REW Beta Release Changes in V5.19 beta 7

[John Mulcahy]

There are quite a few changes in V5.19 beta 7, so I'll make some posts in this thread to explain them.

First up is the ability to make offline measurements. Some systems cannot be connected to a computer to capture their responses - they may not have input connections, or may not be accessible to someone with REW and the knowledge to use it. Offline measurements can handle that. As long as the system is capable of playing back a WAV file, and the system's output can be recorded to a WAV file, REW can generate measurement results. The steps are straightforward and are described in detail in the beta version help for offline measurements, but in essence the sweep signal is saved to a WAV file using REW's signal generator, played back on the system by whatever means it supports while recording the response, and then both files are imported to REW to generate the new measurement. It isn't necessary to be too precise in the timing of the recording, just start the recording, play back the sweep, then stop the recording. The sweep WAV file needs to be generated in V5.19 beta 7 or later, as the first few samples of the file contain information about the sweep that REW uses when generating the measurement. The import window supports drag and drop for the sweep file and the recordings, and multiple recordings can be dragged onto the window at once, producing a separate measurement for each.

 

[John Mulcahy]

Next up, a couple of changes for the Stepped Sine distortion measurements, which if you haven't come across can be found in the RTA window. Stepped Sine measurement is pretty time consuming as it is much, much slower than sweep measurement, but the payoff is much higher signal-to-noise ratios, revealing distortion details that aren't visible in sweep data.

For this beta version the sine frequencies that are generated have been adjusted so that they correspond to frequency bins within the chosen FFT length. That ensures that the fundamental and its harmonics lie exactly within individual FFT bins and don't suffer from scalloping loss, which is a term that describes the reduction in level that occurs when a frequency lies between bins.

The second change is to add an option to capture the full spectrum at each test frequency, which can then be plotted as a spectrogram or waterfall. This makes for big measurement files, but having the spectrum data means distortions other than the harmonics can be seen. As an example, the plot below shows the fundamental of the test frequencies as the red line running from bottom left to top right. The distortion harmonics are the faint blue lines that run parallel to fundamental. There is another faint pair of lines that arc up from the bottom of the plot at 1 kHz, fanning outwards - those are intermodulation products, in this case with the 1 kHz frame rate of the USB connection to the soundcard that was being measured.

The next image is more revealing. Although this second card has similar harmonic distortion levels to the one above, it suffers from drastically more intermodulation with the 1 kHz USB frame rate and harmonics of the rate and also has much higher noise levels, especially at low frequencies. The spectrum data that was captured shows a lot that isn't evident from just looking at harmonic distortion levels.

 

[John Mulcahy]

Another change is support for higher sample rates when using Java drivers. Up to now, the only way to get REW to run at sample rates other than 44.1k or 48k was to run on Windows and use ASIO drivers. With this version sample rates up to 96k can be selected across all the supported operating systems. There is an important caveat: although REW requests data at the selected sample rate, it is important to make sure the soundcard is configured to operate at that rate in the OS, otherwise the OS will resample between the selected rate and the rate at which the input or output device is actually running and the results will not show what you might expect.

 

[John Mulcahy]

It has always been possible to do some limited analysis of audio files in REW by using the impulse response import - REW doesn't know whether the imported file is really an impulse response, so with some fiddling frequency content could be shown. It was far from ideal, however. In this version a specific import option for audio files has been added (shortcut Ctrl+Shift+U). REW treats audio files differently, windowing them to show all the available content and generating waterfalls and spectrograms across the whole range of the data. Here is an example spectrogram of a sweep file which includes a timing reference. The plot takes advantage of another new option in this version, moving frequency to the vertical axis and having time run along the horizontal axis.

Just for comparison, here is the same plot with 3D enhancement active - the shading technique used for the 3D enhancement has been improved (and I turned off the grid):

 

[John Mulcahy]

The spectrogram has had several other improvements for this beta version.

• The characteristics of the frequency-dependent windowing have been modified to improve the time resolution of the wavelet spectrogram to make it more useful for driver time alignment
• Linear amplitude scale options have been added (percent of peak and percent of full scale), which are helpful for highlighting timing variations in wavelet mode, and another log scale for dBFS has been added which may come in handy for audio files
• The spectrogram time span controls have been revised to provide separate controls for the spans before and after the IR peak
• There is a Fill spectrogram floor option, when selected REW uses the bottom of the spectrogram colour scheme to fill areas that have levels below the floor of the graph
• The peak energy time is now a separate trace on the graph with time values shown in the graph legend
• The window type used for the spectrogram can be selected directly from the graph controls rather than having to change it in the Analysis preferences
• The Y axis zoom out button on the spectrogram graph automatically limits to the full span of the spectrogram
• There is an option to use colour bands in spectrogram rather than a graduated colour

Here is an example of the colour bands, on a 3D enhanced plot:

 

[John Mulcahy]

The waterfall has also had a few changes. It is now possible to have up to 501 slices in the waterfall - bear in mind that using more slices will increase memory use and the time taken to generate the plots. Waterfalls have also gained spectrogram-style colour schemes, rather than being plotted in the measurement colour. The line drawn over the plot that shows the cursor frequency has also been made easier to see.

 

[John Mulcahy]

There have been a number of smaller changes:

• When using ASIO automatically reload the ASIO driver after the sample rate is changed - this should get around errors that could otherwise occur when trying to measure after making a sample ate change
• A and C weighted input level figures are shown on the RTA below the main (unweighted) figure
• FBQ1000 has been added as the new name for the DSP1124P
• The room simulator allows room dimensions up to 35 m, in case you have a really big home theater
• I've put in some additional input level checks during SPL calibration to try and prevent incorrect calibration results
• Trace arithmetic treats inputs as incoherent if they do not both have phase data - this affects how they add
• I have reintroduced the file overwrite check on OS X as the OS no longer performs it (it was previously removed because there was one warning from REW then another from the OS)
• I have revised the default trace colours to try and provide a bit more variation amongst them
• There is a "separate traces" button for the All SPL graph (as there is for the other overlay graphs)
• Units labels have been adedd to the graph limits dialog
• Measurement notes no longer appear as a tooltip in the legend area of overlay graphs, they just got in the way
• The Measure menu entry has moved into the Tools menu, didn't really belong in the File menu
• The various import options have their own sub-menu in the file menu
• The graph cursor has been made easier to see, especially on high resolution screens
• The state of the measurement panel expand/collapse button is remembered for the next startup; the default setting on first installation for screens of 1200 pixels or less is collapsed
• The default for showing graph scroll bars on screens of 1200 pixels or less is now not to show them to provide more room for the graph, but they can be turned back on using the button above the graph

There have also been some bug fixes:

• 5.19 beta versions 4 to 6 were not compatible with Java 7 (they used a Java 8 method)
• The spectrogram contour interval was not remembered for the next startup
• Soundcard cal could fail to complete if the measurement sweep had been set to a length above 256k
• Traces with colours that changed along the trace were not rendered correctly on retina macs

And that's pretty much it. Now back to the ToDo list

 

[Sonnie]

This is all super outstanding ... nice work John

 

[John Mulcahy]

Known issues in V5.19 beta 7:

• Layout is tricky on smaller screens - use the collapse button above the measurement panels and turn off the graph scrollbars to make more space for the graph, improvements are coming in beta 8
• Import audio files doesn't work for mono WAV files, use stereo (the channel to import can be selected when loading the file)

 

[jrubins]

Very cool ! Is there any recommendation on how many seconds to run a sweep from 20hz to 20khz?
Thanks!

 

[John Mulcahy]

REW defines its sweeps by their length (in samples), a 256k sweep is typically suitable for room measurements at 44.1k or 48k.