spotify quality vs "hi-fi" lossless options, i cant tell a difference.

[antcollinet]

This guy is a rather talented Studio Engineer, that I have 'followed' for some time.
He does some great 'Honest' reviews of real Hardware, Plug-Ins and stuff and this is an iinteresting comparison

He is not listening blind - which basically invalidates his listening comparisons.

I'm also not convinced by his null test. When comparing two files with a null test it is not sufficient to simply invert one. If there is even a tiny timing difference in the two files, it will cause a phase mismatch at the higher frequencies that cause a big difference in the null that is not actually there in the music. It seems to me his high frequency differences in his null comparisons might be mostly - exactly that.

Our member @pkane 's "Deltawave" goes to great lengths to eliminate this issue by time aligning and phase aligning the two files - even accounting for clock drift in the two files. You won't get this in a simple invert of the waveform.

DeltaWave Audio Null Comparator | DeltaWave documentation

deltaw.org

 

[Ze Frog]

A real hifi streaming service would select and curate their library manually ...

And that's the problem, profit comes well and above all else. Only way to curate your library in a way to achieve knowing you have the best is to do it yourself. It's kinda funny really, 'audiophiles' like to think for example class D or extra conversion is sapping realism or whatnot, all the while getting rid of all their music and placing on a harddrive or streaming. It's a whole hobby full of really odd contradictions.

As for A.I, that's likely the enemy of music going forward, although compared to a lot of stuff today it can't really get much worse. Still, digital and streaming isn't about the music, it's about profit and monopoly. Also gives great control over artists as well in a bad way, soon you will start seeing artists getting dropped for views they may have that don't fit with the streaming companies ideals.

 

[NiagaraPete]

Hi, I have tried it but I haven't tested it. For my comparison, I was using the regular Apple Music app.

I don't listen to classical unless its mixed with something progressive. But I'd heard Apple bought some top notch Classical collection and that they had rolled it out. I do use Apple music and for the most part I'm very happy. I've also used Tital and Spotify. Quite frankly I'd doubt anything but a measurement mic could actually hear a difference. The key is level matching.

 

[kemmler3D]

On the contrary I would be very surprised if they lie about the bitrate. Could you imagine when it get's out? -That the largest streaming service lie to their customers?

Well, you are right, that's a pretty big accusation. But, the quality control I can find in public documentation for Spotify uploads is this: https://support.spotify.com/us/artists/article/audio-file-formats/

They simply require you to upload WAV or FLAC. I imagine there is no quality control whatsoever on the contents of those WAV or FLAC files. So if you happen to be a lazy or incompetent distributor, you could end up converting a lower bitrate MP3 to FLAC which is in turn converted back to OGG.

Anyway, probably doesn't happen often... just speculating as to how a 320 OGG (or 265) could measure and sound non-transparent.

 

[Deleted member 48726]

Well, you are right, that's a pretty big accusation. But, the quality control I can find in public documentation for Spotify uploads is this: https://support.spotify.com/us/artists/article/audio-file-formats/

They simply require you to upload WAV or FLAC. I imagine there is no quality control whatsoever on the contents of those WAV or FLAC files. So if you happen to be a lazy or incompetent distributor, you could end up converting a lower bitrate MP3 to FLAC which is in turn converted back to OGG.

Anyway, probably doesn't happen often... just speculating as to how a 320 OGG (or 265) could measure and sound non-transparent.

I found that "guide" as well after posting. But then thought that it probably doesn't happen much. Why would it? What would the benefit of the artist/mixer be to upload their tracks in ****** quality? I think maybe this is a situation where self government works.

But it could happen, it seems.

 

[kemmler3D]

What would the benefit of the artist/mixer be to upload their tracks in ****** quality? I think maybe this is a situation where self government works.

Generally artists aren't uploading their own stuff, it goes through a distributor of some kind. Most artists are as superstitious about bitrate and stuff as anyone, but if someone responsible for uploading finds themselves without the original files, they might just upload some crappy transcode anyway.

Also, not all artists are very technically adept and might just mess up the process at some point.

And labels themselves have made some pretty egregious mistakes when it comes to masters over the years.

I really wouldn't expect this to happen often, but who knows?

 

[stoo23]

If there is even a tiny timing difference in the two files, it will cause a phase mismatch at the higher frequencies that cause a big difference in the null that is not actually there in the music

If that were the case, he would NOT get a complete NULL in Any of his examples !!
It has to be one or the other.

I could do this Way back myself, in the Studio, coming from Tape back in Analogue days.

 

[antcollinet]

If that were the case, he would NOT get a complete NULL in Any of his examples !!
It has to be one or the other.

I could do this Way back myself, in the Studio, coming from Tape back in Analogue days.

He would if the files he are comparing are from an identical source - which I can only assume is the case with the complete Null. If the files are different in any way, you won't get a -infinity null.

 

[stoo23]

Well then Surely, that is the Whole 'Point'.
He is comparing the Original, (which He Recorded & Mastered), with the 'Streaming' version !!
Some cancel perfectly, some Don't !!

To suggest there would be 'timing and/or phase differences Only on the examples that Don't Null completely is perhaps too convenient, when the other examples would Not seem to suffer from that and Do Null completely.

Anyway, I don't wish to argue or care whether you believe my thoughts or his tests, as stated, I thought it was simply an interesting comparison and knowing his 'Technical' ability, electronically etc, figured there Was some 'Merit' in his comparisons !!

 

[antcollinet]

Anyway, I don't wish to argue or care whether you believe my thoughts or his tests, as stated, I thought it was simply an interesting comparison and knowing his 'Technical' ability, electronically etc, figured there Was some 'Merit' in his comparisons !!

Fair enough.

I, on the other hand do not.

 

[Palladium]

Lossy compression can be very good. I've never done any ABX tests but I have an iPod full of MP3s ("V0") and every time I thought I was hearing a compression artifact, the CD has turned-out to have the same "defect".

I have heard some lousy sounding low-bitrate MP3s. I chose V0 because it's the "best" variable bitrate setting and it seemed good enough. I could probably get-by with more compression but I have enough storage and I don't need to squeeze-down the file size.

I probably should have made a FLAC archive but when I started ripping CDs for the iPod
I don't think I had the extra disk space. Nowadays there's usually no need for lossy compression unless you are streaming and trying to save Internet bandwidth.


I don't think that Spotify uses MP3. Some of the "newer" compression schemes are supposed to be better but it also depends on the amount of compression (bitrate) and once you can't tell the difference between the lossy copy and the original you can't say that another format is "better".

From what I read from Hydrogenaudio, 192kbps VBR Opus is completely transparent from loseless in ABX.

The ironic part is their listening sample suite that gives current lossy codecs the most trouble is electronic music, which is like the furthest genre you would associate audiophiles with.

 

[Mnyb]

From what I read from Hydrogenaudio, 192kbps VBR Opus is completely transparent from loseless in ABX.

The ironic part is their listening sample suite that gives current lossy codecs the most trouble is electronic music, which is like the furthest genre you would associate audiophiles with.

It’s a bit contradictory, for example in the beginning of codecs Suzanne Vega’s Toms Dinner record the song with an acapella solo voice in a dead acoustics environment gave codecs grief

Large orchestra is alllegedly easy to convert to lossy , it’s almost self masking as there is so much going on simultaneously and there is always some acoustic echoing all the time . And the lush soft spectrum of it ?

I can figure that some electronica has very high levels of near 20kHz content and sub 20Hz content

 

[Palladium]

Don't visit this site I you want to get good at recognizing lossy artifacts: https://web.archive.org/web/20111005054145/http://ff123.net/training/training.html

So according to this page, Xing 128kbps on castanets is supposed to be immediately obvious, and Xing is also supposedly to be a really bad MP3 encoder.
Unfortunately the MP3 link in the archived page was dead, so I downloaded and installed the 1999 version of Xing encoder to check for myself.
To my surprise I couldn't tell reliably tell a difference between the encoded 128kbps and the original FLAC/WAV castanet sample from Hydrogenaudio at a comfortable ~70dB SPL. Spek definitely puts the MP3 brickwalled at 15KHz though.

 

[Eric Natural]

I find myself so relieved that the extra lossless resolution I'm led to believe exists - doesn't.
Very happy to have found this forum/group of rational music lovers to push back the nonsense I've read elsewhere.
Thanks !

 

[danadam]

So according to this page, Xing 128kbps on castanets is supposed to be immediately obvious, and Xing is also supposedly to be a really bad MP3 encoder.
Unfortunately the MP3 link in the archived page was dead, so ...

This one: https://web.archive.org/web/20111005054145/http://ff123.net/training/castanets_xing128.mp3 ?
Works for me. Here, I put it in the attachment.

The link to the original doesn't work actually, but for that it's not necessary to use archive.org: https://lame.sourceforge.io/download/samples/

 

[Waxx]

I hear a difference, and did a blind test (someone else in an other room was playing the music, we just saw the speakers and the amp), but i can enjoy higher bitrate lossy music also. Lossy encoding came a long way, and on a lot of systems that are used a lot, you and i won't hear the difference. I hear mainly the difference on higher volume, and with music that is not overly compressed. With modern pop, i sometimes hear it when the music has very deep bass, but mostly not.

But i stil prefer lossless, so i collect my music in lossless format. I don't stream much, and always on free streams (youtube) that are lossy. But when i listen to a quality system on higher volume, the difference is obvious for me. If it isn't for you, that's good, you can use lossy for me. I also still enjoy vinyl (that is lossy also, but in analog format) so i can understand it does not matter for you, but it does for me in critical listening situations or on high volume (dj's and so).

Btw, most radio stations, even over FM, stream in lossy formats. The radio i broadcast myself on (Urgent.fm, the universaty radio of the Universaty of Ghent) use 256mps AAC streams to go to their antenna system (that is in a different location than their studio), and i listen also to it. In Belgium all stream in lossy format but Klara and Musiq'3, both radio stations from the public broadcast company (VRT/RTBF) that only play classical and jazz music. And they do that because the listeners complained that the sound quality was gone when they tried lossy streams in the past.

 

[Ambient384]

From what I read from Hydrogenaudio, 192kbps VBR Opus is completely transparent from loseless in ABX.

The ironic part is their listening sample suite that gives current lossy codecs the most trouble is electronic music, which is like the furthest genre you would associate audiophiles with.

You mean MDCT codecs since Subband ones like Musepack/MP2 lack that issue. Folk there were laughing that MPC at ~170kbps was transparent on a old killer sample that AAC/MP3/Vorbis struggled on. It seems like in heavy Electronic samples Musepack behaves more like Faux-Lossless codec since It leaves stuff It can't compress at 450 ~ 1411kbps.

If musepack somehow became popular and was used at 256kbps VBR we wouldn't be here arguing what lossy codec the best, But considering how Musepack & Opus are transparent at 192kbps, It like that Two strong Guys beating up box Meme for AAC/Vorbis/MP3.

 

[pablolie]

Bandwidth and storage are so cheap these days that I fail to understand why we're trying to save a few kbit/s here or there. OTOH I think 24/192 is wasteful.

 

[Zapper]

Bandwidth and storage are so cheap these days that I fail to understand why we're trying to save a few kbit/s here or there. OTOH I think 24/192 is wasteful.

Cheap isn't so cheap when you have a catalog as big as Spotify's. A lack of profitability doesn't help. Even in tech there's a limit on how much of other people's money you can spend before you have to earn some of your own.

 

[pablolie]

Cheap isn't so cheap when you have a catalog as big as Spotify's. A lack of profitability doesn't help. Even in tech there's a limit on how much of other people's money you can spend before you have to earn some of your own.

Last I checked, Spotify turned a profit last year, though. I haven't checked what percentage of their cost network bandwidth is, but I'd bet their bigger cost factor is content rights.

Content distribution over a network is how companies like Akamai make their bucks. It's not a linear algorithm between end user bandwidth vs the bandwidth Spotify buys from the BTs, ATTs, Verixons etc on the planet. A lot of it is cached.

In any case, how Spotify gets their business done is of no interest to me, as long as I get my 320k stream for 9.99 a month I am happy.