How Well Can You Hear Audio Quality? (2015)

When I was at Berklee [0] we had considerable ear frequency response training that would involve weekly hearing quizzes where they would boost/buck certain frequency bands of either pink noise or music and you had to pick out which band was made louder or softer. Additionally, we had a quiz where you had to listen to lossless vs mp3 at different bit rates and determine which was which. We also did neat things like trying to pick out what type of guitar and pickups someone was using in a recording simply by ear.

320 kbps mp3 vs lossless 16bit WAV (cd quality) is extremely difficult to hear the delta without training, but if you know what the source instruments are supposed to sound like, such as hats and you can focus on their frequency band, you can hear compression artifacts and comb filtering.

This course really changed how I take in music and sounds in general.

[0] https://online.berklee.edu/courses/critical-listening-1

Let me put it this way: I prefer Music-A so much more than Music-B, that I'd choose Music-A playing on an AM-radio over Music-B playing on an FM radio.

Fidelity is indeed swell, but how important is it really? A lot of the music I like best, I first heard on a crappy recording or a shitty sound-system. There were A LOT of hit songs recorded before FM came along, and yet that 10k AM bandwidth (e.g. when the Beatles were big) didn't keep people from buying copies by the millions. Gasp! How can that be?

How much of the time I'm actually listening to music I like do I actually care about the fidelity? If band A's crappy cover is available in FLAC, will I choose it because band B's brilliant cover is only 128K mp3? NO.

This eternal argument has little to do with music. I still like 'Louie Louie' by the Kingsmen (a recording that cost $36), never mind how shitty the studio was. I'll still choose it over a FLAC from the studio master of "Never Gonna Give You Up". I can hear around the 'flawed' repro. Two kinds of 'purism' at work here. Pure genius beats pure fidelity.

I picked 4 out of 6 and never picked 128kbps. I am pretty sure I can reliably weed out 128Kbps, but WAV vs 320Kbps is more or less luck. Jay-Z was the easiest as the chiptune sample is a rectangular wave and hence has a very high frequency component that actually makes sound less palatable, not more. "Hissiness" of strings helps on classical and Neil Young, but pop music (Katy Perry and Coldplay) with its flat wall of sound is much harder - I was listening for "z"s and "s"s and high hats. Acapella (Suzanne Vega) was the hardest for me and I think I just lucked out.

In addition to headphones, and hearing, there is also training.

There are a number of audio artifacts in MP3s that are much more noticeable after training. However, many of them become much less discernable with variable compression rates and lower compression (>256kbps). It's possible to train on those artifacts, but I've always avoided it since I don't want the "projectionist effect" of making standard recordings less enjoyable.

This is not the case for many audiophiles. They want to fix all the bugs!

That Coldplay song: the uncompressed one is clipped[1].

[1]: https://imgur.com/a/XSCw824

…

[2]: https://en.wikipedia.org/wiki/Dynamic_range_compression

[3]: https://en.wikipedia.org/wiki/Loudness_war

Just to add to stats (which really, how useful is it with the variables of audio setup and subjective preference): 5/6 with Mobvoi TicPods using aptX HD. My wrong answer is the Jay-Z one and I picked 320kbps.

I listened to the samples multiple times for each test. As I went through, I kind of used a strategy to skip to and compare the busiest parts of the songs for "muddiness". Still, it was always very tough to tell between 320kbps and WAV. Most of them were decided on a subjective "it just sounds slightly clearer and fuller". I think I had the easiest time with the vocals only track; can attribute that to being a vocally-focused listener.

Side note on that last point: Only recently have I noticed how little I heard of drums in my music listening. Picked up edrums recently and realized how lost I was at drum beats "ideas". For melodies I feel like I have a large library of ideas to draw from. On the flipside, now I notice drums a lot more. Hopeful that others can relate to this too?

I got 6/6, you can select the right question based on how long it takes to start playing audio initially :)

I listen to a lot of music and got 1/6.

I did a similar experiment to myself years ago with a MiniDisc player and found that 128kbps was indeed the sweet spot for myself. Especially when space mattered.

While I am sure that more expensive equipment or years of dedicated study can help me discern the difference, I don't feel like my lack of refinement hampers my enjoyment of music in any way. If anything, I have no interest in training myself how to notice peas under mattresses.

I got 4/6, missing on the last two. Interestingly, I picked 128kbps for those, and they were more "pop"-y songs. Maybe I'm trained to like some more compression on pop songs?

I was also listening on my work setup, which isn't ideal: Bose QC35s plugged into an O2 amp and D50 DAC. I'd like to think I would've gotten more right with some flatter headphones.

The main things I noticed when comparing were:

* low end was smoother on the bass

* clarity of high end hits ("pss/tss" sounds)

* some vague sense of better imaging, like the sound stage was better

I didn't notice much difference in the midrange or vocals, but that might be because of the gigantic V-shape of my headphones.

This kind of test is really tricky because you are either trained to recognize the artifacts of MP3 compression, or you are not in which case you just "trust your gut". And if you are used to listen to compressed music (e.g. youtube), you will probably pick that.

And that is not even taking into account that MP3 (despite being really old and not the most efficient at this point) is good at what it does, especially at 320k.

4/6

For every case, I picked the one I feel subjectively the best. Interestingly, I still prefer the 128k version from the "Tom’s Diner" example, even after knowing the "correct answer". This is amusing because according to the descriptions, this song was chosen as the "benchmark" for the algorithm, which could have resulted in some overfitting ;)

Wow I kept picking the 128kbps MP3. Im using Sennheiser HD58X. I'm plugged into the headphone jack of my Desktop monitor which itself is getting audio via USB-C. Maybe the amp is bad in the monitor? Wonder if an external DAC would be any better.

So, this is a combination of 6 tests, and each has 3 variants of the answer. That's definitely not enough to make useful statistical conclusions about the ability to detect bad 128kbps audio, simply because it is rare.

Indeed, the probability of choosing all the best clips by chance is 0.13%, and the probability of never choosing the worst (also by chance) is 8.8%. Choosing almost all the best clips, with exactly one mistake, would have the probability of 1.6%, and choosing exactly one 128k mp3 has a 26% chance.

Disclaimer: I have not actually performed the test.

I could easily tell the difference between 128kbps and the other two. Telling the difference between the 320kbps and uncompressed was close to random at first. Once I know which is which I can pick each out. So the problem isn't being able to tell that there's a difference, but telling which is closer to the original. Each piece can have a clear tell, like the timbre of a particular note that's played in a busy soundscape. I used to have excellent hearing and now my high frequency hearing drops off at about 15kHz. I think because of this sometimes content that has already had some low-pass filtering may seem more natural where I can hear all that is being reproduced. I could tell the differences using an LG G8 QuadDAC on JVC HA-FXD80 earphones or the Surface Go speakers 3/4 volume at arms length.

[Edit: I'd also been drinking tonight which raises my blood pressure and I can hear slight ringing in my ears triggered by bloodflow or viscosity change. I can try the test again tomorrow.]

The best indicator for sound qualify is how long I enjoy actively listening to music. With a good source and sound production I can literally do nothing other than listen to the music for hours and not fidget with or read anything. Same with club systems. If it's not good I won't really have a good time for long. The Alpha Dynacord at Stereo in Montreal was amazing the time I heard it with 45 vinyls being played.

4/6, never picked 128kbps.

Used $50 Audio Technica ATH-M20X on $90 Behringer UMC202HD.

Did audio production in the past (amateur level).

Failed Coldplay, and Neil Young, the latter I argue has such poor source quality that it's hard to tell a difference.

Tip: listen to the "air" of you want to spot the difference, that's the first stuff the compressors throw away. "Air" is the ambient sound of the room, the tail of the reverb, the stuff you hear when you don't hear anything if that makes sense.

I got 5/6 with a Schiit Magni/Modi stack and Beyerdynamic DT 770s.

On the flip side of the coin, my wife regularly listens to music just out of her phone speaker and claims to not notice the difference.

Everyone enjoys music in a different way, and I'm glad there are plenty of options!

6 for 6 listening directly from laptop speakers. That being said I really had to listen intently to discern any difference between 320 kbps vs uncompressed. Everything sounds ever so slightly flat or muffled.

Oh, how well can I hear the difference in quality, from my laptop speakers? Geez, I'd have a better chance on my phone, and that would be no chance at all.

I listen to Tidal through a USB DAC connected to a preamp -> amp -> respectable speakers.

And I can definitely tell the difference. But no one can tell the difference on laptop speakers, the idea is nonsensical, or worse, spreading the idea there is no difference between uncompressed audio and highly compressed audio.

6/6 on poor laptop speakers, but based on a side-channel - the one that takes a second to load is the WAV. Actually listening, it's tough to make out 128kbps MP3 vs WAV on this setup - there's a slight difference, but only when listening closely and directly comparing the two samples.

Generally I can usually (depending on the song) pick out 320kbps MP3 vs WAV on reasonable headphones (ATH-M50x), but I don't have a pair to hand to test these samples.

The only one I got dead wrong was Coldplay. That also happened to be the furthest from my usual listening habits, I don't know if that had any effect.

For the others I picked at random one of the two that sounded the most similar. In no instance could I tell 320Kbps and WAV apart. Could have been just luck that I didn't pick 128Kbps more than once.

Hardware is Bose QC II on low noise cancellation, connected by wire to a MacBook Pro and ears, one of which can hear crickets, the other not.

5 out of 6, and the wrong one was 320 kbit, but that one was very close between them.. between wav and 320 mp3, it's difficult enough that I can't put firm words on it.. the wav is "more" or "larger" or "fuller" without there being anything specific I can point at.

So how many of these selections are wry jokes?

The inclusion of the Neil Young got a smirk from me because of his past antics with promoting his “Pono” high-res music service. (2012-2017)

The Jay-Z is there for similar reasons, explained in the opening: he was promoting his own high-res music service. (Still theoretically extant but he’s gone back to streaming on Spotify.)

And the Suzanne Vega track is there because it was one of the three or four songs that was used as a benchmark for the development of the MP3 format; if there is any song that should sound good as a shitty low-res mp3, it is Tom’s Diner.

But what about the Coldplay? And the Katy Perry? And the Mozart - and maybe this particular recording of Mozart? Are any of these entwined with the history of lossy audio compression, or attempts to sell lossless compression to a wider market than audiophiles?

Some past threads:

How Well Can You Hear Audio Quality? - https://news.ycombinator.com/item?id=9654758 - June 2015 (71 comments)

How Well Can You Hear Audio Quality? - https://news.ycombinator.com/item?id=9743877 - June 2015 (1 comment)

How Well Can You Hear Audio Quality? - https://news.ycombinator.com/item?id=9688095 - June 2015 (2 comments)

Can't say I did well on the test. Actually made a decision on 2 of them and got the 128kbps on one and WAV on the Suzanne Vega (which must have been luck). The rest I just clicked something to read the snippets.

And I've got HD598s and was listening through a UMC1820 interface.

That said, I'll take lossless FLAC files for a download any day, simply because it's future-proof in a way compressed audio isn't. I'm quite happy to listen at 320kbps on Spotify most of the time, or transcode to 96kbps Opus any FLAC I have when I load it on my phone.

One thing to consider is that this for listening. However as soon as the audio needs to be manipulated e.g. use it in a film, sample it, EQ it heavily, adjust the dynamics DJ with it, low quality audio (especially mp3) gets exponentially more noticeable.

In the end that compressed great song will still beat the uncompressed lame one, but how noticeable compression gets also depends on what you do with the material.

Music is engineered to sound good on every quality, up until the point when it is heard on low quality devices and adjusted so it is still perfect. It causes finding differences so difficult. It is not only that compression is so good, or that we don't hear the difference. It is also about the music being made in a way we shouldn't hear it.

This test result is a tribute to sound engineers who make music good on every device.

So despite a lot of comments here saying that can't tell the difference, there are still people out there determined to pay more money for Tidal's "HiFi" sound quality and with both Amazon and Spotify (Apple also?) also now coming out with "lossless" quality options. It may be considered snake oil by some but snake oil sells (not news).

Happily surprised I managed 5/6 on the sony xm1000m3 (man I hate their naming) over bluetooth. Didn't expect that!

(my audiphile dad did 'train' me... made me listen to the highend stuff he's been building/crafting since I was little. I owe him disliking 128kbps on almost every setup ]: I'm fine with 320kbps for most things though)

I was able to distinguish the 128 kbps file every time (they sounded "muddy" to me), but the 320 / uncompressed wasn't distinguishable.

In particular, the coldplay song is one I've listened to alot, and I was able to instantly tell the difference in the instrumentation.

Listening on Apple Airpods Max, audio is getting piped through a virtual audio device via Loopback.

From what I recall about compression, the artifacts are at the high frequencies which also tend to get muted with age / sound exposure. Rather than play the game of determining if you can hear the difference, are there any diagnostics that can work backwards and determine the max bit-rate one could benefit from when listening?

Interesting, whenever I failed it was between 320kbps and uncompressed wav. I know that I did a blind test about 10 years ago and could distinguish between ~192kbps variable bitrate and lossless audio with my left ear, mostly influenced by percussion (left ear for some reason can hear 21kHz frequencies).

I do experience differences on my HifiBerry + MFB (motional feedback) speakers between 320bps mp3 and flac. Mainly the crispness of e.g. percussion and voices is subjectively much, much better.

This might of course be the IKEA effect, but I'm quite happy nevertheless with my flac collection.

one subtle point they missed: if your headphones are wireless (as it seems many are these days, as are car audio interfaces), there's a fair chance the audio is getting compressed again by a different lossy codec.

I got 4/6 with a pair of Takstar Pro 82's (nothing special) connected over HDMI from my monitor.

I missed Neil Young and, surprisingly, the classical music, although I suppose that classical music is used extensively when testing codecs.

Listening on my bluetooth Plantronics headphones in a fairly noisy environment, I can't tell them apart at all.

That said, maybe the headphones compresses the wav as well...

Curiously I picked 320kbps all but one occasion. Perhaps all those years of listening to MP3s have conditioned me to prefer the MP3s?

I used a pair of Powerbeats (Bluetooth earbuds) and picked either the 320 or the Uncompressed WAV (3/3) on all of them.

I found a cheat. At least on my system, the WAV version takes ever-so-slightly longer to begin after you click play.

I am sitting in a commercial studio today, working on a mix. I have been sitting here for most days over the past 8 years. This is my job.

The equipment in front of me in the control room alone is worth around 90000 (SSL board, RME conversion, Neumann kh410s for monitoring and serious acoustic treatment) — nothing super crazy when it comes to studio setups but certainly in the 'above average' area of things, even in this profession and certainly when it comes to anything normies are gonna experience. My most expensive headphones (I own around 10) are Sennheisers for ~1.5k, powered by a Phonitor, which cost around the same.

To toot my horn, I have been fairly successful at this. I work with chart pop/alt/indie musicians on a regular basis. I am actually in the rare position to have made a good living off of this.

While working on a Mix I am intensely focused on listening, to the point where I will not notice when being talked to. I am listening to and passing judgement on over 100 individual sources of audio running at the same time. I will hear if one of the 16 background vocals tracks is slightly out of tune. I will hear if there's an unpleasant resonance on a guitar somewhere for a split second. Absolutely nobody, including the artist, will ever be as acoustically intimate with the whole thing as I am during this time (this is probably true for any notable mix engineer).

Considering all this, the highly intensive and focused process, the considerable amount of money carefully spent (in contrast to audiophiles just throwing money at something because of what I would imagine are insecurities and boredom, I have to make a profit and thus spent my money carefully even) and my experience, I rank the difference between high quality cbr/vbr mp3 and lossless as so fucking negligible, that I will randomly chose either when making exports to compare my future working status to.

To elaborate on what this means: I will make an export, place it in the project, continue working on it, and then at a later point solo (= switch to) to those exports to check if I am going in the right direction, if I improved on the critical parts, if something got lost on the way.

To be absolutely clear on this: I would cost me an extra half second to make sure I always have a 24/96khz export of that mix as my reference instead of a maybe-i-dont-care mp3. I do not invest that half second. I don't know how to be any clearer than this.

Do not confuse this with me not being able to probably spot the difference. It's just that it is about a skill as spotting a difference between two apparently white walls. I am sure with enough effort and attention to the task anyone would be able to make their brain tune in on the the differences. But why the fuck would anyone do that? There is no upside. No, it will not make you a better music listener or the music more enjoyable. If anything you are doing more focusing on the wrong things.

Every now and then I stumble across this topic online and what used to be insecurity has turned into perplexion. I have no hope for or illusions about that the people who have trained their brains to search for mp3 artefacts (well done, I suppose) but maybe I can reassure some who seem to be on the fence on this topic: Do not dwell. This is a great way to waste time and money and get absolutely nothing of value in return. Aim for headphones you enjoy listening on. They are probably not that expensive. Make sure they have replacement parts available, specially for the ear pieces and cable.

The Neil Young track sounds muffled even in the uncompressed to me.

Doesn't it depend on decoder used by your browser? Will these sound different if downloaded and player with a music player software?

Difference is too hard to notice in voice only audio example because it does not have that variation/detail in it anyway.

3/6 they all sounded exactly the same to me. I just guessed.

How is this still a debate? Just keep a .wav backup and convert it to modern formats as they come. At this point I can only assume people listening to actual mp3's pirated the songs.

A worthless experiment. Audio is only as good as the weakest link. All phones will be insufficient for this test. All computer jacks and all monitor jacks will be insufficient. Most headphones will be insufficient.

Only upon using a dedicated headphone amplifier with a moderately good pair of headphones should the answers become easy.