HydrogenAudio

Lossy Audio Compression => MPC => Topic started by: mithrandir on 2002-07-19 03:43:35

Title: Sample: more transparent in 1.01j than 1.06
Post by: mithrandir on 2002-07-19 03:43:35
Here's a short clip where I noticed a loss of transparency using 1.06 --standard. The same clip encoded with 1.01j --standard offers improved sound. Pay attention to the cymbal hit at about 2.4 seconds. The original WAV sounds smooth here while 1.06 --standard adds a silibant edge to the attack. It seems 1.01j --standard isn't audibly transparent either but it's much harder to notice (good AB/X test).

Although this is just one case, I'm leary about using any of the post-1.01j encoders.

Original FLAC (http://mywebpages.comcast.net/briangabel/reason.flac)

1.01j MPC (http://mywebpages.comcast.net/briangabel/101j.mpc)

1.06 MPC (http://mywebpages.comcast.net/briangabel/106.mpc)

Yes, I am using the "bug fixed" version of 1.06.
Title: Sample: more transparent in 1.01j than 1.06
Post by: mithrandir on 2002-07-19 04:07:07
Because I know you'll be asking:

1.06 vs original - 16/20, 1% guessing chance
1.01j vs original - 22/30, 2% guessing chance
Title: Sample: more transparent in 1.01j than 1.06
Post by: Frank Klemm on 2002-07-19 09:40:13
Quote
Originally posted by mithrandir
Because I know you'll be asking:

1.06 vs original - 16/20, 1% guessing chance
1.01j vs original - 22/30, 2% guessing chance


1st)  statistically not significant (ask ff123)
2nd)  Try 1.06 --ms 0
3rd)  I can easily ABX 1.01j (8/10), but not 1.06 (12/20).
Title: Sample: more transparent in 1.01j than 1.06
Post by: Garf on 2002-07-19 10:12:33
Quote
Originally posted by Frank Klemm

1st)   statistically not significant (ask ff123)


Uh?

giancarlo@garf:~/r3mix > ./confid
Ok's / Trials ?
16 20
1000000 iters, 6010 success, 0.601000% luck probability
giancarlo@garf:~/r3mix > ./confid
Ok's / Trials ?
22 30
1000000 iters, 8079 success, 0.807900% luck probability

--
GCP
Title: Sample: more transparent in 1.01j than 1.06
Post by: Frank Klemm on 2002-07-19 11:41:50
Quote
Originally posted by Garf


Uh?

giancarlo@garf:~/r3mix > ./confid 
Ok's / Trials ?
16 20
1000000 iters, 6010 success, 0.601000% luck probability
giancarlo@garf:~/r3mix > ./confid 
Ok's / Trials ?
22 30
1000000 iters, 8079 success, 0.807900% luck probability

-- 
GCP


16/20 vs. 22/30        => 0.6% vs. 0.8%      (1.01j is better than 1.06)
15/20 vs. 23/30        => ?% vs. ?%            (1.01j is ?        than 1.06)

BTW I have serious trouble with my sound card with this signal.
It sound completely different with -6.07 dB PCM attenuation (*0.497) and
setting volume control from 82/100 to 91/100 (Linux OSS). °)

K-20 replaygain reduces level by NEARLY 19 dB.
This is new world record!


°) I have calibrated my sound card using a loop back connection.
Title: Sample: more transparent in 1.01j than 1.06
Post by: GeSomeone on 2002-07-19 12:48:04
Quote
Originally posted by Frank Klemm
2nd)  Try 1.06 --ms 0

Frank,
in the (recent) past you have always condemned the use of --ms 0 . Now you suggest trying it (also in the thread "stereo separation").
Do you mean that the quality problems that were introduced by --ms 0 in previous versions are now solved?
--
Ge Someone
Title: Sample: more transparent in 1.01j than 1.06
Post by: Frank Klemm on 2002-07-19 18:31:12
Quote
Originally posted by GeSomeone

Frank,
in the (recent) past you have always condemned the use of --ms 0 . Now you suggest trying it (also in the thread "stereo separation").
Do you mean that the quality problems that were introduced by --ms 0 in previous versions are now solved?
--
Ge Someone


--ms 0 is currently unmaintained and with the current encoding theme difficult
to handle. With --ms 0 you can perfectly encode 2 directions (-45°,+45°),
with --ms 15 4 directions (+/-90°,-45°,0°,45°). An omnidirectional encoding
would be better than both. A way would be to extend --ms 0.
Title: Sample: more transparent in 1.01j than 1.06
Post by: user on 2002-07-20 12:07:33
"With --ms 0 you can perfectly encode 2 directions (-45°,+45°),
with --ms 15 4 directions (+/-90°,-45°,0°,45°). An omnidirectional encoding
would be better than both. A way would be to extend --ms 0."


Does this influence encoding of surround signals and their resolution ?

I would prefer, that every encoder (mp3, mp2, mpc, ogg) should be able to reproduce surround signals very well.
And not only "old" Dolby ProLogic Surround (one mono rear channel),
I like much more: DS2, separated rears, full frequency.
Title: Sample: more transparent in 1.01j than 1.06
Post by: David Nordin on 2002-07-20 12:21:00
Quote
Originally posted by user
"With --ms 0 you can perfectly encode 2 directions (-45°,+45°), 
with --ms 15 4 directions (+/-90°,-45°,0°,45°). An omnidirectional encoding
would be better than both. A way would be to extend --ms 0."


Does this influence encoding of surround signals and their resolution ?

I would prefer, that every encoder (mp3, mp2, mpc, ogg) should be able to reproduce surround signals very well.
And not only "old" Dolby ProLogic Surround (one mono rear channel), 
I like much more: DS2, separated rears, full frequency.


D. PL is just sheisse - just a shortcircuited mess of phases.
I say ambisonics!
the only way to go, especially for such a great format as MPC.
Title: Sample: more transparent in 1.01j than 1.06
Post by: user on 2002-07-20 12:32:26
"1. D. PL is just sheisse - just a shortcircuited mess of phases.
2. I say ambisonics!"


Hi,

1. I agree, but don't want to make angry those persons still using DPL, not having/using Logic7/DPL2/DS2

Do you have listening  experiences with DPL2 ?

2. What is it ?
I have found www.ambiosonic.net (http://www.ambiosonic.net)
But do you think of any usage of it in real life in future ?
DD5.1 and dts, DPL2 are common, not ambiosonic.
Title: Sample: more transparent in 1.01j than 1.06
Post by: Frank Klemm on 2002-07-20 12:46:15
Quote
Originally posted by user
"With --ms 0 you can perfectly encode 2 directions (-45°,+45°), 
with --ms 15 4 directions (+/-90°,-45°,0°,45°). An omnidirectional encoding
would be better than both. A way would be to extend --ms 0."


Does this influence encoding of surround signals and their resolution ?

I would prefer, that every encoder (mp3, mp2, mpc, ogg) should be able to reproduce surround signals very well.
And not only "old" Dolby ProLogic Surround (one mono rear channel), 
I like much more: DS2, separated rears, full frequency.


Dolby ProLogic I and II is less critical than Stereo.
There's a VCA in it. Speakers with reduced signal are additional
damped. This starts at -3 dB. Stereo issues are at least 20 dB
below this point. I measured 50 dB attenuation for standard and
60 dB attenuation for xtreme before this become relevant.
Title: Sample: more transparent in 1.01j than 1.06
Post by: David Nordin on 2002-07-20 12:48:56
Quote
Originally posted by user
"1. D. PL is just sheisse - just a shortcircuited mess of phases.
2. I say ambisonics!"


Hi,

1. I agree, but don't want to make angry those persons still using DPL, not having/using Logic7/DPL2/DS2

Do you have listening  experiences with DPL2 ?

2. What is it ?



IMO we should all abandon AC3/DS - well Dolby period.
they actually make lousy formats, whereas the smaller better formats, such as ambisonics are hidden in the marketing of Dolby's solutions.

No, I have little experience of DPL2, but sincerely doubt there's much of interest.

I'm planning of building a Ambisonics setup at home, so I guess there will be little time for anything else

cheers,
David