HydrogenAudio

I want to find out the average commandline of the MPC users, I expect it to be around 6.
The choices are the above HQ modes.
--standard = --quality 5
--xtreme = --quality 6
--insane = --quality 7
--braindead = --quality 8
If you use something else, please comment on what is used.
Examples:
--thumb
--radio
--quality 1-4.99
--quality 8.01-10
tweaked or bla. bla. 

How about this:

--xtreme --xlevel

?

(The "--xlevel" part is to avoid internal clipping, works in most cases.)

Originally posted by kjempen
How about this:

--xtreme --xlevel

?

(The "--xlevel" part is to avoid internal clipping, works in most cases.)

thank you

Read this (http://pimp.gotdns.com/MPC/--xlevel.htm) btw.

Originally posted by MTRH

Read this (http://pimp.gotdns.com/MPC/--xlevel.htm) btw.

I think Frank is a little overcautious about --xlevel's backwards compatibility. One one hand it's understandable, on the other hand it should be safe to use, given that one uses a reasonably up-to-date decoder / Winamp plug-in.

From the information i could gather, --xlevel expands the range of usable scale factors with relative addressing (without --xlevel, 6 bit scale factors are used). In the SV8 test versions, 7 bit scale factors are used for a possible dynamic range of 190 dB. I don't know the advantages of the SV8 method, maybe Frank can explain that. But i'm pretty sure that --xlevel doesn't decrease sound quality. It seems to be just a compatibility concern: Andree's old decoders are not compatible (but they're also more than a year old).

--quality 10 for me...alittle overkill I assume But I can spare the hardrive space.
Plus I have to assume that this is the highest quality possible Lossy.  My files seem to average 295Kbs - 355Kbs.  If  am assumeing wrong PLEASE someone fill me in as to why my thinking is wrong.

Originally posted by smg
--quality 10 for me...alittle overkill I assume But I can spare the hardrive space.
Plus I have to assume that this is the highest quality possible Lossy.  My files seem to average 295Kbs - 355Kbs.  If  am assumeing wrong PLEASE someone fill me in as to why my thinking is wrong.

Hm yeah, it's overkill... i don't even think you're getting much theoretical improvements above --quality 8. Audible improvements already become negligible above --quality 6 for most people.

I was using --extreme for my own CDs ; but after experimenting some *little* problems on castanets like samples, I am now using --insane (the new one, --quality 7). It's overkill, I know it, but, 240 kb/s are not a problem for me. It is just for a maximal security.
But I am using --braindead too, and even --quality 9, for foreign CDs.

I am using --standard for my own cd's.
For foreigns cd's, I just copy them:D

--standard for mppenc 1.02 [for a few albums --standard --ltq fil where --xtreme sounds aggressiv in high frequencies]

--quality 6 for mppenc 1.1 [--standard is not enough on most albums because 1.1 sounds complete different than 1.02]

for daily use on work and home.

CDDA for archiving byed cds, FLAC for archiving all other things like lended cds or own music.

Dezibel

Originally posted by Dezibel
--quality 6 for mppenc 1.1 [--standard is not enough on most albums because 1.1 sounds complete different than 1.02]

??? Can you ABX a difference with "--standard" on most of your albums?

...nope.

i don't did any abx tests. the results are just from on the fly hearing using xmms.
so don't rate my results.

Dezibel

I encourage you to ABX to verify it's not placebo.
ABX is the path of the Klemm side...

Originally posted by Dezibel
--quality 6 for mppenc 1.1 [--standard is not enough on most albums because 1.1 sounds complete different than 1.02]

Interesting. I find --standard to be improved (albeit in tiny ways) in 1.1 over 1.02.

With mppenc 1.1, I am using --standard --ltq 30. That's a very insensitive ATH curve in extreme HF, but I'm deaf in that region anyway.  Even with +30dB@20KHz, the #30 ATH curve still encodes frequencies over 18KHz or so, if they are strong enough. I find this arrangement plenty sufficient and cannot ABX between --ltq 30 and --ltq 35.

Since I want to burn all my MPCs to CD-R, I'm leaning toward bitrate savings over the last word in transparency. --standard generally does a superlative job so I don't want to "invest" disk space for --xtreme and above.

mppenc --braindead --quality 10 --ms 15 --forcewrite --verbose --verbose %1.wav %1.mpc

Note: --braindead is now redundant with 1.1, I guess

Originally posted by audiophile
mppenc --braindead --quality 10 --ms 15 --forcewrite --verbose --verbose %1.wav %1.mpc

Note: --braindead is now redundant with 1.1, I guess

Holy sweet jesus... :eek: :jawdrop: that's like... killing baby jesus
what are you trying to accomplish?

and the KITTENS!! don't forget the KITTENS!
God kills atleast one kitten everytime you encode something like that...:evil:
  :cry2:

Originally posted by MTRH

and the KITTENS!! don't forget the KITTENS!

The kittens will survive, he's effectively using --quality 10 --forcewrite --verbose --verbose. Still overkill though..

default (standard or quality 5)

Originally posted by Dezibel
...nope.

i don't did any abx tests. the results are just from on the fly hearing using xmms.
so don't rate my results.

*sigh*

<RANT>

I certainly won't.  Please don't bother posting results though if you're not going to attempt to verify them in any fashion (sample or abx).

It's simply useless.  It wastes people's time chasing problems that possibly don't exist, and increases the likelyhood of  spreading misinformation and rumours.

Sorry, but we really have enough of this "well I must have superhuman hearing, I can differentiate MPC in all these circumstances, BUT... I won't abx/don't believe in abx, etc, etc, etc" kind of ideology.  It seems to be on the rise to, which is disturbing, because the whole point of HA is to try and be the 1 at least remotely objective site out there where actual useful measurements are performed, and tests have some meaning.

Just one last bit to add, if so many of you guys can supposedly pick out MPC so well even well beyond --standard (and I'm not talking about 4 samples, I'm talking about all the time like some seem to claim or imply), how come you aren't doing something useful by helping Frank to tune the codec and fix these alleged "problems"?

</RANT>

I have recently converted from my heathen MP3 ways and have graced the MPC format

I have never tried an ABX, but I think I will certainly try one in the near future once I have some free time. I am using --standard, and --xlevel since many of my encodes report internal clipping errors. So far I have one track that reported clipping with xlevel turned on, so I did it again using the recommended --scale.

Anyway, I am very satisfied with MPC. I didn't have any problems with --alt-preset fast standard on Lame, but I find that the Muse encoder is on average twice as fast as Lame with the settings I was using, and the files are 10-20% smaller.

Also, I must say that the APE2 format is far superior to the ID3v2 hack.

The one thing that is slowing me down from re-ripping my cds and using Muse is that I have a really great general-purpose plugin for WinAmp called EAR... it reads ID3 tags and displays the information in the MiniBrowser. The format of the output is customizeable using XML. It can display file information on MPC files (bitrate, streamversion, encoder, ...), and displays ID3v1, but doesn't support APE tags. So I have turned EAR off, and am slowly trying to get over it. Perhaps I will find a replacement in the future.

Just for anybody who is curious, I wrote a small software in C++ that creates a batch file for encoding ripped wavs. I originally wrote it for my Lame settings, called it Blame (batch lame) ... encoding settings are completely hardcoded, so if anybody wanted to use it, they would have to tinker with code and recompile.

The way it works is that you tell EAC to use a user-defined encoder, and pass the title, artist, album, year, and track information as command-line arguments. The program uses the arguments for several things... it finds and writes the file names using my naming convention, and also uses the encoder's built-in tagger. It generates a batch file which can be run later on. This is helpful because your cpu isn't being used to encode files while it is trying to rip cds. Also, it allows you to rip many cds during the day, then run a single batch file at night, requiring no extra work later on.

Today, I rewrote Blame so that it would encode mpc for me also, and named it Amuse (auto muse) ... heh i'm having fun with these names.

Ahoy fellow batch file user! I made this one awhile back. It only works with 8.3 format filenames.

@ECHO OFF

DIR /B *.WAV > FILES.LST

FOR /F %%F IN (FILES.LST) DO MPPENC --standard --forcewrite %%F

REPLAYGAIN --auto *.MPC

DEL FILES.LST

Your solution is probably better, but I like keeping everything simple

Originally posted by Dibrom


*sigh*

<RANT>

I certainly won't.  Please don't bother posting results though if you're not going to attempt to verify them in any fashion (sample or abx).

It's simply useless.  It wastes people's time chasing problems that possibly don't exist, and increases the likelyhood of  spreading misinformation and rumours.

Sorry, but we really have enough of this "well I must have superhuman hearing, I can differentiate MPC in all these circumstances, BUT... I won't abx/don't believe in abx, etc, etc, etc" kind of ideology.  It seems to be on the rise to, which is disturbing, because the whole point of HA is to try and be the 1 at least remotely objective site out there where actual useful measurements are performed, and tests have some meaning.

Just one last bit to add, if so many of you guys can supposedly pick out MPC so well even well beyond --standard (and I'm not talking about 4 samples, I'm talking about all the time like some seem to claim or imply), how come you aren't doing something useful by helping Frank to tune the codec and fix these alleged "problems"?

</RANT>

ok... don't be annoyed. i will say nothing more about codec quality in future. i just use what sounds good for me and be happy with the good work from frank klemm and co.

the reasons for me not doing any abx tests are:

- have not much time to make aby tests and play with tweaked commandlines
- short abx samples says nothing about sound quality for me. only hearing a whole album on a real player let me enjoy the music or even not when sounds quality sucks

i can listen to lame --aps songs without having problems with quality. they sounds a little bit warm, a little artifact here and there, but the quality is listenable. i can listen to mppenc 1.02 --standard songs without any problem. i can listen to mppenc 1.1 --quality 6 without any problem. sound is very good. but i can't listen for example oggvorbis at any bitrate. anything disturbs me. can't discribe it.

tuning and listening tests are your job. i can't make expressive listening tests so i don't do them.

can't translate this without be misunderstood:

Ich wäre froh es würde nicht jedes Wort von mir auf die Goldwaage gelegt und meine Höreindrücke lediglich als Erfahrungen eines normalen Anwenders gesehen. Ich würde mich gerne an der Entwicklung bzw. tuning des ein oder anderen Encoders beteiligen. Allerdings fehlen mir die passenden Kenntnisse sowohl im Audio Coding, als auch im Programmiertechnischen Bereich dazu. Ich als Hobbymusiker "höre" sowieso "anders" als normalsterbliche, und wäre damit im tuning Bereich eh Fehl am Platz. mppenc 1.1 mit --quality 6 bringt für mich die beste Qualität bei akzeptabler Dateigrösse die man beim lossy-coding bekommen kann. Allerdings Fehlen hier und da noch ein paar Minimalitäten um wirklich "transparent" zu sein wie Ihr es nennt. Die Änderungen von 1.02 zu 1.1 die gemacht wurden gingen definitiv in die richtige Richtung. Ein klein bischen mehr tuning und das Teil klingt nahezu perfekt.

Ausserdem bin ich der Meinung dass die ganzen Qualitätseinstellungen unnötig sind. Wenn der Encoder laut Entwickler und Testhörer bei --quality 6 stabil arbeitet, d.h. gute Qualität bringt, warum dann nicht diesen Level als Standarteinstellung einrichten, und die übrigen Optionen nur den Entwicklern und Testhörern zur Verfügung stellen? Damit kämen auch nicht mehr diese hirnlosen Bemerkungen der Anwender wie: --standard sounds good for me but i using --braindead just to be sure! Unnötiger Ballast der dem Anwender wie dem Entwickler abgenommen werden könnte.

Lediglich meine Meinung zum Thema lossy Audio-Coding.

Dezibel

Originally posted by Dezibel

- short abx samples says nothing about sound quality for me. only hearing a whole album on a real player let me enjoy the music or even not when sounds quality sucks

I'm working on a small ABX-style file copier/randomizer program, in order to allow those type of tests. Don't know when I'll have it ready, though.

Originally posted by paranoos

The one thing that is slowing me down from re-ripping my cds and using Muse is that I have a really great general-purpose plugin for WinAmp called EAR... it reads ID3 tags and displays the information in the MiniBrowser. The format of the output is customizeable using XML. It can display file information on MPC files (bitrate, streamversion, encoder, ...), and displays ID3v1, but doesn't support APE tags. So I have turned EAR off, and am slowly trying to get over it. Perhaps I will find a replacement in the future.

For instance I use only ID3v1.1 (in EAC) and there're no problems for me... I don't see why you won't choose it over APE.

Originally posted by Dezibel

Ausserdem bin ich der Meinung dass die ganzen Qualitätseinstellungen unnötig sind. Wenn der Encoder laut Entwickler und Testhörer bei --quality 6 stabil arbeitet, d.h. gute Qualität bringt, warum dann nicht diesen Level als Standarteinstellung einrichten, und die übrigen Optionen nur den Entwicklern und Testhörern zur Verfügung stellen? Damit kämen auch nicht mehr diese hirnlosen Bemerkungen der Anwender wie: --standard sounds good for me but i using --braindead just to be sure! Unnötiger Ballast der dem Anwender wie dem Entwickler abgenommen werden könnte.

I don't know. I think the quoted statement is more honest than stating that --standard is completely insufficient without providing evidence of this. It doesn't say anything is wrong with the codec and thus is not leading people to stay away from certain settings without good reason ("standard can't be enough, because it's only standard").

Originally posted by lucpes


For instance I use only ID3v1.1 (in EAC) and there're no problems for me... I don't see why you won't choose it over APE.

Most of the artists are trunca

Originally posted by Continuum
I don't know. I think the quoted statement is more honest than stating that --standard is completely insufficient without providing evidence of this. It doesn't say anything is wrong with the codec and thus is not leading people to stay away from certain settings without good reason

I don't read German, so I wouldn't know.

From a crude Babelfish translation I get the impression that there's the statement in there that the inclusion of the other settings must mean that they are somehow useful/necessary.

Well, first of all, Buschel, the original MPC author, has stated many times that --standard is supposed to be transparent.  Beyond that I've never seen a clear reason for the existance of xtreme, and insane has always been unnecessary and is just a mode useful for comparing graphs in cool edit.

This is really pretty much similar to the --alt-presets in LAME.  --alt-preset standard is what you're supposed to do.  --alt-preset extreme offers no clear benefits, and insane is totally unnecessary.  Almost any sample which screws up standard is also going to screw up insane, just maybe to a tiny degree less.

Back to the original point, if someone says they can hear a difference at such a level, and at a point to where it goes beyond what the majority of this board have found to be the case, then I expect some sort of proof of this.  I'm not looking to say that everyone is wrong, I'm looking to find the truth of the matter.  I'm tired of people jumping to stupid conclusions based on false information.  There's enough of that out there on other sites, we don't need it on HA as well.  I'm a little more agitated about this kind of thing than normal given some of the comments made in regards to ff123's test on other sites.  It's clear that most people don't give a damn about objectivity or proper testing at all.  Well, as the founder/owner/admin of HA, I've stated many times that this is not what the forums here are about.  Instead, the idea is to create progress through assembling a knowledge base founded upon reason, verifiable data, and tendency towards objectivity.  I don't want to see these continue to be dilluted by the current trendy nonsense which has been popping up around MPC (ie, the need to encode with --braindead because --standard just isn't good enough, etc, without performing any listening tests or providing any test data).

("standard can't be enough, because it's only standard").

This assertion has not a single shred of correlation to actual measured audio quality.  It's simply a misunderstanding of terms.  Perhaps --transparent would be a better name than --standard and then everything above could be named --bitwastelevel1...level2... etc.  It really doesn't matter what you call it though.

Everyone should feel free to use what they want to encode.  I'm not so pretentious as to tell people what they can and can't do.  However, if you plan on recommending settings to other people (or even implying that they are necessary) on this forum and you make an outrageous statement, be prepared to back it up.

Originally posted by Dibrom

This assertion has not a single shred of correlation to actual measured audio quality. It's simply a misunderstanding of terms. Perhaps --transparent would be a better name than --standard and then everything above could be named --bitwastelevel1...level2... etc. It really doesn't matter what you call it though.

Everyone should feel free to use what they want to encode.  I'm not so pretentious as to tell people what they can and can't do.  However, if you plan on recommending settings to other people (or even implying that they are necessary) on this forum and you make an outrageous statement, be prepared to back it up.

I'm not sure, whom you meant by this. "standard can't be enough, because it's only standard" was meant as an example of unsubstantiated critique against certain encoder settings. I certainly don't agree with it! (If this was unclear in my previous post, shame on me)

I'll try again: IMHO it is more honest and (fair to the developers) to say "I use --braindead, although --standard is already transparent for me, just to be sure" than to say "I feel something is wrong with --standard, but I don't bother to abx it".

Originally posted by Frank Klemm


Most of the artists are trunca

That is pure comedy.

Disclaimer: This is completely unscientiffic and I don't recommend this setting to anyone!
I currently use --quality 5.6 and am thinking about adding --xlevel to that commandline. Why 5.6? That gives me filesizes in the middle of standard and xtreme (q5.5 is closer to standard). q5 is transparent to my ears (except 2nd_vent, but this info is close to irrelevant). I just felt that i could spend a little more bits to help me get that warm fuzzy feeling. I guess in such a situation others go ahead and use q10. I try smaller steps. The quality scale is real nice since I don't have to tweak the command line.
Again, I am aware that what I'm doing is not recommended, but hey, you asked for it.

Little anecdote: yesterday I encoded the Autechre Album "Confield". I thought: hehe, codec killer material! Took the track that seemed most difficult: "Pen Expers" and encoded with mpc q5, mp3 aps, ogg q6. I investigated the blips that sounded most wicked but I couldn't abx any of the encodes! I have lost some confidence in my ears but have gained trust using mpc.

Thanks all who work so hard to tweak these formats! And thanks for that laugh Frank ("Most of the artists are trunca"). :rofl:

Originally posted by Dibrom
Well, first of all, Buschel, the original MPC author, has stated many times that --standard is supposed to be transparent.  Beyond that I've never seen a clear reason for the existance of xtreme, and insane has always been unnecessary and is just a mode useful for comparing graphs in cool edit.

This is really pretty much similar to the --alt-presets in LAME.  --alt-preset standard is what you're supposed to do.  --alt-preset extreme offers no clear benefits, and insane is totally unnecessary.  Almost any sample which screws up standard is also going to screw up insane, just maybe to a tiny degree less.

I think ff123's recent 64kbps tests have helped put things in perspective. The 5 encoders performed so widely that comparing the quality of --standard, --xtreme and --braindead seems like an exercise in futility. Comparing 64kbps AAC with 64kbps Ogg -q0 often reveals dramatic differences, but it is an extremely rare sample indeed where the same thing happens between MPC --standard and --xtreme.

When ff123 rolls out the high bitrate test that he's already hinted at, it would be interesting to include both MPC --standard and --braindead samples. I wouldn't be surprised if people give both encoder settings a 5.0 on the ABX/HR test. I gave a few of those 64kbps clips a 4.0+ rating here and there. Triple the bitrate and use an encoder hellbent on transparency and you are looking at a bunch of 5.0's.

Originally posted by Gecko
I currently use --quality 5.6 and am thinking about adding --xlevel to that commandline. Why 5.6? That gives me filesizes in the middle of standard and xtreme (q5.5 is closer to standard). q5 is transparent to my ears (except 2nd_vent, but this info is close to irrelevant). I just felt that i could spend a little more bits to help me get that warm fuzzy feeling. I guess in such a situation others go ahead and use q10. I try smaller steps. The quality scale is real nice since I don't have to tweak the command line.
Again, I am aware that what I'm doing is not recommended, but hey, you asked for it.

Gotta love that warm fuzzy feeling.

But it's understandable why it's not recommended. You aren't getting poor quality, of course, but you're simply wasting bits. I suppose the whole purpose of --standard/--quality 5 is that it represents the point where anything higher is hitting the point of excessive diminishing returns. How does 2nd_vent ABX between --quality 5 and --quality 5.6? Maybe you need --quality 8 to gain transparency on this clip, but I guess the trouble is that you'd be "forced" to use --quality 8 all the time to "rule out" the risk of non-transparency and meanwhile you are boosting bitrate by 40% or so.

If you plan to store music on a hard drive only, you can use --quality 10 all the time and realize it's completely unnecessary but you wouldn't have to justify it because hard disk space is so cheap nowadays. But my big thing now is burning my lossy files to CD-R and I'm stuck with a 700MB budget per disc. --quality 10 would mean about 3 albums per CD-R and that rather defeats the purpose. Now that I am using --standard --ltq 30, I can fit roughly 10-12 albums per CD-R...at a quality level is pragmatically transparent.

How does 2nd_vent ABX between --quality 5 and --quality 5.6? Maybe you need --quality 8 to gain transparency on this clip, but I guess the trouble is that you'd be "forced" to use --quality 8 all the time to "rule out" the risk of non-transparency and meanwhile you are boosting bitrate by 40% or so

Around --quality 7 is transparent for me and others who tested it, I use --quality 7.4.

LOL @ Frank's comment...

Originally posted by Frank Klemm


Most of the artists are trunca

Yes, my benchmark for a tag is Weezer's song "The World Has Turned and Left Me Here" ... ID3 leaves out the last two words. Yes, it is adequate for most uses, but I do prefer the flexibility of a variable-length tag. And as I understand it, the WinAmp MPC plugin that Frank wrote (or optimized?) seems to support only one tag format per file... either way, I technically COULD use both APE2 and ID3v1, one for flexibility, the other for compatibility with EAR, but that isn't a very smart solution.

I tried the LCD plugin yesterday, but I find the design of the on-screen option to be fairly limited. I do like EAR, and I would be more than happy to rewrite it to allow for APE2 tags, but I am quite sure the source is closed. And I have never seen an update for this program... it has been version 1.01 since I first got it.

btw, we need more smilies! i could really use one like "shake head" to show disappointment.

Originally posted by Dibrom

Perhaps --transparent would be a better name than --standard and then everything above could be named --bitwastelevel1...level2... etc.  It really doesn't matter what you call it though.

Apparently not., or people would not use the settings with names as 'insane' and 'braindead'.

They're called that way for a reason.

--
GCP

Hmm...
A slightly better tuned --standard as --transparent would be quite nice.
Gives an entirely other impression compared to plains standard...
psychological factors here, seriously it sounds great.

Nice to see the confidence in MPC. 50%+ using the --standard profile.
Even though this forum has many extreme users...
wow

--quality 7.4

--standard

Because its enough for me, and its trusted.

Ruairi

I somehow missed this thread...

I use --xtreme to have some extra headroom over --standard for jobs like re-encoding. I wouldn't mind at all using --standard, I don't doubt for one moment it's transparent on most samples. But I don't know how much worse it would perform when re-encoding.


Dezibel: I don't believe a single word. I agree entirely with all of Dibrom's comments.

What you say about only being able to judge quality when listening to "a whole album on a real player" is simply wrong. You can only judge the quality of audio codecs using relatively short samples and a blind test.

If things like "warmness of sound" or whatever suddenly vanish, it's not due to the test environment, it's because these things simply don't exist with psychoacoustic audio compression. The only deficiencies that can exist are artifacts. If you don't detect any of them, the sample is transparent to you, it's as simple as that. If you feel there's something wrong with a sample you couldn't ABX, it will most likely be the placebo effect playing up on you. (Remember Beatles?)

I had a discussion like this recently, also with a musician. Here's an excerpt:

> Ich bekomme Kopfschmerzen, wenn irgendetwas am Sound nicht stimmt.
> Trotzdem ist Musik für mich nicht nur Technik- ich höre in die Musik
> hinein- bei 10sec Samples ist das gar nicht möglich- viel Elemente,
> Details erschließen sich erst in der Gesamtheit.
> Klang ist nicht nur die Summe von Tönen, sondern auch deren gegenseitige
> Beeinflussung. Zum Beispiel spielt der Konzertsaal eine entscheidende
> Rolle- wie harmonisch das Klangbild am Ende ist.

Diese Dinge haben mit dem Beurteilen der Qualität von psychoakustischen
Audiocodern *nichts* zu tun. Entweder sind Artefakte da oder nicht, Punkt.
Ob die Qualität der Aufnahme gut ist oder wie das Klangbild wirkt, ist
völlig nebensächlich - es kommt 1. darauf an, gut trainiert im Wahrnehmen
von Artefakten zu sein und 2. alle Details des Samples zu erkennen (ja, das
muss auch bei kurzen Samples möglich sein) - also auch Instrumente, die man
normalerweise, weil sie so stark im Hintergrund "versteckt" sind, gar nicht
wahrnimmt -, um sie dann auf Artefakte "untersuchen" zu können.

Wenn man einen ganzen Song als Ausgangsbasis dafür nimmt, kann schon mal der
letzte Punkt nicht erfüllt werden. KEIN MENSCH schafft es, sich wirklich
*alle* klanglichen Eigenschaften eines ganzen Liedes so gut einzuprägen,
dass er, wenn erst das Original und dann der Encode abgespielt wird, sofort
zuverlässig sagen kann, wo er Unterschiede gehört hat. Das kann er auch
nicht, wenn er Original und Encode mehrmals nacheinander durchhört.

Testsamples von maximal etwa 20 Sekunden Länge sind die einzige Möglichkeit,
sehr subtile Unterschiede wirklich zuverlässig aufzuspüren.

Diese ganzen Audiophile, die sich einbilden, sie könnten die Qualität von
Codecs im normalen ABC-Test _mit einem gesamten Lied als Ausgangsbasis_
aufgrund ihrer langjährigen Erfahrung im Hifi-Bereich beurteilen (Redakteure
von Hifi-Zeitschriften sind das beste Beispiel), würden im ABX-Test mit dem
gleichen, gesamten Lied kläglich versagen.

Das merkt man spätestens dann, wenn man sich ihre Beurteilungen der Codecs
durchliest (es wird nie ein Artefalt wie Klirren oder Flanging oder Pre-Echo
etc. erwähnt). Das Beurteilen von Audiocodecs ist EBEN NICHT das gleiche wie
das Beurteilen von Hifi-Systemen und Lautsprechern, hier kann man sich EBEN
NICHT auf das generelle "Gefühl" verlassen, das der Sound angeblich
vermittelt - man kann sich nur auf Artefakte stützen. Ein "Gefühl" gibt es
bei psychoakustischen Audiocodern nicht. Es gibt nur Artefakte (oder auch
nicht), und die kann man im direkten Vergleich von zwei ganzen Liedern nicht
feststellen. Dafür sind kurze Samples und ein ABX-Test notwendig.

CU

Dominic

At the risk of sounding like a newbie, what is an ABX test? How do they standardize file comparisons?

--Rizban

ABX is a type of blind test, that is applied here to audio.

See:

www.pcabx.com (http://www.pcabx.com), which also includes links to some computer-based ABX comparators.

http://users.htdconnect.com/~djcarlst/abx.htm (http://users.htdconnect.com/~djcarlst/abx.htm) , more info about ABX procedure & test results.

...i can listen to lame --aps songs without having problems with quality. they sounds a little bit warm, a little artifact here and there, but the quality is listenable...

You must have unbelievably extreme hearing capabilities or...you live listening to critical samples.

A little warm , a little artifact here and there , with --alt-preset standard?.And you hear these artifacts "on-the-fly" without direct comparison to the original audio?.Come on...

Someone should release a new lossless codec and make people believe that's lossy.I'm sure a lot of them will claim they can hear artifacts "here and there" , "it's too bright" , "too warm" , "synthetic , lifeless".Without providing any proof , of course."I can't prove it , but I can feel it in my bones man".

       

Unfortunately I do not posses the technical skill or free time to help make the –standard setting work better in MPC.  However I think it sounds awesome right now the way it is.  –quality 6 I can understand as well I guess, but the very few sound files I’ve had problems with (with –quality 5), also I have noticed they suffer somewhat as well with quality 6.  People encoding at quality 10 are just trying to feel good about themselves seeing a high bit rate # in winamp…  I personally don’t use lossless, but I’ve played around with them and they work great, I’d use Monkey or Flac in a heart beat if I could afford all the extra HD space and DVD-Rs..

Back to my point, and I think Dibrom would agree:  The point of a lossy codec is to remove what we don’t hear and preserve the rest to get the best transparency possible.  The key here being that all of this should be done at the –lowest- bit rate possible, but still retaining transparency.  Quality 10 still loses some quality, it’s just the nature of lossy codec’s.

</rant?>… lol
--standard sounds awasome... try it more and see..  don't look at bitrate, it'll only make you think you're missing out on some quality.  It is possible to have a lower bitrate than LAME alt standard and still have awasome(actually better) quality.



On a final note, I wish there was a good file sharing program that supported MPC.... :-(

 I admit to like high bitrates being displayed on winamp (I keep the update speed at 1, so it's faaaaaast) !!! 
I just directly started with MPC and kept using a exaggerated "security" margin like --quality 8.5 --xlevel, as its average is always LESS than my lame --insane (320kbps) encodes.
Overkill? Bah! I'm not sharing my files@56k modem    . Slow for that!
And the best! It goes up to 340 or more for about a bit more than one second on some samples!!! Me likes ultrasonics!!!! 

--quality 7 --xlevel

Insane! Overkill? Maybe, but the file sizes are still smaller than aps :]

--quality 7 --xlevel

Insane! Overkill? Maybe, but the file sizes are still smaller than aps :]

I assume you mean --alt-preset standard?

I've encoded around 180+ CDs so far with MPC Insane, which replace my current --alt-preset standard CDs and I've yet to see a MPC Insane encode come out smaller, granted I've not checked every single album though.

--quality 5, or "standard", along with --xlevel, of the 100+ CDs I've encoded, it was transparent 100% of the time, even after exhausting amounts of ABX testing, I've yet to find a result that wasn't transparent at --standard

--quality 7

standard is good enough for me.

--extreme. Chose it over --standard just to be on the safe side. It yields filesizes of about 190-210kbps on average so it's a nice balance between qualty and size.

If you want to give your MPC settings, start a new poll, this one has been completed one year ago !