Mastered CDs vs 44/16 FLAC Downloads

[Oklei]

I always wondered if I get the same mastered version when I buy 44.1/16 FLAC files or a CD. Are these files in fact ripped CDs?
To find out I bought the new Avenged Sevenfold album twice (totally worth it ). The 44.1/16 FLAC @ 7digital and the CD via Amazon. The CD was then accurately ripped with EAC and to my surprise the two versions aren't the same.
I think different filters were used to downsample to 44/16. Here is the Delta Spectrogram of the first track:

These differences might not be audible at least for me (most Delta above 19k). But I'm still a bit disappointed. I always thought I would get the same version when I buy CD quality FLACs online. If it's not the same there potentially could be all sorts of changes (like watermarks etc). For my peace of mind I guess I'll stick to ripping CDs for now.

 

[MacCali]

I always wondered if I get the same mastered version when I buy 44.1/16 FLAC files or a CD. Are these files in fact ripped CDs?
To find out I bought the new Avenged Sevenfold album twice (totally worth it ). The 44.1/16 FLAC @ 7digital and the CD via Amazon. The CD was then accurately ripped with EAC and to my surprise the two versions aren't the same.
I think different filters were used to downsample to 44/16. Here is the Delta Spectrogram of the first track:
View attachment 290569
This differences might not be audible at least for me (most Delta above 19k). But I'm still a bit disappointed. I always thought I would get the same version when I buy CD quality FLACs online. If it's not the same there potentially could be all sorts of changes (like watermarks etc). For my peace of mind I guess I'll stick to ripping CDs for now.

Oddly, and probably because I’ve never owned a quality CD player. I have never enjoyed a CD over digital streaming via a streamer. And perception wise have had the best jump in performance from going to a PC with a DDC to a Bit Perfect streamer.

Not sure if it’s the streamer and or bit perfect.

I know some people have told me that the mastering is crucial. However when I got my UB9000 I tried multiple cds and did not enjoy listening to any of them in comparison to streaming via Quboz through Roon with no dsp

 

[recycle]

The real progress of the flac files on the market is in their 96/24 resolution, when I want to buy flac music, that's what I choose. Now we could start an endless discussion about how audible is the difference between the new formats (Hi res Flac, MQA, DSD etc..), however I think the distribution of music in 44/16 resolution belongs to the past

About your test: often, a dithering using outdated methods produces audible artifacts, you should compare various CDs/files to get more reliable data

 

[Ricardus]

About your test: often, a dithering using outdated methods produces audible artifacts, you should compare various CDs/files to get more reliable data

Does it? Ethan Winder did an AES talk about audio myths and he played dithered and truncated files and I bet no one can hear the difference double blind. So unless the dithering algorithm was broken, I'm not convinced.

 

[Ricardus]

That said, when I'm mastering, I always dither when going from 24 to 16 bits. Because why not...

Different dithering algorithms will certainly output different values if you're doing a hash comparison, but I wouldn't expect the spectra to be dramatically different.

 

[recycle]

That said, when I'm mastering, I always dither when going from 24 to 16 bits. Because why not...

Different dithering algorithms will certainly output different values if you're doing a hash comparison, but I wouldn't expect the spectra to be dramatically different.

I believe that the difference found in the OP between CD and Flac of the same song is due to a bad downsample, I can't imagine any other explanation. That's why I wouldn't draw superficial conclusions but would continue to do tests on different songs to get a more in-depth view of the problem

 

[DVDdoug]

I always wondered if I get the same mastered version when I buy 44.1/16 FLAC files or a CD. Are these files in fact ripped CDs?

You never really know. IN the early days I had the impression that Apple and Amazon were ripping CDs (to AAC and MP3) but nowadays I assume in most cases the record company supplies the same master as is used for the CD.

Sometimes there are multiple-different masters, especially for older recordings or when it's available in different resolutions. But I think your differences are too subtle for that.

Here is the Delta Spectrogram of the first track:

Are they time-aligned to the exact-sample?

 

[Oklei]

Are they time-aligned to the exact-sample?

They are subsample aligned:

Spoiler: DeltaWave Settings

DeltaWave v2.0.8, 2023-06-06T15:18:19.8367180+02:00
Reference: Avenged Sevenfold - 01. Game Over.wav[L] 9969540 samples 44100Hz 16bits, stereo, MD5=00
Comparison: 01 Game Over.wav[L] 10005996 samples 44100Hz 16bits, stereo, MD5=00
Settings:
Gain:True, Remove DC:True
Non-linear Gain EQ:False Non-linear Phase EQ: False
EQ FFT Size:65536, EQ Frequency Cut: 0Hz - 0Hz, EQ Threshold: -500dB
Correct Non-linearity: False
Correct Drift:True, Precision:30, Subsample Align:True
Non-Linear drift Correction:False
Upsample:False, Window:Kaiser
Spectrum Window:Kaiser, Spectrum Size:32768
Spectrogram Window:Hann, Spectrogram Size:4096, Spectrogram Steps:2048
Filter Type:FIR, window:Kaiser, taps:262144, minimum phase=False
Dither:False bits=0
Trim Silence:True
Enable Simple Waveform Measurement: False

Initial peak values Reference: 0dB Comparison: 0dB
Initial RMS values Reference: -14,949dB Comparison: -14,965dB

Null Depth=9,039dB
Trimming 44101 samples at start and 20881 samples at the end that are below -90,31dB level

X-Correlation offset: -597 samples
Trimming 3 samples at start and 0 samples at the end that are below -90,31dB level

Drift computation quality, #1: Excellent (0μs)


Trimmed 7759 samples ( 175,941043ms) front, 19552 samples ( 443,356009ms end)


Final peak values Reference: 0dB Comparison: 0,826dB
Final RMS values Reference: -14,925dB Comparison: -14,925dB

Gain= -0,0003dB (1x) DC=0 Phase offset=-13,530927ms (-596,714 samples)
Difference (rms) = -69,29dB [-84,71dBA]
Correlated Null Depth=86,48dB [92,44dBA]
Clock drift: 0 ppm

but nowadays I assume in most cases the record company supplies the same master as is used for the CD.

Do you think they provide master files for every sold resolution or do they just provide a master file for the maximum (for example 96/24) and each seller does the downsampling to the bought format at his own? This would explain the changes in the upper frequency range if they use a different Anti-aliasing filter than the original CD mastering process.

 

[recycle]

Any file available on Amazon Music or any other platform is provided by the record label and sold online -as is- I really don't think Jeff Bezos sits down at his computer after dinner to rip newly released CDs. This means that any format and resolution is the responsibility of the label.
I am discovering that many vintage releases that are sold as Flac are actually upscaled mp3's, sometimes they are also recordings from vinyl and not from the master tape

 

[Oklei]

This means that any format and resolution is the responsibility of the label.

I think you're right with this one. But I don't understand why Warner is sending two different sets of 44/16 files of the same new album to the online seller and to the replicator factory.

 

[audio_tony]

The real progress of the flac files on the market is in their 96/24 resolution

Except that most of those so called 'HiRes' tracks are simply 44/16 tracks upsampled.

I have bought 'HiRes' (96/24) tracks off Amazon and run them through a spectrum analyser, there is nothing but noise above ~20kHz. Why? Because there was no audio there in the first place - there can't be if the source track was 44/16.

@amirm did a couple of videos on this - you should watch them.

 

[recycle]

Except that most of those so called 'HiRes' tracks are simply 44/16 tracks upsampled.

I have bought 'HiRes' (96/24) tracks off Amazon and run them through a spectrum analyser, there is nothing but noise above ~20kHz. Why? Because there was no audio there in the first place - there can't be if the source track was 44/16.

@amirm did a couple of videos on this - you should watch them.

Obviously, selling an upscaled file is scam: the 96/24 file should be generated from the original master. You deserve a refund

 

[Ifrit]

Except that most of those so called 'HiRes' tracks are simply 44/16 tracks upsampled.

Naxos in their requirements for submitted masters clearly states that upsampled material would not be accepted for hi res release. And they cancelled the hi res release for a couple albums I had with the mixed source material.
Warner requires the 192 (96 acceptable) kHz / 24 bit version files of the masters for hi res releases along with 44/16. And they also QC every version.
Same with a couple of other labels.

Do the online sellers upsample the materials that they are given? Unlikely. Do some labels shoot off the upsampled material? Probably.
There is a probable situation when the release/remaster is done in analog domain and redigitized in hi res, but the source only exists in digital, something like DAT/DASH. It will be shown as SD in all the analyzers, can it be considered hi res?

I think you're right with this one. But I don't understand why Warner is sending two different sets of 44/16 files of the same new album to the online seller and to the replicator factory.

They do not. To the factory they send DDP fileset, or “image” of the CD, not actual audio files.

I always wondered if I get the same mastered version when I buy 44.1/16 FLAC files or a CD. Are these files in fact ripped CDs?

Highly unlikely, as it makes absolutely no sense to rip CDs for file distribution when you just generate another version in Album Publishing software with literally one click.

 

[BJL]

I have purchased downloads occasionally when the physical CD is not available, or if the download was significantly less expensive. The problem that I always have with downloads is that the mastering is generally not documented, there are no liner notes. This can make a difference most frequently with older recordings where there have been several different versions, and one might be significantly better.

As an example: I wanted remastered copies of Miles Davis "Agharta" and "Pangaea" (mid-70s electric band), to replace the original LPs, which were not fantastic, despite being on CBS/Sony. Qobuz has both, but there is no information whatsoever to identify which master were used for the digital files. The best version, (in my opinion, obviously), is the 1996 Sony reissue. As it happens, those reissues also restore some minutes missing from a couple of tracks, so it was easy to determine that the download versions were not the ones that I wanted, and I bought the CDs. But if the track lengths weren't a giveaway, I would have no idea which version I was buying from the download.

Another example: The Tom Petty "Wildflowers" release. This was available in multiple formats, LP, CD, download at both 44.1 and 96kHz. With a bit of research, I found a long comment by one of the engineers involved explaining which version had the best audio (it is the 96/24), meaning, best dynamic range, minimal compression. But if you were to look at the download from any of the services, you would never know that (because, after all, the CD could have been released without compression had the label desired). Why should I have to work so hard to find information that should have been published?

I could give many more examples, if you wanted to become really bored (downloads encoded HDCD without any indication, downloads that are excessively loud compared with other CD mastering, etc.) It should not be this difficult. Downloads should at minimum indicate the year of the mastering and the engineer and studio responsible. At least then we would have a fighting chance of making an informed choice about whether to buy the download, or perhaps a CD release. It would be even better if there was more detailed technical information, but this, of course, is just a dream.

I have purchased excellent, even superb downloads, and I have purchased excellent, even superb CDs. I have also purchased absolutely horrible downloads, as well as CDs on occasion. But with the downloads, (unless there has been a review with authoritative information) it is a crapshoot because there is literally no information on the sources of the recordings, or anything about the mastering. I personally would not trust general record company or download service policies for particular recordings. Personally I do not believe that there is a quality difference between downloads vs. CD as such (assuming the same DAC); there can be substantial differences between the mastering used that determine which version is superior.

 

[Audiofire]

I think the distribution of music in 44/16 resolution belongs to the past

I don't care about my cat's listening enjoyment that much.