FLAC vs. 320 Mp3
Jul 24, 2023 at 9:53 PM Post #1,396 of 1,406
You're comparing different masterings, not different data rates. To do a proper comparison, bounce a single HD track down to high data rate lossy and compare apples to apples. The same song may be different sounding on every streaming service or physical media release regardless of format.
 
Jul 25, 2023 at 6:28 AM Post #1,397 of 1,406
Once I got better gear which improved clarity and details (high end headphones, DAC, amp, streamer, power conditioning, better cables) I did actually hear a difference between even relatively good lossless (Spotify highest quality https://support.spotify.com/us/article/audio-quality/ ) and lossless like FLACs from HDTracks and streaming from Qobuz. I was actually surprised by this!
Unfortunately, this is not just a common misunderstanding/error but so common it’s almost ubiquitous and indeed, much of the audiophile world relies on it! Namely:

You change one or more components in your setup, say a “better cable”, DAC, etc., and you hear a difference or even a surprising difference, sometimes even when you don’t expect to. It’s the most basic logic/“common sense” to conclude that the change you made must be the cause of the subsequent difference you hear. We learn and experience such “cause and effect” almost from the moment we’re born to the moment we die and it’s so obvious we hardly ever or never even think about it. However, this “common sense”/logic/reasoning can be fooled, almost always because there’s some other variable (other than the obvious change we’ve made) that has also changed, which we’re unaware of or haven’t considered and is the real cause of the effect (difference we hear).

Almost always, there are a number of these “other variables” and audiophiles typically either aren’t aware or don’t consider any of them, or consider them briefly and erroneously discard them as the cause. In this case, the most likely variable you haven’t considered is that you’re actually comparing different masters (as @bigshot mentioned) but there are potentially others, such as different loudness normalisation between the two services and the ever present perception biases.

Your link to spotify’s “audio quality” is very informative, apart from the fact it doesn’t give any indication at all of how that audio quality relates to audible differences in quality. With most material it should be easy to hear a difference with 24kbps but by around 160kbps it’s virtually always audibly transparent, let alone at 256kbps VBR. So this variable (256VBR vs Flac) cannot be the cause of audible differences, no matter how good your reproduction equipment/environment.

G
 
Jul 25, 2023 at 10:29 AM Post #1,398 of 1,406
Unfortunately, this is not just a common misunderstanding/error but so common it’s almost ubiquitous and indeed, much of the audiophile world relies on it! Namely:

You change one or more components in your setup, say a “better cable”, DAC, etc., and you hear a difference or even a surprising difference, sometimes even when you don’t expect to. It’s the most basic logic/“common sense” to conclude that the change you made must be the cause of the subsequent difference you hear. We learn and experience such “cause and effect” almost from the moment we’re born to the moment we die and it’s so obvious we hardly ever or never even think about it. However, this “common sense”/logic/reasoning can be fooled, almost always because there’s some other variable (other than the obvious change we’ve made) that has also changed, which we’re unaware of or haven’t considered and is the real cause of the effect (difference we hear).

Almost always, there are a number of these “other variables” and audiophiles typically either aren’t aware or don’t consider any of them, or consider them briefly and erroneously discard them as the cause. In this case, the most likely variable you haven’t considered is that you’re actually comparing different masters (as @bigshot mentioned) but there are potentially others, such as different loudness normalisation between the two services and the ever present perception biases.

Your link to spotify’s “audio quality” is very informative, apart from the fact it doesn’t give any indication at all of how that audio quality relates to audible differences in quality. With most material it should be easy to hear a difference with 24kbps but by around 160kbps it’s virtually always audibly transparent, let alone at 256kbps VBR. So this variable (256VBR vs Flac) cannot be the cause of audible differences, no matter how good your reproduction equipment/environment.

G
You think Spotify has different masters of Taylor Swift albums than Qobuz does? I really doubt that. Streaming companies take whatever files the record company sends them, and recent pop albums aren't old enough to have remastered versions from different years. I think the more likely explanation is that Spotify and Qobuz receive the same lossless master files from the record company, but Spotify compresses them to save on bandwidth costs and Qobuz doesn't.

Volume normalization can be different between services, but I adjust my amp's volume knob as I switch between versions.

by around 160kbps it’s virtually always audibly transparent

Where do you get this number from?

I've heard of informal experiments where listeners compare MP3 to lossless and can't hear a difference, but that was done without audiophile equipment (ex. headphones from Best Buy plugged into a computer headphone jack).

If you do a spectral analysis of FLAC and MP3 versions of the same song with a tool like Spek or Audacity you do see there is additional information within the audible spectrum.
 
Jul 25, 2023 at 11:00 AM Post #1,399 of 1,406
You think Spotify has different masters of Taylor Swift albums than Qobuz does? I really doubt that.
Why do you doubt that? It’s very likely, with such a popular streaming artist, that a compressed master was created specifically for streaming. It’s even possible that different masters were created for different streaming services as well as for different distribution channels, such a one compliant with TV broadcast specifications. At some stage later in time, it might also be profitable to create a re-mix or re-master.
Volume normalization can be different between services, but I adjust my amp's volume knob as I switch between versions.
Most of the streaming services have different loudness normalisation. Do you know what they all are and therefore how much to adjust your amp? If not, then even relatively small differences in volume (1dB or less) can be perceived as a difference.
Where do you get this number from?
About two decades of extensive testing with numerous test subjects. HydrogenAudio details this, as that is the site used by the developers of some lossy codecs.
I've heard of informal experiments where listeners compare MP3 to lossless and can't hear a difference, but that was done without audiophile equipment …
I’ve heard of those too. However, I’ve also heard of such tests being done on much better equipment than audiophile equipment, with more highly trained/skilled listeners, both informal (uncontrolled) and formal (controlled) testing. I’ve done such test myself and of course there’s all the formal/controlled testing mentioned above.
If you do a spectral analysis of FLAC and MP3 versions of the same song with a tool like Spek or Audacity you do see there is additional information within the audible spectrum.
Yes, a very significant amount, even at the highest bit rates. However, it is “masked” (see Auditory Masking - Wikipedia), so that “additional information” is inaudible, you can easily see it on a spectrogram but you can’t hear it. That’s how lossy codecs work, using a “perceptual model” which removes inaudible frequencies, significantly reducing the number of bits required to encode the audible signal.

G
 
Jul 25, 2023 at 11:05 AM Post #1,400 of 1,406
Until you can confirm they’re the same, you shouldn’t just assume they are to back up your ideas.

Volume adjustment unless rapid, precise, and automated, invalidates a listening test IMO.

Seeing a difference on the spectrum has nothing to do with audibility. Auditory masking is real and lossy encoding relies on it. https://en.m.wikipedia.org/wiki/Auditory_masking

As advised, if you do care to verify your conclusion about the lossy vs lossless, you should take lossless tracks and convert them to lossy yourself so you know what you’re testing.



Oh well, gregorio posted and I’m saying the same, so forget it, I was never there.
 
Jul 25, 2023 at 2:18 PM Post #1,401 of 1,406
You think Spotify has different masters of Taylor Swift albums than Qobuz does?
Yes. Different streaming companies have different requirements for submitting tracks, and the distribution arms of the labels have different mastering for different intended venues... for instance a streaming master for a service that streams to portable devices may be quite different than one intended for home systems.

If you want a fair comparison, you have to start with the same master. It's easy to do that. Just take an HD Audio track you like and bounce it down to 16/44.1 and high data rate lossy and compare the them all against each other with a blind, level matched, direct switched comparison. If you do that, I think you'll find that the differences disappear, even with a high fidelity system.
 
Aug 7, 2023 at 3:33 PM Post #1,402 of 1,406
Funny how the --allshort switch in Lame dev encoders actually fixes pre-echo issues. Since I still view It LAME struggling to switch to short blocks fast enough in noiser & complex sections than MP3 as a whole failing. It makes LAME at V2 ~ V0 & 320Kbps very competitive to AAC & Vorbis.
 
Aug 7, 2023 at 3:49 PM Post #1,403 of 1,406
LAME can achieve transparency, just like AAC and Vorbis. It just does it at a higher data rate. It really isn't complicated at all.
 
Aug 7, 2023 at 4:14 PM Post #1,404 of 1,406
Yep. That why I use V2 since I find It much more robust than 192kbps AAC/Vorbis for Dark Ambient & Noise when A/B'd samples in the past. Still view my hearing suits the perceptual model that LAME uses over what AAC & Vorbis use.
 
Aug 7, 2023 at 4:19 PM Post #1,405 of 1,406
If you used AAC 256, you wouldn't hear a difference, and it would probably be a smaller file size than transparent LAME. I think your perception of what suits your hearing is expectation bias. Transparent is transparent. You can't prefer one transparency over another.
 
Feb 2, 2024 at 6:53 AM Post #1,406 of 1,406
If you used AAC 256, you wouldn't hear a difference, and it would probably be a smaller file size than transparent LAME. I think your perception of what suits your hearing is expectation bias. Transparent is transparent. You can't prefer one transparency over another.
That why I switched to Musepack at 175kbps since I've never found a sample that breaks --standard setting. Yet folk at HydrogenAudio love telling me that MPC at 170 ~ 240kbps should be easy despite never posting any samples/ABX logs especially at Q7(240kbps).
 

Users who are viewing this thread

Back
Top