> Of course, it's not only the compression that's at stake-- it's also the mastering quality.

HN threads on lossless music so often descend into "well, you couldn’t hear the difference". But historically a key reason for people to prefer e.g. SACD releases of recordings is not because one can hear the extra frequencies of this format, but because the SACD release – being targeted at people with a good stereo and silent listening environment – is mastered with more dynamic range. The mainstream release of the same music, on the other hand, often features a different mastering with levels pushed up, because they assume that ordinary people will be listening to music in noisy environments, like through earbuds in the metro. This is often a problem for re-releases of 1960s recordings of jazz where the producer and engineer really exploited dynamic range, but that is lost in the most recent of the digital-era re-releases.

(Another reason to prefer SACDs and Blu-ray releases of recordings is for the possibility of 5.0 surround sound, because of course CD and mainstream downloads or streaming are limited to mono or stereo only.)

Absolutely agreed. People spend so much time arguing about whether we're "past the limits of human perception" on sound or video. They want to quote Nyquist and check frequencies.

Spend a million dollars on audio equipment. Turn on your speakers, play some chamber music in insane multi-channel high-res audio, and invite a friend over.

Your friend does not think you hired a string quartet. Your friend thinks you have nice speakers.

If your friend is naive and easily fooled, do an A/B test by actually hiring a quartet and alternating with the speakers.

There are a thousand compromises between the source of music and your ear. What is worth paying for is a recording where those compromises were made well. The compression level is a very small part of the picture.

Theoretica Applied Physics and BACCH Labs audition this exact demo -- A/B test by actually hiring a quartet and alternating the musicians playing with the speakers silent and the musicians silently air-playing and with the speakers playing. With a perfect image you literally cannot hear the difference.

https://www.theoretica.us/ https://bacch.com/

I found a video in one of those links from Princeton. It's a narrative of this guy developing 3D sound that works in normal 2 channel laptop speakers.

It's quite impressive when they demo it. There is a moment around 1:20 that plays flies circling around and my cat next to me didnt care about the normal recording but he stood up with his ears out when the 3d version played.

https://www.youtube.com/watch?v=SQmQD27uCt0&t=3s

I know it's probably old hat by now (the video is ~10yrs old, yes 2010 was over a decade ago) but it still blows my mind.

This is pretty old hat now. IRCAM was experimenting with spatialized sound on a stereo medium already back in the 1990s. The software they came up with was used e.g. for the Deutsche Grammophon recording of Pierre Boulez’s piece Répons where sound is moved around the hall.

Unless you have bad hearing everyone would be able to hear the difference on speakers and live unless live is defined as someone playing far away on a scene or something, no matter how expensive equipment you have. It is impossible to have it sound like musicians playing in the same room. To even come close you would need several speakers per instrument playing in multiple directions and some playing sounds of the people moving, breathing and talking too. Not even a simple snare drum being hit at intervals can be reproduced by a normal stereo setup to sound as if it were in the same room unless it's at distance. Remember, we play music at home so the definition should be an orchestra two to three meters away, not an orchestra ten to twenty meters away up at a scene.

Someone with something to sell is a very unreliable source btw.

> Your friend does not think you hired a string quartet. Your friend thinks you have nice speakers.

When I first got nice high end speakers (TY Craigslist) I did this with an acoustic guitar CD.

The amount of detail I could hear, and by detail I mean "close my eyes and hear every details of the musican's fingers moving across the strings" detail, was incredible.

That is also the day that I learned MP3 encoders of the time (~8 years ago) still had bad encoding artifacts even at 256kbps.

After hearing enough clicks and pops, I ended up having to re-rip my CDs as FLAC, or in some cases buy the CDs so I could rip them.

Now days encoders are a LOT better. You don't get clicks and hisses except at 128kbps.

Also 90% of my listening is done through a wireless headset (...) through spotify (...) so quality has taken a serious nose dive anyway.

That said if I want to listen to music as an all immersive activity, yeah, lossless through high end speakers. But I do that maybe once every other month for half an hour.

> A/B test by actually hiring a quartet

That's easy. The one with room echos, washed out highs, and missing bass is the live.

This may be true about popular music, because this is how we came to know this style of music - live performances (without proper sound engineering) sound worse. It is completely the opposite for classical music. There's still no way to replicate the sound quality of being close to an orchestra.

>There's still no way to replicate the sound quality of being close to an orchestra.

That's why I always tend to bail out of audiophile discussion but I'll bite: why?

What's so special about the soundwaves going into your ears when you sit close to an orchestra that couldn't be reproduced with good audio equipment?

If you're talking about the experience itself of siting next to performers then I wholeheartedly agree, but that's the problem, it's no longer something that can be measured and objectivized.

And that's entirely fine, but I think there's a trend among some people (and especially the type of people who frequent this forum) that deem that if a feeling or emotion can't be objectivized then it's effectively worthless or irrational or something like that, so you see people grasping at straws to justify their emotions with a pseudo-scientific explanation. I find that frankly sad and quite toxic in a way.

Music is art, the enjoyment we derive from it can't be measured in kilobits per second. That doesn't mean that we need to make up pseudo-facts about acoustics to justify our preferences.

I also thought the same, then I went to listen some orchestral performance at the Rome Auditorium.

During an interval I was sitting in a shitty corner up on some balcony, pretty far; but I could hear the voice and discern the words of a girl chatting down in the platea next to an entrance.

A single mouth, chatting, among hundreds other voices.

During the concert you could hear the individual performers’ instruments, the strings rubbing, the valves clicking. I was blown away

A critical factor in this experience was the actual room. Anyone dealing with audio and hifi will confirm how much great acoustics depend on a well treated room.

While most of us can't live in such a purpose-optimized room I think OP's point was that a bunch of high quality speakers in that very same auditorium might have yielded very similar results to your experience.

> That's why I always tend to bail out of audiophile discussion but I'll bite: why?

Disclaimer: I am definitely not an audiophile.

The low frequencies that we feel with/through our bodies feel entirely different at a non-electric live concert as compared to what we feel through speakers. No headphones can reproduce such effects. For sure, there are speakers that can reproduce such sounds faithfully, but how often do the audio production guys use such speakers, and how often do they try to control for the feeling in their bodies? Not often, I imagine.

> but how often do the audio production guys use such speakers, and how often do they try to control for the feeling in their bodies? Not often, I imagine.

Yes, but we're talking about 5 digit and up audio equipment and listening to the top operas of the world, which definitely do have guys that know how to record. Your 10$ (or 250$) headphones combined with a CD from a smaller producer will not get you all the way there, I agree, but once you're in the insane high-price audiophile world, this becomes possible.

To start replicating the sound of an orchestra, you'll at a minimum need one speaker at each location where a player is, playing a single channel (the instrument).

But even this is simplifying things, because each instrument has a different body shape, different sound projection, which would need to be replicated by specialized speakers. Also, the room has an important effect on how the sound waves travel and hit you from different directions. At the end it is probably cheaper and easier to pay a ticket to have this live experience.

Of course not, what you need is model the way your ear handles sound coming from various locations then model this and reproduce it. It's like VR, but for your ears. That's how ASMR effectively works. If you want full immersion you can add head tracking so that the sound "rotates" around you when you move your head. I know that Apple offers that with their latest headphones.

Admittedly in order to do this you need to have post-processing specific for every person since we don't process sound exactly the same depending on our physiology. I know that Sony attempted something like that with its "3D audio". I think it's a bit of a gimmick myself, but it's technically doable.

> Of course not [you just need individual head related transfer functions for everyone and a fast DSP]

I agree that it is probably possible in theory, but aren't we talking about actual equipment and recordings that people can buy?

You can buy it; the keyword is 'binaural' recording/headphones.

https://en.wikipedia.org/wiki/Binaural_recording

Except this is a long way yet from what the GP was describing, although it is an attempt in that direction.

What if you only want to replicate the sound of an orchestra as heard by an observer at a specified point? Why wouldn't two microphones be sufficient? That's how it's perceived, after all.

You can get really nice sound that way, but to get the full effect you need to take into account the shape of the head and ears of the listener. After all, these are important factors that help us place sound in space.

https://en.wikipedia.org/wiki/Head-related_transfer_function

Your ears are highly directional devices and an orchestra in a concert hall creates a highly complex soundscape that is impossible to replicate with 2 channels. Location awareness (i.e. 2 mics where your ears are) is just a small part of the problem.

So it sounds like the experience can be fully captured with two microphones mounted in the ears of an acoustic dummy head, as long as you're willing to accept a single position and orientation for the observer. Personally, that would be more than good enough fidelity.

I suppose then you would need to have the transfer function of that specific observer's head and ears at that specific point. Perhaps some time in the future there will be personalization of the transfer function so that you can hear the sounds as you would hear it at that spot.

> What's so special about the soundwaves going into your ears when you sit close to an orchestra that couldn't be reproduced with good audio equipment?

An orchestra is a large number of instruments spread across a wide stage. Each of those instruments is its own sound source. The audio cortex in the brain is highly tuned to understand things like the 3D location of sounds from cues generated by factors like the individual shape of our ears and how sounds bounce around and down into the ear canal.

Now I'm sure you'd agree that sitting in front of two speakers X metres apart, or with headphones on, no matter how good those speakers / headphones are is a different set of sound waves.

Some people are absolutely able to determine the difference between those sound sources. They're likely to have listened to a lot of music. I think these discussions get derailed by the blanket "people can" or "people can't" statements rather than thinking about who might be able to make those distinctions. It might be that the majority of people can't make that distinction. But that doesn't mean that no-one can.

It's like folks are talking past each other here. There appear to be a substantial number of (I assume younger)posters that have never been to something like a symphony orchestra and don't see the point when they can geek out with FLAC at home. Maybe the orchestras should advertise "Come see our array of 120 separate high end speakers" and the kidz would respond.

"Why go to a museum when I can look it up on Google?"

> What's so special about the soundwaves going into your ears

tongue in cheek answer: nothing. It's your ears that are special.

Less tongue in cheek, while we have got pretty good at making microphones, they don't behave anything like human audio system; this makes it really difficult to reproduce what we experience when generated a different way (e.g. multichannel playback).

There is nothing "special" about classical music here either, it's just got a lot of complexity (from multiple instruments, and room dynamics) and dynamic range.

I see your point: if we could a) design microphones that capture everything going by them in a neutral way and b) design speakers that precisely reproduce everything they are sent in a neutral way, and c) set up a room to reproduce things in a neutral way for your particular position ... then this would work. We can't actually do any of those things.

If you've never heard an orchestra in an auditorium with proper sound design you're in for a treat. Two or four speakers are simply a poor alternative to dozens of live performers around you.

Maybe if you move your head just a little bit with a live orchestra, the change is different from what happens with a few speakers (and often in a different room)

There is a scientific reason reggae dub shows feature massive woofer stacks built out of specific types of wood…

First time I had been to an actual classical concert I thought the walls need more dampening, mics placement is terrible and couple sliders on the right side of equalizer could be turned down a bit.

By the time It was over I completely "got" it though. I was internally complaining that an actual fishing port at 5AM don't look like a Monet copy, not watered down, skewed, or idealized.

Good point, my exact experience of live concerts

I have been surprised in both directions where what I thought was live music was prerecorded and what I thought was prerecorded turned out to be live music. Which shouldn’t be that surprising, audio equipment has gotten really good and audio standards are based around what people notice.

The more realistic scenario where it might actually make a difference is DJs playing on huge high-end club soundsystems where compression artefacts might not only effect what you can hear, but the reverberations you can physically feel, and where there might be post-processing effects applied to the source material that show up the compression artifacts.

Yep, it was a sad day when they switched from vinyl 45 singles to digital.

Meanwhile I'm listening the output of an amplifier playing a distorted guitar. I might listen to an acoustic guitar, and there might be a small horn section. Of course there's going to be drums. So it's super-simple to tell if it's live or a recording - the volume levels are going to be insanely different!

> Your friend does not think you hired a string quartet.

To be fair to the audio system, a significant reason why they'd sound different is because of room acoustics—and we rarely listen to string quartets playing in residential living rooms.

What you said is totally true but it's not relevant here.

Apple already have all the lossless files on their server. Just that before they converted them to lossy format before sending to the user, and now they serve the original.

There is no new master / new re-release involved in all these.

> Apple already have all the lossless files on their server. Just that before they converted them to lossy format before sending to the user, and now they serve the original.

Do you happen to have a citation for when this changed? According to the "Apple Digital Masters" Technology Brief[1], the masters delivered to Apple have historically been AAC files.

[1] https://www.apple.com/itunes/docs/apple-digital-masters.pdf

I might be misunderstanding, but I think that PDF is about how to master your music such that it will sound good after they convert it to AAC. It doesn't actually speak to the audio that's sent to them.

Can't source anything except our own deliveries to Apple, but we've been delivering PCM since ~2018. Likely earlier, that's just when the code got moved into source control. That actually predates the copyright on that PDF.

In general, of the maybe 150-odd DSPs we deliver to, as far as I know all but a handful have us deliver PCM. The remainder have us delivering flac.

> Can't source anything except our own deliveries to Apple, but we've been delivering PCM since ~2018.

Cool, I take your word for it. Thanks!

The referenced document only talks about AAC masters, and the tools at https://www.apple.com/itunes/mastered-for-itunes/ only create AAC masters. The simplest explanation is that the public materials and tools are just sadly out of date.

I'm pretty sure nucleardog is correct and that Apple wants 24/96KHz minimum sent to them; the PDF you linked to actually says that under "Best Practices" -- "To take best advantage of our latest encoders, use only 24-bit sources and send us the highest-resolution master file possible, appropriate to the medium and the project." Then Apple encodes their AACs using the tools and methods described in that document. If I understand it correctly, they're basically trying to ensure that the downsampling from 24/96 to 16/44 preserves as much information as possible. (Does it really make a difference? No idea. But it does mean that Apple has a whole lot of music they can re-encode at varying quality levels.)

It's odd because I've read that before and made an assumption that it was talking about AAC mastering, but it only implies that. There's nothing that comes out and says what they want! There are some parts where they suggest higher fidelity sources, but mostly it comes down to:

    Apple Digital Masters Droplet 
    You can use the Apple Digital Masters Droplet to automate the creation of 256 kbps 

AAC encodes.

There isn't actually anything about what you're supposed to send them, only the recommendation to use their tooling which seems geared to produce 256k AAC files.

Ah, that makes sense, thanks!

For anyone bored and looking to explore: https://dr.loudness-war.info/

Note that all entries that have the format Vinyl on that site can be ignored, as the dynamic range calculations don't work well on audio recorded from vinyl records.

Well, I'm glad I've been using it to pick out which pressings to target haha.

I agree. For me the limiting factor is not necessarily whether it is poodles high quality encoded or lossless. For "normal" music the bottleneck is often the recording. With good headphones it is easy to differantiate good from not so good recordings. I For me a lot of recordings on Tidal/qobuz did not sound better than HQ on Spotify

> I and many others running good high end gear can repeatedly score well on these tests.

I, too, run high-end gear and cannot.

The problem scope isn't simply reproduction accuracy, it is also one of taste: Different compression can elevate/alter parts of the sound, and your opinion may not lean to the highest bit rate but rather some other coloring of the music.

The fact you've run this test "repeatedly" actually hurts your argument, since you may just be learning what to expect in your environment. The crux of the article/discussion is that people cannot blindly pick the highest quality, which they cannot (since the different colorings of the music are subjective anyway).

This is crux to the argument. If you know what to listen for, you can pick out the "bad bits", since you know that is where the issue lies. Once you know how to identify bad bits, then you'll be able to "hear" compression artefacts because you'll be primed for them.

In A/B tests this is true, I _can_ reliably (not 100%, but way better than random) tell you which of A/B is compressed, and I can normally do this on somewhat crappy headphones, since I _know_ where to listen to pick out the compression. And this despite being partly deaf.

However, if I am just listening to music I never notice these artefacts, since when listening to music, I'm not trying to determine which A/B stream is wrong. Most of the time, the music is just playing in the background anyway.

Problem is, sometimes when you learn how to spot a deficiency, you can't unlearn it and stop spotting it. It's like with kerning[0] or image compression artifacts. Everyone has a different level, and I haven't measured mine, but even when casually listening to music - including music from bad speakers in a waiting room somewhere - I can identify too badly compressed music and get immediately annoyed by it.

--

[0] - https://xkcd.com/1015/

    It's like with kerning

Kerning is a great comparison, I think.

Studies have shown that well-compressed lossy audio is nearly indistinguishable from the originals. This is certainly true for me. I can sometimes pick out the compressed version if I know the track well and can look for specific details, but generally I'm absolutely not consciously aware of any compression artifacts.

Similarly, I doubt most people could pick out small problems with kerning and other typographic issues.

And yet...

I strongly suspect folks would enjoy a book with excellent typography more than they would have if it had slightly less-good typography, despite not being consciously and acutely aware of the differences. I believe there may be parallels with lossy vs. lossless compression when it comes to music, and with many other experiences as well -- you may not be able to tell the difference between cookies baked with cheap butter and grass-fed butter in an A/B test, but might you enjoy one cookie more than the other anyway?

But the question is do they enjoy the book with excellent typography because they're told it's excellent or can they perceive the higher excellence and derive enjoyment thereof.

I can speak to this as a graphic designer.

In the Battle of Naboo in the Phantom Menace (2001, 115m budget,) you see a CGI battle between a million zillion CGI robots and CGI aliens with CGI force fields, etc. It's a proud display of state-of-the-art CGI capability, circa 2001.

To this day, when you watch Jurassic Park (1993, 65m budget,) you're looking at a fucking dinosaur.

Great typography is like great special effects in a movie (most of the time,)— the viewer shouldn't realize it's there. "Good typography is invisible" is the most commonly repeated maxim for typographers. Typography is good if it's readable, legible, all elements on the page serve their purpose without needing to be labelled or puzzled over, and the most noticeable thing on the page is the information being conveyed. (Postmodernism challenged this in some interesting ways but for most practical purposes, this still holds true.) A designer might notice the typeface choice, paper choice, tracking, leading, margins, kerning, paragraph 'rag,' line length/column width element hierarchy and placement. The user probably won't notice the greater reading speed, lessened eye fatigue, fewer incidents of losing their place, easier scanning for pertinent bits, and things like that. But it will certainly be there.

I remember the first time I listened to a perfect-condition vinyl of something that I had always only streamed— I was immediately struck by how much more spatially-open it was. I imagine that was pretty deliberately done by the producer and with the necessary high-end loss in lossily compressed music, it just doesn't make it through. Not sure if I would still be able to tell as much of a difference in my 40s as I did in my early 30s, but I think most people would be subtly more engrossed in the music, even if not consciously. That was also sitting at home directly in front of my modest but competent stereo. The real question is context. Will the listening devices convey those subtleties? I know a lot of people are using streaming services as sources for home audio these days, so maybe it would to them? Certainly wouldn't to me as I was listening to music while riding the subway. I might enjoy having it available for a nice sit-down critical listening session.

A famous 1932 essay on the matter— The Crystal Goblet, or Printing Should Be Invisible by Beatrice Warde: https://veryinteractive.net/content/2-library/52-the-crystal...

Great question. There's no way to know outside of big randomized trials but I'd bet thousands of dollars it's the latter. (edit: surely this has been studied, right?)

Though, I'm certain it would be less "perceiving excellence" like a connoisseur, and more like reaping the practical benefits of solid typography.

While there's certainly a lot of showy typography in the world, the sort found in the body copy of books is generally focused on readability, in other words, letting the eye/brain more easily recognize letters and words. Do we enjoy a book more when our eyes and brains aren't overtaxed? It's hard to imagine otherwise -- who enjoys eye strain?

I'm less sure of it but I suspect this is true for lossy vs. lossless music as well. For example, one telltale sign of badly compressed music is when the percussion sounds more like blasts of white noise than the actual instruments -- these sharp transients are hard to losslessly compress efficiently. (It's somewhat similar to how the sharp edges of text and line art are not handled well by some lossy image formats like JPEG)

Is a drum that actually sounds like a drum going to be more pleasing to the ear than something garbled and white-noisy, even if nobody tells you anything? Generally yes, I am fairly sure.

I'd say: before being aware of how typography should look, the main difference is that reading a badly-set book will feel tiring for some reason that's hard to pin, but it's definitely not content. After being made aware, it's just annoying - if you can name the thing that's wrong, you recognize it's wrong.

> Similarly, I doubt most people could pick out small problems with kerning and other typographic issues.

The XKCD comic I linked makes an important point though: someone else pointing out the issues to you can make all the difference. At least with my experience, it literally took that comic for my brain to suddenly start recognizing kerning as a concept, and subsequently see issues with it everywhere.

> I strongly suspect folks would enjoy a book with excellent typography more than they would have if it had slightly less-good typography, despite not being consciously and acutely aware of the differences.

That I agree with, and I think in those cases, giving someone even the most basic framework to understand what is "good" and what is "bad" is enough to make them much more aware of the issues in the works they're enjoying.

Being partly deaf might be what lets you hear the compression artifacts, because a lot of it relies on masking effects.

Why does any of this matter? The point of a blind ABX test (and the whole question) is to see whether you can tell the difference between two formats. If you can, you can. Personally, I stop being able to tell somewhere around 200 kbit MP3s (with cheap equipment), anything above that is wasted space.

You can run the tests as much as you want, if the hypothesis is that "people can't tell the difference between lossless and 320 kbit MP3", even one person demonstrating that they can, disproves the entire hypothesis.

> I stop being able to tell somewhere around 200 kbit MP3s (with cheap equipment)

Yups. Me too. Even with cheap professional equipment I cannot hear the difference of lossless vs >= 192kbps MP3.

> anything above that is wasted space.

Unless you want to mix/remix/etc. Changing the pitch brings out the artifacts.

I did a blind test with a friend once, 192/320/lossless, and we both could definitely hear which file was 192 - but not which was 320/lossless.

Most of the compression in 320 mp3s is also in areas humans can't hear. I once created a 320 mp3 with a 16kHz cutoff instead of the 20kHz lame uses by default, and to me there was no audible difference - even inverting one and listening to the difference didn't produce anything audible (to me).

Of course nobody uses mp3 anymore, not even iTunes, they encode as 256 aac, which is a far superior codec and near lossless. I doubt anybody can hear the difference between the lossless file and the iTunes file.

> Unless you want to mix/remix/etc. Changing the pitch brings out the artifacts.

Certainly, you probably need lossless if you're going to do any sort of editing.

Lossless, and possibly upsampling before editing, but high end gear probably does that by default. As for bitrate compression, although I have a good musical ear, I can't find any appreciable difference from a high bitrate (320) mp3 and the corresponding lossless track, so no problems using mp3s to listen to music.

I also managed to resample my entire portable playlist (yeah, making mp3s out of mp3s, I know in some circles I could be killed just for thinking about that:*) to overcome that dreadful loudness craze that makes pretty much every track produced in the last 20-25 years destroy your ears if you happen to listen to it just after something 10 years older. Unfortunately all methods to level the tracks (mp3gain, replaygain etc.) don't work properly because they should analyze all tracks dynamic content, not just the average/maximum level, then first adjust their dynamics and later normalize their level. In other words, level compression should be applied more to softer tracks and not the other way around. So I faked a solution to the problem by reconverting my playlist after applying compression to an extremely low threshold, making clear to myself that none of those tracks would ever leave my player. Some sound really horrible, but I spared my ears.

As far as I know, mp3 frames have a "volume" field, which is what mp3gain changes to losslessly change the volume of the entire track. Could you not take advantage of that too?

The problem is the huge difference in dynamics between "old" and "new" recordings (that is, not just their maximum level) which makes really hard if not impossible to have more tracks sounding near the same level.

This Wikipedia article about loudness wars explains the problem in details. https://en.wikipedia.org/wiki/Loudness_war

Not sure you're saving your ears by doing this compared to say, turning the volume down.

You can't 'uncompress' an MP3 back to having the dynamic range it had before.

"You can't 'uncompress' an MP3 back to having the dynamic range it had before."

True. In fact I would rather compress old music to bring it to the same dynamics of the new one, then and only then normalize everything. I know it's almost a crime; I did it just on my player and would never give these tracks to anyone.

Well… maybe some cool AI would help here… We already have it for images!

We can take a jpeg and run it back to RAW?

Because I don't think we can.

>anything above that is wasted space

Having lossless copies is still useful for storage and later transcoding. Even if you can't hear a difference between a 192kbit/s MP3 and a higher bitrate lossless file, you might be able to hear one after reencoding the MP3 using another lossy format.

Yes, I should have said "for casual listening".

Considering the proliferation of music remixing social media services (i.e. Tiktok) and the cheap price of storage, maybe even casual users can benefit from lossless these days.

I think the Key point would be, could you hear any difference in those test?

If you could, then yes there is a quality differential, rather you got it right or wrong may be subject to taste. But at least we established there are differences.

But when we say 128Kbps are sort of CD Quality or CD being no-difference to 320Kbps mp3, they generally refers to vast majority of consumers, not audio experts. I think most consumers could tell the difference between 256kbps AAC and 128Kbps MP3. But beyond that it sort of goes into consumer dont give a damn territory. Although most Pros would want the highest quality possible and there are no reason why we cant deliver lossless with today's infrastructure.

I am wondering if there is a market for an even cheaper Music Streaming Services that target lower quality, 128Kbps music.

> I am wondering if there is a market for an even cheaper Music Streaming Services that target lower quality, 128Kbps music.

If you control the client and are able to update it, I think you're better off switching to opus or similar.

96kbps opus is actually quite good for music already: https://wiki.hydrogenaud.io/index.php?title=Opus#Music_encod...

I guess the question then is whether there's a market for 48kbps opus music. But that'd be assuming that a big part of the costs are bandwidth. I'd bet that's not the case.

Also imagine how little traction you'd get once people associate your brand with poor music quality. I don't think you can compete by being 2€/month cheaper.

Mobile data is quite expensive in many parts of the world. Quality is just one factor.

128kbs is absolutely noticeable drop imo.

I can't see any reason for 128kbs service unless you're absolutely starved for data. There is no other advantage.

>I can't see any reason for 128kbs

A lower price tier? Or even a Freemium model. Considering most of them are already available on Youtube. As strange as this may sound I dont think Streaming is the business model that will save Music industry. And despite the headlines ( whatever MSM likes to print these days ) most artist's real income are from somewhere else.

128k Opus is quite a bit better, if for some reason that bit rate is crucial.

Streaming definitely wont 'save the industry', it woefully underpays the artists.

You're only making that problem even worse by offering a low quality, even lower price model.

Frankly I can't imagine there would be many artists particularly happy with having huge swathes of people introduced to their music at 128kbs.

We shouldn't destroy the industry even more just so that some people can get it even cheaper than they can already.

MP3 compression is not the same thing has compression the audio effect. Compressed music in the effect sense can be enjoyable and is a matter of taste, but hearable MP3 compression artifacts is nearly always crap.

For those unfamiliar with the other definition referenced here: dynamic range compression (https://en.wikipedia.org/wiki/Dynamic_range_compression)

Been out of high end audio for few years, trying this test for past 30 minutes, and can't seem to get past 2/6 with wireless or wired or dynamic or BA or through proper amps or out of a laptop jack. Constantly falling into 320k. Maybe source isn't great in the first place, equalizer slightly suppressing low-mid and noise at the high is doing it.

I use to be a sound engineer. Those "test" usually with gear they owned and you can of course tell the difference with gear you are use to. Most gear has anything but clarity. Most people will dislike clear sound.

The POINT is blind test 128 MP3 and lossless. Still the vast majority of people don't have the equipment to hear any difference.

The Second POINT is most people prefer to have boomy colored sound and again that ruins the need for lossless. This is why people prefer old vinyl records. It compresses the sound and makes it sound better in most people's opinion. If people would listen to 80s tapes I bet they would love that sound also.

The Third POINT music is heard with background noise all the time. No way while listening to music people could tell the difference and those that can it would require great concentration. There is a limited return of investment for most people.

Personally I own a DAC and a preamp and have my hundreds of dollars headphones and my Genlec speakers that cost thousands of dollars. I hear a SLIGHT difference between 320 MP3 and FLAC and that is 100% concentration and I can hear a difference in only certain types of music. Classical Jazz and Classical music. I can't tell with pop and electronic music which is what most people listen to.

> I and many others running good high end gear can repeatedly score well on these tests.

Every time this comes up someone says this. I've never seen anyone actually doing it. e.g. a video of setting up a double blind test of this and someone actually passing consistently.

When I've challenged people IRL who claim the same thing the response is always "Well, obviously I can't do it with any of the audio gear that either of us have access to... but if I did have high end gear I could tell for sure!"

My buddy can do it. I didn’t believe him at first but he scored better than me, consistently. We’re both musicians but he’s more into lossless audiophile stuff.

We tested it out in our college dorm freshman year. Speakers were the built in ones on my MacBook Pro. I will say that the difference is extremely minor and he has said he only cares about lossless for the sake of having original media in high quality, so he can make compressed versions in whatever format he wants without losing quality. He rarely actually listens to lossless audio. What he recognized was minor clipping on some high-hat hits, and after he pointed it out I could notice it too if I really tried.

I think this is what confuses many people when they do their own tests, they don't know what to listen for and expect a more general sound quality difference over all parts of the recordings. The differences are actually in particular kinds of sounds, like high hats or piano, and only at specific moments, and only in specific music. For music that is all distortion all the time likely no one can tell the difference. Some people are irked at the slight changes in sound at specific moments compared to listening to CD's or vinyl, some don't even notice it because it is brief or those instruments/moments are not present in a particular recording.

Yeah, it's not a general loss of quality. He actually showed me a really cool YouTube video that explained what is lost with compression, where the person making the video played the same passage of a song compressed and uncompressed, and then showed visually the parts of the waveform that were clipped, and then played the clipped parts isolated. It was a lot of high-pitched, tinny noise, presumably from the upper end of the high hat cymbals and stuff. Wish I still had it handy.

So many of these posts aren't specifying the exact compression format / bitrate they used for the test though. It's obviously possible to pass at some level of compression, but what I really want to know is if anyone can pass a double blind test at a high bitrate on a modern codec vs uncompressed.

The codecs obviously represent human-audible information much more efficiently than raw PCM samples. I'm not sure why the quality couldn't scale up (by increasing the bitrate) until it surpasses 100% of human hearing limits (since there's a lot of headroom to increase the bitrate and still be smaller than uncompressed). Maybe some formats have hard-coded frequency cutoffs (that don't change with the bitrate) that are too low for 1% of the population, but there's enough format choice that it would be surprising if they all had that problem.

Factors: - Quality of audio equipment - The experience of the listener - The nature of the audio

It’s quite obvious that the average listener listening to a 128kbps podcast stream is not the issue here. Nor is the average listener listening to a 320kbps stream of a loud, noisy pop/rock album.

But there are certain types of music where it’s much more apparent. I have a trained ear and so I can’t speak for everyone. There are certain types of music where I can’t tell the difference. But some classical piano albums on Spotify sound terrible to me. The artifacts of the conpression are more obvious on sensitive, detailed music.

All this to say that a random seection of people failing an online test to some random audio isn’t a comprehensive answer to this question.

If you're interested, check out the listening tests section at https://hydrogenaud.io where experienced testers contribute results to devs trying to improve various lossy codecs.

It often depends on the source material -- some sounds are harder for some algorithms to encode, leading to increased bitrate if the codec supports VBR, or perhaps artifacts of not, or if there's insufficient bitstream available.

Unsurprisingly the newer codecs usually manage to sound better than MP3 at significantly lower bitrates. Remember the only point in lossy encoding is to save space / bandwidth -- it's literaly why MP3 et al were invented.

Not only that, biology is also against them. Somewhere between 25-30 years of age, your hearing already dips <17KHz. Where does lossy do the vast amount of its information discarding? >17KHz.

I am almost 40 and can easily hear pure ones over 17khz but can't differentiate pure noise with or without anything above 13khz.

Right around 18.5khz there's a sharp falloff on pure tones no matter how much I pump the volume. I use a loudness meter that shows me the freq spectrum to confirm that my speakers and headphones are producing a pure constant tone.

On the tests at: "https://www.audiocheck.net/blindtests_frequency.php?frq=14" I can get 10/10 on the 10khz and 11khz test, 9/10 on the 12khz test and then it's just guessing at 13khz and higher.

High hats tend to fall around 16khz and I have been able to differentiate on audio tests just by listening to high hats but that has more to do with how aggressive the codec is and not necessarily my hearing capability. I enjoy piano music and using the Apple afconvert tool I've been able to get piano tracks down to under 100kbps and they sound fine to me which is pretty amazing.

I cannot use some of the more popular Berydynamic headphones because they sound too trebly and some pop songs are physically painful for me to listen to (eg the marimbas at the start of Ed Sheeran's Shape of You are like ear daggers on the 880 headphone and unlistenable on the 990's). I wonder if some of that brand's popularity comes from older ladies/gentlement being able to hear treble in the music that they otherwise wouldn't hear at all.

Beyerdynamic headphones have a lot of fake detail (actually sibilence) that comes from echoes at high frequencies as well as their treble tuning. It makes songs mastered for radio sound more detailed in the same way a sharpness filter would work on images.

I can't tell you anything except that you have exceptional frequency reach for your age. ISO has actually standardized age-related hearing loss (yes, really), and they point to the same 17KHz point for even young adults. It actually makes me quite envious of you. I have sensitive ears - I can pick up very faint electrical circuitry sounds in screens or the delivery driver scootering into the beginning of our street despite our windows being closed, but I'd happily trade that away for higher frequency range.

Ha! I literally just sat at my PC and saw your response at 3m old. Yea I've noticed I tend to test pretty well on all the official charts but really can't hear a difference in a lot of "audiophile" stuff people claim night/day differences. I had a beat up, 10+ year old pair of sennheizer hd600's that I never got tired of listening to. I finally had them refurbished and they don't sound as "easy" to listen to.

One thing is I suffer from tinnitus or something because at night I am always aware of a constant noise or "roar" in my hearing as well as my visual noise in my eyesight. I have pretty terrible eyesight so my hearing circuitry may have just amped the volume a bit to compensate (like upping iso on a camera).

In both our cases I wonder how much of this is just our brains have learned to pick things out of the signals are ears give vs our ears actually being more/less capable than other people. I think my brain has just had a lot of practice. I've always loved music, practiced classical piano hours a day when I was younger, and cut my teeth programming learning DSP algos for audio processing. I recently did a hobby project involving reproducing NES music from ROMs and that involved just listening to raw square/triangle waves and literally just following intuition to get everything to sound right. I'm willing to bet my case is one more of training than raw ability.

> I tend to test pretty well on all the official charts but really can't hear a difference in a lot of "audiophile" stuff people claim night/day differences.

Part of that is also your brain. Master sommeliers can tell you which vineyard the grapes in a wine came from, sometimes even which side of the vineyard. Yet give those sommeliers a white wine with red coloring mixed in, and they won't immediately notice.

I've actually met one other person that said what you say, if they really try they can hear a minute difference with cymbals and hi-hats in regards to lossless vs high bitrate lossy, but they also said they had to strainingly focus on the listening and they would never listen nor enjoy music that way.

> In both our cases I wonder how much of this is just our brains have learned to pick things out of the signals are ears give vs our ears actually being more/less capable than other people

A little bit of both, I think. Your ears are naturally more sensitive, so you put the volume on your speakers a lot lower, which teaches your brain to pick up low volume stuff better. I'm always worried the downstairs appartement can hear my 'loud' TV at night, but they've told me they can't hear a thing.

I highly suggest that people check out/try out "ABX comparator" plugin for Foobar2000 player. You give it two files (say a 128kbps mp3 version and a lossless flac verison) and it runs you through a double blind test of the two versions, quizzing you along the way. It's not just "listen to A" "listen to B", it gives you lots of options to listen to them and compare them in different ways, including obfuscating what track is A and what track is B.

If you think you can really hear fine differences in audio, it can be very eye opening (ear opening?).

+1 I have nothing against NPR but that has like 5 samples and absolutely nothing about methodology.

As far as I know, the goto test is: http://abx.digitalfeed.net/

They have a few other tests heres: http://abx.digitalfeed.net/list.html

I wish people would stop posting this useless test.

It compares FLAC with 320 Kbps AAC which no streaming service uses (Apple uses 256 Kbps) and is a very high bit rate.

And its choice of indie tracks isn't suited for demonstrating codec artefacts. At least have classical, EDM etc as well.

There's a great tool for creating online ABX tests: https://abxtests.com/

The guy who made it, jaakkopasanen, is also the author of other great audio tools - AutoEQ and Impulcifer.

There is also the theoretical case of double lossy encoding, e.g with Bluetooth headsets.

Something not perceivable when listening to AAC decoding directly punched into a wired headset vs reencoding for BT transport which could transform previously unhearable compression artifacts into the perceivable range (think editing a jpeg and resaving as jpeg often produces artifacts)

I have not seen any tests for that.

(Of course if the device is able to push the AAC stream to the headset this does not happen. This kind of thing is quite opaque and hard to get info about as to which hardware or combination of hardware would support that)

> So a very well mastered track with plenty of dynamic range will still sound noticeably better on high performing speakers or headphones

I'm quite convinced that a good part of vinyl being reputably better comes from that, especially with the virtuous loop of audiophiles being the main audience for modern vinyl, which calls for good mastering.

I've done blind tests and I can tell if I try really hard but I wouldn't even say that the mp3 sounds worse, just different. I focus on hi hats.

I do collect lossless audio for archival reasons and the ability to convert from a non lossy source.

I believe that high bitrate audio is entirely bullshit though- as Chris Montgomery has shown:

https://xiph.org/video/vid2.shtml

> high bitrate audio

Based on the opening part of that video, I think you must have meant to say high sampling rate.

Yes. Thanks. It's been awhile since I watched.

>I wouldn't even say that the mp3 sounds worse, just different.

If you're comparing against lossless, sounding different means sounding worse.

The goal is transparency: Preserving the qualities of the original audio that you can actually hear, while occupying less space.

This is largely because MP3 is a crappy, old format. Try high-bitrate Opus on most samples and you're a lot less likely to get above-chance results on an ABX test, pretty much ever, at high bitrates.

Ultimately though, lossy compression is always program-dependent ("program" here means the song, not the app). The encoders use psychoacoustic algorithms, and there is no way to prove they will perform well on all samples. So you will always be able to find corner cases where an encoder or format does worse and results in more audible artifacts; some of these might be encoder bugs, some might not. Therefore, it is fair to say that you might not want to rely on lossy compression for archival. But it is perfectly reasonable to use it to put music on devices to actually listen to.

You say good mastering makes it easier to notice encoding artifacts, but actually, bad mastering can do that too. Over-compressed stuff (in dynamic range) with inter-sample peaks above 0dB can clip and distort after getting run through a lossy codec.

As for lossless formats, though, anything above 48kHz 16 bits is complete nonsense for final delivery to consumers. All the "high resolution" stuff with higher sample rates and bit depths is just marketing bullshit, and this has been repeatedly demonstrated in trials (the well-designed ones anyway; there are plenty of terrible ones - no, comparing the 96kHz download to the 48kHz download of a song is not how you test this properly).

Bit depths above 16 are useful during production (because quantization noise accumulates and is boosted by things like dynamic range compression); in practice you want to record at 24 bits for headroom reasons, and process in 32-bit float because there's no reason not to with modern computers (and many practical advantages, e.g. ~infinite dynamic range). Sample rates above 48kHz are less so; the main reason to use them is to avoid aliasing artifacts, but it is usually much more effective to do that with well-designed DSP algorithms that include internal oversampling, as opposed to just doing everything at a higher sample rate. Unless you're doing extreme pitch shifting; then higher sample rates may make sense.

Videos everyone interested in audio production should watch:

https://xiph.org/video/vid2.shtml https://www.youtube.com/watch?v=-jCwIsT0X8M

The little tidbit at the end of the second video about how 48kHz is, in some respects, twice as good as 44.1kHz, is quite neat and not something I'd realized before.

> "This is largely because MP3 is a crappy, old format. Try high-bitrate Opus on most samples and you're a lot less likely to get above-chance results on an ABX test, pretty much ever, at high bitrates."

I'll add to this that "high bitrate" for Opus can reasonably be considered as 128kbps or above, generally 160 or 192kbps are comfortably beyond what is needed to achieve audible transparency.

For anyone who's used to MP3 and just doing everything at -V0 or 320kbps, it's extremely impressive to hear how good Opus is, and how low bit rates you can get away with. Even at 24kbps it's serviceable for music, it sounds no worse than a slightly worn cassette tape.

It's a seriously impressive codec, both for speech and musical content.

> it's extremely impressive to hear how good Opus is, and how low bit rates you can get away with

24kpbs is perfectly adequate for voice/podcasts/audiobooks

I wish Apple devices support Opus.

> This is largely because MP3 is a crappy, old format. Try high-bitrate Opus on most samples and you're a lot less likely to get above-chance results on an ABX test, pretty much ever, at high bitrates.

This is such a key point: MP3 has known weaknesses with certain common sound profiles (e.g. sharp transitions such as a cymbal or snare drum) which aren't really fixable. For me the threshold was AAC 256kpbs where I stopped noticing artifacting.

> This is largely because MP3 is a crappy, old format. Try high-bitrate Opus on most samples and you're a lot less likely to get above-chance results on an ABX test, pretty much ever, at high bitrates.

On the whole your comment is great, but I think this point might be revealing of either your age or the last time your looked at mp3 encoders. LAME has been great for at least a decade now. Other than (possibly) extremely rare killer samples, you're unlikely to be able to ABX a LAME encoded track at 256 Kbps. I certainly can't, and I have pretty decent gear and have been listening to (and working on) music for a very long time.

Back in the 90s, all mp3 encoders were shit. Xiph's Monty, in the post you link, mentions being able to distinguish between them using only their results as a party trick. Even in the early 2000s you needed 256-320 Kbps to have a shot at transparency. But now, most music is transparent at `lame -V2` settings, which gives approximately 192 Kbps results. I can still ABX a very small number of songs (last I tested), but even these go away by `lame -V0`, which is the highest quality VBR mode. HydrogenAudio, a trustworthy source as far as "audiophile" claims go, says that anything from `-V0 to -V3` should be transparent under most conditions. https://wiki.hydrogenaud.io/index.php/LAME

AAC and Vorbis are certainly better codecs, achieving transparency most of the time at closer to 160 Kbps. Opus is even better. I've never been able to ABX a 128 Kbps track, and music sounds great even at 96 Kbps (for stereo!). There are supposed to be killer samples for all four, forcing them to require a higher bitrate to achieve transparency, but interest in finding these seems to have dropped.

LAME is great, and in fact it's better than e.g. the ffmpeg AAC encoder at equal bitrates, which is telling. But the MP3 format has fundamental limitations that no encoder can fix, and Opus is unquestionably better in every way. What I'm saying is I'm pretty sure you can find killer samples for 320k LAME (under very good listening conditions), while that's a lot less likely for something like Opus.

Yeah, 96kHz/24bit is only really useful for production, since it makes it more foolproof and harder to mess up your recording.

But storing music over 48kHz and 16bit for comsumption is really just waste of storage space and pure snake oil. 16 bit give a dynamic range of 96dB and I'm not aware of any recording outside of experiments that take advantage of it. Even the most expensive speaker / headphones will distort terribly if you play them on the loudness level where this dynamic range would matter...

And most "high quality" recordings with a high frequency range (96kHz+) only really add ultrasonic sound and mostly noise artefacts that don't correlate to the music recorded (and that no speaker / headphones can even reasonably, without high distortions, reproduce, if they even pass the LPF of the reproduction chain).

So yeah, I also think that 48/16 is all we need for optimal consumption. If the format is lossless, it's also nice for archival reasons.

You may find this interesting. The author argues against 96khz+ sample rates, but pegs the optimal rate to be around ~60hkz (which of course doesn’t exist in practice).

http://www.lavryengineering.com/pdfs/lavry-white-paper-the_o...

MP3 with a good compressor and reasonable bitrate has repeatedly shown to be indistinguishable from uncompressed. The reason to avoid mp3 is to be able to squeeze the bitrate and save on storage bandwidth - but that has mattered less year on year.

I'm personally baffled and irritated by this move from Apple. They'll be selling healing crystals next.

> MP3 with a good compressor and reasonable bitrate has repeatedly shown to be indistinguishable from uncompressed.

With most samples, yes. The issue, as I said, is that psychoacoustic codecs are basically heuristic by definition, and so you can pretty much always find counterexamples. It's possible to say 320kbps MP3 is indistinguishable from uncompressed for most samples; it is not possible to say it is indistinguishable for all samples, present and future.

It's also the case that lossy encoding is a bad idea if you're going to be further processing the audio in any way; repeated transcodes definitely start bringing out the artifacts more. So lossless music is always something worth having as an option.

Now, when you start talking about "high-resolution" 96/192k 24b stuff... yeah, that's just as good as healing crystals.

> it is not possible to say it is indistinguishable for all samples, present and future.

Are there likely to be real-world cases where this happens - and the effect is detrimental enough to actually matter to anyone?

Surely we're debating "when is it good enough for anyone sane and reasonable?" not "when could a hypothetical oracle be fooled 100%"?

I've commented in the past about a song that is annoyingly altered in 320k MP3: https://hn.algolia.com/?dateRange=all&page=0&prefix=false&qu...

Do you have a FLAC of that around or know where I can buy one? I'm quite curious to try it myself, and see if I can figure out what happened. I wonder if it's an encoder bug, a real format limitation, or a problem with the original song (e.g. a clipping ISP issue, which would be solved by slightly reducing the volume before encoding).

That's actually quite interesting.

It is useful to make the distinction, because some people do want to have the highest available quality (and this is not, inherently, a bad thing) - having a hard line that separates out the minor, but nonzero effects, from the pure snake oil, is useful.

And as I said, really, the main reason to avoid lossy compression at this point is due to generational loss and post processing. Lossy compression should be considered a final processing step - what you do to store music that is then going to be delivered directly to a listener, unaltered. If you're going to do anything else, you'd do much better with a lossless version.

That is not to say, of course, that if all you have is a lossy version, it is a major problem :-)

For the record, my lossy format of choice for e.g. putting stuff on my phone is 96kbps Opus. Even that much is excellent for casual listening. But I much prefer to keep lossless FLACs as my primary archival storage - not just because that way I can take advantage of better compression formats as they come (e.g. how I moved from ~130k Vorbis to ~96k Opus), but also because compression makes a massive difference with certain kinds of processing and editing which I sometimes enjoy doing. All the psychoacoustics go out the window if you start doing things like subtracting instrumental mixes from full mixes to get vocal tracks out.

While I can understand the frustration, it’s worth noting that they’re not increasing the price of subscription and giving people additional options. To me, it’s more like trying to tell people with 4K screens to watch 8K content, but what do I know.

These sorts of tests are BS in my opinion. How often are you listening to the same track 3 times across different quality settings in the same listening session?

The only "real" test for these sorts of things is

> You've got a new device. Your streaming app defaults to low-quality upon installation. How long until you notice this?

There are plenty of things that can poison the user experience that don't rise above the sub-conscious. Just because you don't notice something doesn't mean that if that thing is improved, you won't have a better experience. E.g. an added 15ms latency on a website.

I'm actually quite skeptical random people with high end gear can hear the difference between 320kbps and lossless. It has more to do with ear training, mp3 compression is not really on the same axis than what high end gear give you (basic equipment can have a large frequency response).

On the topic of ear training and hearing differences most people don't hear, I had a recording I was mixing down that featured a squeaky kick pedal on the bass drum. I spent more time than I should have trying to scoop that out of the mix, and in the end, I could still hear the squeaky kick pedal, but no one else in the band could.

Later that week I put in a CD with songs I'd heard hundreds of times, and to my absolute horror, I could now hear the squeak of that drummer's kick pedal, like an icepick jabbing me in the ear. It had been there the whole time, pressed to millions of CDs, broadcast over countless radio stations, but only became apparent to me after intense focus on that particular sound.

I imagine that people who claim to hear the difference between 320kbps and lossless experience something like that. There may be a particular, narrow band of frequencies that sound slightly more modulated in 320 when compared to lossless, but unless you make a concerted, drawn out effort to find that, you'll never know. And if you do make that effort, you can end up ruining your otherwise pleasant experience for absolutely no good reason.

Anyway, after the kick pedal squeak incident, I switched to listening to wholly electronic music for the next month or so as a palette cleanser, and I haven't picked up on the squeak anywhere else since.

Can people who claimed that they can distinguish 320Kbps MP3 and lossless start posting their ABX log already?

This is really interesting! I did the test you linked, and for all but one I picked the lowest bitrate version.

I'm using WH-1000XM3 headphones and am a lifelong musician with pretty mild tinnitus.

Same headphones and tinnitus and picked out WAV 5/6 times. I did spend some time to listen tho, not pick blindly.

I did have aptX enabled at some point in Macos, not sure if it's still working.

Similar thing happened to me. What sounded "best" is subjective and I really had no idea what to listen for.

Listened using Beyerdynamic Custom Studios with a cheap-ish DAC on a windows PC.

When you say "score well" on that test, what does that mean? From the marketing around lossless, they make it sound like most folks will get 6 out of 6. I've got good monitors and good hearing (in the audiology sense), and I can mostly rule out the 128kbps, but it's a tossup between 320 and lossless.

It really depends on the encoder and song; I think the vast majority of people would have a very hard time getting above-chance on an ABX test vs a properly encoded MP3 at 320kbps, but I would not discount specific instances where things fall over and it still introduces artifacts that you can learn to recognize. Lossy compression is, unfortunately, not predictable in this sense, so there are always corner cases (or even outright bugs in encoders).

Here's one result which lines up pretty much with what I'd expect - if you get lucky with very specific songs/styles, you can pick out a 320kbps mp3, just barely, with good playback hardware.

https://www.head-fi.org/threads/abx-test-of-320kbps-vs-flac-...

There are a very small % of people that can. You can check HydrogenAudio's double blind test for explanation on how to do a proper test.

https://wiki.hydrogenaud.io/index.php?title=Blind_test

Being able to discern a difference in a side-by-side test doesn't really mean much.

You can tell which of 2 similar shades of blue is lighter next to each other, but if you were to look at one and then another a day apart you probably couldn't say which was lighter

It's a good point about optimization of certain things in general. Like, yeah, someone might post a study that in an anechoic chamber you can tell the difference between compressed and uncompressed audio, but how much of a difference is that making for people in the real world? Often when I listen to music, I'm at the gym, or the office, surrounded by all sorts of background noise that is going to impact my experience way more than compression.

I just got 5/6 with just my MacBook Pro speakers (but am a musician). "Bigger and stronger" is kind of an asinine way to describe the difference though. Was "you hear more subtle details" too many words?

Test using pixel 4 xl built in speakers at 2/3 volume. You don't need high end gear, you can pick artifacts in half the test cases almost more easily on low end speakers. A few of the tracks only become noticeable once theyre busy.

https://photos.app.goo.gl/Bx1XZP5DKpMvvmHf8

On NPRs site I can tell which track is the wav file by observing the load time of each track.

You may be confusing data compression with dynamic range compression, since you mention dynamic range. Dynamic range compression (limiting / brick wall limiting) has nothing to do with data compression, which is just referring to the removal of unnecessary data that is not noticeable from a psychoacoustic perspective (eg humans can’t hear above 24khz so that data isn’t needed, etc.) here is more info on DRC and MP3: http://www.tonestack.net/articles/digital-audio-compression/...

I took the test, and it pretty conclusively proved to me that I cannot detect the difference.

And even if I could detect the difference, I am not convinced that the improvement in bitrate would actually result in any increased enjoyment.

I tried it with my decade old entry-level Seinheiser headset plugged into my desktop computer.

4/6 with two errors where I confused 320Kbps and raw wav.

But to be honest, I could hardly distinguish between the samples and I mostly guessed.

That test was amazing. You made me really excited and I even guessed the first one correctly by seemingly hearing the difference. So I spent minutes on each of the other questions, listening and listening, but I just couldn't hear it and sure enough I guessed wrong on every other question after that.

And I have a mid-range mini amp that fits in between my books, and two mid-range dali speakers hooked up with a regular 3.5mm outlet to my Ryzen MSI PC board, volume up to max on the computer but not on the amp.

>even 320kbps MP3

Not "even" as if that was any good. MP3 320kbps is a very low bar.

OPUS vbr averaging 128kbps is much harder than mp3 320kbps to ABX against FLAC.

MP3 simply is a really bad codec by current standards.

Yet if you open Soulseek 99% of tracks are mp3.

Guessed WAV's 5/6 on bluetooth headphons, but took a lot of attention. Also - the samples can steer you in many directions.

Not sure how reflective that test is, since my perception is already primed by noticing the loading time per sample

If they are delivering a compressed file and an uncompressed file, that already disqualifies the test. The only way to do ABX comparisons is to round-trip one version through the encoder and back to the original format. Anything else introduces uncontrolled dependencies (e.g. on a particular device's decoder implementation) and side channels that unblind the experiment (like loading time).

This is a common mistake people make, e.g. comparing 48k vs 96k files. What you need to do is take a 96k original, downsample it to 48k, upsample it again to 96k (both using very high quality algorithms), then compare it to the original 96k file again. Otherwise you're relying on your playback software or hardware's resampling algorithm, and I guarantee that's a compromise between quality and performance, and not valid for a scientific test.

It's technically possible to distinguish between 320kbps and loseless but it's nothing to do with perceived sound quality or not but more of whether you're trained to hear high-freq sound domain which tends to more suffer from compression.

Audio is a deep rabbit rabbit hole. There is always a next level of quality. You can hear the difference, so always a reason to spend just a bit more. Then you have a 100k setup and someone is wondering why you just book the band directly.

I run what I consider to be "good high-end gear" and am surprised that I cannot even tell the difference between 128kbps and 320 kbps, let alone lossless...

I could hear the difference on the first question on my pixel 5 phone speakers.

Try a proper ABX audio test like: http://abx.digitalfeed.net/

It is a bit lengthy, takes 5-10m, but it is a true blind A/B test with enough samples to be significant.

I gave these and I can't distinguish between 320kbps and uncompressed.