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: How to read this spectrumgram ? How and Why happen ? Is it HARMFUL? (Read 1523 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

How to read this spectrumgram ? How and Why happen ? Is it HARMFUL?

Recently I'm collecting my favorute songs. Nowadays, with the stream platforms grow, hires audios are more easy to collect than lossy ver. But the many platform may add audio watermark into the audio, so I prefer Mora/HIGHRESAUDIO than other platform(Qobuz/AmazonMusic...).

But here  a song from Qobuz and Mora confused me. See the spectrumgram below:

This is the song from Qobuz:


The same song from Mora:


Some background:
This song [もう少しだけ] released at 2021-05-10. The artist is Yoasobi, from Japan Sony Entertainment. It's a top popular group in recent years.
Mora is a web music store of Sony(of subsidiary).

my question is :
1. Why the spectrumgram of Mora looks so unnatrual?  Do they upscale or somewhat "remake" with the origin audio which is not hires?

2. If Sony offer a same nonhires audio source to Qobuz and Mora , is it mean the "remake" method of Qobuz is better than Mora?

3. If Sony did the "remake" with the audio to make hires, and then offer Qobuz/Mora, is it mean Qobuz did sth  to make  audio more natrual?

4. The last common sense question. I have seen so many spectrum like Mora ver. above from hires audio. Are those hires audio made from nonhires by deliberately? Did these "remake" work harm to the audio quality?


Sorry for my bad English and  barren knowledge about audio. I was confused for which ver. lossless audio should be collect.

Re: How to read this spectrumgram ? How and Why happen ? Is it HARMFUL?

Reply #1
I don't know how they were processed but the differences appear to be above the audible range.

There may be other differences that you can hear and in that case it's up to you to decide which one you prefer (by listening).

The spectrogram is NOT the best way to judge audio quality.  You can have a "pretty spectrogram" that sounds bad.   It's probably easier to make a nice looking spectrogram than to make a good-sounding file that may, or may not, have a good-looking spectrogram.    Sometimes the spectrogram MIGHT help with diagnosing why it sounds bad (or why it sounds good).

Re: How to read this spectrumgram ? How and Why happen ? Is it HARMFUL?

Reply #2
You're right, the quality deciding by listening  in the end.
Because I cant hear their difference, so I convert them into opus on local device for listening. The hires one was keep on cloud as collection.
In fact, by reading many article on this platform, I have keep listening test as a quality test method in my mind.

Now what I wonder is,
How this spectrumgram comes from? By some "remake" like 44.1/48 upsample to 96?
what's the effect behind the "remake" theoretically?
The forum as I saw was so professional, I believe friends here must know the answer.

The other question are optional, might get answered if lucky.

Re: How to read this spectrumgram ? How and Why happen ? Is it HARMFUL?

Reply #3
How this spectrumgram comes from? By some "remake" like 44.1/48 upsample to 96?
That "line" appears to be at 24kHz, so it could be something to do with resampling at 48kHz without adequate prior bandwidth limiting to ensure the sampler is not presented with frequencies in excess of 24kHz.

If that bandwidth limitation is not respected, the effect of sampling is to mirror frequency components which are over half the sampling frequency back into range.  For example: with 48kHz sampling, 0-24kHz input frequencies are reproduced as-is, but 24-48kHz become 24-0kHz.  A 25kHz frequency component ends up as 23kHz in the output.

There is a superb article here:
I wouldn't worry about that.
It's your privilege to disagree, but that doesn't make you right and me wrong.

Re: How to read this spectrumgram ? How and Why happen ? Is it HARMFUL?

Reply #4
That "line" appears to be at 24kHz, so it could be something to do with resampling at 48kHz

Thanks for your explanation, it's clear!
But what's that thing ?
As I  strongly suspect it's resample from a CD(44.1khz) audio content.
Because its purpose is to make [24/96 hires] audio,
if the origin one is 192, the "line" shouldn't be at 20-24khz,
if the origin one is 96,  no need for "remake"(unless that's the origin form and qobuz clean that line....Not possible at all)
if the origin one is 48, no need to resample to 48 to make a "line" there, as you said,
if the origin one is 44.1, that makes senese. Mora firstly upsample to 48, then double to 96. 
(so 44.1 resample to 48 will make such a "line"?By some software? This remind me mora said they use k2hd  )
My guess need your professionals confirm, pls.

By the way, thanks for your link from xiph.org.
In fact, I have read many contents from hires argument , many experiments. But I have no digital  audio background, so
mostly I only read the summon part and experiment conclusion parts.

---------
mora use k2hd  in old songs old songs
but this song didn't mention on web page this song

-------
here is a k2hd mora spectrumgram from old songs

It's different from the song above. The spectrumgram doesnt have a  clear line .

Re: How to read this spectrumgram ? How and Why happen ? Is it HARMFUL?

Reply #5
If no sound difference, it doesnt matter. But if i was forced to decide i'd take the version on qobuz. It does fit my definition of lossless audio a little more:
- energy is distributed over the whole spectrum
- energy has a rolloff towards the high frequencies (not like the mirroring around 22 khz on mora)
- no visible artefacts (like that shelf on mora)

That is very strange indeed. My guess is that it started as a 44.1 khz file, then got resampled to 48 khz and then to 96 khz.
 
Edit:
Quote
if the origin one is 44.1, that makes senese. Mora firstly upsample to 48, then double to 96.

Oh wow i didnt even realise we had the same thought before i typed this. :D

Re: How to read this spectrumgram ? How and Why happen ? Is it HARMFUL?

Reply #6
As Octocontrabass's linked article makes clear, it is disastrous to sample any signal which has not been sanitised by a bandwidth limiting filter to cut out frequency components greater than half the sample frequency.  There is no point whatsoever resampling a signal which has previously been sampled at 44.1kHz, no improvement is necessary and resampling at a higher frequency will not improve it.  A 48kHz file will only be better if the original analogue recording was passed through a 24kHz filter on the way to a 48kHz sampler (or a 96kHz filter on the way to a 192kHz sampler and then DSP'd to 48kHz).

The 20kHz+ components might well be inaudible, but they occupy bit rate and they have the possibility of becoming audible as intermodulation, according to the characteristics of the reproduction equipment.  All this is thoroughly spelled out in the article mentioned above.

It's different from the song above. The spectrumgram doesnt have a  clear line .
It does, at 21kHz.  It looks like that file has been properly mastered, with a 21kHz filter protecting the sample frequency.
It's your privilege to disagree, but that doesn't make you right and me wrong.

Re: How to read this spectrumgram ? How and Why happen ? Is it HARMFUL?

Reply #7
Thx you guys, many thanks.
I'll make a summary about the theory/tenique of the strange spectrumgram in my word later, to make sure I dont misunderstand.