could disable log spaced data by deselecting the Allow 96 PPO Log Spacing option in the Analysis preferences and re-applying the IR window.

 

[John Mulcahy]

Do you have an example of a file that is slow to load Jerry? One that you haven't already saved in the new format.

 

[AustinJerry]

Do you have an example of a file that is slow to load Jerry? One that you haven't already saved in the new format.

Sorry for the delay, John.

Here is a file that loads slowly: https://www.dropbox.com/s/ks46rs6s84yxst5/Final Sweeps.mdat?dl=0

Another peculiarity I noticed with beta 5 is that the RT60 graph displays very slowly.

And a comment on a recent enhancement that is not a bug: I would prefer that the Waterfall and Spectrogram graphs not be automatically generated, or at least have an option in Preferences where this new feature could be turned off for those who don't like it. Why? Because if I click on the wrong measurement, I would prefer not to wait for the graph to be displayed before moving on to the correct selection. JMO.

 

[John Mulcahy]

Thanks for the file Jerry, I'll look at RT60 as well. You don't need to wait for the graph to display before selecting another measurement.

 

[John Mulcahy]

A first observation on that file: the impulse responses are very long (10.9 seconds), the default for the Analysis preference is to truncate the impulse response after 1.7 s as there is nothing of interest beyond that point unless the measurement is of a very large venue like a stadium. Selecting that will reduce your file sizes and load times, and the RT60 generation times.

 

[John Mulcahy]

I saw a file from another user on the AVS thread that had the full IR retained. I suspect users may be selecting that based on this section of your guide Jerry, though that setting is not what is being discussed. Probably worth updating that image.

 

[AustinJerry]

Thank you for the response, John. I have updated the guide with your recommendation.

 

[AustinJerry]

I am experiencing another severe performance issue with a large measurement file (48MB). When I load the file (which is extremely sluggish), then open the overlays window, and select the ETC tab, the ETC window is unresponsive. For example, "Clear Selections" does not clear the screen. Selecting and de-selecting indifidual measurements has no effect. And "Controls" will not open.

IOW, I am unable to do any analysis with this measurement file. I am running beta 6. File download link:

https://www.dropbox.com/s/h1fky2y69rg5byv/Screen up down a-b comparisons.mdat?dl=0

 

[John Mulcahy]

File opens OK for me, ETC overlay displays normally. Were there any other steps to show the problem? If you look in the REW log files has anything been captured? Very high noise level (or very low signal level) in the "C SW ONLY" plot, maybe because it was measured to 20k. Still have 10.9 second impulse responses in that file btw.

 

[AustinJerry]

Thanks for looking at this issue, John. I did some more testing. I have two REW computers, a laptop, and a more powerful desktop. As for the file loading times, the file loads in 30 seconds on the desktop, and 1 min 53 seconds on the laptop. This may be normal, I don't know.

But the more serious issue is the unresponsiveness when accessing the Overlays screen. On the laptop, when I open the Overlays screen, it takes ~30-60 seconds to respond to commands, which is better than the desktop, which after opening the Overlays screen becomes completely unresponsive. I tried the laptop a second time, and this time the Overlays screen was normally responsive.

I have attached two laptop log files, Log1 is when the Laptop was slow to respond, and Log2 is when the laptop responded normally. The single desktop log is for when it was complexly unresponsive.

I see multiple entries "Exception in thread "AWT-EventQueue-0" java.lang.OutOfMemoryError: Java heap space". Both the desktop and the laptop are running Windows 10 Home Edition 64-bit, patched to current levels. I looked at previous REW logs, and see none of the errors that are shown in the current logs. The errors started happening only with the file I shared with you yesterday, so my theory is that there is something wrong with the file, or that its size is causing an issue.

 

[AustinJerry]

Update: I deleted the "C SW ONLY" plot, saved the file with a different name, and re-opened it. The Overlays window is now normally responsive, and the log file shows no errors. Is there something wrong with that plot?

 

[John Mulcahy]

No, those times aren't normal, file loads in 8 seconds on my desktop. Which graph was open in the main window when you were loading the measurements? Were any other REW windows open?

None of the measurements give particular problems I can see but generating the RT60 overlay does get the processor fan running due to the length of the impulse responses, depending on how many cores you have that could cause responsiveness issues and with those impulse lengths does use a lot of memory, so could also cause out of memory problems. It would also lead to slow loading if an RT60 graph was open when loading the file.

I've said this before, but I strongly suggest you set IR truncation to its 1.7 s default value.

 

[AustinJerry]

No, those times aren't normal, file loads in 8 seconds on my desktop. Which graph was open in the main window when you were loading the measurements? Were any other REW windows open?

None of the measurements give particular problems I can see but generating the RT60 overlay does get the processor fan running due to the length of the impulse responses, depending on how many cores you have that could cause responsiveness issues and with those impulse lengths does use a lot of memory, so could also cause out of memory problems. It would also lead to slow loading if an RT60 graph was open when loading the file.

I've said this before, but I strongly suggest you set IR truncation to its 1.7 s default value.

The file we are using here is not a file I created. If the user who created it was not using the IR Truncation setting, then it would continue causing an issue on my PC, regardless of my setting, correct?

Several of my friends have asked, "What memory option did you choose (512KB, 1GB, 2GB) on install/update?". I don't see a memory option when I install REW. What are they talking about?

Anyway, I suspect the issues I am seeing are restricted to this one file. Since I am not seeing similar issues with the measurement files I create myself, we can move on.

 

[John Mulcahy]

Sorry Jerry, thought it was one of yours, shouldn't have assumed.

I added a screen to choose memory allocation in the 64-bit installer, with options up to 2 GB. For machines with at least 4 GB of memory the 2 GB option should be OK, previously REW allocated 1 GB. The versions with the 32-bit JRE tended not to start with a 2 GB allocation so I have made a screen with options up to 1.5 GB for those, that will be in the installers for the next beta.

 

[AustinJerry]

Mystery solved. Without paying attention, I have been running 32-bit REW with the 32-bit Java JRE. I uninstalled this version, downloaded the 64-bit REW and the 64-bit JRE, and I am now prompted for memory usage. I chose 2GB, and now have no trouble with that measurement file from several days ago. Thanks again for your guidance.

 

[John Mulcahy]

That's good. I spent the day taking another look at file loading and worked out a few ways to speed things up, that file now loads in under 2 seconds on my desktop. Also managed to speed up the RT60 generation pretty dramatically, both those changes will be in the next beta.