Some people say that there are artifacts present in some standard encodings that are not in extreme. I always thought that extreme was a waste of space. And for transcoding, your gonna loose quality with any lossless, so no sense kidding yourself.

In my experience with LAME, the few samples that have tripped-up --aps didn't really sound much better using extreme, so I just stick with standard for my MP3 needs. Same for MPC standard. From what I've read from other users, extreme and insane are probably a waste of space on all but a few samples, and even then, it's questionable as to whether they will sound all that much better due to the inherent limitations of each format. As for transcoding (which is evil ), I'd rather see people go from MPC --quality 5 or 6 rather than from LAME --ap anything.

{Edit: clarity}

MPC's "standard" setting was designed for aural transparency. However, in real-world experiences, it has been found that occasionally certain passages prove very difficult to encode. Therefore, very occasionally the "extreme" setting can be used to provide more headroom, to make sure that these passages are encoded without artifacts. Note that it isn't normally essential, and that most music won't need any more than "standard," even on high-end systems. Also note that *unless* you have a nice system or headphones, you won't be able to tell the difference anyway, because "standard" is so close already.

I use quality 7 (insane). Average bitrate is ~225kbps which is ok by me.

Yea, but why? Just 'cus that's the bitrate you wanted? I think that setting encodes the entire frequency range, which I think is kinda pointless.

My only problem is that when I go to transcode for my portable, if I want a good MP3 I go with --alt-preset 160. Many times this gives me a LARGER file than the MPC. That kinda irks me a bit.

Although there is no data or test results to back up this statement (been encoding with MPC for only a few days) - I'm not happy unless the stream rate is pushin' up into the lower 200 kbps range. Let's not get too efficient now..!!


Enjoy...

quality 5 should be enough for 99 % of the music, the other 1% maybe needs more quality (like quality 6 or even quality 7). I'm just use quality 5, and I couldn't hear any artifacts yet, I think that mpc is the best lossy format.

It's possible that you might notice an artifact with standard but not xtreme. However, these situations are very uncommon and if you hear problems with standard you'll usually also hear them with higher bitrate presets. When MPC artifacts, it doesn't have so much to do with the ATH thresholds and noise maskings but rather because of flaws of the format (SV7 has some "incurable" though relatively minor weaknesses).

I discovered recently that I'm able to ABX some harpsichord tracks at -q 5 and at -q 6, but with some effort. With insane, I'm KO

Hmm... what do you think about transcoding? Could the extra bits make that a little better, or is that a waste as well?

As I desperately tried to find a compromise between lossless and lossy compression I just realized the only reasonable way is using BRAINDEAD setting. It sounds decent and acceptable even on the HiFi equipment around $ 5000 And as for transcoding, when using lossy compression, there's hardly better alternative... I made some CDs from MPC Braindead and passed it around between my friends telling them not what's goin' on... Everybody was satisfied, definitely

None of this means that --braindead is required for "good sound" though.

Personally I'd be surprised if most of these people could tell the difference between --standard and the original (yes, even on a $5000 system -- the system used is probably the least important factor in hearing artifacts), let alone --xtreme or anything higher.

Yes, you might respond with a "But I use --braindead for peace of mind, etc, etc, space is not an issue, etc, etc", which is fine, but that still doesn't mean that it's necessary for "good sound", again. Personally I find the fact that people go for such a high setting in a sort of knee-jerk reaction kind of odd. I also find it odd when people recommend increasing the lowpass to levels which are commonly beyond the hearing abilities of most people, or using other tweaks which offer dubious benefits. If you consider the fact that these lossy codecs are already discarding 5x-10x the amount of data in the original .wav file, it seems kind of foolish to trust them in those regards but not in such a trivial and vehemently unverified issue (ie, "Oh my.. the encoder didn't encode all the way up to 22.5khz on this one section.. it only went up to 18khz.. there must be something wrong! I can hear a lack of bass and a loss of high frequency "feeling" now!"). That's not to say that all tweaks are useless, but most of the tweaks people seem to be applying as a "necessity" for "good sound" are based their misunderstandings of and flawed notions regarding audio compression, or even just plain stubbornness.

As a final note, I see that you referred to --braindead as "decent" and "acceptable", but not "transparent". Decent and acceptable certainly don't imply even "good" performance... so does this mean that you and your friends found --braindead to be flawed and not transparent (even if you considered it a worthy tradeoff)?

At any rate, I certainly don't think that it's reasonable to use --braindead without any sort of verified need, and certainly not the only reasonable choice even then.

Btw, I'm not trying to attack you or anything, but I am getting a little tired of seeing the notion that people must use presets like --braindead, etc, especially when there is no sign of evidence or testing of any kind to support this. If you want to tell people that you use these settings because you get the "warm fuzzies", then that's fine. Please don't tell people they need this though, especially when most evidence points to the contrary.

All the recent problem samples that have been reported, artifact over a broad quality range. The audibility of the artifact will become less if you increase quality of course, but if you listen closely it will still be there even with q7. That q7 cures all artifacts is a myth.

When you transcode from q5 to mp3-aps you will get a few mild artifacts here and there (So that mpc-q5 --> mp3-aps is ABXable from original-wav --> mp3-aps). q6 cures most of them. For use in my portable mpc q5 sourced transcodes are quite sufficient.

Btw, you are not limited to only standard or xtreme. You can use anything in between like --quality 5.3. Make some of your own tests and see where those transcoding artifacts disappear.

@torok: insane/q7 no longer encodes the full spectrum (which was pointless anyway).

@budgie: I suppose your friends are as audiophile as you are, and they are fully satisfied and enjoy brilliant sound. You however refer to mpc-q8 only as "decent" and "acceptable". I believe that you would enjoy your mpcs as much as the originals if you didn't know, what you were actually listening to.

Bottom line: if it artifacts at q5 (which is rare), it will aritfact at q6. You will not have less artifacts by increasing the bitrate. q6 offers better transcodability.

Well, maybe we use just a bit different words... It may be, because English isn't natural for me. Of course it's transparent enough and that's why it's acceptable



If I can remember well, it is written in the manual to MPC where it reads "...Braindead... for maximum headroom and reserve enough for transcoding" or something like this...



It may be, because English isn't natural for me. Of course it's transparent enough and that's why it's acceptable... And of course I can't praise it extraordinarily, for what good there would be wav files then? It sounds great, if this is it, what you want to hear... and I personally think it is great!



Can I then assume that -q8 will be even a bit better?

Believe me, it was painful and really long way for me from being "savage audiophile" to become just a quite "normal music consumer" with a bit more matter-of-fact view on things... Maybe is MPC -standard really good enough, but I simply cannot believe it from "one day to another"... I still need to know there is a (huge) excess of bits, just when (maybe) there's a need for it... B)

This is what VBR and an advanced and highly tuned psymodel is for. Second guessing that based on a high level, clumsy, inaccurate, and arbitrary measurement of some sort is foolish.

In places where MPC needs more bits, it uses them. If you want to see this in action, you can get utilities that will show you the bitrate that MPC uses down to a very small and precise section of audio. MPC is a very flexible format with a huge upper bitrate limit. Forcing the bitrate to be higher across the board is not going to make much of a difference in the critical areas of the file because MPC already scales up quite massively in those sections.

Quit making assumptions and start doing listening tests..

Where can I get these utilities and is there any deep-in source about MPC? I couldn't find anything...



It won't help... definitely not now, yet. I'm not quite cured from being an audiophile, y'know... And, btw, I guess there's plenty of others who are more suitable to make these listening tests... I want to taste/luxuriate/savour (don't ´know, what's more appropriate) MUSIC and not SAMPLES... some knicks and knacks here and there surely won't knock me down if I got no original near at hand...

Who talked about knick and knack or samples? Do the tests you want, but do them blind. That's what we call a listening test.

I don't think I need any blind tests, it's a time waste for me, especially when they were made already in the tuning process by more competent people (in this matter)... I just stick for the time being with MPC -braindead and AAC -archive... Howgh!
And btw, I think the whole thing should be addressed to somebody else... I didn't asked for any advice on transcoding... if I can remember

What is strange though is that you would have this view (that these people are competent and know what they are doing), and yet you wouldn't trust them at the same time.. thinking that presets like --standard or --xtreme are not enough (of course, I'd be perfectly happy if you were to prove this.. but just claiming that it is so is not enough). Seems kind of contradictory, doesn't it?



Of course, but one of rules of this community is that if you make a claim about something which is contraversial, that you must also back it up. If you have no plan on doing this, then please stop recommending things which you haven't tested properly yourself and/or which have been shown to be largely unnecessary or for which go against previously established and commonly held conceptions within the community.

Where can I find a utility that plots (or writes to a txt file) the bitrate of each mpc "frame" (if that's what they're called). This would be very interesting. Frank has already suggested to me that I can reduce the averaging of the display in Winamp to see the instantaneous bitrate changes, but a bitrate-vs-time graph would be better!

Cheers,
David.

"braindead" = "reasonable"? I sense contradiction..

seriously, is it so hard to believe that this preset has been called "braindead" by the codec dev / tuner for a reason?

I don't know about plotting a graph, but Case made a program awhile ago called mpcbits that allows you to view the bitrate on a frame by frame basis. I've never actually tried it though, and I don't think it is "publically" released. Not sure if he even wants me posting about it (though I don't see why not).

Perhaps he can shed some light on the matter.

I've already suggested long ago a Visualization plugin or option for Winamp, in addition to the SpecAn and O-scope, an autoscaling/resetable VBR histogram display. In this way, the upper & lower limits, as well as the density of sample rates could easily be observed.


Cheers..!!

I downloaded it from someplace...it's in the public domain AFAIK. I think it's available on Case's MPC page.

I've have seen situations where even standard MPCs swell up to 500+ kbps over very short periods of time (<0.1 seconds). mpcbits allows you to specify the time window for analysis. If you chose 1 second, you'll almost never see anything above 300kbps with --standard, but knock that down to 0.25 seconds and you'll see a surprising amount of "frames" that go over that bitrate threshold. I think in theory SV7 can go up to 1411.2kbps...which is the same bitrate used by an uncompressed WAV.

This has nothing to do with the trust or something like this... I just don't like the feeling it's such a low bitrate. And till there is a possibility to use highest rate available, I will always use these redundant bits option, be it MPC -braindead or AAC -archive or in some exceptions LAME -insane...



I don't think it's controversial, but I respect you and your opinion and your right to have an opinion as you want... I thought it was tested properly and thoroughly by others (maybe MPC developers)... Why does this possibility exist, then, when it's controversial and maybe unnecessary? When something is not needed or just waste of bits or space, why they let such possibilities in the encoder(s)?

Your view is controversial because the general community understands through personal experiences that --standard is indeed tuned for transparency. MPC's developers made standard to be more than just "good enough" or even "very close to transparent". Of course standard can fail to produce transparency but that's not so much from a lack of tuning or testing as it is from the fact that no perfect digital model for the human ear exists. --braindead isn't perfect nor is AAC --archive...for the same reason.

--braindead's existence is borne from possibility not from necessity. You are free to use it if it floats your boat but it is not acceptable to claim that its usage is somehow required to maintain a certain level of transparency.

It can be found here:

http://www.saunalahti.fi/cse/mpc/files/

Allright. So, 5 is of course transperent (to me and I'm sure everyone I know or will know), but 6 will give me better transcoding for my portable? That's what I wanted to know. -q 6 it is! Thanks.

PS: And anything above -q 6 was never an option.

Yeah, I know

The reason I didn't post the link is because I wasn't sure if it meant for public consumption, hence "publically"

While I applaud the efforts made by Dibrom et al to take this argument seriously, I simply cannot. From what I understand you base your choice of preset on how high a bitrate you get...?

If I have understood you correctly, to me this is like saying: "I drive a Formula One car to work. With a speed limit of 50 km/h, I know it's a big waste and all, compared to a fine car like, say, a BMW, but there might be a theoretical situation where the F1 will give me the edge I need. I have done plenty of research on the matter, and have concluded that the F1 car is the best choice because it looks to be a lot faster than a BMW. My experiences tell me that the physical appearance of the F1 indicates it is indeed faster and better than any other car out there. A BWM simply looks to be slower. My argument implies that the rest of you should also get a Formula One car to drive the 6 km to work every day."

And if you like high bitrates, why not go lossless? Or keep the .wav? Or even better, upsample and convert to 24 bit/96 kHz! Absurd? Indeed. Perhaps you prefer absurdity within "reasonable" limits. In this case, I suggest using --xtreme.

No offense, but let us use our ability to reason to the greatest extent possible, shall we? Either that, or you could exercise your right to form and have opinions, but not clicking the 'post' button and venting your opinions to the (perhaps learning) public.

Cheers,
Uosdwis

I chose -q 7 because it is what's recommended on MTRH's *. I don't have golden ears or anything, but I figured I'd trust his guide rather than my ears.

After reading this thread though, I think I'm going to switch to -q 5.


MOD:* no links or names of ripping groups please.

Hello all, I'm a newbie to this forum.

So what exactly is the difference between standard and extreme?

I'm only familiar with the basics of MP3 technology, and not with the intracacies of LAME and the alt-presets.

I know of the difference in bandwidth, but how does this actually factor in? Is the higher bandwidth due to a lower threshold of hearing, or...?

Gecko, is this your experience when encoding certain test samples, or just various samples from your music collection? I tested fatboy, castanets, and drone_short a while back and could either ABX both ("q5 -> mp3" vs. original) and ("orig -> mp3" vs. original), or I could ABX neither. But then again, my suite of test clips was only a pre-echo stress test. I don't recall whether I could hear a difference between the two mp3's, but in any case the difference must not have been very big.

Also, for samples where you found "q5 -> mp3" sounds different from "original -> mp3", were both mp3's non-transparent, and was the source mpc non-transparent?

In cases like this, wouldn't it be more efficient and quality-oriented to store the information as lossless for these few frames? Is this a planned feature or anything? Or would this be too difficult to implement?

Yeah, its unfortunate that a guide for newbies directs them to basically waste bits on mpc's, for no real good reason. I will say the rest of the * guide is well done, though


MOD:* no names or links to ripping groups.

When mppenc encodes it thinks it is being transparent 100% of the time (standard profile and higher). It doesn't know when it is producing artifacts. And you can't assume that artifacts are likely only on segments with high bitrates. It's not a solution to say: if a frame needs over [x] kbps, switch to lossless mode.

That's not what I was saying. What I meant is that when a frame of MPC encoded data is larger than the equivalent losslessly encoded frame, wouldn't the latter be preferable? Perhaps even if the MPC frame is only a little smaller? Or is this technically possible?

When the bitrate approaches or exceeds 700kbps, ... That's about the standard size for lossless. But it's probably not technically possible with the way MPC is designed.

AFAIK MPC includes (a form of) lossless compression, a Huffman algorithm. Up to SV 7.1 it is not designed to be efficient above about 240 kbs, if I understood correctly. In SV 8 this would be addressed.
This is BTW one of the reasons why using mpc with settings that give a lot of > 240 kbs frames is not optimal.

Have a nice day.
--
Ge Someone

But that's on average, not necessarily on a frame by frame basis.
--
Ge S.

I randomly chose some songs I liked and was familiar with from my CD collection. Only a few I picked seemed to me like they could be hard to encode. The majority was average stuff at least to the extent that I could tell. Some of the problem cases had to do with tonality problems. I'm not sure if pre-echo was ever a problem.

I am unsure about my concrete procedure. I didn't test the direct mp3 encodes for transparency, that I know. I think I tried to abx the original vs the transcoded mp3 and if I found a problem I tried to abx the original-mp3 vs transcoded-mp3 to see if the problem was a result of transcoding or due to the mp3 format in general. I assume that the mpc source was transparent since I have all the songs encoded on my PC and never found a problem, but I didn't investigate that.

Even when transcoding from standard, the differences were quite small. Not very hard to abx though, and except for one sample, definately not annyoing. If you didn't have the original I doubt you would notice. I came to the conclusion that I can be happy with mp3s transcoded from q5 mpcs when put on my protable.

I posted some results in a thread started by liekloo about transcoding. The thread is still there, but the data was lost during the forum software conversion (was encapsuled in [CODE ] brackets, I believe).

This goes for all posts addressed to me in this thread:

It is really not important, what general community understands and accepts... In former Czechoslovakia general community accepted Communist Party and all the shit involved and after it fell down almost nobody accepted it then If there's a bunch or majority of people who think THIS IS ALRIGHT it doesn't mean it's alright automatically... It doesn't mean a thing (at least as for it's concerning me) when somebody says this is good, THAT IT IS REALLY SO GOOOOD... I am enough experienced to judge, what is good for me and what cannot be accepted, i.e. must be ruled out... Everything in real life, what has a "standard" mark, is suspicious for me. I was never a standard man by no means... I have never accepted any "standards" set by anyone else...

I just expressed my opinion on the forum, that's all. When somebody or anybody else puts down anything, even nonsense, it's okay but when I just express what I think it's immediately a sort of flame war

I just wanted to say to all you "-standard" MPC people out there that I have hacked the MPC encoder. I have enabled the ability to encode at -q50... giving me ultimate sound quality, even better than lossless, yeah you heard me....

I am currently wrapped in a blanket listening to music.. I feel warm and fuzzy all over..

lol

Edit:
File size for a typical song is now over 100 megs. Keep in mind everyone that I'm trying to make the sound quality more better-er. I'm currently working on new ways to make the filesizes bigger, because we all know bigger is more better.. better that is than being more better than bigger-er at being more better-er.

If you can't hear it, you can't hear it!
(in style of Zork Grand Inquisitor Hades phone, it was originally:
)
Just go lossless (read away) then...

Just back up your opinion with some blind tests.
I recommend a program called ABX or similar one.
If you think it's a waste of time, then don't post on this forum, okay?

Wow... any download link? (w00t)

(Maybe you should update your signature )

Actually, the discussion here is about MPC rather than LAME.

Basically, you should be safe using --standard (MPC) or --alt-preset standard (LAME).

Anything more is an additional safety margin for peace of mind, or to improve performance a tiny bit on very difficult samples.

Yes, the difference of bandwidth is mostly due to more precise quantization of samples (ie: more average bits per sample), thus assuming a lower threshold of hearing, and a higher SMR and SNR.

Welcome to this forum.

Cheers

The creator of mpc wasn't satisfied with the de facto standard mp3. So he created something better. MPC --standard is not a lie. It was never the intention to fool people about it's level of transparency. No one denies that even mpc fails from time to time. The percentage of good ears and critical listeners is high among the users of mpc, because these people were dissatisfied with the performance of other lossy encoders and searched for a better alternative.

The whole thing is not about opinions, it's about facts. Placebo effect = fact. Hearable difference or not = fact.

The general public of the former Czechoslovakia didn't have much of a choice. All of us here are free to use whatever quality setting they want. Using --standard is not forced upon us, but many find it ideal. Some, including you, don't seem free enough to trust their ears and need an extra safety cussion to make them feel happy. That's fine, just don't assume others need the same amount of cussioning.

YEEEEAH... right!!! Everything's easy when you understand the essence...

A true rebel, right here on HA! Sadly, the only thing you appear to be rebelling against is common sense.

Regardless of your impled accusations (and insults) towards the members of this forum, and the creators of Musepack, I stand by my point: you have every right to have your own opinion on these matters, and every right not to let facts or expertise influence your decisions. This, however, does not automatically mean that you have to express your opinions in a forum created for informative and rational discussion. Keep encoding music the way that suits you best, but please refrain from posting if you have nothing to bring to the discussion but unsubstantiated claims and sophistic opinions.

-- Uosdwis

The essence of psychoaccoustic audio compression is surely not political debate. If you think what I said is wrong, please argue with me and help me overcome my ignorance instead of saying "YEAH RIGHT". I am sure you are very happy using quality 8 and your music sounds wonderfull. There's nothing wrong with that. However your decision isn't based on blind testing or other scientiffic methods but rather opinion and thus isn't suitable to be debated in a scientiffic environment where facts should rule instead of opinions.

In my remark "Some, including you, don't seem free enough..." I was adressing you specifically because I didn't want to include the sarcastic undertone by making a generalizing comment about "some ominous people (you know who), I don't want to say their names, but it should be obvious who I'm talking about, maybe they're even reading this..."

You can tell me what you think is wrong with what I said and we can continue from there. Otherwise we should not continue this debate.