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: What kind of music/samples is considered "problematic"? (Read 7594 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

What kind of music/samples is considered "problematic"?

Hi everyone

in the context of lossy audio codecs, I often read the term "problematic samples".

For example here:
https://de.wikipedia.org/wiki/LAME#Qualit.C3.A4tsstufen

The german Wikipedia article about LAME is talking about the quality settings -V0 and -V1, as they are considered to be transparent for trained ears, but maybe problematic samples might be improved.

I also read the term on other sites all over the internet, and also in context of other codecs/encoders than MP3/LAME.

I would like to know which kind of audio is mostly problematic to lossy codecs? Are there differences between codecs (like one codec can deal with problem XY better than another)?

I know about the phenomenom "pre-echo", and I often heard it myself in low-bitrate MP3 files. So I guess, one common problem for lossy codecs are sudden transients, like the attack of a drum for example.

But especially, I would like to know, what kind of audio can be improved with regard to quality, if I for example switch from LAME -V2 to -V1 or even -V0.

Generally asked, in other words:
If a quality setting of any lossy encoder (may it be Lame, Vorbis, whatever) is considered to be transparent, what kind of audio sources could bring that encoder down, so that there are actually audible differences to the original source, that could be eliminated by selecting some higher quality setting?

Re: What kind of music/samples is considered "problematic"?

Reply #2 –
I wish I had titled that thread differently. 

If I'd thought it through more, I'd have realized that what I was looking for was some general *principles* that would explain what makes a track is 'problematic'.  And by that I meant, hard to encode transparently *even at the highest quality settings* . 

Some of the answers came close to supplying answers, most were simply anecdotal reports reflecting the (to me) weird abundance of metal and electronica lovers on HA.

Re: What kind of music/samples is considered "problematic"?

Reply #3 –
Sharp near transform limited transients to be overrepresented in problem samples.

Re: What kind of music/samples is considered "problematic"?

Reply #5 –
a) harpsichord music is hard to encode. Worst sample I know is harp40_1. At very high bitrate however (>=220 kbps) quality is very good though not transparent  even with old mp3.
b) Extremely tonal music can be hard to encode.  Worst sample I know is lead-voice which is very problematic at least for standard Lame. It is an electronic sample however, and IMO electronic music should be considered especially. Encoding quality of at least very special kind should not be mixed up with music originating from acoustical instruments including vocals.
c) transients in the music are often considered problematic. The most outstanding problems however also are with electronic music. For acoustical instrument music it takes extremely pre-echo sensitive people to hear an issue.

As for Lame I've done some efforts to bring down known issues like the ones I mentioned.
My lame variant lame3995o with setting -Q1 (~224 kbps) taggles them in a way which is most satisfying at least for me.
lame3995o -Q1.7 --lowpass 17

Re: What kind of music/samples is considered "problematic"?

Reply #6 –
Quote from: bennetng on
Quote from: krabapple on
*even at the highest quality settings*
https://hydrogenaud.io/index.php/topic,39970.0.html
Can't do that again, maybe my hearing is not as good as 11 years ago.

Right, I know that problematic samples exist for high quality encoding -- HA is a great resources for identifying those --  but again,  what I was seeking to know was, what characteristics make them problematic.

Answers from this and the previous thread seem related to transients, and/or music being 'tonal'  (a word being used in a way I'm  not familiar with -- to me 'tonal' music is the opposite of 'atonal', it means music with a stable tonal center, a definite 'key' like A minor, Bb major, etc. )

I don't know what 'sharp near transform limited transients' means exactly, either.  To me transients are very short, relatively loud sounds with lots of high frequency content, at the leading edge of an acoustic event.  Castanet clicks would be an example.  I suppose the keyboard clicks of solo harpsichord would qualify too.

Re: What kind of music/samples is considered "problematic"?

Reply #7 –
Quote from: krabapple on
'tonal'
If I'm not mistaken, I believe it means content with sustained tones (notes) with a pitch that doesn't modulate or does so in only a very shallow and subtle way.

Quote from: krabapple on
To me transients are very short, relatively loud sounds with lots of high frequency content, at the leading edge of an acoustic event.  Castanet clicks would be an example.  I suppose the keyboard clicks of solo harpsichord would qualify too.
Correct.  This generally includes any form of percussion that occupies a significant amount of spectrum with a sharp attack.

Re: What kind of music/samples is considered "problematic"?

Reply #8 –
Quote from: krabapple on
I don't know what 'sharp near transform limited transients' means exactly, either.

https://en.wikipedia.org/wiki/Bandwidth-limited_pulse

Quote from: krabapple on
  To me transients are very short, relatively loud sounds with lots of high frequency content, at the leading edge of an acoustic event.  Castanet clicks would be an example.  I suppose the keyboard clicks of solo harpsichord would qualify too.

Unless you are just listening to an infinite sin wave or something, music is full of transients and tones, almost all of which are easy to encode.  At least in my opinion (and it is possible someone with more experience will disagree) ones that tend to cause problems are the ones that are close to being limited in their time/frequency compactness.  So a sudden click like castanets with a distinct tone to it is more likely to be noticeably "wrong" after bad encoding then an equally short blast of white noise.  Both are "transient" in time, but one is broad in frequency and so less distinct. 

Re: What kind of music/samples is considered "problematic"?

Reply #9 –
Another kind of problematic audio are transitions between tracks, which result in gaps or glitches when converted to lossy.
Here's an example. Last time I checked Vorbis was the only codec that didn't produce an audible glitch in that transition.

Re: What kind of music/samples is considered "problematic"?

Reply #10 –
Broadband transients, especially added to long-term stationary tones (Track 11 of "We shall be happy" by Ry Cooder, album "Jazz").

Broadband signals comprised of separate tones that are not so close or incoherent as to represent noise.

Signals with low pitch rate and broad spectrum.

Squam 40, German Male Speech, Castanets, glockenspiel, suzanne vega "Tom's Diner" (a capella version), some test signals
-----
J. D. (jj) Johnston

Re: What kind of music/samples is considered "problematic"?

Reply #11 –
One of the most intresting examples I found was track 12 from this release https://itunes.apple.com/us/album/final-fantasy-xiii-2-original/id528529390

I own physically the disc and while trying to rip the disc into a lossy audio format for my portable audio player I came to that conclusion.
The track consists mostly of sound elements that most lossy audio formats struggle to encode or have a distinct way of encoding them. Synths with weird harmonics, lots of pure tones, stereo image that kinda confuses encoders and the list goes on.

It's intriguing to find a tracks like this one, it's just filled with moments that expose a weakness to the codec.

Re: What kind of music/samples is considered "problematic"?

Reply #12 –
I was going to say it would be intriguing to compare any of the CD releases to the original game, since at least the Xbox360 version used 96kbps WMA Pro.

Re: What kind of music/samples is considered "problematic"?

Reply #13 –
Quote from: kode54 on
I was going to say it would be intriguing to compare any of the CD releases to the original game, since at least the Xbox360 version used 96kbps WMA Pro.
I did that already.
The PC version uses OGG at 96kbps and sounds average to good, they did a decent job.
The PS3 version uses MP3 at 128kbps and sounds way inferior, some of the tracks struggle to reach 14kHz.
My biggest issue is that vgmplay does not support the 4 channel interleaved format they use on most of their tracks (general issue with scd format) and on most of the tracks you will hear like 500ms of each track jumping back and forward from the regular version to the remixed version.

PS: I wish I could find a way to know what encoder Square Enix uses for mp3. It sounds terrible, like something from 1999.