Acoustica Premium 7.0.51 - Analysis > Spectrum

[Paulo-Brazil]

Mr. Stian,

I refer to Acoustica Premium 7.0.51, running on Windows 10 Professional 64-bit.

I am trying to compare two (or three) files in relation to tonal characteristics.

It's is very difficult to use the spectrum analysis (menu Analysis > Spectrum) to make such a comparison. We cannot superimpose two spectra, because they have solid (non-transparent) backgrounds. Looking at the spectra side by side is not possible, for although I have a large computer screen, I cannot reduce the size of the spectrum windows sufficiently to place them one right above the other, so that they share the same horizontal (frequency) axis, for comparison. Anyway, it would be difficult to properly compare two spectra that way, looking at different windows. As things stand now, it seems that the spectrum is of little practical use for actually comparing two (or more) files.

I think that some way should be devised in Acoustica to put two (or more) spectra, in different labelled colours, on one same window, under the same set of coordinate axes. This would make comparisons possible in a proper way. Maybe, a choice could be somehow offered to the user: either each spectrum on its own window, or all spectra superimposed on one same window, for comparison purposes.

Also, maybe someday, the spectrum of one file could be put, in a special colour, on the window of the Equalize module opened for a second file, so that the spectrum of the first file could serve as a reference in the equalization of the second file.

About the spectrum itself: a key characteristic of the spectrum is the degree of smoothness. Too little smoothing, and we have a spectrum too full of ups and downs, where the general pattern across frequencies can be difficult to discern. The degree of smoothness seems to be controlled mainly by the (Settings) Block Size, but also by the Resolution, but this is not stated in the Help file. The very idea of smoothing the spectrum, an important one, is not presented in the Help file.

And something seems strange in the definition of Block Size found in the Help file, in the second sentence, marked with asterisks:

"Block size (samples): The duration of the time slices that are used create the spectrum (see Time and Frequency Domains for more information) in samples. ***The spectra from the analyzed time sliced are averaged to create the shown spectrum***."

The point is, for a given (fixed) Resolution, the spectrum, as things stand now, is actually the more smoothed the smaller the Block Size is. In the calculations, it seems to me, if we were to average something over smaller blocks, we should expect to find a wilder and more variable average across blocks. That is, the smaller the Block size, we should expect a less smoothed spectrum, not a more smoothed one as we actually find now. The present definition of Block Size in the Help file does not seem to agree with the observed relation Block Size x degree of smoothness, for a given Resolution . Something seems to be amiss here. Or am I misunderstanding something?

Also, this is not clear to me: In the Spectrum, the vertical axis is in units of "Spectral level (dB)", while in the Spectral Histogram the vertical axis is in units of "Intensity (dB)". Shouldn't the two vertical axes have the same units?

Regards,

Paulo

[stian]

Hi Paulo,

It's is very difficult to use the spectrum analysis (menu Analysis > Spectrum) to make such a comparison. We cannot superimpose two spectra, because they have solid (non-transparent) backgrounds. Looking at the spectra side by side is not possible, for although I have a large computer screen, I cannot reduce the size of the spectrum windows sufficiently to place them one right above the other, so that they share the same horizontal (frequency) axis, for comparison. Anyway, it would be difficult to properly compare two spectra that way, looking at different windows. As things stand now, it seems that the spectrum is of little practical use for actually comparing two (or more) files.

I think that some way should be devised in Acoustica to put two (or more) spectra, in different labelled colours, on one same window, under the same set of coordinate axes. This would make comparisons possible in a proper way. Maybe, a choice could be somehow offered to the user: either each spectrum on its own window, or all spectra superimposed on one same window, for comparison purposes.<

Thanks, that's a good suggestion and we've received similar requests before. It's already on the "to do" list.

Also, maybe someday, the spectrum of one file could be put, in a special colour, on the window of the Equalize module opened for a second file, so that the spectrum of the first file could serve as a reference in the equalization of the second file.

Yes, we might add something along those lines in the future.

About the spectrum itself: a key characteristic of the spectrum is the degree of smoothness. Too little smoothing, and we have a spectrum too full of ups and downs, where the general pattern across frequencies can be difficult to discern. The degree of smoothness seems to be controlled mainly by the (Settings) Block Size, but also by the Resolution, but this is not stated in the Help file. The very idea of smoothing the spectrum, an important one, is not presented in the Help file.

The block size affects the spectral resolution vs. time resolution. This is a bit intricate to explain, but it's a side effect and the purpose isn't really smoothing, but to be able to find an optimum between spectral and time resolution (which depends on the audio material).

And something seems strange in the definition of Block Size found in the Help file, in the second sentence, marked with asterisks:

"Block size (samples): The duration of the time slices that are used create the spectrum (see Time and Frequency Domains for more information) in samples. ***The spectra from the analyzed time sliced are averaged to create the shown spectrum***."

The point is, for a given (fixed) Resolution, the spectrum, as things stand now, is actually the more smoothed the smaller the Block Size is. In the calculations, it seems to me, if we were to average something over smaller blocks, we should expect to find a wilder and more variable average across blocks. That is, the smaller the Block size, we should expect a less smoothed spectrum, not a more smoothed one as we actually find now. The present definition of Block Size in the Help file does not seem to agree with the observed relation Block Size x degree of smoothness, for a given Resolution . Something seems to be amiss here. Or am I misunderstanding something?

The help file is correct -- Acoustica slices up the audio into fixed blocks if a specified size. These are analyzed and the results averaged to display the spectrum. In general, smaller blocks will decrease spectral resolution -- so that part is also correct. Thorough understanding of the Fourier transform and it's applications is necessary for a complete understanding, which would unfortunately be outside the scope of the Acoustica user's manual.

Also, this is not clear to me: In the Spectrum, the vertical axis is in units of "Spectral level (dB)", while in the Spectral Histogram the vertical axis is in units of "Intensity (dB)". Shouldn't the two vertical axes have the same units?

Yes, you're right -- it would be better to call it "Spectral level" in both analyses. I've changed that in the internal code base already.

Best,
Stian