Noise in spectogram

[microcontrollerpro]

Hi,

This is a question about the 24bit and 16bit versions of a soundtrack having different spectrograms.

The soundtrack is the Xenoblade 2 game soundtrack. There were 3 editions of this soundtrack released: Type A (Limited USB edition - 24bit), Type B (Deluxe CD Edition) and Type C (Standard CD Edition). I have the Type B soundtrack and I was curious if there were any differences with the Type A soundtrack.

After getting the tracks for the USB edition I loaded them up tracks into a spectrum analyzer and found that many tracks in the Type B edition have noise in the spectrogram starting at around 18khz (photos are attached at the end of the post). This does not occur in the USB edition.

A couple of questions:

• Is this noise called high frequency noise or is there a more accurate term?
• My understanding is that music is mastered at 24 bit and then converted to 16 bit so there shouldn't be that many differences in the spectrogram. Would a poor conversion result in the noise seen at 16bit or is there another reason (can't imagine that this is done on purpose)?

Thanks,

Spoiler: Spectograms

Type A (Source: USB 24-bit files)	Type B (Source: CD)

 

[JSmith]

Would a poor conversion result in the noise seen at 16bit or is there another reason (can't imagine that this is done on purpose)?

Huge Noise Band Between 18KHz and 21KHz - Why?

I was checking if the albums I bought are affected by the UMG Watermark, luckily they are not. But compare to HR version, the CD version seems to have a huge noise band between 18KHz and 21KHz. I'm wondering why this is... I have seen this in other audio files as well. Google search returns...

www.audiosciencereview.com

Aggressive noise-shaping of the dithering used for downsampling from the HiRes version.

JSmith

 

[microcontrollerpro]

Thanks for the thread link that answers my questions.

 

[Christoph-ASR]

Can the noise came from the Nyquist-Frequency restriction from the different sample rates?

A little bit of advertising must be allowed, you can also try it with (my) freeware:

GitHub - Christoph-Lauer/Sonogram-Visible-Speech: A speech and sound anallysis tool.

A speech and sound anallysis tool. Contribute to Christoph-Lauer/Sonogram-Visible-Speech development by creating an account on GitHub.

github.com

--
Chris

 

[AnalogSteph]

You may want to fix the About for your "speech and sound alalysis tool", just saying.

I tried your tool with JDK 20 on my Windows 10 machine. It's been kind of a rough time so far.
1. I have been unable to get the "open remote" function for the example files to work. Says it "can recognize this Media File but cannot read out the samples from this Media File. Perhaps the Media File is Broken". (BTW, why do you capitalize Random Words? Rule of thumb for English, nothing gets capitalized without a good reason, e.g. start of sentence, name, or you're writing a heading. And why does "Media File" appear no less than 3 times in this brief error message?)
2. The local file picker (courtesy of Java?) isn't exactly fun to use. Sometimes directories must be opened by double click and Enter does not work, it can get into a state where it will not (re)open specific folders for love nor money (particularly after using Backspace to go up, it seems) and only an application restart helps, selecting a folder in the dropdown field will have it try and open it as a file, and on this German language version of Windows it is partially localized (with the OK button switching between "Okay" and "Öffnen" depending on what is currently selected). The dropdown field can also go out of sync with the file view, i.e. not display the folder the contents of which are currently shown.
3. It refuses to open specific files, saying "The JMF-Processor cannot convert this file to specific format !". One I tried was 96 kHz at 24 bits, maybe it's because of that? (BTW: Bitte nicht plenken.)
4. The Perspectogram function will just crash the app without comment, regardless of whether or not it previously complained about it being "very large". (Maybe a question of graphics adapter? UHD Graphics 750 courtesy of an i7-11700 here, with a fairly recent driver. GPU-Z says it supports OpenGL 4.6, OpenCL, DirectCompute, DirectML and Vulkan, and does not support CUDA, Raytracking or PhysX.)

 

[Christoph-ASR]

LONG ANSWER: You get it for free, and the errors affect one percent of all use cases. The core is, anyway, what distinguishes the program from others, the Wigner-Ville time-frequency analysis. The error with the perspectogram, as I called the "new" analysis which tries to cross the nyqist boundary by trying to approximate in higher dimensions, similar to the wavelets, is described on the website. Btw: you should read the error message, you can try to fix it with the VM memory parameter, which gives the VM more memory, as it is described in the readme.txt. All other bugs are also known, but I have no time to create issues on github.
SHORT: No Time

here the "new" time-frequency algorithm by the example of a blackbird.

CONVENTIONAL:

mit PERWOLL gewaschen: (pseudo smoothed wigner ville transformation)

--
Grüße lo unnen hin, no bayan, aus new york,
Chris