Comment by lloeki | Hacker News Reader

Comment by lloeki | Hacker News Reader

> AAC and other newer audio codecs can produce better quality than MP3, but the difference is only significant at low bitrates. At about 128 kbps or greater, the differences between MP3 and other codecs are very unlikely to be noticed, so it isn’t meaningfully better for personal music collections.

This is patently false unless you listen on bad audio gear (bad DAC, bad cans).

I don't even have serious equipment (MBP+iPhone so guess about the DAC, as for headphones: AKG k514 mkII, Sennheiser Momentum (over ear), SoundMagic E10 (intra), Marshall Monitor (not mine), BeyerDynamic DT 770 250Ω (not mine)) and the difference is downright obvious going from 128k to 256k almost whatever the place and definitely noticeable on select features from 256k to 320k.

Some corner-case audio patterns are just known to make MP3 encoders choke whatever the bitrate (a higher bitrate merely mitigates the problem for MP3), while they pass with flying colours on AAC. Sadly those corner cases are not just theoretical and have a real effect for some songs. At 320k though it won't make the song any less enjoyable (which is another metric entirely) but it can definitely be noticed (which is what the article claims).

> This is patently false unless you listen on bad audio gear (bad DAC, bad cans).

Double blind tests suggest this is true for somewhere between 128 kbit and 256 kbit (256 kbit is CD quality, while 192 kbit is near CD quality, and the difference between 128 kbit and 192 kbit is miles compared to 192 kbit and 256 kbit; hence I suggest 192+ kbit. However, its 2017 and 320 kbit or lossless isn't a huge issue anymore; its even OK over 4G). Also, my quote is from 2000; ever since we swapped to LAME and ABR > VBR > CBR; ABR isn't even a static (constant) bitrate.

Quoting r3mix.net from 2000:

Facts:

128 kbit/s is not cd quality

256 kbit/s is cd quality (x) (in case of Lame or some Fraunhofer, not Xing)

In february 2000 c't magazin organised a blind listening test. 300 Audiophiles were involved, finalists tested 17 1-min clips from different artists (classic and pop):

original CD recording

128 Kbit/s Joint Stereo [MusicMatch (FhG) v4.4] encoded PC decoded Mac

256 Kbit/s Joint Stereo [MusicMatch (FhG) v4.4] encoded PC decoded Mac

all on cdrs and played in a Recording Studio on:

B&W Nautilus 803, Marantz CD14 with amp PM14 (Straightwire Pro cabling and extra's) [DM30000- so bit more than $15000]

Sennheiser Orpheus Electrostatic Reference-headphones with tweaked accompanying amp (digital and analog out) [>$10000]

Conclusions:

90% of the 128 Kbit material was picked out

MP3@256 was rated to have the same music quality as cd!

If you find MP3@256 to be of inferior quality compared to the original cd, you're very likely to be doing something wrong with the test (correct decoder, no objective double blind testing, DSP filters distorting the process, ...) Maybe this is something for you. You can always read the article in the german c't 6/2000 on p92.

The treshold of mp3 transparency lies somewhere between 128kbit/s and 256kbit/s, depending on the kind of music and your hearing and equipment.

English translation of the c't article: https://web.archive.org/web/20140217151727/http://www.geocit...

"If you find MP3@256 to be of inferior quality compared to the original cd, you're very likely to be doing something wrong with the test"

Try running that 256kbit MP3 through a pitch shifter. The quality loss becomes very apparent very quickly.

"Try playing a random binary blob as a video file, it might play, it might not"

Mp3 is a clever compression algorithm made for consumption. It isn't intended to be pitch shifted...

FLAC claims to be lossless/archival yet the same artifacting appears when pitch-shifted.

Listening through a pitch shifter is wrong, unless you are evaluating the codec for the purpose of listening through a pitch shifter. That's the thing - in normal use, high bitrate MP3 still sounds "the same enough" a lot of the time.

The author claimed that codec choice matters little >128k, not that 128k is of equal SQ to 256k.

BTW, if you can easily discern 256k from 320k, I'd venture to guess that you are a 1% "golden ear" listener. OTOH, even my 60yo ears can tell 128k from 256k.

I understood that but didn't made it point clear: my point is that it matters in a bit more nuanced way. I'll use [0] as some sort of visual support to help transcribe how I perceive it:

- 128k MP3 is like watching a Q=10 JPEG (artefacts and blockiness all over the place) while 128k AAC is much more like Q=50 (artefacting is much much more local, like the little grass patch on the left side of the stone)

- 256k MP3 is generally great but still has artefacts happen occasionally in some local, specific areas, similar to Q=50 (but not in the same way as 128k AAC), and definitely not on all songs, while 256k AAC is very close to perfect in any situation.

- 320k MP3 is very close to perfect save from very specific (and rare!) pathological corner cases that throws encoders off (like you have to have some specific bass line and some footwork on the bass drum and some very specific cymbal then the cymbal gets subtly crashy and noisy while the bass work gets subtly glitchy and muddled), which doesn't happen with AAC. Yes I've double blind-tested this with a friend who did not believe me, and used multiple MP3 encoders for that.

So to sum things up:

          MP3        AAC
    128k  annoying   enjoyable
    256k  excellent  awesome
    320k  superb

Those being ordered this way: annoying < enjoyable < excellent < superb < awesome

It's not much that I have some "golden ear" that I've been trained into looking for some patterns when I was studying DSP and now that I know what to look for they just annoyingly pop out. It's a bit like watching logs fly by just fast enough that you can't read them yet spot some oddity and just go "wait, there was something right there", and should you scroll back, sure enough, here's the deal standing out like a sore thumb.

Taking a bet that audio equipment gets better and cheaper with time I'd rather have a reasonable quality as a source, especially in light to the storage space and general bandwidth we have at our disposal. Then again in an increasingly streaming world we don't get much in the way of a choice whether it's about formats themselves or encoders.

[0]: https://en.wikipedia.org/wiki/JPEG#Sample_photographs

In a 2014 listening test, 136kbps MP3 was generally rated as somewhere between "Perceptible, but not annoying" and "Imperceptible": http://listening-test.coresv.net/results.htm

It really does depend on the music you're listening to IME. Some stuff sounds ok on 128k, but a lot of the electronic music I listen to is horrendous on 128k. Which is a real shame as a lot of sets I have I can only find in 128k mp3.

Spot on. This is something that is rarely mentioned, but having an extremely varied taste in music - I've noticed this exact thing.

http://abx.digitalfeed.net/list.lame.html

I have never seen anyone pass that bullshit-detecting test. 128kbps encoded with lame is barely detectable, and anything higher just isn't.

This test is relative to the sound system and listening environment in use at the time of testing. It's not useful as support for absolute statements like "128kbps encoded with lame is barely detectable, and anything higher just isn't."

I'm not a snake oil audiophile, but it's obvious that conducting the test on a train with crappy $1 earbuds and an old Realtek DAC will yield results that are considerably different from e.g. http://www.jeremykipnis.com/Jeremy_Kipnis_-_Home_Page.html

Reply on news.ycombinator.com