Hacker News Reader

Comment by lxgr | Hacker News Reader

> Accordingly, the company also says "the difference between AAC and lossless audio is virtually indistinguishable."

Then how come they are offering lossless audio on Apple Music, and even (actuall physically!) imperceptible "Hi-Res Lossless"? Are enough dogs and bats paying for it to justify the extra bandwidth required by 24 kbps/96 KHz audio?

When catering to pseudoscience, arguably one really needs to go all the way. Do it like Sony, for example; go full "audiophile" and invent your own Bluetooth audio codec [1]. Or digitally sign your bits to make a special light go on on expensive hardware [2].

Delivering lossless 24/96 to the device and then downsampling it to 16/48 and AAC compressing it for the last meter over Bluetooth is just sad.

[1] https://en.wikipedia.org/wiki/LDAC_(codec)

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

View on HN

> Then how come they are offering lossless audio on Apple Music, and even (actuall physically!) imperceptible "Hi-Res Lossless"?

Because people, despite repeated double blind testing, claim they can hear a difference. That’s the only reason. Same reason anyone offers an output format beyond 16-bit/48k for audio playback. It’s pointless, but people think it matters for stupid reasons.

Yep. People will pay for it even if there’s only a perceived benefit. Everyone who has hung out with audiophiles has known at least one who is convinced they can hear the difference. But if that person was tested in a blinded format would they actually succeed?

Lossless music is great. As an archival tool, to transcode into something smaller on your actual listening device. It’s not a technology designed for streaming or listening in today’s world.

> But if that person was tested in a blinded format would they actually succeed?

I always think of that famous coathanger post: https://forums.audioholics.com/forums/threads/speakers-when-...

And the immediate nitpicking following it: "Is your speaker system time coherent?"

Meh. Some people might truly be able to hear the difference. A very small amount. I have some visual perception differences, even participated in a Stanford study on high fidelity visual perception.

Example - on a 2880x1800 15" screen up to 10 feet away, I can see a quarter pixel (subpixel, really) difference in position. Can also tell a 3% difference in HSL values. New colleagues like to "test" me to see if that's actually true by moving something by 0.25px to see if I can actually tell, and I always can.

Being pretty good at visual design/ having freakish perceptual abilities lead me to meet a lot of music producers at the top of their game/ musicians that sell out stadiums, etc. I've tried testing a bunch of them on lossless versus lossy. Something like 2 out of 5 of them can repeatedly tell the difference.

That said, all but three of those people said they can't actually tell the difference unless they're paying really close attention. One of them, one of the top grossing producers of all time, said lossy compression gives him a migraine.

Just saying that dismissing ANYONE having the ability to tell the difference seems inaccurate. Though I'd bet the number of people, globally, who can tell the difference numbers in the thousands

I've been doing a bit of research on the limits of human vision today, as I was wondering if any ancient humans (say, at Newgrange) would ever have been able to discern the spiral of a galaxy with their naked eye...

According to what you're saying, I think it's just about possible. M51's spiral features are 20 arcseconds, and you can apparently resolve 15...

And turns out M51, the easiest spiral galaxy to view, would have been directly overhead Stonehenge (the best viewing position).

So... Might be somethin' for you to try, if you ever get somewhere extremely dark and perfectly clear?

M51: https://duckduckgo.com/?q=%22m51%22&t=ffab&atb=v340-1&iax=im...

Newgrange: https://duckduckgo.com/?t=ffab&q=newgrange+spirals&atb=v340-...

Can most people not see the sort of briefly appearing cloudiness/ cloudy streaks of light (idk how to describe it)/ galaxy arms in the sky, if you're in the middle of a forest/ with no light pollution??

I think people can see the milky way in the right conditions alright.. We know from ancient records that people could see it as well.

As far as I can tell though, making out the actual spiral features of another galaxy is generally assumed to be impossible... And I don't think it actually is, for someone like yourself.

I’m not contesting there is a difference between lossless and lossy audio in specific instances. It’s small but sure, I’m sure some people could tell. My contention is with the “Hi-Res” audio (which, for Apple, means 24-bit and 192 kHz sampling rate).

Nobody, and I mean absolutely nobody, will be able to hear a difference, assuming there aren’t mastering differences in the tracks. If you take a 24/192k track and downsample it to 16/48k, nobody will be able to tell. Ever.

Apple Music will not stream 24/96 to you if you are not using compatible equipment

But it’s nice to have the option because even my wife, who doesn’t care about audio at all, can immediately tell if my home theater is playing a DTS-MA track or an AAC track. We have a pretty modest 5.1 setup from a aperion audio. High quality audio formats are not flim flam in and of itself.

LDAC is actually a great codec and I understand it’s available to every android device, so thank you to Sony for inventing it. It’s indistinguishable to me against a flac file over wired connections. And I actually can tell the difference with lesser codecs.

I cannot tell the difference above CD quality audio, and even still not reliably at lower quality. That doesn’t mean it’s worthless.

Last I checked, Apple's iTunes store is still stuck on 256k AAC "iTunes Plus", first introduced in 2007. Sounds OK, but doesn't transcode well. I just want regular audio samples as a future proof format that doesn't require physical media.

Cymbals used to be the hallmark of terrible MP3 encoding artifacts, but they seem better on AAC.

I have half decent equipment i.e. Schitt Lyr/Bifrost and Focal Clear.

For me it's very easy to tell the difference between regular AAC and Hi-Res lossless.

And I have been blind testing every day for years now as my streams switch quality when I go from Cellular to WiFi.

Oh, I don't doubt that people can distinguish AAC and 44.1 kHz at 16 bit! Anything beyond that is what I find questionable.

> Then how come they are offering lossless audio on Apple Music, and even (actuall physically!) imperceptible "Hi-Res Lossless"?

Overkill makes good marketing. People largely just want to feel like they're getting the best.

Calling it pseudoscience is a little simplistic IMHO.

There are double-blinded studies which shows experts can distinguish between 44khz and 88khz sampling rates. The Pras & Guastavino paper [1] is the most convincing and offers insight on the difference between content types. In particular, for the recording test, the differences only held for orchestral content, not other types of music.

There are earlier studies which shows the general public cannot discriminate e.g. Meyer and Moran [2], which kicked off the entire controversy back in the day. And even the Meyer and Moran paper points out in the conclusion that the introduction of higher bandwidth audio in turn has enabled recording engineers to spend more time on producing quality recordings, and that has increased listening quality for all users, versus recordings made just to maximize loudness.

[1] https://www.researchgate.net/publication/257068631_Sampling_...

[2] https://drewdaniels.com/audible.pdf

The statistics in the Pras & Guastavino paper seem a bit suspect. They do a lot of significance tests with no multiple comparisons compensation, and find few statistically significant results.

The problem is that with enough testing, you're very likely to get statistically significant results just by chance.

Reply on news.ycombinator.com