Can You Hear the Difference between Uncompressed and MP3 Audio?

[Operandi]

Almost anyone can tell the difference if you pick the right material. I have 96kbps of simple pop/techno that sounds as good as FLAC, I have filled-out rock that sounds like it's lost a lot even at 320kbps

Just like with image and video compression, some things compress fantastically without losing much quality, others barely compress and look like shit if you try.

Thats pretty much nonsense, those types of analogies don't work, and loading a track in Audacity and looking at the waveform alone dosn't tell you how something is going to sound.

Did you run though the test samples? I am able to hear differences between lossless and and 320K on the Jay-Z track. Walk into a random audiophile room or even pro sound room and there is a good chance you'll hear Lorde used as a demo track. 128Kps is going to sound like crap with any kind of music, dosn't matter what the genre is.

Indeed, but would the digital audio from an electric set be based off a recording done via ADC, or is it possible to correctly synthesise each note directly in PCM?

Different amps with electric guitars produce different results, as much as microphones.

A synth or drum machine could be going straight to digital medium but I doubt thats really a thing (but maybe?). Even if it did though does that really matter?, the track still has to be edited and manipulated in the mastering process in a recording session, or at the sound board in the live show.

More to the point though PCM isn't something you can listen to without going through the reconstruction filter of the DAC and thats really a approximate best guess by the DAC designer as how to recreate the (analog) sound. Its not like the DAC is taking the bits, processing them, and putting the waveform back together the exact same way it was encoded.

 

[lexluthermiester]

128Kps is going to sound like crap with any kind of music, dosn't matter what the genre is.

That's absolute nonsense. 128k sounds fine for most kinds of music. Not perfect, but doesn't stop the experience from being enjoyable. The situation where 128k is going to stand out is in high quality HIFI type audio systems in perfect(or near perfect) listening environments. Listening in your car, on your phone, on your laptop or even a PC, speakers or headphones, will sound just fine as ambient environmental "noise" will drowned out the imperfections of the 128k bitrate anyway. Sure, higher bitrates are preferable, but 128k is FAR from the end of the world for music listening.

 

[Chrispy_]

Thats pretty much nonsense, those types of analogies don't work, and loading a track in Audacity and looking at the waveform alone dosn't tell you how something is going to sound.

You're misquoting me; I'm not the waveform inspection guy.

I'm just saying that not all content compresses equally and in some cases 128k is good enough to be imperceptible. If you're going to try and defend that with some subjective opinionated bollocks, let me remind you that at 128bps (yes, a thousandth of the 128k you're scoffing at), most codecs can compress a single pure square wave losslessly. That's not subjective, that's objective fact. Obviously a square wave synth tone is not music, but it provably establishes that compressibility is scalable depending on content - just as common sense dictates.

 

[Aquinus]

Warning, this is a highly subjective comment. I personally find certain sounds to sound better when lossless compression is used. The best way I can describe it is that individual sounds are more distinct and don't "blur together." That is, it's easier to isolate particular instruments for example. However, that gap really depends on the bitrate of the MP3 and the encoder that was used. It doesn't really matter if you're at 320Kbit constant bitrate if the encoder is garbage to begin with. A 128kbit MP3 produced many years ago is not necessarily the same beast as one encoded with a modern encoder.

So, unless you're an audiophile who wants to bask in the clarity of every tiny little sound, you probably don't really care about the difference between a reasonable bit rate, like 192kbit MP3s, and lossless. However if given the choice, I'll always opt for a lossless codec.

 

[Operandi]

That's absolute nonsense. 128k sounds fine for most kinds of music. Not perfect, but doesn't stop the experience from being enjoyable. The situation where 128k is going to stand out is in high quality HIFI type audio systems in perfect(or near perfect) listening environments. Listening in your car, on your phone, on your laptop or even a PC, speakers or headphones, will sound just fine as ambient environmental "noise" will drowned out the imperfections of the 128k bitrate anyway. Sure, higher bitrates are preferable, but 128k is FAR from the end of the world for music listening.

When I say sound "like crap" I mean you will hear the crap that 128Kps compression is doing regardless of the music. Certain genres or tracks might highlight it more but its noticeable 100% of the time, and once you notice it its impossible to not be aware of it. However I also agree compression dosn't ruin the enjoyment of music, wasn't trying to say that at all. YouTube music sounds just fine on my dedicated two channel setup, not as good as high bitrate rips or lossless but it still works.

You're misquoting me; I'm not the waveform inspection guy.

I'm just saying that not all content compresses equally and in some cases 128k is good enough to be imperceptible. If you're going to try and defend that with some subjective opinionated bollocks, let me remind you that at 128bps (yes, a thousandth of the 128k you're scoffing at), most codecs can compress a single pure square wave losslessly. That's not subjective, that's objective fact. Obviously a square wave synth tone is not music, but it provably establishes that compressibility is scalable depending on content - just as common sense dictates.

Yeah I kinda combined those two things as I've seen people look at the waveform of a highly compressed (lack of dynamic range) modern pop, rap, or electronic and draw the conclusion there would be nothing to gain from a higher bitrate or lossless file and thats not the case at all.

 

[HD64G]

Very easy to seperate the low quality from high or uncompressed when the track's uncompressed is of good quality. Some of them are terrible even uncompressed though.

 

[Chrispy_]

Very easy to seperate the low quality from high or uncompressed when the track's uncompressed is of good quality. Some of them are terrible even uncompressed though.

Yes. Some tracks are just mastered so badly that compression is never going to be the biggest problem.

 

[lexluthermiester]

When I say sound "like crap" I mean you will hear the crap that 128Kps compression is doing regardless of the music.

While that's a fair point, it's still very situational and depends on personal "taste".

and once you notice it its impossible to not be aware of it.

Only if you really care and it bothers you. Some people, like myself, care only about enjoying the music. When listening through my home HIFI, yeah, I want the best experience and high bitrates rule the day. But in almost every other situation, the differences simply don't stand out.

However I also agree compression dosn't ruin the enjoyment of music, wasn't trying to say that at all.

Fair enough. That's kinda the way it came off. No worries.

Yes. Some tracks are just mastered so badly that compression is never going to be the biggest problem.

This is true!

 

[Bomby569]

it doesn't depend on tastes or subjectiveness or even any compression quality, but what you are using to hear the sound.

 

[Operandi]

While that's a fair point, it's still very situational and depends on personal "taste".

Only if you really care and it bothers you. Some people, like myself, care only about enjoying the music. When listening through my home HIFI, yeah, I want the best experience and high bitrates rule the day. But in almost every other situation, the differences simply don't stand out.

Yeah, just within the context of this topic is low bit rate "crap", otherwise I think our view points are pretty much the same. I mean if you are trying to hear differences high bitrate / lossless and 128Kps or similar the low bitrate stuff is objectively crap, whether you care or not is different story.

If I'm in the car, on the bike, or running I really don't care. When I'm at home listening on two channel livingroom setup I do legit enjoy higher quality stuff more but even so I'll still listen to YouTube and it dosn't ruin it for me. The question now that I know I can hear a difference is do I start keeping my own music in FLAC instead of alt EXTREME Lame MP3s? I'd say yeah but the only downside is storage on phones is still a consideration with FLAC and high bitrate MP3 always seemed like a good compromise.

 

[lexluthermiester]

Yeah, just within the context of this topic is low bit rate "crap", otherwise I think our view points are pretty much the same.

Ah that cleared up what you meant and agreed.

I mean if you are trying to hear differences high bitrate / lossless and 128Kps or similar the low bitrate stuff is objectively crap, whether you care or not is different story.

Right.

The question now that I know I can hear a difference is do I start keeping my own music in FLAC instead of alt EXTREME Lame MP3s? I'd say yeah but the only downside is storage on phones is still a consideration with FLAC and high bitrate MP3 always seemed like a good compromise.

For me, the file sizes are a factor and I don't like how large they are comparitively. A secondary concern is support. Everything supports MP3, not so much with FLAC. I realize today that's not such a factor anymore, but 10 years ago when I was remastering my entire music library, it was important. I settled on 320/256 MP3 which wasn't just good enough, but is still excellent in 99% of music listening situations and have stayed with it.

 

[SchumannFrequency]

The difference between mp3 @ 320 kbps and FLAC 24 bit can often/usually be heard directly on some systems.

The OS with the best audio quality is FreeBSD in bit-perfect mode with vchans disabled so you'll hear the difference more easily there, especially in certain apps like Audacious.

For example on the F&D F550X in combination with FreeBSD you can already hear the difference, the F&D F550X are rather high end speakers despite their price.

Here is a sound demo of some FLAC songs on the F&D F550X:

untitlednn.wav

drive.google.com

 

[Operandi]

So I tried this again the other day on my main system (Singularity tower speakers, Pioneer Elite amp/DAC)... and got 5/6.

I have to say this process is pretty crazy different depending on what I'm using. Last time when I was using my headphones (Sony MDR-V6) was just listening to how clean different parts of the song sounded not particularly how good I thought the song itself sounded. I don't know if thats just my take on headphones but the headphone experience was more like blunt instrument than something I would experience music though in contrast to listening to the Singularities sounded way more engaging and the 320 and lossless tracks just having that much more life vs. just sounded cleaner on the headphones.

Also, I'm not sure what the proper way to do this with statistical meaningfulness is but I went though each track probably 2-3 times just to hear what each one sounded like and in the process got 320 and lossless mixed up again. After 20 mins of that and getting used to what I was hearing I sat down and tried a complete run and got 5/6 correct and got the last or second to last one wrong and actually picked 128K but I think I may have psyched myself out I also really had to go to the bathroom, lol.

 

[Chrispy_]

The only one I struggle with is Tom's Diner, which is just vocals.

There's not a lot of audio data there, so 128Kb/s seems to be good enough that I cannot tell the difference, so this one trips me up.

My worst device is a pair of Sony MDR1A headphones which are by no means shit, but they're a bit bassy and I guess this accentuates differences lower in the range than I'd normally notice. Whilst I can pick up the 128Kb version of Tom's Diner with those reliably, I'm picking it not because it sounds worse (it actually sounds cleaner to me) but because it sounds different to the other two - and I'm making the assumption that the the 128kb version is most likely to be the odd one out. I'm comfortably middle-aged so my high-frequency sensitivity is definitely not great any more.

My other stuff isn't exactly high-end but the desktop I normally use has a pair of half-decent Eris E8 monitors with room-mode correction through an external DAC+DSP and acoustic corner traps to ensure they're as close to perfectly flat as they can possibly be. My living room has some pretty average Q-Acoustic speakers but I cut the sub out for music and the main drivers aren't terrible for accuracy - especially given that the Yamaha receiver has a dedicated mic and (basic) DSP to do room calibration and compensate for the worst of the room distortion.

 

[Operandi]

The only one I struggle with is Tom's Diner, which is just vocals.

There's not a lot of audio data there, so 128Kb/s seems to be good enough that I cannot tell the difference, so this one trips me up.

Thats probably the song I was most familiar with before doing this test and I thought it would be hard too for the same reasons. Its also one of the tracks used to test during the development of MP3 so you think if any track would be hard it would be that one. I think 128/320/lossless all sound good with this track but I can pick out 128 pretty easy and while I was pretty reliably picking out lossless on both the MDR-V6 and the Singularities I had absolutely no idea what I was listening for on either setup, just which sounded the most natural and going with my gut.

My other stuff isn't exactly high-end but the desktop I normally use has a pair of half-decent Eris E8 monitors with room-mode correction through an external DAC+DSP and acoustic corner traps to ensure they're as close to perfectly flat as they can possibly be. My living room has some pretty average Q-Acoustic speakers but I cut the sub out for music and the main drivers aren't terrible for accuracy - especially given that the Yamaha receiver has a dedicated mic and (basic) DSP to do room calibration and compensate for the worst of the room distortion.

I have no experience with PreSonus but I have KRK V4 S4s on my desk and not considering what the setup cost I'm not super blown away. I think the KRKs are pretty accurate but surprisingly not all studio monitors are, its more about their settings that are able to compensate for room and location and their pro interface than being absolutely 100% accurate. I also really have to wonder what kind of design compromise they are doing when they design a 8" two-way like that.

I have heard good things about Q Acoustics though, and if your sub is giving you trouble you could consider adding second one opposite to cancel out the room modes?

 

[mechtech]

Interesting

https://kth.diva-portal.org/smash/get/diva2:1701966/FULLTEXT01.pdf

Basically says if 256kbits/s then at diminishing returns. Could do 320kbit/s vs flac or wave.

Said was encoded/ripped with LAME. I wonder encoder type makes a difference when making mp3 files from flac or wave?

 

[Operandi]

Said was encoded/ripped with LAME. I wonder encoder type makes a difference when making mp3 files from flac or wave?

Probably back in the day but 20+ years later I doubt there is much you can do to optimize things. That said I'm pretty sure LAME is still in active development.

 

[ir_cow]

So I redid this test now with some decent headphones and a sound card. 320 and uncompressed sound the same to me. Just shows what a good setup can do. Before it was easy to pick out the 3. One got me* good and I picked 128 because the hissing I thought was a compression error. 128 was so compressed it cut it out completely.

 

[mechtech]

So I redid this test now with some decent headphones and a sound card. 320 and uncompressed sound the same to me. Just shows what a good setup can do. Before it was easy to pick out the 3. One got my good and I picked 128 because the hissing I thought was a compression error. 128 was so compressed it cut it out completely.

Can you try 256 and see if there is a discernible difference between that and uncompressed?

 

[ir_cow]

Can you try 256 and see if there is a discernible difference between that and uncompressed?

I could make my own from a FLAC 96/24 track and see. Here is the thing though. I do this test on my home theater setup and I can clearly hear the difference. By which I mean, the clipping is easily disguisable. So is the sound card actually really the magic here?

 

[Ferather]

You would need 1:1 digital up to the speaker, and then the best possible speaker to determine some factors. Then you can fairly compare original digital, and then compressed digital.
Not all music is produced using actual instruments, so in various cases the audio can start as digital, an analogue output in this case would be un-original.

You can also say the same for games and also movies, where the audio is produced digitally, not recorded sounds.

You also cant compare headphones to a proper speaker system, even the most expensive headphones cannot physically output a full range at the correct dB.
There is certainly no way to compensate for a proper LFE unit with headphones, if you could TV speakers would be awesome.

Different amps work in different ways (class), and have varying frequency ranges.

 

[unwind-protect]

Said was encoded/ripped with LAME. I wonder encoder type makes a difference when making mp3 files from flac or wave?

Yes, it does. But most so-called other encoders use lame inside anyway.

Same for video encoding. The hardware encoding offered by GPUs is inferior to what a software encoder with arbitrarily complex code like ffmpeg does.

(and to clarify this, ffmpeg uses liblame for mp3 encoding)

 

[Belfaborac]

I did a comparison of a handful of my best (clean, simple, high DR) recordings some 15+ years ago, encoded with Lame to 128, 192, 256 and 320kb/s. The basic takeaway then was that I could always pick out 128; often 192, depending on the type of music (more difficult on rock and pop, easy on jazz and classical); only sometimes at 256 and only on a single track at 320.

Since then I'm sure that Lame and other encoders have been substantially improved and I've seen it claimed that in stereo 128kb/s AAC is transparent. I kinda doubt that, except that it might well be for your typical use case scenario (mediocre in-ears with a mobile as source). I suspect that anyone with a decent stereo or a serious headset set-up will need rather more than that.

 

[Easy Rhino]

Interesting

https://kth.diva-portal.org/smash/get/diva2:1701966/FULLTEXT01.pdf

Basically says if 256kbits/s then at diminishing returns. Could do 320kbit/s vs flac or wave.

Said was encoded/ripped with LAME. I wonder encoder type makes a difference when making mp3 files from flac or wave?

I think most people who listen to enough music realize it has more to do with your speaker/amp setup than anything else. I can defin hear the difference between lossless and 192kHz/24bit audio tracks on my KEF speakers at higher volumes. I have done my own blind tests. On simple speakers I cannot tell the difference between a CD and 256Kbps rips.

 

[Ferather]

It also does not help when Windows is being a male chicken. There is plenty of points where the Windows audio engine gets in the way of 1:1 audio.