V5.20.14 early access build

[JoinUs]

REW can't do that. You are probably best setting the Analysis preferences option for imports to "For imports set t=0 at first sample" and when exporting use "Place t=0 at sample index" with an index of zero, then truncating the export to your required sample count in post-processing.

Okay! I will import it like that from now on.

No, because that would mean there was a discontinuity at the merge point.

That's a shame. But that's all right.
What do you think about doing the 1/a band limit that I wrote above, or if I don't, the response is weirdly reversed?

 

[Bentxu]

@John

Works as intended for me. The default target type in the Equaliser preferences is used for new measurements. Changing the default will not affect the target type for any existing measurements.

Sorry for never replying to this. You are right. Works for me now too. But as you said, does not work on old measurements.

Another feature request:

Would it be possible to put a freehand value (or some more fixed ones like 50, 75, 100) after which to stop an TRA measurement? Think the max value for that is 32 but I think a lot of users do 100 to get it stable. Particularly in a car where you you often have a less controlled environment than in a home cinema with cars driving by etc.

If that already works, would be great if you could point me towards how it is done. The ones I know who could potentially know also think it is not possible at the moment.

 

[JoinUs]

There is a difference between the target and almost 3db.
Yes, it can be adjusted with a shelf filter.
But it's buggy because it wasn't like this before....

This is exactly the result of LowShelf +3db at 1000hz.
There seems to be a bug in the calculation method of 1/a.

 

[John Mulcahy]

I also tried A/B on flat delta targets without putting a frequency band limit.

There was a bug in the A/B result if the SPL offset of measurement B was smaller than the SPL offset of measurement A, I have fixed it for the next build. Your 1/A results were mainly due to using a very large regularisation, and choosing a target level higher than the response in the inversion range. The smoothed responses conceal how large the differences are and hence the gains needed to invert.

 

[John Mulcahy]

This is exactly the result of LowShelf +3db at 1000hz.
There seems to be a bug in the calculation method of 1/a

Please attach the mdat file if something is to be investigated. Images do not provide enough information about the type and source of the measurements.

 

[JoinUs]

Please attach the mdat file if something is to be investigated. Images do not provide enough information about the type and source of the measurements.

Yes. I attach the mdat file.
The order of the work I did is as follows. (You can see it from Mdat.)
A is Original Messurement, B is Target

1. A/B (Apply to A - FDW15) (For To apply a little smoothing)
2. 1/A (Upper frequency limit 1000 Hz)
3. 1/A minimum phase
4. Original file * 1/A minimum phase (Before applying the multiplication arithmetic, the FDW was released on the original file.)
5. Short 3db response generated in shape of lowself from approximately 1 kHz
6. So, I Apply last lowself 1000hz +3db

There was a bug in the A/B result if the SPL offset of measurement B was smaller than the SPL offset of measurement A, I have fixed it for the next build. Your 1/A results were mainly due to using a very large regularisation, and choosing a target level higher than the response in the inversion range. The smoothed responses conceal how large the differences are and hence the gains needed to invert.

Very Kind. Thanks.

 

[John Mulcahy]

Yes. I attach the mdat file.

I think the issue was the level shift of A/B and the need to choose a target level for 1/A because of that. With the A/B fix and a target level of 0 everything lines up as it should.

 

[JoinUs]

I think the issue was the level shift of A/B and the need to choose a target level for 1/A because of that. With the A/B fix and a target level of 0 everything lines up as it should.

Thank you for looking into my case. So that means, Do I need to properly adjust the level of A/B every time? I understood that 1/A can be checked as target level 0.

 

[John Mulcahy]

Builds updated today (ea74) with these changes:

• Fixed: Match response to target did not work with a configurable equaliser with fixed Q
• Fixed: Array index out of bounds could occur when exporting impulse as WAV with certain t = 0 index values and data lengths
• Fixed: The result of A/B trace arithmetic had a level shift if the SPL offset of measurement B was smaller than the SPL offset of measurement A

 

[jeppekristoffer]

1. Many times when i open a measurement with eq-filters in it(all pass filters) and when looking in the overlay tab "Predicted phase" and "unwrap" engaged i have to disengage/engage the filters for the phase change to take place.
2. And if the phase trace is wrapped and i press unwrap then not all traces selected unwrap. To be able to make them unwrap i have to switch to another of the overlays tabs and back to "predicted phase" tab and then all traces selected are unwrapped. The same applies when goind from unwrapped the wrapped.

Regards Jesper

 

[JoinUs]

Builds updated today (ea74) with these changes:

• Fixed: Match response to target did not work with a configurable equaliser with fixed Q
• Fixed: Array index out of bounds could occur when exporting impulse as WAV with certain t = 0 index values and data lengths
• Fixed: The result of A/B trace arithmetic had a level shift if the SPL offset of measurement B was smaller than the SPL offset of measurement A

Never mind. maybe my fault. It seems to work now

 

[JoinUs]

A/B disappearing from the graph altogether seems to have been an error because there was a part that goes down to -300 at the beginning and end of the SPL I made virtually.
So I fixed it and now the graph is visible.
But it still shows a different calculation than the old Trace. Inconsistent with target...I don't know what version I used to use a lot of inversion using the target (a few months ago), but it worked well no matter how curved the target is. At the moment, I don't think I'm taking into account the shape and level of my target at all.

View attachment 64735

Blue is Target. Green is Result.

 

[serko70]

A/B disappearing from the graph altogether seems to have been an error because there was a part that goes down to -300 at the beginning and end of the SPL I made virtually.
So I fixed it and now the graph is visible.
But it still shows a different calculation than the old Trace. Inconsistent with target...I don't know what version I used to use a lot of inversion using the target (a few months ago), but it worked well no matter how curved the target is. At the moment, I don't think I'm taking into account the shape and level of my target at all.

View attachment 64735

Blue is Target. Green is Result.

It works fine with me...

Keep target level at auto during 1/A operation and apply a negative offset equal to the target level calculated by 1/A before you generate MP of 1/A and convolve it with the original response:

 

[Lowclock]

"Scale IR data..." seems to not work properly sometimes. Here's is an example.

test1.mdat is the starting measurement.

test1.wav is the wav that gets exported from it. Apply IR window and 4k sample count checked.

test2.mdat is the measurement saved after using the Scale IR data action set to -6db and nothing else. The SPL & Phase tab frequency response changed, but the Impulse tab looks the same as before.

test2.wav is the wav that gets exported from it. It's the same as test1.wav.

test3.mdat is test2, but after changing any IR windowing setting. Start, end, FDW, anything, even if you change it right back. The SPL & Phase and Impulse tabs snap to looking how they should, though the magnitude of the Impulse tab's frequency response plot does not, but I'm not sure if the latter is by design.

test3.wav is the wav that gets exported from it. The result is properly scaled.

It seems like Scale IR data only gets partially applied to the actual data unless anything about the windowing is actually changed. I spent way too long trying to figure out why I couldn't export a scaled impulse response wav properly before I finally figured out how to make it work.

 

[John Mulcahy]

"Scale IR data..." seems to not work properly sometimes.

If an FDW was applied the windowed data wasn't being updated, I've fixed it for the next build.

 

[JoinUs]

Can you add the default path setting option for exporting Wav files?
Once specified (for example, in a particular folder) it will remain in place until you turn off the REW, but restarting the REW will reset the download path and will require you to specify it again.
Of course, it's not difficult, but I always save it in the same folder, so I was wondering if this was possible. As if you can specify the default download path in a web browser such as Chrome or Edge.

 

[John Mulcahy]

Can you add the default path setting option for exporting Wav files?

The folder last used for WAV export is always remembered between startups for me.

 

[JoinUs]

The folder last used for WAV export is always remembered between startups for me.

Umm... Not this version i guess.. This is the latest REW beta and is automatically assigned to the "My Documents" folder. I usually save it on my desktop folder.

 

[John Mulcahy]

I am using the latest build and the folder is remembered. Perhaps there's some issue with saving preferences on your system.

 

[Loupy]

I am having a small issue, when I am in the "RTA" page, I cant see the "average the Responses" tab, Which is meant to be at the bottom left corner of the graph
Is there a setting for this in "Preference", I am running v5.20.14

Cheers
Peter

 

[powerplaymike]

@John Mulcahy :
Not sure if this is technically the correct thread to ask, but was wondering if the current version on REW website (5.20.13) is ok to download for MacOS 14 (Sonoma).

The link only shows macos 10.11 - 13, and says the build is from September 2022 which pre-dates Sonoma that was released October 2023.

If not, should we run the beta 5.20.14 instead?

Background: just got a new macbook pro 14.....previously used REW for many years but was on an older mac that couldn't upgrade past Monterey (macos 12)
Also, I've only ever used the full version on the REW site, not the beta versions, so I'm not sure what (if any) differences I may expect if I need to download the beta, but based on the AVS thread the beta appears to be pretty bug-free.

 

[sm52]

John, I don’t know what build this started happening with, but the measurement window resets the number of measurement repetitions after closing the window. I set it to repeat 2 times, take a measurement, but next time the number of repetitions is 1 again.
And. During the impedance measurement calibration steps, a message appeared and REW had to be closed.

 

[John Mulcahy]

was wondering if the current version on REW website (5.20.13) is ok to download for MacOS 14 (Sonoma).

Better using the latest 5.20.14 build, the older build may have issues installing (requires theme set to Light) and may have mic access permission issues.

 

[John Mulcahy]

the measurement window resets the number of measurement repetitions after closing the window

Please post a screenshot of the Measure dialog as you have it set before starting to measure

During the impedance measurement calibration steps, a message appeared and REW had to be closed

Please post a screenshot of the Measure dialog as you have it set for the calibration.

 

[sm52]

Please post a screenshot of the Measure dialog as you have it set before starting to measure

I forgot to say that I meant the number of repetitions of measurements when performing calibration steps before measuring impedance.