I would like to ask a general question and get people's opinions on which setting is better, --standard or –xtreme. I have read though general guides and I know the basic differences between the two.

I have somewhat decent speakers for my PC now and can’t notice a difference between the two settings. I will most likely be getting a much better system in the future.

Basically… please answer:
Is “--standard” good enough for you? Do you personally hear the difference between the two settings? Would you recommend that I use the --xtreme setting so that everything will sound good on my new system in the future?

Also... if file sharing programs start supporting MPC, would you look for standard files or xtreme?

thanks!

I've said it before and I'll say it again, I see no objective point in using 'extreme' over 'standard'

The only thing it gains is a warm fuzzy feeling for some people.

I use standard myself (it's transparent to me), but as Garf said, if extreme gives you the warm fuzzy feeling - go ahead and use it.

There are some point where --standard is objectively under --extreme (for me)

· Strong attacks (like castanets.wav), where --standard generate a strange distorsion (not really pre-echo for me).

· On some difficult to encode passage, it's possible (but really hard : it depend on moments, at least for me) to ABX --standard, but not --extreme. Difference is really subtle, and ABX often fails.

· less significant, when you listen at high volume very silent passage (a piano extinction for exemple), tonality is really different than with --extreme, much closer (identical ?) to the original sound.

=> that's my personal experience (objective I think, if succesful ABX tests are assimilate to that word)

I agree with guruboolez. There's definitely ABXable difference between standard and xtreme sometimes. Depends then on the people, who really can hear this and who can't. castanets.wav is definitely a sample which is quite easily ABXable with standard and improves with xtreme.

I wouldn't consider this an objective disadvantage.

Same thing for castanets. Are there similar problem with music (not known testsamples)? If so, shouldn't --standard be fixed or improved as it's then obviously falling short of it's goal?

The last problem sounds strange, almost like a bug?

I answered to the question : « Do you personally hear the difference between the two settings ? »
But you're right : it's not really a disadvantage. It's not annoying at all, just a very small difference.
But I find it really annoying for castanets.wav

For the tonality on very low volume, you're mabe right. I noticed it yesterday, with the 1.14 mppenc version. I should test it with previous versions.
Note that presets under --quality 5 are obviously distorded in the same condition (I suppose it's PNS).

Over PC speakers, I often have a challenging time telling --radio apart from --xtreme. Listening over headphones is a different situation.

The 10+% size penalty that --xtreme imposes is generally not worth the rare improvement it offers over --standard. I don't think any preset has been more tuned than --standard.

Although some people say they note a difference, this is not likely in most cases.
[BTW Other people want an extra safety margin and are willing to pay for that, yeah, why not?]

Standard is a nice profile, but the eternal problem with psychoacoustic models is that the model might underestimate the needed bitrate for transparency (it fails) and in such cases standard turns out to be insufficient.

If the underestimation is not exaggerated, you won't hear any problems in standard, or sometimes you will hear them in standard and they disappear if you throw some bits at it (xtreme). But many failures will still be noticeable, you really gotta throw buckets of bits in order to eliminate them. This is not so efficient, and that's why F. Klemm concentrates on improving the model itself (making it smarter).

So, if your purpose is eliminating those failures, xtreme is not very suited (hardly better than standard).

--standard will *always* sound "good", no matter what you throw at it. Cases where you can ABX --standard are very rare as guruboolez pointed out, and without a direct A/B comparison, you wouldn't even notice there's something wrong with the encode.

I used MPC --xtreme for nearly a year just for the sake of feeling better , but --standard is just fine.

Volcano: and without a direct A/B comparison, you wouldn't even notice there's something wrong with the encode.



*cough cough* (sorry, I can't just ignore this)

Right, but transparancy is probably our aim, isn't it?
(If not, what's the point of using Musepack? 192kbps MP3 will sound good too, untill you compare to the original...)

honz318712: so that everything will sound good on my new system in the future?

The quality of your stereo (or speakers, on condition they are good quality) will make much less difference than you would expect... Maybe you could verify this by listening through a stereo (probably your stereo is better quality than your decent PC speakers)

I agree. Standard profile MPC doesn't degrade like mp3 does but it becomes painfully obvious *in places*. Most people wouldn't notice it but because I encode I know what to look for, so it becomes apparent. 99% of the tune is no different but the occasional bit sticks out. Dont know why.

seany

I know that's nosense, but I started directly to encode with --quality 8.5 (braindead+overkill). Transparency problem ended.

liekloo:



Oh come on, that's comparing apples and oranges. MP3 @ 192 exhibits obvious flaws that can easily be picked up without a direct comparison.

To make you feel a little better, I'll add this to my above post: "If you want to be absolutely certain, use --xtreme, since file sizes produces by this preset are still very acceptable. Anything above that is a waste of disk space though IMHO."

Standard sounds fine to me on a pretty decent hifi system. I used insane for awhile, but I can't honestly tell the difference. Possibly I could if I abx'd them, but I generally listen to my music casually, and if I can't pick out artifacts on casual listening, thats good enough for me (often I can pick out such artifacts with low-bitrate mp3 though).

I'd use higher than -standard only if you anticipate transcoding (though you are better off just encoding the original source to both formats while you have it)

If the model underestimates the bitrate needed for transparency using the standard profile then there is a reasonable chance it won't allocate enough bits using the xtreme profile either. Certain artifacts (pre-echo, for instance) can't be weeded out simply by making the model a little more sensitive and it seems the only real problems MPC has left to eradicate are those that cannot be simply removed by using a higher quality profile.

Standard has been tuned so that very few people will have a legitimate complaint about its performance. In fact, perhaps a lot of people would find --quality 4.5 just as transparent. I think --standard has enough of a safety margin where going to --xtreme or higher is more of an exercise of feel-goodness than necessity.

Some people won't use --standard because the word doesn't sound stringent enough. "Standard? Like default? Bah, that's not good enough for my ears. I need something above the standard."

There is much different in many areas. Read the excellent topic by Artemis3.

Lately it seems as if some albums I had encoded with --standard and even with --xtreme some time ago now sound just a bit scratchy in the higher frequencies. I'm still using 1.01j, maybe it is too outdated. Or more likely the music encoded is too demanding.

I have yet to ABX, so until then I'm just another afflicted with over-sensitive warm-fuzzy syndrome. But my opinion is there is less high-frequency scratchiness in --xtreme on some grindcore material I have.

Oh, come on guys, don't be so perfectionistic ! A lossy codec will never never generate absolutely perfect music-files. And meanwhile I can't hear anymore the pretentions "I can hear a difference between --standard and --xtreme" without any real proof ! What is much more important than the last 0,005 % of quality-reach, is, that a lossy music-file has to be short, short and short - so that it can be transferred acceptably fast thru the always and forever slow internet. Finally this "lossy-coding"-technique was invented only for this reason. Take a look on those file-sharing-services like Direct Connect ++. If you have ever been logged-in to a hub like aquaudio, you know what I'm talking about. Even if you have a so-called broadband-connection (which are mostly asymetric with a ridiculous upstream-speed), you will have to wait for ages until a complete music-album is transferred completely from another ADSL 128kbit-upstream-line (which you have to share often with 2 or 3 other downloaders) to your computer. ....And then these people you download from also have had the "great" idea, to encode this album in --braindead. Iiiiiii.... que locco!
What I want to say with this: Please, stop encoding in --braindead, if you want to share music over the internet, and I'd recommend also to stop encoding in --xtreme, even if it's just for that "warm fuzzy feeling", cause this is just for nothing. For archival purposes I also can't see a reason, why to encode in --xtreme or --braindead, as they are both lossy. Better use a lossless-codec for perfect quality, even if the files are 3 or 4 times larger. But as the harddisk-space is very cheap in these days and meanwhile the HD's are huge, that also shouldn't be a problem.

i use mpc --standard, mostly for casual listening and to conveniently skirt around the need of having to browse for the cds i want to listen to all the time.

it really does the trick for me..


for close listening on really good equipment i use the cds themselves, and i have better things to do than trying to spot subtle differences between a lossy file and the original wav all day. you most likely won't hear those without the direct comparison anyway..


as far as file sharing is concerned: any version >= --radio will do, --standard preferred.

I whouldnt say I hear difference between standard and extreme, exp on my pcspeakers, but if i burn a cd, i dont want to hear diference either on a good stereo, or with headphones. and, I am transcoding the filez to my mp3player, so some extra headroom might be cool. And I dont care for the few extra bits.

hehe very funny
that's exactly how I feel...sometimes

i use --xtreme for the following reason:

a few weeks ago i wrote Frank Klemm an email. i asked him how to tweak the encoder for more "fuzzy" and "warm" feeling. he just answered that i should play with the --nmt and --tmn values for "my_perfect_sound". i did many encoding sessions (encoder was mppenc 1.02 under linux) on very different music genres. most techno and house music sounds good at --standard for me. most metal and hardcore music sounds good at --xtreme. but female voices (tori amos, sheryl crow) and guitar solos (steve vai, joe satriani and yngwie malmsteen) sounds not really "live" for me, even at --braindead.

i've encoded steve vai's "tender surrender" at --standard, and increased the --nmt and --tmn values in little steps. then i did the same operation with the --xtreme, --insane and --braindead profile. my experience was that a higher --nmt value than 8 let high frequencies sound aggressiv, and a higher --tmn value than 20.5 let the whole song sound fat and unprecise (fuzzy, blurry). if i remember correctly the old encoder (1.02) used --nmt 8 and --tmn 20 in --xtreme profile.

i have not tested the newer ones, but since these tests, --xtreme sounds adequate to me, even on dream theater

BadHorsie (alias Dezibel)

And to no surprise, I'm sure many of the board members would like to see your ABX results for the fuzzy and non-fuzzy settings vs the original sample. Perhaps even with a verdict on each setting's own fuzziness. There is a chance it could lead to the profiles being renamed: --notsofuzzy --kindawarmandfuzzy --totalfuzz and --guitarpedal) B)

Cheers,
Uosdwis

--cathartic
--mindblowing
--heroinfortheears