Can You Hear the Difference between Uncompressed and MP3 Audio?

[MarsM4N]

FYI, the tracks are randomized each time you load the page.

Well, I did only one run. IMO it's pointless. Randomized samples would be enough.
Didn't test it, but I hope the samples are randomized, lol.

 

[QuietBob]

An interesting experiment! I attempted it with the knowledge that I cannot reliably distinguish between WAV/FLAC and 320k MP3. My results partly confirmed this, but there were a few surprises. I listened to every sample twice using headphones and a DAC at low volume. I got Mozart and JayZ correctly, and with Vega I chose 320K MP3. So more or less what I was expecting.

However, I picked the lowest quality for Katy Perry, Neil Young and Coldplay That Coldplay track sounded terribly compressed regardless.

I believe that in the age of streaming media most consumers do not care for audio fidelity. But blind listening tests such as this one can prove that even people who claim to hear the subtle nuances in music would have a problem telling compressed from uncompressed.

 

[Ferather]

"That Coldplay track sounded terribly compressed regardless" - that's why I said its better for me to hear the original and then compare.

 

[Operandi]

Encoding to a lossy format can and does change the sound. Per the statement above lossy also means gainy. Intentionally limiting dynamic range by pushing everything upwards to near clipping is one way to hide noise and other interference that does find its way in.

Lossy media encoding doesn't the dynamic range of whats being encoded.

MP3 have trouble storing frequencies above 16KHz which eventually resulted in some answers with higher bit rates. Effectively this means they allow unwanted elements to pervade across a further range while losing more information (lossy) as well. Size was their major benefit over lossless formats. What I don't want to leave unexposed is the human element, artistry, that can be poured into making the best use of a chosen media format. Someone who makes MP3 to their utmost is not worried about the constraints or benefits of other options they didn't choose.

Hearing acuity and full range response are audiophile qualities. Some people still have better than 20/20 eyesight at a rather advanced too. This is where the normal person starts thinking they are crazy and believe me it drives those inflicted with such high level sensitivity in one sense crazy as well! Audio pros are a mixed bag depending on how well they protected their hearing. The great leveler being what is locked inside their heads doesn't actually require that great of hearing to conceptualize inside a high level of experience.

By this point you should realize I broadly generalized where it was safe to do so. For the purpose of exposing what will determine success in this social experiment. Limiting discussion out of the gate with high level maths and theory serves nobody.

192Kps MP3 covers up to 19Khz, 320 covers 19.5Khz, encoders like LAME might cover the whole thing. Its removing information everywhere along the spectrum though and thats evident in the Susan Vega track which apparently the OG developers used as was pure example of the human voice which is right in the middle of the auditory range. That said I found it one of the easier ones to pick lossless on, though if I tried to pickup on the differences I'd sometimes physc myself out and pick 320 but if I tried less and just picked the one that sounded more correct I tended to pick the lossless one, kinda weird.

Local playback compared to streaming essentially. There are many reasons this might be the case.

When you listen to these high bitrate MP3 you open a player program which nowadays does a fair amount in cooperation with the OS to lessen the effects of other processes etc. Chances are youtube follows a much simply chain to reach your ears. Upload quality is another example already mentioned. Don't overthink it but realize your hardware and software chain can be optimized for one or the other. Doing both with realistic means? Third party software that handles both streaming and local playback produce a more even output.

In my case I'm playing them back on the same HTPC so it should be the same audio stack. Could be a different master for sure but its the same mix in the tracks I used. Still too many potential variables involved to really use to compare to anything I guess.

 

[dgianstefani]

Yes - but you need appropriate playback hardware - I use a schiit Hel and M100 Master headphones, but my Nommo Pro also picks up quality well.

High end earbuds or wireless headphones can also play a good source in better resolution than mp3.

Tidal Hi-Fi is a a decent high quality source, and what I use.

 

[VSG]

4/6 on TV speakers lol, and 6/6 using a good set of headphones and DAC/amp. I am typically quite good with this, and I'd also suggest the Harman training software for those interested: https://harmanhowtolisten.blogspot.com/

 

[Ferather]

That's one of, or the main, reason THX certificated audio devices existed, shame it still exists but not exactly used. There are plenty of high end devices without THX.

"THX Certification is a globally recognized assurance of uncompromising quality, consistency and performance.
We blend art, technology, and the dynamics of real-world listening and viewing environments to ensure products deliver the artist's true vision."

Would be nice if THX or even a new audio certificate that specifies a device can do true lossless (without change).

 

[nomdeplume]

Lossy media encoding doesn't the dynamic range of whats being encoded.

192Kps MP3 covers up to 19Khz, 320 covers 19.5Khz, encoders like LAME might cover the whole thing. Its removing information everywhere along the spectrum though and thats evident in the Susan Vega track which apparently the OG developers used as was pure example of the human voice which is right in the middle of the auditory range.

Lossy encoding resculpts the file contents which changes many aspects. What and how much vary. Dynamic range can certainly be heavily impacted. Mp3 are known for being bass heavy and having reduced highs even if they technically have data as high as CD. You are losing data points to establish the delta between softest and loudest elements dynamic range calculations use.

Along with quality and aptitude of equipment (classical may sound awful on equipment that pop amazes on). Consumer equipment can have large swings in how it presents sound. Typically lower end equipment degrades with heat and better built stuff gradually improves with it building up within electronic components.

The kicker is while most people will think the lossy file sounds good across multiple types of playback. If they get propped up in front of behemoth speakers with large amps. A high quality lossless version of that same file will be almost unrecognizable. Unless you have especially good hearing the furthest it is advisable to go into audio reproduction is quite low. Where it is still enjoyable, where most music still sounds good.

4/6 on TV speakers lol, and 6/6 using a good set of headphones and DAC/amp. I am typically quite good with this, and I'd also suggest the Harman training software for those interested: https://harmanhowtolisten.blogspot.com/

Very good. I removed a few bits about trained listening in my post above. That site can be highly valuable if you are serious about establishing technical criteria that educates how you interpret sound quality and changes to it.

Which will indirectly move awareness towards understanding everything has a sound. This is why people keep and repair/replace equipment once they know it works for them. At some point better becomes more a point of being different and roughly equal.

 

[Ferather]

A bit off topic:

Even they way the system is connected can cause loss, from research the only known way to transmit without loss is optical (digital). Conductive circuit has resistance and EMI (and requires more parts).
More parts normally equals more THD, THD-N. I notice a difference between analogue (which is electric) and optical, it also depends on DAC's (receiver).

For me the best way would be optical right up to the speaker, and just in front of the driver, a high output PowerDAC.

----

Just remembered, it was not that long ago your TV audio system would pickup and playback mobile phone signals. I guess phones used analogue signals back then.

====

https://www.techpowerup.com/forums/attachments/soundbar-png.240662/

https://www.techpowerup.com/forums/attachments/cables-png.240661/

 

[VSG]

Very good. I removed a few bits about trained listening in my post above. That site can be highly valuable if you are serious about establishing technical criteria that educates how you interpret sound quality and changes to it.

Which will indirectly move awareness towards understanding everything has a sound. This is why people keep and repair/replace equipment once they know it works for them. At some point better becomes more a point of being different and roughly equal.

Given I review headphones and sources on TPU, I'd better have a good and trained set of ears haha.

 

[delshay]

@Thread ..YES

Chord Hugo with my Shure SE-545 or 846 earphones.

 

[dirtyferret]

4/6 on TV speakers lol, and 6/6 using a good set of headphones and DAC/amp. I am typically quite good with this, and I'd also suggest the Harman training software for those interested: https://harmanhowtolisten.blogspot.com/

Doesn't Harmon have an "official" sweet spot for their speakers? I recall they did a user study sometime ago.

 

[nomdeplume]

Doesn't Harmon have an "official" sweet spot for their speakers? I recall they did a user study sometime ago.

I believe you are referring to the Harman Curve.

Which intends to create a flat profile response curve that in effect bumps certain bass frequencies and lowers the highs to be more pleasing on modern hardware. There are separate loudspeaker and headphone curves.

 

[R-T-B]

On better equipment you will begin to find it is slightly inferior to wav.

Calling bullshit. FLAC is literally used 99% of the time as lossless compression. Meaning it's the same.

 

[dgianstefani]

Goldensound on YT has some good videos about lossless standards and audio hardware. Tidal has a "master" quality that is technically worse than their hi-fi due to altering the source.

He specifically tested their MQA standard by uploading a test file, then analysing the MQA file after they hosted it, in comparison to his original.

 

[Sabotaged_Enigma]

Actually for most cases most people can't. But for those low quality mp3 such as 128 kbps it clearly has differences that we can hear.
Look at the frequencies of a HQ 320 kbps 44100 Hz mp3 which almost fill the whole graph all the way to 20000 Hz. A mp3 file like this sounds almost no different at all from a flac or ape or wav.
Though human ears have a range of 20 Hz to 20000 Hz but in reality your ears are very good if you can actually hear sounds above round 16000 ~ 18000 Hz.
And yes of course for audio engineers who have the right speakers and their excellent ears, it'll be easy and necessary for them to tell the differences.

 

[ChosenName]

• Flac is versatile. You can compress files to be smaller or not at all. On better equipment you will begin to find it is slightly inferior to wav.

.... only if the equipment is broken in some way.

 

[lowrider_05]

I got 4 out of 6 correct with a cheap Logitech H390 headset on my company laptop. Jay Z (128kbps) and Coldplay (320kbps) were my errors.

 

[qubit]

I might give this test a go later. Shame they don't have any electronic music, like melodic trance.

Flac is versatile. You can compress files to be smaller or not at all. On better equipment you will begin to find it is slightly inferior to wav.

Sorry, I must challenge you on that. FLAC is lossless so there can't be any difference from WAV, by definition. Why do you think there is?

 

[lexluthermiester]

My 2 cents, at or above 256kbps it is very difficult to hear a difference between MP3 and uncompressed audio. 192kbps & 212kbps are very high quality and while telling the difference is possible to a discerning ear, such requires high end equipment to hear the differences. For most people, 192kbps is very good quality, 256 is very high quality and 320 is HIFI audiophile level.

While this is just my opinion, it is based on 35 years worth of AV equipment use. I'm one of the few people I know who actually have a quality component HIFI system(Onkyo). With my system, the differences can be heard, but as the bitrate increases, the differences become less distinctive. I personally stopped using FLAC years ago because 256 and 320 was excellent.

 

[nomdeplume]

Calling bullshit. FLAC is literally used 99% of the time as lossless compression. Meaning it's the same.

.... only if the equipment is broken in some way.

Sorry, I must challenge you on that. FLAC is lossless so there can't be any difference from WAV, by definition. Why do you think there is?

Since we are in season. Say I take a tomato and split the seeds between two gardens in reasonably close proximity tended by separate people. To but a few intimately aware of them both will it be apparent one set of plants received better soil prep and nurturing. Across a wide range of uses they portray less unwanted flavoring or other characteristics. It is right that anyone should be happy with and freely mix the two.

They are highly similar. Not the same. Regardless, my intent here was to provide a basis of examination that furthered interest. Flac is a highly useful backend industrial streaming format. Meaning it can be quickly converted to any number of other formats or compressed for delivery. Portraying the term lossless as an absolute is a falsehood. It describes meeting a level of mathematic proof having no bearing on any other element that determines quality of what is stored. Both tomatoes proved insect resistant enough to ripen without flawed skin.

Actually for most cases most people can't. But for those low quality mp3 such as 128 kbps it clearly has differences that we can hear.
Look at the frequencies of a HQ 320 kbps 44100 Hz mp3 which almost fill the whole graph all the way to 20000 Hz. A mp3 file like this sounds almost no different at all from a flac or ape or wav.

Repurposing my example above, who can taste the bugs or that black pucker on the skin where they pooped from eating too much. On a sandwich with so many artificial flavors and preservatives.

Though human ears have a range of 20 Hz to 20000 Hz but in reality your ears are very good if you can actually hear sounds above round 16000 ~ 18000 Hz.
And yes of course for audio engineers who have the right speakers and their excellent ears, it'll be easy and necessary for them to tell the differences.

Acute hearing ability and sensing air movement in other ways can have very similar effects in real life. This applies at both ends of the perceptible frequency range. Being aware of a low frequency rumble that registers over a longer period than is typically considered sound is the most common example From two miles away you can tell if Spring snowmelt has arrived at a waterfall or not.

This information didn't arrive within the few milliseconds or less that high frequency does. These frequencies are easily blocked and reflected. Losing all energy before they can reach your ears. They are nonetheless detected as less threatening and more easily disregarded. A tingle.

 

[ExcuseMeWtf]

Since we are in season. Say I take a tomato and split the seeds between two gardens in reasonably close proximity tended by separate people. To but a few intimately aware of them both will it be apparent one set of plants received better soil prep and nurturing. Across a wide range of uses they portray less unwanted flavoring or other characteristics. It is right that anyone should be happy with and freely mix the two.

They are highly similar. Not the same. Regardless, my intent here was to provide a basis of examination that furthered interest. Flac is a highly useful backend industrial streaming format. Meaning it can be quickly converted to any number of other formats or compressed for delivery. Portraying the term lossless as an absolute is a falsehood. It describes meeting a level of mathematic proof having no bearing on any other element that determines quality of what is stored. Both tomatoes proved insect resistant enough to ripen without flawed skin.

This is not what "lossless" means.

Lossless compression by definition must yield exactly the same output as was before compression. If there is any difference whatsoever, it is lossy, end of story.

If FLAC is lossless, it must sound exactly the same as uncompressed audio. If it does not, then it is possible that there was some lossy method involved with FLAC input.

 

[qubit]

Since we are in season. Say I take a tomato and split the seeds between two gardens in reasonably close proximity tended by separate people. To but a few intimately aware of them both will it be apparent one set of plants received better soil prep and nurturing. Across a wide range of uses they portray less unwanted flavoring or other characteristics. It is right that anyone should be happy with and freely mix the two.

They are highly similar. Not the same. Regardless, my intent here was to provide a basis of examination that furthered interest. Flac is a highly useful backend industrial streaming format. Meaning it can be quickly converted to any number of other formats or compressed for delivery. Portraying the term lossless as an absolute is a falsehood. It describes meeting a level of mathematic proof having no bearing on any other element that determines quality of what is stored. Both tomatoes proved insect resistant enough to ripen without flawed skin.

Rubbish. What the tomato analogy has to do with it, makes no sense at all.

The fact is that:

1 Take a random WAV file
2 Losslessly compress it with FLAC
3 Decompress it with FLAC
4 Compare original file with the FLAC processed file: difference is zero and hence will sound exactly the same

It's as simple as that. That's precisely what lossless means. Please don't try to muddy the waters on something so basic and so settled. Google and Wikipedia are your friends if you're still unsure of this.

 

[Vario]

One thing I have noticed about modern digital music is the replay gain is often excessive, I usually store music at 128 because I have a tin ear and want smaller file sizes, and normalize everything to replay gain of 89 decibel.

 

[qubit]

One thing I have noticed about modern digital music is the replay gain is often excessive, I usually store music at 128 because I have a tin ear and want smaller file sizes, and normalize everything to replay gain of 89 decibel.

Sacrilege: off with his head!