I was encoding one of my CDs today with mppenc 1.14 --standard .
I began to listen to the encoded track and woah i picked my first mpc artifact .
I reencoded the track with --extreme still the same result .
I cant explain what i hear , but it feels like the upper frequencies are garbled .
I thought i might make an important contributions to the hard to encode sample
so here it is .

Sample


[B] edit :- An ape sample has been uploaded .

i thought the error might be coming from 1.14 "beta" only so tried out
old reliable mppenc 1.01j but alas still same result .
also note that i was able to abx (easily) till --quality 7 on mppenc 1.14
and the encoder says that this is insane .
i get perfect files with --braindead .
also note that i was also able to hear the same effect with other tracks of the
album , but for them i only tried till --standard .

Interesting. But what's needed is a short sample of the original file, either as plain WAV or losslessly compressed (MA, FLAC..). Providing an already encoded sample doesn't help at all, i fear... sorry.

I can't download. Please re-up.

Here's a trick for downloading non-HTML files from Geocities (using MSIE, and it may work with Netscape as well):

(1) Click on the original link to the file. You'll get a page that says "Page Not Available" or something like that.

(2) Click on the "E" icon to the left of the URL in the address bar (where you type in URL's and where they're displayed), drag it to the right a little and drop it "back" into the address bar.

Bingo!

fewtch

Funny thing. My dl manager failed but this trick worked wonders. thx fewtch

What song and album is your sample from? Perhaps someone owns the cd and could provide the sample in its original form.

Sure thing... Geocities only allows links to HTML files from other sites (checks for 'referrer' tags, I think) so what this does is make it appear like you typed in the URL directly instead of following a link. Not sure why your d/l manager didn't work (I've never used one), but Geocities might be blocking those too.

Stan: Original file is really needed. Please try to provide one.
Also, does this happen with previous mppenc versions?

I have upped an ape sample ( took great pains ) .
Hope this might help .
Ape Sample

I can download the file when i am logged in geocities otherwise i cant but
you can always get it using fewtch's trick.

@fewtch
Man your trick does work .

@atherean
The album was Backstreet Boys' Backstreet's Back .
And the track was " All i have to give (Conversation Mix) "
Song position 0 seconds to 6 seconds .
Its a Pop Album .


edit -

OT: Why cant i upload my avtar
And if anyone could suggest a good place where i can upload my files say images and
samples like this

Could this be one of those "protected albums" or some type of flaw in the cd itself. It was mentioned that it occured throughout the disc. Sounds odd to me. I don't have that album or I would rip another copy for you. I wonder if that song is on the net - to use as a comparison at the particular time frame

@xbenchman

The album is from 1997 so no way it can be protected .
The problem does not occur throughout the CD .
But on quite a few tracks .




I have already uploaded an ape sample . Look my previous post .

I have tested with mppenc 1.01j and mppenc 1.14 .
I am going to soon test with a few other versions and post here .

Also If people find it difficult to download via fewtch's trick you can visit
this page :
Ape Sample

And download via any download manager , i think it should work now .

I tested with MPC 1.14, --quality 5 and --quality 7. The first tone (the high part of the "ooooh" chord) seems to end earlier at both quality settings, but it was much more evident in the -q5 encode. It seems indeed to be a problem with the high-frequency content. I can report this to Frank Klemm, but it would be better if another person could confirm this.

Wouldn't be able to abx without CiTay's tip, and even then the problem wasn't that obvious to me. this is --standard.

Note that --quality 7 is --insane

Is it possible internal clipping could cause this?.

I prefer: Right click->'Save As'

I think clipping means the things you get when a sample's gain is badly amplified .
If this is the case i dont think the problem here is with the sample clipping .

What does internal clipping mean ?
i seen it mentioned quite a few times before and it seems like a restriction of SV7 .

So if you are trying to say that SV7's restriction caused this artifact .
I must mention that i have tried lame --aps as well and the problem is badly evident there as well .


OT : The new forum looks good . Thanks Dibrom .

No. BTW, i already informed Frank Klemm about this sample, currently waiting for reply.

I'm not an expert on this, but from what I've gathered some modern albums are mastered to be as loud as possible (to try and sound the loudest on the radio, I suppose. I've heard pop albums are the worst offenders) and this generally results in clipping in lossy encoders. This might account for any distortion you hear in the high frequencies, but not the first "oooh" ending early.

Have you tried running the mpc file through replaygain and checking the difference? When I did, it tagged the file with a -3.28db value.

Yes; in fact, they are often mastered louder than the dynamic width of the media in which they are stored! Sounds foolish, I know... but this is done by deliberately clipping the peaks of the waveform in order to exceed the maximum decibel level of a standard Red Book audio CD. Some of the worst offenders have been mastered at Sterling Sound in New York (although not all of Sterling Sound's efforts exhibit this flaw...); for a quick look at an obvious example of such clipping, check out any of the Jimi Hendrix reissues by Experience Hendrix. If you open a WAV file from any of the Experience Hendrix/Dolly Dagger CDs in a visual wave editor, you will easily be able to see the clipping.

To John Edwards (John33): This prompts a speculative project which might be right up your alley! Would it be possible for WaveGain to analyze the clipped portions of a waveform and attempt to reconstruct the natural curvature of the peak? Mathematically we know this can be done, and visually it's easy to imagine the appearance of a corrected waveform, but so far there is no audio software capable of interpolating a peak from the slope of the sides. Would you know how to approach this problem?

- M.

Cool Edit Pro does this with its clip restoration transformation.

Hmm... didn't know this. Still, CEP is not freeware; WaveGain is. If it can be done, I'd still like to see it... but I may be the only one. Who knows?

- M.

I just opened Beck's "Get real paid" into cooledit, after finding that the replaygain values were set at -8db's for the entire album. Wow, does it look awful!

Will using clip resoration make it sound better, or just distort an already distorted wav?

BTW Cdnow lists it as an HDCD, is there any reason why this would cause clipping, or is it simply the mastering?

Getting OT: Wow - it's scary how many CDs I have that are mastered by them (I'd guess at least 10). I remeber seeing "Sterling Sound" mentioned in many of the album liners. Almost all of them clip when encoded into MPC .

Yes, many of the latest rock & pop albyms are clipped. it likes "the new standard" for the label industries. The last Christina Aguilera's album sounds awful, have hard clips, the intention is make to sounds louder in a home system, not in the radio, I'm work in a radio station and the audio processors do the same works, then the music sound with the same loudness, applying agc (automatic gain control),compression, expansion and 4 band limiters, the final mix is similar to these albums. It's so ugly listen to the music this way and provoques several audition fatique.
Sorry for my poor english. I'm speak english just a little.-