With the many versions of mppenc coming out, I'm simply wondering about what version to use.

I just read that 1.05e is not a realese version, as " /.../ I expect some quality problems now, find out where /.../" indicates...

How are the test and release versions numbered?

Johan

Test versions, such as 1.05e, are released internally, and normally won't appear on the website (http://www.uni-jena.de/~pfk/mpp/#encoder-binaries).

The last "stable" release, so to speak, was 1.0. I suggest using 1.04, or alternatively 1.01j if you still have a copy of it.

Wasn't there a rather big drop in bitrates for the 1.04, causing some people to question quality?

I find it a little strange that 1.0 isn't downloadable on the main sites. I still have it, I think, but have been using 1.02 since it appeared. Are there some bugs that one should keep in mind using this version?

Johan

Here you can look up what has changed since 1.0. I also added downloads, if available.


1.01a: http://www.hydrogenaudio.org/forums/showth...=&threadid=1512 (Download not working)

1.01g: http://www.hydrogenaudio.org/forums/showth...=&threadid=1583 (Download not working)

1.01h: http://www.hydrogenaudio.org/forums/showth...=&threadid=1613 (Download not working)

1.01i: http://www.hydrogenaudio.org/forums/showth...=&threadid=1644 (Download not working)

1.01j: http://www.hydrogenaudio.org/forums/showth...=&threadid=1773 [Download]

1.02: http://www.hydrogenaudio.org/forums/showth...=&threadid=1843 [Download]

1.04: http://www.hydrogenaudio.org/forums/showth...=&threadid=1980 [Download]


1.01j and 1.02 should also be a safe choice.

The quality affecting changes started in v1.02. The gapless issues were fixed in v1.01g. Therefore v1.01g-j are IMHO good choices if you want to stick with the old and proven quality settings.

I've made a simple Windows frontend for people who want to play with the different mppenc versions. If you keep them in one directory named something like: mppenc101j.exe, mppenc104.exe, then you can select the encoder in this frontend. Download here.

Thanks for that, Citay, that cleared things up (I was getting a bit confused myself with all this MPC talk)...

I may be blind (I'd better say I haven't been keeping track of the forum lately ), but do I see correctly that there is still no "real" answer to the bitrate drops in mppenc 1.04? ???

CU

Dominic

I just verified that 1.01j produces the same output as 1.0 final (at least with --braindead), provided that the very rare gapless bug doesn't occur. The unique sound data has the same CRC. That makes 1.01j probably the best "safe choice" for now.

Thanks for clearing things up a bit.

I think this information is something that should exist on the mpc sites. But maybe it's hard even for the site administrators to keep track of the releases - I just noticed that the version to download on musepack.org changed from 1.05e to 1.04 in the latest hour or so. Good - it could hurt the use of mpc if test versions are the most easy to find.

And /still/ nothing "meaty" regarding the bitrate drop?

At least I haven't seen anyone reporting that they could verify a change with abx or similiar. That's a good sign, but I hope Frank will explain a little of what is happening soon.

Anyway, thanks again! Johan

Yup. But on the other hand, no one actually found any problems (not yet at least).

It sounds like most of the Musepack users are deaf. Datarate of mppenc drops
around 20% on average and noone found the places where are bits saved ;-)

On the other hand we have "golden ears" which can distinguish between insane
and braindead.

I re-encoded 4 CDs with 1.04, they were previously encoded with 0.9x-1.0. All of the new files were larger than the old encoder had produced. I only got bitrate drop when comparing against 1.02.

Well for now I am sticking with MusePack Encoder version 1.04, I hear no difference between this version and 1.01j.

Cheers
AgentMil

Yeah, all right, we're bad boys and should be testing more. I'd love to do so, BUT...

You still haven't given any information on why the bitrate actually does drop that far, and what you actually want us to find out. It's no good keeping on saying "you should be able to hear a difference" and keeping everything else kind of secret.


That was just one case, and not even a member of this board, as you know.

You saw what CiTay said in another thread - testing with --standard* makes no sense in many cases, because it is already transparent very often. And, according to you, testing with --radio* doesn't make any sense either, because hardly any changes were made to that preset. (And testing with --xtreme* would not make any sense either, would it?) So please tell us, what exactly do you expect us to test for? You know the community here is willing to help the development by testing, but we do need some information first.

u.A.w.g.

CU

Dominic

*: or the corresponding quality settings

If Frank told us what he did to produce the bitrate drops, we would "hear" these changes. Placebo effect. Let's say he did something with the low frequencies and published that info here. You'd get a slew of responses saying "oh, yea, the bass is not right".

Mind you, I DID write here that I noticed something audibly different with mppenc 1.04 but have been unable to convert subjective perceptions into hard objective proof. So I do think that whatever Frank changed in mppenc resulted in audible differences but they are tough to track down (which means they may or may not be real). Since the biggest bitrate reductions have been on electronic music (think Moby), my hunch is that the changes apply to the treble region of the psymodel.

I am almost certain that Frank made a change that negatively affects quality (from a measurement perspective), but the question that remains is: is the reduction in quality equivalent to dropping from, say, a 100 to a 99.7 or a 100 to a 97.5? If you can reduce the bitrate by 10%, for instance, and only a few people can hear the degradation on just a few clips then that's probably a good change.

I can understand that, but there's a simple solution to that, just demand blind tests. ABX will at least prove that there is a difference, whether people would then describe it correctly is another question.

Anyway, my main point is that none of the presets (or corresponding quality modes, whatever) seems worth testing, for the reasons I mentioned: Standard is transparent in most cases already, xtreme certainly is (perhaps even if there is a more or less large quality degradation in the new version), to say nothing of insane or braindead, and radio was not modified heavily.

Anyway, I'll start testing soon, perhaps the quality degradation is very audible after all (which would make all I've written so far absolute bullshit )

CU

Dominic

I changed:
- stereo imaging for all modes
- things which introduce ringing
- ATH has changed by really a lot (>30 dB for xtreme)
- TMN/NMT handing over SPL (sound presure level)
- bandwidth handling
- PNS usage (not --standard and above)
- pre-echo handling

HQ mode Testing should be done
- --standard --nmt [1 dB less than the default] --tmn [2 dB less than the default]

MQ modes:
- --radio
- --quality x.xx, so you got the same bitrate as the (old) encoder you
want to compare

LQ modes:
- --thumb

Thanks.

So meanwhile it has been figured out which ENcoder to use.

What are the facts about decoders ?


Are there different ones available ?

In the case of "yes", if the latest deocder is used, is it able to decode perfectly even mpc-files, which have been encoded with older encoders ?

Yes, there are. There are normal decoder + decoders with 16bit, 24bit and 32bit outputs that can dither and noiseshape.


Yes, all official stream versions SV4 - SV7 will work.