Streaming Services Are Comparatively Different, subjectively…

[audio2design]

You don't have to draw a graph, but it would be great if you could explain how two-channel audio can have differing vertical soundstaging.

Once Spotify goes FLAC (very soon), It's game over for all the above mentioned.

Amazon was big in music Spotify or not. Qoboz has a good European following. Tidal is probably not going to last.

 

[Belker]

I have compared digital streams from Spotify Hifi, Qobuz, Tidal and Deezer with my own CD rips (dbpower with checksum verification). I run these through MasVis (Swedish Audio Society software to analyze mastering quality and more) to do a course check that masters are the same. They usually are, because I have selected albums that are a bit “obscure”, where risk of remasters are lower. Of course I don’t pick MQAs and HiRes albums. Then I run selected tracks through Delta Wave. Result so fas has been 100% bit identical, regardless of service.

To investigate if the player client is doing some funky DSP, I continued with a not so scientific test, but still. I used an iPhone 10 as platform for Tidal, Qobus and Spotify apps. All playing CD quality. I used Roon app as reference for the CD rip. I can’t get the digital stream out here, so I used a Focusrite Scarlett 2i2 to capture the analog signal. The AD converted tracks were run through DeltaWave. Again, as far as one can tell from analog rips, they were identical. Not bit perfect, of course, but close enough. No level difference. No apparent EQ.

This was done a few month ago, and I haven’t done any more research into this with different platforms, as I regard it as a non issue. With that said and done, there is still the issue of how different services differentiate themselves with different masters. If you like what HiRes or MQA is doing to the sound, go for it. Don’t expect Redbook CDDA, to sound different, though! On that note, analyzing and comparing Redbook and HiRes releases of the same album from Qobuz, I have seen a trend that new releases, originally made for HiRes, usually looks better. A bit more headroom and DR. The opposite can be said about re-issues of old albums, now in glorious HiRes. More mastering engineers should just sit on their hands….

 

[kyle_neuron]

Just a little fun I've thrown together if anyone fancies trying their luck...
https://neuron-av.com/webMUSHRA/?config=ASR_streaming_quality.yaml

This includes CD and 'high res' quality versions of a track from the Cachaito album, from Amazon, Tidal and Qobuz. I truncated each file to the same length of 3 minutes via sample trim using ffmpeg, so there's no re-encoding of any kind taking place. Of course, I haven't level matched the files, but that shouldn't be too much of an issue.

Just for verification, there's also a 320 kbps CBR MP3, and a 110-150 kbps VBR MP3 in there

Unfortunately, I don't have access to Apple Music Lossless in a way that makes it easy to 'obtain' the original source file without re-recording via cable connection.

I'm working a show tonight, but I'll share the results from the CSV file the web app creates sometime over the weekend.

 

[audio2design]

I have compared digital streams from Spotify Hifi, Qobuz, Tidal and Deezer with my own CD rips (dbpower with checksum verification). I run these through MasVis (Swedish Audio Society software to analyze mastering quality and more) to do a course check that masters are the same. They usually are, because I have selected albums that are a bit “obscure”, where risk of remasters are lower. Of course I don’t pick MQAs and HiRes albums. Then I run selected tracks through Delta Wave. Result so fas has been 100% bit identical, regardless of service.

Ditto, I have done formal, but not so formal comparisons of Amazon, Tidal (non MQA) and Qoboz and the results, on the same master, were identical using a home grown difference comparison tool. Even claims of audible watermarking appear to be overblown based on my testing.

 

[kyle_neuron]

I should mention that the order of samples is randomised on that test, so you’re welcome to take it multiple times to see if you’re consistent in your rankings.

 

[Headchef]

So, as I’m fed up with assessing the different streaming services, which are fundamentally just mechanisms of convenience and screw the artists I’m ditching the lot and just buying the music I like direct from Bandcamp, as .wav, then the artist/label get the money without the streamers cut, I get the unmolested file as they produced it and can listen to it should the net fall over.

 

[Headchef]

Just a little fun I've thrown together if anyone fancies trying their luck...
https://neuron-av.com/webMUSHRA/?config=ASR_streaming_quality.yaml

This includes CD and 'high res' quality versions of a track from the Cachaito album, from Amazon, Tidal and Qobuz. I truncated each file to the same length of 3 minutes via sample trim using ffmpeg, so there's no re-encoding of any kind taking place. Of course, I haven't level matched the files, but that shouldn't be too much of an issue.

Just for verification, there's also a 320 kbps CBR MP3, and a 110-150 kbps VBR MP3 in there

Unfortunately, I don't have access to Apple Music Lossless in a way that makes it easy to 'obtain' the original source file without re-recording via cable connection.

I'm working a show tonight, but I'll share the results from the CSV file the web app creates sometime over the weekend.

nice one, should be interesting, will give it a go

 

[audio2design]

So, as I’m fed up with assessing the different streaming services, which are fundamentally just mechanisms of convenience and screw the artists I’m ditching the lot and just buying the music I like direct from Bandcamp, as .wav, then the artist/label get the money without the streamers cut, I get the unmolested file as they produced it and can listen to it should the net fall over.

My offline cache only needs to connect to the web every 30 days. If the Internet is out for > 30 days, I have far bigger things to worry about then if I have access to my music.

 

[Headchef]

Just a little fun I've thrown together if anyone fancies trying their luck...
https://neuron-av.com/webMUSHRA/?config=ASR_streaming_quality.yaml

This includes CD and 'high res' quality versions of a track from the Cachaito album, from Amazon, Tidal and Qobuz. I truncated each file to the same length of 3 minutes via sample trim using ffmpeg, so there's no re-encoding of any kind taking place. Of course, I haven't level matched the files, but that shouldn't be too much of an issue.

Just for verification, there's also a 320 kbps CBR MP3, and a 110-150 kbps VBR MP3 in there

Unfortunately, I don't have access to Apple Music Lossless in a way that makes it easy to 'obtain' the original source file without re-recording via cable connection.

I'm working a show tonight, but I'll share the results from the CSV file the web app creates sometime over the weekend.

I think there difference I heard were mostly down to volume levels,

Your results:
cachaito (MUSHRA)
C3: 100
C1: 89
C6: 97
C4: 100
C2: 86
reference: 0
C5: 100

C1 & C2 seemed to have slightly less bass definition but again, this could be down to volume?
one of them didn't play?!? (I guess that was the reference one?)

but C3 & C4 seemed (louder?) to have a little more in the hf/lf range.

Probably not the best that I've been cooking under a loud extraction fan for the best part of the day, be interesting to try it again after a nights sleep and a morning off.

 

[kyle_neuron]

If you'd like to have another go, then please do so. For when you're curious, the condition correlations were as follows:

Spoiler: Conditions

reference: conversacion_amazon_uhd.flac
stimuli:
C1: conversacion_amazon_cd.flac
C2: conversacion_320cbr.mp3
C3: conversacion_qobuz_cd.flac
C4: conversacion_tidal_hifi.flac
C5: conversacion_tidal_master.flac
C6: conversacion_128vbr.mp3

The reference track is the one that plays by default, it has its own Play button but is also in there again for checking.

 

[Headchef]

If you'd like to have another go, then please do so. For when you're curious, the condition correlations were as follows:

Spoiler: Conditions

reference: conversacion_amazon_uhd.flac
stimuli:
C1: conversacion_amazon_cd.flac
C2: conversacion_320cbr.mp3
C3: conversacion_qobuz_cd.flac
C4: conversacion_tidal_hifi.flac
C5: conversacion_tidal_master.flac
C6: conversacion_128vbr.mp3

The reference track is the one that plays by default, it has its own Play button but is also in there again for checking.

thats interesting, I’ll see if I can source the original.wav file from the label and I’ll also try with both the chord cutest and the Aune s8, I’m picking up a new handheld db meter tomorrow (more convenient than rigging up pro mics etc) to ensure the volume levels are matched, perhaps once I get you the .wav you could add that to the selection and randomise their placements?

thanks

 

[kyle_neuron]

Sure thing. It might be worth picking another tune or album? It’s relatively easy to add more once you get the hang of the language for the scripting.

It automatically randomises the presentation order every time you run the test, including the hidden reference. So no worries with that! I can also level match them in software, but it would mean ‘touching’ the files with a re-encode to do so with a DAW. It’s definitely possible with ffmpeg, I just need to remind myself of the flags to get the right results.

I would have thought that the purchasable WAV from the likes of Qobuz is ‘from the label’ but without confirmation, I guess it’s worth testing another source. I can always do a CRC check and null test too.

 

[Headchef]

An interesting development...

Audirvana Studio has updated their software to now include Frequency response analysis which works during playback, this. is great as it not only shows up if you're actually getting so called "hires" files but also where they are not being particularly honest. Well worth checking out, It's looking like I'll be ditching Roon as this does all Roon does but with a far more useful tool set (inc DAC management)

 

[Headchef]

Tidal V’s Qobuz….

What’s interesting is that the Tidal “Master” is little more than a 16bit file yet it transports as higher, where as the Qobuz is closer to 24bit yet transports at a lower rate than Tidal.

hmmmm

 

[audio2design]

Audirvana cannot fully unfold MQA so I don't think what you are displaying above is accurate/relevant.

 

[Headchef]

Audirvana cannot fully unfold MQA so I don't think what you are displaying above is accurate/relevant.

ah, thanks, I had no idea that it only did the first unfolding