Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Recent Mpc Convert - Me! (Read 5487 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Recent Mpc Convert - Me!

First off, I would like to thank whoever has been developing and tweaking MPC.  My kudos to you , whoever you all are.  I am glad to see that someone made a codec that has transparency in mind.  You would think that it should have been since day one?

I've gone down the road of FhG, then LAME, then Ogg, tested some others and Monkeys and Flac, and now MPC.

I've read quite a few threads here before doing my own tests, and I found I am quite happy with the results.  I know the developers and users here enjoy a good review, especially if someone is willing to describe why they feel the way they do, so here's a link to one I did for 3DSoundsurge's forum.  (It's my "home" and I like posting there first.)

I welcome any comments on the review, either here or there.

Chastity's MPC review
In Case Of Bose, Break Glass
Flac yuo in teh ASIO!

Recent Mpc Convert - Me!

Reply #1 –
I don't see anywhere the version of the MPC encoder you used.

IMO --standard is in that kind of quality level, that if doing a "review" you can't really say that DBT is not needed there. Even if it's not needed by you, it's certainly needed by your "audience", at least here, in order to have any credibility. 
Juha Laaksonheimo

Recent Mpc Convert - Me!

Reply #2 –
Oops.  Details..  The encoder I used was 1.1b.  But since you seem to enjoy statistics, I'll be more than happy to supply them.  In fact, I'll rerun them so that you may enjoy the actual log.
In Case Of Bose, Break Glass
Flac yuo in teh ASIO!

Recent Mpc Convert - Me!

Reply #3 –
This has nothing to do with 'ejoying statistics'. Double blind testing is _required_ if your claims are to be taken seriously.

Recent Mpc Convert - Me!

Reply #4 –
Hi, ath, nice to see you here. 

What ath says is true, it's more a question of removing listener bias playing tricks on his perception, also making sure that he hears what he claims, and is not just random guessing.

Recent Mpc Convert - Me!

Reply #5 –
WinABX test report
09/28/2002 01:09:04

01:11:03    1/1  p=50.0%
01:11:11    2/2  p=25.0%
01:11:20    3/3  p=12.5%
01:11:35    4/4  p= 6.2%
01:12:26    5/5  p= 3.1%
01:12:45    6/6  p= 1.6%
01:13:02    7/7  p= 0.8%
01:13:28    8/8  p= 0.4%
01:13:51    9/9  p= 0.2%
01:14:01  10/10  p< 0.1%
01:15:09  11/11  p< 0.1%
01:15:31  12/12  p< 0.1%
01:15:52  13/13  p< 0.1%
01:16:42  13/14  p< 0.1%
01:17:31  test finished

Song:  "Paint It Black" Rolling Stones, Live  Flashpoint
Encoded:  MPC v1.1b, at -quality 5, decoded using mppdec v1.1

I would have to say it wasn't easy, and I have more respect for level 5. 
In Case Of Bose, Break Glass
Flac yuo in teh ASIO!

Recent Mpc Convert - Me!

Reply #6 –
Quote
Encoded:  MPC v1.1b, at -quality 5, decoded using mppdec v1.1


Where did you manage to find v1.1b? You probably meant v1.11b. Just curious, Chastity... 

Recent Mpc Convert - Me!

Reply #7 –
1.11b 

The alpha listed on Case's site.
In Case Of Bose, Break Glass
Flac yuo in teh ASIO!

Recent Mpc Convert - Me!

Reply #8 –
Chastity, could you describe what kind of differences could you appreciate in the MPC file at your test, and in which part of the song you could hear them?

Recent Mpc Convert - Me!

Reply #9 –
I was paying attention to the lead guitarist.  The recording was very detailed on his play, with the mastering done quite well (1990 DDD recording/mastering), and one can listen to the individual strings being strummed.  In the intro, this is particularly noticable since he is playing solo, and at 0:19 to 0:24 (0:21 to be precise), this riff has a particular point that 2 string get plucked just offset of each other.  In the -quality 5 setting, the 2 strings blend together.  At -quality 6 the seperation was true to the wav, being distinct.  I did not try adding -ms 12 to the -quality 5 string since I was testing -quality 5 as is.

Is it being picky?  Yes, I think so.  I would like to add that it took awhile to find a piece of the composition where I could readily determine the difference, but I believe the piece represents a "typical" recording, as opposed to castanets.  :x 

The files at -quality 6 are still smaller than my -alt-preset extreme mp3s, saving about 8-12M of space per CD.
In Case Of Bose, Break Glass
Flac yuo in teh ASIO!

Recent Mpc Convert - Me!

Reply #10 –
Quote
Is it being picky?

Never mind, most times you have to be quite picky in order to find artifacts with good quality encoding. But the thing is that you actually heard a difference.

Thanks for your info.

Recent Mpc Convert - Me!

Reply #11 –
You're very welcome!
In Case Of Bose, Break Glass
Flac yuo in teh ASIO!

Recent Mpc Convert - Me!

Reply #12 –
I've uploaded the clip here. It's from 15 sec to 29.5 sec into the song (the entire solo). The file is self-extracting WavPack, so you don't need a decoder.
 

Recent Mpc Convert - Me!

Reply #13 –
Oops, I forgot about non-Windows users. Here's the FLAC file too.

Recent Mpc Convert - Me!

Reply #14 –
in reference to your posting on the 3dsoundsurge forum, why are coolplayer users "fuxored"? you mention that a winamp plugin exists then apologize to the coolplayer users, which doesn't make any sense to me. sorry, i'm more than a little thick-brained.

Recent Mpc Convert - Me!

Reply #15 –
Coolplayer supports wav, ogg, and mp3 (MAD) natively.  Unfortunately it used to have support for WinAmp plugins, but that is broken, so unless the developers code it in, it won't be able to recognize mpc files, thus fuxored. ("Fuxored" is my way of saying f*cked.)  Coolplayer's development rate has been, well, slow.

I mentioned it because Coolplayer is a favorite amongst the users of 3DSS, especially  mp3 users, since it's MAD + Wave Out was much better to Winamp's, until very recently.  Now they are on par.  And thanks to a kmixer.dll update in SP1, XP users can now enjoy improved SSRC resampling on AC97 cards.  (Live, Audigy, TBSC, GTXP, etc.  All the consumer cards.)  Hopefully, other Windows will get the update, probably in a WMP upgrade.  I haven't done any 44.1/48K comparisons since the update, except a rudimentary one, so I have no opinion on the matter yet.
In Case Of Bose, Break Glass
Flac yuo in teh ASIO!

Recent Mpc Convert - Me!

Reply #16 –
After having been told exactly where to listen and exactly what to listen for,  I abx'd the --quality 5 (--standard) file.  Thanks, Speek, for providing the wav.

-------------------------------------
WinABX test report
09/28/2002 22:25:05

A file: C:\My Documents\test\paintitblack.wav
B file: C:\My Documents\test\paintitblack.std.wav

22:25:36    1/1  p=50.0%
22:25:51    2/2  p=25.0%
22:26:05    3/3  p=12.5%
22:26:27    3/4  p=31.2%
22:26:35    4/5  p=18.8%
22:27:06    4/6  p=34.4%
22:27:14    5/7  p=22.7%
22:27:23    6/8  p=14.5%
22:27:30    7/9  p= 9.0%
22:27:39  8/10  p= 5.5%
22:28:03  8/11  p=11.3%
22:28:22  9/12  p= 7.3%
22:28:36  10/13  p= 4.6%
22:28:52  11/14  p= 2.9%
22:29:36  12/15  p= 1.8%
22:29:47  test finished

Chastity, you gots sharp ears.  But, I gotta tell ya, I would never ever have found this distinction on my own.  Hearing the detail is a blessing and a curse when it comes to lossy.  This sure won't make me decide to go --xtreme as a matter of course, but I guess if I acquire this tasty track, I'll be compelled to extreme it.  And maybe I'll have to start listening closer 
Yeah, when you call my name
I salivate like a Pavlov dog...

Recent Mpc Convert - Me!

Reply #17 –
Damn, that sample is a bitch. I'm not switching to extreme either. :-)

(mppenc 1.1)

ABX Results:
Original vs D:\Recordings\listening\paintitblack_mpc.wav
    13 out of 16, pval = 0.011


(mppenc 1.1b)

ABX Results:
Original vs D:\Recordings\listening\paintitblack_mpc11b.wav
    11 out of 16, pval = 0.105

  1 of  1, p = 0.500
  2 of  2, p = 0.250
  3 of  3, p = 0.125
  3 of  4, p = 0.313
  3 of  5, p = 0.500
  4 of  6, p = 0.344
  5 of  7, p = 0.227
  6 of  8, p = 0.145
  7 of  9, p = 0.090
  8 of  10, p = 0.055
  8 of  11, p = 0.113
  8 of  12, p = 0.194
  9 of  13, p = 0.133
10 of  14, p = 0.090
10 of  15, p = 0.151
11 of  16, p = 0.105

I used ff123's ABC/HR, I hope that's ok.

Recent Mpc Convert - Me!

Reply #18 –
This thing was so damn hard, I nearly gave up. Used KikeG's winabx ABA and mpc 1.11b standard. Result: 32/57. Winabx says that the propability I was guessing is < 0.1%. Heh. I was under the impression I was guessing most of the time. During the test I focused on different positions; in the end mostly on 1.5 - 2.5 secs. I absolutely can not describe the difference since there are so many things I tried to get a hold on, and no individual point made me go: "ah, that's it!" So I basically listened to the three aba samples and went by feeling which one did not sound like the other two. A thing that helped me stay focus: since the sound is concentrated on one channel, I turned around my headphones from time to time. After doing that, I could identify the correct sample during the next few rounds better. Guess this has to do with fatigue and memory effect of the ear.

Now does that mean there is a difference? You tell me, I would be very interested what others think about a result like this. I trust the guessing propability shown by winabx is correct. I do believe however that other listening factors account for much more of a difference than this little oddity. I am one step closer to the insanity afer this sample.

Recent Mpc Convert - Me!

Reply #19 –
Quote
Used KikeG's winabx ABA and mpc 1.11b standard. Result: 32/57. Winabx says that the propability I was guessing is < 0.1%

For an ABA test, the probability to score at least 32 of 57 trials by guessing is 0.0003342..., so it's extremely unlikely. This value is not taking into account that you might have aborted the test at an earlier stage, but even then the probability should stay well in the 1% range. (I'll calculate a more accurate value later).

Recent Mpc Convert - Me!

Reply #20 –
Quote
Result: 32/57. Winabx says that the propability I was guessing is < 0.1%. Heh. I was under the impression I was guessing most of the time

Well, if you had been really guessing you would have got something like 19/57 (1/3 times correct answer).

Quote
I trust the guessing propability shown by winabx is correct.


Thank you  The probability is calculated using same method that "traditional" ABX comparators use, so I believe it is quite trustable.