Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Artifact war: AAC max bitrate vs mp3 max bitrate (Read 22229 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Artifact war: AAC max bitrate vs mp3 max bitrate

For artifacts on mp3 @ 320 bitrate settings, are these artifacts also detectable (abx-able) on the MAXIMUM bitrate and quality aac/m4a settings (either quicktime or nero), and do they seem noticeably better or worse (harder or easier to detect) than the mp3 max bitrate version. (I can't detect, but would like to know if sensitive listeners can detect.)

Thank you--I know this is usually not the most pleasant listening test. Though I'm looking in particular for instances of classical music (--do any exist for aac max bitrate?), any examples would be helpful (especially if any are non-synth music). There have been (very) few confirmations of this before; but an update and more confirmations are helpful.

Here are some confirmed mp3 @ 320 artifacts:
http://www.hydrogenaudio.org/forums/index....showtopic=70598

(btw, I assume that there has never been any example of an artifact detectable from any recent aac/m4a encoder at max bitrate that is not also detectable in mp3 encodings -- I mean leaving aside bugs that have been fixed.)

Artifact war: AAC max bitrate vs mp3 max bitrate

Reply #1
Yes, this is because mp3 is a very bad (comparatively, at least) lossy audio codec. You're almost certainly never going to run into a sample that still produces notably audible artifacts for 256kbit CBR AAC (assuming a good encoder such as Nero's, of course). The other thing is that you're almost certainly not going to have classical music be such a sample if any such sample exists: classical music lends itsself very well to being compressed by lossy psychoacoustic transform-based codecs, as the instruments themselves tend to be very harmonic in nature (i.e. they're not, say, drums, which instead tend to be made principally out of hitting things), and computing absolute threshold of hearing functions for individual instruments/voices is relatively straightforward.

What you want is stuff that tends to cause pre-echo (castanets or percussion), stuff which is harmonically complex (distorted guitar is a good place to start), stuff that masks in a very weird and inconsistent manner (symphonic metal tends to be really good at this, oddly enough), and anything that has a lot of stuff going on at once in the frequency domain, preferably also involving all of the above to begin with (for instance, the last 30 seconds or so of Ennio Morricone's "The Ecstacy of Gold", which features bells, drums, woodwinds, violins and an operatic voice, all at approximately the same volume).

Again, though, I'd be surprised if such a sample exists, and even if it does, it's probably barely perceptible - most of the ones for 320kbit CBR mp3 are barely perceptible themselves, for that matter.

Artifact war: AAC max bitrate vs mp3 max bitrate

Reply #2
Thank you, that is consistent with what I've seen (note the conclusion from tests here is that mp3 is comparable to aac at moderate to high bitrates). I'm surprised however that you suggest Morricone might fail in an abx test. I'd certainly consider it an issue, and very unexpected, if anyone can ABX any part of Ennio Morricone's piece (or any similar work) at the max bitrates for AAC/M4A quicktime or nero (--higher than 256, since I'm interested in failures under the best settings possible for the codec).

Artifact war: AAC max bitrate vs mp3 max bitrate

Reply #3
classical music lends itsself very well to being compressed by lossy psychoacoustic transform-based codecs, as the instruments themselves tend to be very harmonic in nature (i.e. they're not, say, drums, which instead tend to be made principally out of hitting things), and computing absolute threshold of hearing functions for individual instruments/voices is relatively straightforward.


  What about Harpsichords & Co?
... I live by long distance.

Artifact war: AAC max bitrate vs mp3 max bitrate

Reply #4
Comparisons like this are skewed because mp3 problem samples are better known. (Though of course AAC is a better codec specification than mp3 too!).

There are a (very few) AAC problem samples identified in the official tests (e.g. Pitch Pipe supplied by Dolby IIRC), but I've never seen these made available on HA.

Cheers,
David.

Artifact war: AAC max bitrate vs mp3 max bitrate

Reply #5
...
There are a (very few) AAC problem samples identified in the official tests (e.g. Pitch Pipe supplied by Dolby IIRC), but I've never seen these made available on HA.


Thanks for pointing that out.  Are those problems still problems, i.e. ABX-able, with the latest encoders at the MAXIMUM bitrate and quality settings?  and are there samples of classical or some traditional or non-synth music.  (I take it the pitch pipe is like a tuning fork.)      --If there's a skew against mp3, one way to balance it is to compile a list of the problems that break aac, but I had thought that the principles between the two were similar enough, except where aac removes an mp3 design limitation, that it would be very unlikely to have an aac flaw that doesn't also affect mp3 (except those caused by bugs).  Are those samples problems with aac only, or do they also affect mp3?  (I'm looking for _any_ aac problems at max quality, but ones that don't also affect mp3 would I guess be impossible to find now--apart from bugs.)

Artifact war: AAC max bitrate vs mp3 max bitrate

Reply #6
I suggest you go through this thread (which is less than one week old) and all the referenced threads therein.

To me, AAC problem samples are well known, and all types of these samples are linked to in the above thread. Unfortunately, the pitch pipe recording is not publicly available, but you can get an idea of the artifacts with the first few tones of the SQAM harpsichord arpeggio (harp40_1).

Chris
If I don't reply to your reply, it means I agree with you.


Artifact war: AAC max bitrate vs mp3 max bitrate

Reply #8
I suggest you go through this thread (which is less than one week old) and all the referenced threads therein.  To me, AAC problem samples are well known, and all types of these samples are linked to in the above thread. Unfortunately, the pitch pipe recording is not publicly available, but you can get an idea of the artifacts with the first few tones of the SQAM harpsichord arpeggio (harp40_1).
Chris


Thanks, that is a helpful list in those threads, and I think I had all of them downloaded.  But of those samples, how many were detectable (abx-able) at MAX aac bitrate and quality settings?  As far as I can see, at the max bitrates, there were no traditional, non-synth, non-metallic examples of music pieces (let alone classical).  I know that emese breaks aac at any setting, but emese is far outside what strikes me as music; and almost all of the samples are in some striking way "non-traditional."  (Let me add again that generally _I_ can't abx them, but I'd like to see whether proficient listeners can abx.)

I didn't see where harpsichord failed at aac max bitrate (harp401.flac)--I may have missed some posts while searching.  Was that a barely noticeable or a sharper, less subtle artifact?  As far as the traditional music pieces, I have written down that there were issues (maybe only mp3?) with a Sarah McLachlan sample (ice ringing), a snippet had a sound a bit like a bad microphone's squeal, but don't show issue reported at max aac bitrate.  Same true with Camille's Là Ou Je Suis Née.

Artifact war: AAC max bitrate vs mp3 max bitrate

Reply #9
Thanks, that is a helpful list in those threads, and I think I had all of them downloaded.  But of those samples, how many were detectable (abx-able) at MAX aac bitrate and quality settings?  As far as I can see, at the max bitrates, there were no traditional, non-synth, non-metallic examples of music pieces (let alone classical).  I know that emese breaks aac at any setting, but emese is far outside what strikes me as music; and almost all of the samples are in some striking way "non-traditional."

Correct. I should have mentioned that. The focus of that test was 128 kbps. Except using "transient-noise-like" items like emese, I haven't found any speech or music sample which I can ABX reliably above 256 kbps. The harpsichord is transparent to me above around 192 kbps (tested with a Fraunhofer encoder, though, haven't tried the nero).

User /mnt reported being able to ABX some songs from Kraftwerk's Electric Café at quite high bitrates, but I forgot the details and think it wasn't higher than 320 kbps.

By the way, according to the coding standard, the highest possible AAC bit rate per channel is 264 kbps for 44.1 kHz sampling rate and 288 kbps for 48 kHz sampling rate.

Chris
If I don't reply to your reply, it means I agree with you.

Artifact war: AAC max bitrate vs mp3 max bitrate

Reply #10
Except using "transient-noise-like" items like emese, I haven't found any speech or music sample which I can ABX reliably above 256 kbps. The harpsichord is transparent to me above around 192 kbps (tested with a Fraunhofer encoder, though, haven't tried the nero).  User /mnt reported being able to ABX some songs from Kraftwerk's Electric Café at quite high bitrates, but I forgot the details and think it wasn't higher than 320 kbps.

By the way, according to the coding standard, the highest possible AAC bit rate per channel is 264 kbps for 44.1 kHz sampling rate and 288 kbps for 48 kHz sampling rate.

Chris


So 528kbps total for cd stereo--nero encoder goes up to ~420 max, and quicktime to ~320 max.  I guess that if anyone discovers such a piece -- a "non-noise" and non-synth* music sample that breaks recent aac encoders at such high bitrate settings -- it would become notorious here.  (If I were promoting lossless audio codecs, I would certainly publicize such items.) 

(*re Kraftwerk: ok I confess I did like trans-europe express at one point, though not the sorts of pieces in which artifacts matter to me.)

Artifact war: AAC max bitrate vs mp3 max bitrate

Reply #11
If I were promoting lossless audio codecs, I would certainly publicize such items.
I'm not aware of anyone "promoting" lossless audio codecs here, and I don't think most people who use them do so because they can hear artefacts in the music they listen to using 320kbps mp3, never mind 320kbps AAC.

But I think you're right that almost any member who found a sample that produced artefacts at high bitrates wouldn't be able to resist posting it to HA.

Looking at the polls though (FWIW!) I don't think that many people here are using mp3 at 320kbps, and even fewer are using AAC "maxed out". Some notably artefact-sensitive members switched from lossy to lossless years ago, so they are unlikely to contribute new problem samples.

Cheers,
David.

Artifact war: AAC max bitrate vs mp3 max bitrate

Reply #12
I don't think of publicizing as in itself bad, and I agree there are excellent reasons to use lossless apart from the guarantee of transparency. 

I should explain my rationale.  There are some people who want to maintain only one digital copy of their music, and there is simply no lossless format that is standard or widely compatible across devices (Apple lossless comes closest because of the ubiquity of iTunes and i-gadgets, but even that won't play on car receivers).  Given that and perhaps to save space & battery power also, standardizing on mp3 or aac at high bitrate lossy formats could make sense.  (For aac there's a question of 256 or going somewhere higher to assure absence of artifacts across all listeners, leaving aside artificial noise or high synth samples.)  That wouldn't seem to be as good an idea if people were occasionally detecting artifacts in aac in regular music samples even at the highest bitrates.

Are there signs that lossless formats are becoming more popular?  I would think lossless would be an option at itunes and amazon and other places if there was demand (unless the music industry is against it).

Artifact war: AAC max bitrate vs mp3 max bitrate

Reply #13
That wouldn't seem to be as good an idea if people were occasionally detecting artifacts in aac in regular music samples even at the highest bitrates.
Normal people aren't "occasionally" (or ever!) detecting artefacts in mp3 in regular music samples at modest bitrates. Most of the audible problems that lame -V2 ever produced are documented here on HA. It's not a long list. It's not compiled by "normal people"

"Regular Music" already comes with far bigger sound quality problems that we can do absolutely nothing about because they're on the CD.

Only mp3 plays everywhere. If that is your goal, you don't have a choice. You have to put up with the possibility of artefacts on some rare samples. You have to put up with inconsistent handling of gapless playback on many devices. etc etc etc.

People who care more are either stuck with restricting the devices they can play their music on (which IMO is unwise - you never know what cool gadget won't play your chosen non-mp3 format in a few years time), or stuck keeping a lossless archive and producing lossy as-needed. Or dual archive. A fast computer and a fast encoder makes the latter less and less important IMO.

The even easier solution is to join the other 99% of humanity and train yourself to not really care

Cheers,
David.

Artifact war: AAC max bitrate vs mp3 max bitrate

Reply #14
It can be a problem just to know that there are artifacts that a few listeners might detect, but thanks for pointing out that these are overshadowed by the limitations in the quality of the recordings (at least for pop music). It looks like I'll stick with recommending, for a one copy solution, max bitrate mp3 or aac (which is almost as universal as mp3 now on recent hardware, though a good chunk of car receivers still won't play it). Dual lossless/lossy would be better of course, and as you say it has gotten easier to do, but it's still a hassle and things get out of sync.

Artifact war: AAC max bitrate vs mp3 max bitrate

Reply #15
It looks like I'll stick with recommending, for a one copy solution, max bitrate mp3 or aac ...


There is little evidence that many problem samples are cured at any bitrate.  Using a sane VBR preset will likely add little to the "appearance" of artifacts and will cut average size potentially in half.


Dual lossless/lossy would be better of course, and as you say it has gotten easier to do, but it's still a hassle and things get out of sync.


Tag sync issues are solved by only keeping lossless on your PC and transcoding to lossy as part of the transfer process to your external device.  For most people in most situations storage speed is their bottleneck and the CPU effort of converting during transfer causes marginal, if any, slowdown.
Creature of habit.

Artifact war: AAC max bitrate vs mp3 max bitrate

Reply #16
There is little evidence that many problem samples are cured at any bitrate. Using a sane VBR preset will likely add little to the "appearance" of artifacts and will cut average size potentially in half.

I admit we're talking about extremely few cases of artifacts, but higher bitrates have been shown to eliminate them or make them much softer in those handful of cases where they remain (synth/artificial).  The situation is not one where saving space is a primary concern, but trying to get to one transparent universal format.

Quote
Tag sync issues are solved by only keeping lossless on your PC and transcoding to lossy as part of the transfer process to your external device. For most people in most situations storage speed is their bottleneck and the CPU effort of converting during transfer causes marginal, if any, slowdown.

That is effective and I would recommend that first, but there are some for whom it's not worth the hassle.  You still end up spending more time with double copies than single--even aside from storage speed and transfer speed issues, you have synch issues.  For one thing, there are cases where it's more convenient to have the lossy versions on your hard drive also (depending on the devices), and there are also lossy purchases that don't have lossless that need to get backed up; then various music management software can get confusing with duplicates. 
(As a side note, for some unfathomable reason, apple lossless uses the same extension as lossy aac files.)