Lossless mp3 (44.1 KHz 24bit 2ch 244kbps) vs. FLAC

Roon processes the MP3 file into those parameters. What you’re seeing is Roon’s decompression of the lossy MP3. Sounds pretty good to me.

1 Like

Yep Bill is correct. All lossy codecs such as MP3, oggvorbis or AAC will get upsampled to 44.1/24 in Roon so it can handle them better so that’s what you are seeing. If you rip flac lossless then you will notice a difference in sq compared to these regardless of upsampling g. Garbage in Garbage out is always the rule.

Roon does not “upsample” MP3 and AAC to 24 bit. It decodes those lossy files to 24 bit.

AJ

3 Likes

What’s exactly the difference?

1 Like

When something is lost, it’s is lost forever. No matter how much you upsample it.

1 Like

Check out this thread: A way to decode lossy files to >16 bits per sample [Delivered in build 333]

1 Like

Here you go screengrab of the signal path for MP3 files:
Roon-MP3

I never realised they had no apparent bit depth . Learn something new all the time.

Think it’s answered above. It’s just the way Roon decodes it.
The lower quality is represented by the amber light.

Never seen it before as I have never owned any mp3 files.

The input to the lossy encoder does possess a specific bit depth and almost always is 16 bit – the most notable exception to that may be Mastered for iTunes AAC.

AJ

I find it hard to distinguish high bitrate MP3 from lossless FLAC. Would I rather have my music in lossless FLAC ? Yes. Do I stop listening to something because it’s in MP3 ? No.

8 Likes

There’s a greater ease and sparkle to lossless for me - a sense of space and air. (The kind of almost-too-subtle difference - like moving from 16/44 to hi-res.) But a great recording, well encoded in MP3 still sounds great - and I’ve decided not to sweat it for the bits and pieces I have in MP3 (unless I can get a cheap CD or download). I got five Blossom Dearie albums for something like £8 as MP3 - that would have been nigh on impossible to source on CD or download - sampled first on Spotify first to make sure the transfer wasn’t a dog, That can be very persuasive.

The distinctions are based upon measured “scientific” SQ, which some/many think is at least different from actual SQ, if not dissimilar.

All of us would like to think we can distinguish between the two. Fewer actually can.

1 Like

I had never done any A/B comparisons of mp3 vs. FLAC when I eventually noticed a difference between the two. Back in 2013, when prices for new vinyl were climbing but hadn’t quite reached the current ridiculous numbers, I was buying new releases on vinyl. At that time, almost all new releases came with a download code, and the download tended to be 320kbps MP3. I should have known better, but I downloaded the MP3s, probably out of a sense of miserliness.

I tend to listen to digital mostly by shuffling the entire collection. For a period of about six months to one year, I was skipping tracks that I knew I absolutely loved. Something about those tracks felt too sharp and harsh. I felt like I was headed for a headache. It was a sound thing, not a music thing. I still skip tracks that are recorded like that or that are just too “experimental” for me - especially when it will avoid pain between my ears and eyes.

Once I noticed I was doing it, I started thinking about it, and I remembered my downloads. I started checking, and, sure enough, the tracks I wanted to skip ended up being those 320kbps MP3s. Before that, I’m not sure that I would have been able to recognize the difference between them and FLAC encodes. Once I knew that headache-is-coming feeling, I could tell the difference, even at that really high bitrate. Obviously, bad recording quality could hide that difference (um, Red Hot Chili Peppers, anyone?), but even so-so recording quality could not hide it.

That experience sealed for me the debate about A/B comparisons. I think some things require long-term observation, even non-mindful observation, to discern. One could say that the differences such observations reveal must be too small to be significant, but my experience with those MP3s revealed a difference that was very real and significant to me: the difference between wanting to hear a track and get into the groove of it, and wanting to make that sound go away. Given my unhealthy obsession with music, that’s big.

Then again, context is important, no? I still listen to 192 MP3s in the truck, because more tracks fit on the iPod that way, and I can’t tell a difference there (no headache feeling). My experience back in 2013 was through Aerial Acoustics 7b speakers, driven by VTL amplification, fed by Perpetual Technologies (remember those bastards?) DAC and Squeezebox Touch. Not a great speaker/amp pairing, because the amps just didn’t have the forcefulness the speakers require.

Telling that story has me curious to do an A/B comparison, though… I could enlist my wife - she has the ears of a bat.

7 Likes

Interesting real life test.

My main source of mp3 material is Internet Radio, but I have 162 of 14,784 albums in mp3 also (local and Tidal). I upsample everything to DSD512 and use the xtr-mp-2s filter and DSD 7 256fs+ modulator in HQP. It goes into a NOS R2R DAC and then into tube amplification (Manley Snappers). It’s quite possible that all that film-flam plays a part in my blissful ignorance.

2 Likes

For a long term A/B test, take some FLAC files, make copies, convert those copies to your desired bit rate MP3, then convert those MP3 files back to FLAC. Ensure that all FLAC files – both original and doubly converted – are externally indistinguishable.

Upload to Roon. Albums/tracks will be duplicates of one another. If you believe that long term listening tendencies indicate audible or subliminal preferences, check your play counts between the respective duplicates after a reasonable period of time.

At the conclusion of the listening period, if play counts show any statistical tendencies, run FFT or spectrum analysis on those albums/tracks. Determine whether lossy encoding was applied.

Caveat: If you cheat by making the duplicate albums/tracks somehow distinguishable or you run analysis before listening has concluded, then you are not interested in conducting a legitimate test, only in confirming your own cognitive biases.

AJ

1 Like

I’m a big fan of cognitive biases. They are why I didn’t do A/B tests when I asked my wife-to-be to move in with me, why my stop-the-ball-but-keep-running move works so well on the soccer pitch, and why my mom is one of my biggest fans. I would never sacrifice those judgments to the amoral scrutiny of science. Same with my enjoyment of music. It’s bad enough that I’m starting to enjoy crappy music because it was well recorded.

As the line from the movie Barfly goes, “Truth is for suckers.” Then again, another great line from that movie goes, “The more you believe in, the better off you are.”

Then yet again, I don’t think I would take on that huge long-term A/B test, for the simple reason that it could be considered flawed. It wouldn’t be double-blind, for obvious reasons, and it’s qualification for single-blindedness would be jeopardized by (at least) the fact that I would know which tracks were transcoded. If I was to make claims about the results of that test, the audio-is-objective crowd would have a feast. You don’t want to rile those folks - they are as unkind as the flu and as humorless as Nebraskans.

I think I’ll stick with the results of my seemingly more blind experience with vinyl download codes. Objectivists take note: my satisfaction is not rooted in a desire for scientific certainty in answering a question akin to “which tastes better, white or wheat?” Until we understand the brain better (beyond which zone gets more blood when you make me think of Charlize Theron), I just don’t think those questions are susceptible to rigorous experimental design. There are too many factors at play in music. For example, it doesn’t seem to go too far to claim that a 320kbps MP3 of a Bernie Grundman Mastering recording would sound better, even in the long-term, than anything issued by Dead Moon (and I LOVE Dead Moon!). I don’t think we can design an experiment that will satisfy the objectivist crowd until we can subject to falsifiability a claim along the lines of, “the following factors are exhaustively determinative of the enjoyment of music: a, b, c…” That seems impossible.

@andybob’s blissful ignorance seems so much more comfortable.

6 Likes

Love the line ‘As unkind as Flu and as humourless as Nebraskans… lol :joy:

1 Like

I lived in that unimproved state for thirteen years, so I should know! A buddy on my soccer team is from there, or went to school there, and it is rare to catch him without an “N” on every item of clothing. I tease him mercilessly. The other day, he was lamenting how 40 acres in northern Colorado was selling for $150k when the same number of acres just a few miles to the north (near Cheyenne, so basically the same as all of Nebraska) was going for a fraction of that price. I pointed out that just because all the land in Nebraska is flat and uninteresting doesn’t mean that all the land everywhere should be considered the same. I got that Nebraska blank look from him… I love him, and he’s a great defender, but I’m still recovering from that place. They recently adopted a new slogan: “Nebraka. Honestly, it isn’t for everyone.” I’m not joking. They just did it.

1 Like

Chris

You gotta also love the one above it …

“the audio-is-objective crowd would have a feast. You don’t want to rile those folks”

I like it…

–MD

1 Like