Greetings,

I currently try to decide whether to use extreme or not for all my rips. For that decision I need to know how much std problem samples are out there and to what extent the artifacts disappear at higher qualities. AFAIK there are currently at least two reported problem samples in this forum, the BSB clip and 2nd_vent_proper.wav (which I have no source for) from Citay.

Please consider this thread as a container for problem samples, i.e. anyone who has such beasts please post them here (preferably with comments on which part was ABXable from the original at which quality, etc.)

IMHO this is of public interest here, although - of course - there already is such a thread then please excuse this (and perhaps give me a link)...

Hehe, sorry, i thought you knew how to use the search function...

http://www.hydrogenaudio.org/forums/index....=ST&f=11&t=2048
http://www.hydrogenaudio.org/forums/index....=ST&f=11&t=2079
http://www.hydrogenaudio.org/forums/index....=ST&f=11&t=2122 (uploaded the file again)
http://www.hydrogenaudio.org/forums/index....=ST&f=11&t=2290
http://www.hydrogenaudio.org/forums/index....=ST&f=11&t=2386
http://www.hydrogenaudio.org/forums/index....=ST&f=11&t=2424
http://www.hydrogenaudio.org/forums/index....=ST&f=11&t=2386

This search will give you some more samples too.

I just tried to ABX mpc-std from the original BSB Samples for myself and actually heard no difference at all, not at the "ooooohhh" either. I could not detect an early fading out of, but maybe I just have a bad high Frq. hearing or my Altec Lansing stereo system is too bad for such tests (which I doubt). Can someone tell my "exactly" what I have to pay attention to?

BTW: I use the recommended Buschmann(?) encoder 1.14. Is this still the "best" version? Are there any improvements in Frank Klemms encoders and if yes which version should one use?

ABX tests should preferably be done with headphones.

Quite probably, your speakers and training in hearing artifacts are not enough to notice them. See, most obvious MPC artifacts are obvious just for trained people, usually listening with headphones. More subtle artifacts are very difficult to hear even for many trained people with very good equipment. So, don't worry.

1.14 is the latest beta version of frank klemm's encoder. last buschmann encoder is version 1.79c which is very old.

newest klemm alpha encoder is 1.15r. it was recommened by dibrom somewhere (only downside is, that the alpha encoders show "unstable/experimental" instead of the preset used in the fileinfo dialog, but this is only a visual issue which can be adressed by adding used quality setting in the comment tag for expample).

you can always get the lastest klemm alphas at www.ca5e.tk.

regards; ilikedirt

Newest encoders are available at http://www.saunalahti.fi/cse/. MPC alphas are considered as quite safe compared to lame alphas. 1.15r is used and recommended right now (at least this is what can be read in recent posts).

- edit: too slow 1ce again :'(

1.14 is already a Klemm-optimized encoder, the last Buschmann encoder was from 2001. That 1.14 is the foolproof recommended version, however, the latest 1.15 alpha versions show some quality improvements, so you may want to use that for testing.

hah. and i thought i was slow

Isn't there an alpha version which stores the used quality / isn't it possible to modify an alpha encoded mpc file in order to belatedly store the quality profile? Otherwise I will lose the quality information when I e.g. make an mistake while using various ape taggers... (frontah etc.)

I am confused now... 1.14 or 1.15r.. what should I be using for best quality?

1.15 is work in progress with slight quality improvements over 1.14 (at this stage, i don't think we'll see many groundbraking enhancements anymore). If you think it's worth the upgrade, feel free to use it. Being at "r" already, you can also wait for version 1.16.

Are there any situations where 1.15r is believed to produce lesser quality? Is it a matter of it being an alpha and it may crash occasionally?

My basic question is why isn't everyone using 1.15r? I understand it's an alpha and that may imply there are major bugs, but everyone seems to talk about it like it's stable and ok to use...

If you ever manage to crash the MPC alpha, let me know, and say how the heck .

Does this mean I should use the alpha and feel warm and fuzzy all over?

If it works for you and you like it, and you want to, sure. It does offer sound quality improvements. I didn't compare it to 1.14 or earlier 1.15 alphas though.

edited: maybe i need english lessons?

I think he could have figured that out himself. MPC alphas seem quite stable & there have been claims as to the quality of 1.15r being better quality than 1.14b but as ilikedirtthe2nd has pointed out there are issues with MPC alpha versions & really if the author names the software "alpha" it means only use for testing. Why take the risk when MPC 1.14-1.15 only have slight quality differences?

[SARCASTIC GRIN]

The eternal story... what am I missing if I use 1.14 instead of 1.15?

... "Musepack: Hear what you've been missing!" (former slogan on Spase's musepack.org)

[/SARCASTIC GRIN]

We are all quietly waiting for 1.16 beta. mppenc has been at 1.15r for quite some time now.

The last time mppenc went through a period of alpha turmoil was the 1.02-1.06 period. That's when even some of the betas proved to be problematic.

There is no 1.15 only 1.15r, alpha, testing phase. My point was for those of us who don't do extensive, time consuming testing every time a mpp alpha gets released should wait for a beta release.


That's why I started using musepack.

I don't understand, If you trust your methods & you do extensive testing for each mpp release & want absolute best quality from mpc why doesn't the * standard allow alpha releases? Verified by you of course


MOD:* No links to or names of ripping groups please.

Just kidding, man!


1. I don't do extensive testing for each mpp release, especially not for alphas (that would be more work than I can actually afford...). Any tests I should have 'direct use'(they are meant for those mppenc's I plan to use or am curious about...): I also rely on other people's findings.
2. Still I do, but I second Citay's point that that Musepack's development is already very advanced atm.

It does, to some extent.

IMHO the current problem with * is the required quality setting: --insane.

It's a bit controversial to trust the codec developer on one hand, and to skew the psychoacoustics to reach a hypothetic, "higher" quality level, on the other hand.

I can understand to demand set --quality 5.5 or (as a stretch) 6.0 as a minimum requirement, but not 7.0 !

In my opinion, setting two steps of overkill as a **minimum** requirement, is ridiculous. This would work with MP3 (ie: 256 instead of 128) but MP3 was never designed to be transparent at the advertised typical settings ! So, just increasing the recommended setting by an arbitrary amount, is just nonsense in MPC.

For a high-quality standard I'd rather set 5.5 as minimum, and 7.0 as maximum - which can be overridden on music for which someone proves that artifacts are heard.


MOD:* No links to or names of ripping groups please.

*

why is q7 the standard minimum ?

because:

Most of the few artefacts occuring with mpc, vanish, if you go from q5, 6 to q7, and especially at q8.
Just read some posts of Guruboolez, please.

And yes, a lot people are satisfied by q6 eg., because listening abilities are distributed by a Gaussian (bell-like-) curve, distribution.
So, this means, a majority is satisfied by q5, 6, but very clearly, not neglectable, there are people out there, who demand q7 and even q8.
(This Gaussian-bell-curve means, too, that there are even people out there, who are satisfied by xing-mp3@128 (half kidding ) , or mpc at q4 perhaps, just as example.
So, give everybody the freedom to encode with preferred setting.
I don't bash here, too, regularly those guys, who spread the theorie, that q5 is the 100% perfection, which is obviously not true.
So let people live with their tastes.

* is obviously a high quality standard, meant for people, who don't compromise regarding quality/size.
The philosophy is: "It is better to do too much than to do not enough ! The goal is quality, no compromises !"


The space/time issue of q6 vs. q7:
It is relative, of course, if you compare q5, q6, q7, there is an increase of some percent in space.
But compare the bitrates/sizes of q5, 6, 7 with wave (= 1400 kbit/s) or lossless (around 50-60% of 1400 kbit/s) then you see, that the space difference between q5,6,7 is neglectable, but q7 saves huge amounts of space compared to lossless/wave.


Well, perhaps a matter of taste and personal view,
everybody is free to decide, what he wants to do, what he wants to get.

So, please, don't make a religion or dictate out of the question whether q5, q6 or q7 should be preferred.
I assume, that most people out there, who know about mpc, have enough knowledge, brain-power, to decide themselves, what they need, want, use.



And last but not least a view into future:
the ages of DVD-writer for all (Blue-ray disc even available in Japan this summer with about 40-50 GB per each disc..), or broadband internet, are rising, slowly, but they are...., cable, satellite, ADSL/DSL, UMTS....
---> lossless formats will get more interest, perhaps.


MOD:* No links to or names of ripping groups please.

Already done this, long ago. If you read a bit more from him, you'll see that he recommends -q6.


This is a very cheap excuse. When mpc artifacts, it's because of a mistake made inside the psychoacoustic model. It has little to do with the setting being too low for your hearing capabilities. The recent few samples which fail at all quality settings show this. Also, the 99+ % of samples having no problem, show this.


Well, yes, everyone has the freedom to choose their setting. So it's dumb to force them to -q7 minimal.


Same here - example: you decide that -q5 is enough for this piece of music, you use -q6 for transcoding margin. Even though carefully ripped and encoded in the best lossy format using very good settings, your file is not *-compliant. THAT is the problem.


MOD:* No links to or names of ripping groups please.

"" Well, yes, everyone has the freedom to choose their setting. So it's dumb to force them to -q7 minimal. ""




Well, nobody is forced to anything ?

Who forces you to rip according to * standard, if you don't like it ?

Who forces you to rip according to * standard, if you don't like it ?




Guruboolez recommends q6 probably for the majority of people, himself he uses higher setting, if I recall correctly.






"" This is a very cheap excuse. When mpc artifacts, it's because of a mistake made inside the psychoacoustic model. It has little to do with the setting being too low for your hearing capabilities. The recent few samples which fail at all quality settings show this. Also, the 99+ % of samples having no problem, show this. ""



No, by higher q setting artefacts may vanish, it depends on the type of artefact.
Those few, which don't vanish at even q8, are very very rare.






Then (which i have forgotten to mention as reason in previous post):

* MPC -> High quality mp3 encoding, for HiFi-DVD/MP3-player,
Transcoding. --> you need safety margin.


* Surround content of music, for Logic7 / DPL2 decoding:
This content is listenable better saved in mpc at higher --ms setting, that is either q5 --ms 15 (still less bitrate than q6, but better sound !), or q7 (which uses --ms 13).








"" Same here - example: you decide that -q5 is enough for this piece of music, you use -q6 for transcoding margin. Even though carefully ripped and encoded in the best lossy format using very good settings, your file is not *-compliant. THAT is the problem. ""


Why do you have a problem with *-compliance ?
Do, what you like, like everybody.


MOD:* No links to or names of ripping groups please.

About * and q7:
Imagine the following situation: An ABX tester sits in front of his computer/stereo equipment, wearing above standard headphones with a good isolation from environment sounds. His eyes are closed to increase his concentration. Then he hears a sample, smaller than a second, which he thinks he can distinguish from the original at q5/6, but he isn't all too sure and must hear it several times.

This is reality for 99.9% of non-artificial (just created to produce artifacts in lossy encoders) problem samples.


Imagine another situation:
A *-Freak who has never done ABX tests, but who has read somewhere that * is cool and the only way to go for quality without compromises (lossy means compromise) enters a room where q5/q6 music is playing on a good 5.1/stereo equipment. His hearing skill may be above average and after one song he is absolutely sure that this must be q5/q6 because of all the artifacts.

IMO, this is pure fiction for 99.9% of non-artificial samples.

MOD:* No links to or names of ripping groups please.

I recommand nothing. At least, not on HA board.
I've just found some cases, on special listening conditions, where mpc -q5, -q6, -q7... isn't totally transparent. That's all. If you listen to harpsichord, and listen to the same second hundred times, then mpc --extreme shouldn't be the best choice.

The most interesting thing is maybe that, after one year with mpc, I can hear problems (and ABX them) I couldn't hear one year ago. Same thing with --alt-preset standard : I can hear pre-echo on regular piano music. But I wasn't able to perform a valid ABX test on it (or simply find short samples) last year ; I can do it now. Evolution of hearing.


P.S. I switched from --extreme to --insane preset last summer.
P.S. (2) : the only recommandation I have is to use WinABX or ABC/HR, and to be careful in future.

Actually, if you look at nearly all the recently discovered problem samples, they still artifact even at q7 for quite a number of people. Just search around the forum (see also the links pointed out in this thread). If you are trying to avoid artifacts present at q5, stepping up to q7 won't help much.

On the other side, some small differences (I don't say problems) we can ABX with --standard with a bit concentration doesn't occur at higher settings.
And on the biggest problems, --insane encodings are less annoying than --standard one.

Of course, nobody is forced to use them - but they have big influence on the public.

If I want to rip something very carefully, these "*" and "*" guides are the first I'll find in Google. They are well-written, and give great advice - except, I think, for the quality setting (which is sad).


Seriously, I don't like pro-logic so much. This kind of encoding lowers the quality of the sound anyway.


You're right - but it's just a bit sad to follow no standard, just because of that.


MOD:* No links to or names of ripping groups please.

True. But then you are still making a compromise (which psychoaccoustic lossy compression will allways be). It shouldn't be labeled otherwise. user: "* is obviously a high quality standard, meant for people, who don't compromise regarding quality/size."


MOD:* No links to or names of ripping groups please.

Well, interesting to hear the opinions here.



As far as I can read, * is q7 and higher, this means obviously, lossless encoding is a possibility, too.

So why all the calling of q7.

Guruboolez expressed it exactly, some problems are less annoying, clearly less, imo.

So I understand, why q7 is a minimum switch.





I have friends, who swear on the transparence of q5, and others, who prefer q8, so what's the matter ?

The point is, if a guy (who is satisfied normally by q5) listens to music of q8, he has same enjoyment, fun of music.
The opposite, if the guy, (who prefers q8), listens to q5, he may be disappointed.









QUOTE :
user wrote:
" ...Surround content of music, for Logic7 / DPL2 decoding .... "


Numlock wrote:
" Seriously, I don't like pro-logic so much. This kind of encoding lowers the quality of the sound anyway. "



Sorry to have found probably messing you up with dpl/dpl2/logic7-DEcoding
I hate pro-logic, you don't like it much, here we agree...
You are messing up DPL with DPL2/Logic7, I assume.

I have experience with Logic7 (nice) , bad experience with DPL, cannot speak for DPL2.


I have written it here at HA, I assume, I was one of the first, who published experiences with Logic7 (as saying goes, DPL2 is very similar, but even commercial German HiFi/"Audiophile" magazine "Stereoplay" wrote, Logic7 would have had a slight advantage regarding sound quality).
Logic7 offers great quality, even with stereo music of the 60s and 70s, as nearly no sound engineer during producing of stereo LPs, thought of surround out of stereo source by a matrix encoding......

DPL is clearly crap for music, it sounds dull, not exact, not precise anymore.
You cannot compare this outdated decoding procedure with Logic7, sorry.

Obviously the encoding of music requires more details, so that the Logic7 decoder finds later still enough details to decide, which parts of music is distributed to each of the 5 channels, FL, FR, C, RR, RL.


MOD:* No links to or names of ripping groups please.

Exactly. It is too easy to say that q7 would be q5, plus wasted bits.

Quality improves (codec failure is less striking), but this improvement is not worth the extra bitrate for many people. I think everyone has to make this choice for himself.
Another myth is that if you go lossy, you cannot expect high-quality. Therefore it would be useless to use a high bitrate for lossy codecs...
* is essentially a concept that puts quality above anything else. Best proof: all requirements are minimal requirements. (start = MPC@q7, the sky is the limit (lossless)). For the same reason (quality prevails over speed etc...) the ripping procedure is superior & strict.



This is thinking in black/white. Some people hear artifacts better than others, no matter whether this is due to training (important factor, as most of us will have experienced ourselves:) ) or good ears or good equipment.

MOD:* No links to or names of ripping groups please.

It´s just a kind of crazy, to increase the file-size up to 33 % (= distance from -q5 to -q7) for each file, just for these few problem-samples, which in normal music practically will never be noticable.



And talking about standards: Setting a standard is a good thing. But it should be set so, that everyone can see the sense in it and is because of this willing to fullfill it. The minimum standard for generating mpc for example should be a secure read-in method from CD (best with EAC in secure-mode) and using an up-to-date mpc-codec-version (1.14). And nothing more. - That´s it!

edit: Don't discuss DC hub stuff here, please.

Let's say q5 is pretty good. Proclaiming it to be perfect goes a bit too far IMHO.


Feel free to make such a "min. standard for mpc"