Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: (Free) Software to analyze HD and DSD tracks? (Read 5159 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

(Free) Software to analyze HD and DSD tracks?

Hi,

I performed countless blind tests on myself and finally gave up (re-)buying more expensive HD tracks. To save some space, I want to convert some of my 1-3 GB albums to 44.1 or 48/16. For that I want to uncover all the badly done masters and cut off all the information that is only noise. (Backup the original downloads on an external drive and forget about them.)

I guess the best way to do that is to use a spectral analyzer. Then, look at the information that is happening above 22 kHz and try to see whether it's actual audio (should be a downward slope) or noise (non-dynamic and constant peaks).

To some extent, Audacity is sufficient. But it limits the analysis to 109 seconds and I noticed that the beginning often has less reach so I have to search the file for the part with the most amount of data. Also, it does not read DSD.
BTW, I'm on a Mac so unfortunately a foobar2k plugin wouldn't help me.

If I got anything wrong, please let me know. I am willing to learn!
Thanks!
headphone aficionado, tech fan, realist
 

Re: (Free) Software to analyze HD and DSD tracks?

Reply #1 –
You can use ffmpeg to convert to PCM and then analyze PCM as usual.
Also why do you need anything above 22 kHz? Do you plan slowing them down?
If not, you can simply convert them all to 44100 Hz or 48000 Hz without even looking.
It's 100% okay for good records to have a cutoff at ~20 kHz if they aren't intended for further sampling/creative use, because ultrasonic content can only make the sound worse if there are nonlinear distortions somewhere in the playback pipeline, and otherwise it's just extra wasted power in amplifiers/etc.

…just make sure you aren't introducing clipping when resampling. AFAIK if you do the resampling with sox, it will print a warning if the maximum sample level was exceeded.
a fan of AutoEq + Meier Crossfeed

Re: (Free) Software to analyze HD and DSD tracks?

Reply #2 –
Before I move on, let's see if I am doing this correctly.

So I took a 192/24 track and loaded it with Audacity.


First thing to notice is the big noise wall on the right. This should be the typical noise-shaping of DSD. Or is it normal PCM dithering? Anyway, it's not useful information, it should have been cut off.
Then there is a surprising peak at 15 kHz. Listening to the track, there are no high-pitched instruments (drums, contra bass, banjo, tuba, bari sax) so it has to be hiss from the recording. (I also don't know any instrument that would peak that high.) The same goes for all subsequent peaks.
If I increase the resolution, there is more noise from 0 to 30 Hz (peaks at 12 Hz) at -49 dB, which is just 22 dBu apart from the highest SPL. Not sure why one need 24 bit resolution for that kind of information...

Source: Dr. Chesky's Ultimate Headphone Demonstration Disc
headphone aficionado, tech fan, realist

Re: (Free) Software to analyze HD and DSD tracks?

Reply #3 –
It would be more revealing to look at the spectrogram which is 3 dimension graph (time, frequency, energy). This image doesn't show how it changes over time.
But given the "hill" at the right side on this image I can tell it's a conversion from DSD with high confidence.
Quote
is it normal PCM dithering?
if there are more than 20 bits and you see it on the spectrogram without any amplification then it's not normal PCM dithering.
it's either DSD noise leftovers, or just a noisy record.
Quote
Then there is a surprising peak at 15 kHz
I bet this is just record noise.
a fan of AutoEq + Meier Crossfeed

Re: (Free) Software to analyze HD and DSD tracks?

Reply #4 –
Thanks!
Quote from: magicgoose on
It would be more revealing to look at the spectrogram which is 3 dimension graph (time, frequency, energy). This image doesn't show how it changes over time.
Which brings me back to my first question. Which Software would let me do that? All I want is to analyze my audio, no editing necessary.
headphone aficionado, tech fan, realist

Re: (Free) Software to analyze HD and DSD tracks?

Reply #6 –
Audacity can do it too. Also sox (command line).
a fan of AutoEq + Meier Crossfeed

Re: (Free) Software to analyze HD and DSD tracks?

Reply #7 –
Audacity will do it. Load a track, click on its title bar and select "Spectrogram" from the drop-down list.
Regards,
   Don Hills
"People hear what they see." - Doris Day

Re: (Free) Software to analyze HD and DSD tracks?

Reply #8 –
Quote from: Roseval on
A couple of spectrum analyzers can be found here: http://www.thewelltemperedcomputer.com/SW/AudioTools/Spectrum.htm
Thank you. I chose XiVero MusicScope and it seems to be a very good piece of software. Only problem I have is that it doesn't read AAC (yet). Anyway, I was mostly interested in the HD files and it does that well!
headphone aficionado, tech fan, realist

Re: (Free) Software to analyze HD and DSD tracks?

Reply #9 –
Quote from: 16Bit Audiophile on
Before I move on, let's see if I am doing this correctly.

So I took a 192/24 track and loaded it with Audacity.

First thing to notice is the big noise wall on the right. This should be the typical noise-shaping of DSD. Or is it normal PCM dithering? Anyway, it's not useful information, it should have been cut off.
Then there is a surprising peak at 15 kHz. Listening to the track, there are no high-pitched instruments (drums, contra bass, banjo, tuba, bari sax) so it has to be hiss from the recording. (I also don't know any instrument that would peak that high.) The same goes for all subsequent peaks.
If I increase the resolution, there is more noise from 0 to 30 Hz (peaks at 12 Hz) at -49 dB, which is just 22 dBu apart from the highest SPL. Not sure why one need 24 bit resolution for that kind of information...

Source: Dr. Chesky's Ultimate Headphone Demonstration Disc

The 15KHz peak could also be the result of CRT's present in the range of the microphone. Even at quite long distances these CRT noises come through. Especially in older recordings where CRT's were mainly/only used instead of flatscreens.

Re: (Free) Software to analyze HD and DSD tracks?

Reply #11 –
Quote from: Fairy on
Quote from: 16Bit Audiophile on
Before I move on, let's see if I am doing this correctly.

So I took a 192/24 track and loaded it with Audacity.

First thing to notice is the big noise wall on the right. This should be the typical noise-shaping of DSD. Or is it normal PCM dithering? Anyway, it's not useful information, it should have been cut off.
Then there is a surprising peak at 15 kHz. Listening to the track, there are no high-pitched instruments (drums, contra bass, banjo, tuba, bari sax) so it has to be hiss from the recording. (I also don't know any instrument that would peak that high.) The same goes for all subsequent peaks.
If I increase the resolution, there is more noise from 0 to 30 Hz (peaks at 12 Hz) at -49 dB, which is just 22 dBu apart from the highest SPL. Not sure why one need 24 bit resolution for that kind of information...

Source: Dr. Chesky's Ultimate Headphone Demonstration Disc

The 15KHz peak could also be the result of CRT's present in the range of the microphone. Even at quite long distances these CRT noises come through. Especially in older recordings where CRT's were mainly/only used instead of flatscreens.
Then it would be a clean tone (a sharp peak at 15625 Hz or nearby), while on the OP's image it doesn't look like a clean tone.
But yes, there is such a thing in general.
a fan of AutoEq + Meier Crossfeed

Re: (Free) Software to analyze HD and DSD tracks?

Reply #12 –
This is a few years old but wanted to add some info since I've just been reading up on it -

Ultrasonic noise is a side effect of the way the DSD format works. In simple terms the signal is an unencoded 2.8MHz (or multiples thereof) square wave that increases in density according to the peak value of the signal wave. The way this is decoded is roughly just to low pass filter it, but even with the extremely high frequency of the square wave/sample rate some harmonic noise is present in or near the audible spectrum, so noise shaping is used to shift this further outside the audible range. This results in the very high frequency peak as you displayed in your spectral analysis plot.

My rough understanding is that higher rates of DSD (128fs/x and above) incur less of this effect, but that may simply mean their noise peak is significantly further beyond the audible range.

Hope that helps!