V5.20.14 early access build

[KSTR]

OK, thank you for the detailed answer, appreciated. Will check back with Pavel...

 

[GERONIMO.USMC]

@John Mulcahy

Hi. I've noticed this box rendering error under All SPL for awhile across builds on my desktop, but they have been fine on my laptop. Both run the latest Win 11 builds, 22621.963.

My display settings are attached and I'm running at 120Hz refresh rate.

 

[John Mulcahy]

The difficulty is with the screen scaling, at 300% the effective resolution of the screen is only 1280 x 720, so there is very little vertical height available. That can be helped a little by going to the View preferences and unticking "Show toolbar text labels" and "Show graph button text labels". If that isn't enough you can untick "Show toolbar". Or reduce the screen scaling a little if the text stays readable for you at lower scaling.

 

[dcibel]

Hi @John Mulcahy ,
Kimmo, the VituixCAD developer, has indicated some timing problems with "loopback with merge IR" when there is a long timing delay between loopback and measured signal of >3ms. Please have a look at the VituixCAD thread, link to start of relevant discussion is below:

VituixCAD v2 - HTGuide Forum

For information, VituixCAD speaker simulator has been updated from version 1.1 to 2.0 with the following additions and changes (so far): Crossover schematic changed from formal block ladder to free form supporting unconventional topologies. Predefined passive blocks replaced with individual...

www.htguide.com

 

[JoinUs]

@John Mulcahy
Sometimes, This issue occurs when i try to save using the capture function in Spectrogram or Waterfall. (Other graphs are certain, but I don't think I experienced any errors.)

I copied the error image and the error details in the notepad and attached it.

 

[John Mulcahy]

Thanks, I have fixed that for the next build. In the meantime just untick "Include title".

 

[John Mulcahy]

Builds updated with these changes:

• Changed: Merging cal data into the IR takes account of the "Limit cal boost to 20 dB" Analysis preference setting
• Fixed: Merging cal with IR when using loopback as cal ref would give incorrect IR position if the left window was shorter than the measurement delay
• Fixed: Capturing an image of the waterfall or spectrogram graphs with "Include title" selected would cause an IllegalArgumentException

 

[dcibel]

Builds updated with these changes:

• Changed: Merging cal data into the IR takes account of the "Limit cal boost to 20 dB" Analysis preference setting
• Fixed: Merging cal with IR when using loopback as cal ref would give incorrect IR position if the left window was shorter than the measurement delay
• Fixed: Capturing an image of the waterfall or spectrogram graphs with "Include title" selected would cause an IllegalArgumentException

Hi John,
There is still some problem with a pre-application of windowing to the measured impulse, when using "merge loopback response to IR" is used. I think this is most easily observed by simply measuring background noise (speaker disconnected) and then normalizing the IR level to show the high noise.

Here we have the impulse response of the entire measurement, if "make calibration data from loopback response" is used:

And the same measurement, if "merge loopback response to IR" is used. You can see, the default window as been pre-applied to the measured IR. This is detrimental if I change the default window settings to something like 4ms right window.

 

[John Mulcahy]

That is how it is intended to work. All arithmetic operations use the windowed IR.

 

[dcibel]

Arithmetic for "merge loopback response to IR" when measuring should use the entire measurement length, not whatever the default window setting is, please. I would like to apply the window post-measurement, or at least be able to adjust it if needed when using this measurement method.

 

[John Mulcahy]

I'll change it for the next build.

 

[John Mulcahy]

Builds updated today with these changes:

• Added: An indicator ("T") at the bottom of the impulse graph shows the position of the timing reference when the measurement was made (only for measurements made after this build)
• Added: Noise floor capture before a sweep measurement is now optional, controlled by a check box on the Measure dialog. Capturing the noise floor takes about 1.4 seconds.
• Changed: Minimum frequency for waterfalls and spectrograms is now the axis start frequency set in the View preferences rather than 10 Hz
• Changed: Merging cal data into the IR uses the whole IR rather than the windowed span
• Changed: If "continue anyway" is selected when REW prompts about the measurement input not having SPL calibration REW will not ask in future for that input device.
• Fixed: The signal generator waveform preview might not show a clipping warning when signal peaks exceeded full scale until the signal was played, though the level display would show peak above 0 dBFS
• Fixed: The sweep dither preamble started playing before noise floor capture had finished
• Fixed: Trace arithmetic, vector average and cross correlation alignment would fail if the responses were entirely in negative time

 

[dcibel]

Changed: Merging cal data into the IR uses the whole IR rather than the windowed span

Thank you John. I've only had 1 minute to test before I head to work this morning, but it looks very promising

 

[onlyoneme]

"Treat 32-bit data as 24-bit" seems to no longer work for me. I'm getting low level harmonics artifacts.

 

[John Mulcahy]

"Treat 32-bit data as 24-bit" seems to no longer work for me. I'm getting low level harmonics artifacts.

Can you provide a little more information? What is your test signal, what is the signal path, what does the spectrum look like, does the spectrum change if you change the setting?

 

[onlyoneme]

Can you provide a little more information? What is your test signal, what is the signal path, what does the spectrum look like, does the spectrum change if you change the setting?

Sure. 1k sine played over toslink to RME ADI-2, RTA capture via USB.

Before, with mentioned setting enabled:

Now, regardless of that setting:

 

[John Mulcahy]

What sample formats are shown at the bottom left of the REW main window?

 

[onlyoneme]

What sample formats are shown at the bottom left of the REW main window?

Int32L in and out.

 

[John Mulcahy]

Odd, not something I see using the ASIO Madiface driver and an ADI-2 Pro with its internal loopback. Which 5.20.14 build are you running and what dither is being applied to the generated signal?

 

[onlyoneme]

Odd, not something I see using the ASIO Madiface driver and an ADI-2 Pro with its internal loopback. Which 5.20.14 build are you running and what dither is being applied to the generated signal?

Latest one, ea25. I have 24 bit dither applied.

Edit: I'm getting different results when using an old wav file and the one generated by the current build with the same parameters.

 

[onlyoneme]

New file, regardless of setting mentioned before:

Old file with current build, regardless of setting:

Calculated numbers are also different.


And measurement with the old file on older build with "Treat 32-bit data as 24-bit":

 

[John Mulcahy]

Just seen your edit, you are seeing differences with wav files generated from different versions rather than live data. That will likely be a result of this fix in ea16:

Fixed: WAV files dropped on the RTA could show artefacts at around -162 dB

There was an error in the way received data was scaled that could cause quantisation artefacts. It wouldn't occur with REW's own files, as there was a complementary error on the output side, but could occur with files generated by other applications.

 

[onlyoneme]

Just seen your edit, you are seeing differences with wav files generated from different versions rather than live data. That will likely be a result of this fix in ea16:

Fixed: WAV files dropped on the RTA could show artefacts at around -162 dB

There was an error in the way received data was scaled that could cause quantisation artefacts. It wouldn't occur with REW's own files, as there was a complementary error on the output side, but could occur with files generated by other applications.

Hmmm... but those older files were generated by REW, ver. 5.20.13 probably. And it wasn't loaded into RTA, it was a live measurement using the file generated by the older build.
And what about "Treat 32-bit data as 24-bit" setting with the current build? How it is supposed to work? Previously it could handle those strange artifacts.

 

[sm52]

John, hello. On the ALL SPL tab, I need to move one of the graphs up or down. I can only do this if the vertical axis is SPL. With other units of measurement, the numbers in the Measurement actions window change, but the graph selected in the same window does not move. That is how it should be?

 

[onlyoneme]

Live measurements with j-test file also look different.

Old build, old file:

New build, old file:

New build, new file:

Should I generate all my test files again with the current build?