IPB

Welcome Guest ( Log In | Register )

3 Pages V  < 1 2 3 >  
Reply to this topicStart new topic
MPC Quality: -standard or -xtreme?
mithrandir
post Feb 26 2002, 03:31
Post #26





Group: Members
Posts: 669
Joined: 15-January 02
From: SE Pennsylvania
Member No.: 1032



I stopped using --standard because I've slowly discovered there's something "problematic" about that preset's output. On an ABX test, it is hard to reliably tell the difference between a --standard file and the original WAV. However, after extended listening to MPC --standard encodes of albums that I am intimately familiar with, I notice that there seems to be a difference. Obviously, it's subtle if it doesn't show up in an ABX test (or it doesn't exist at all as the pure objectivist would say). But there just seems to be a lack of resolution, a loss of "air", immediacy and liquidness that you can associate with the original. So it's not an artifact, per se, but just a difference in emotional presentation.

I CAN, of course, find samples where --standard sounds different (worse) than --xtreme --nmt 16 --tmn 32 on an ABX test, so I'm not making this stuff up. And, of course, this should be expected because the ATH model isn't perfect. I just don't think --standard is good enough for truly serious listening with good equipment, i.e. assuming you don't care about consuming more bits as required with a higher-quality preset. For less strenuous listening, --standard is fine and is light on the diskspace. I just don't consider it a panacea or the end-all. It's probably "good enough", but frankly, the high-bitrate --extreme encodes just seem to maintain the emotional content better on some material.

YMMV.
Go to the top of the page
+Quote Post
Gecko
post Feb 26 2002, 14:15
Post #27





Group: Members
Posts: 934
Joined: 15-December 01
From: Germany
Member No.: 662



Mithrandir, I very much understand your concerns about "emotional representation". I feel the same about mp3s (no matter what preset), but can't proove it with abx. The question is: does the encode sound better to us, because there is a percievable difference, or simply because of our knowldege that the encode has got to be of better quality. I do believe that with rising bitrate mpc quality scales along, but how much does it take to eliminate an artifact? After all, your command line is more than "xtreme" and we were comparing the standard vs. the regular xtreme preset. To me it seems this calls for tuning of the psymodel (which is a very difficult, I know) rather than raising the bitrate.

I am very confused. I've been using standard and felt fine, but there has to be a reason why so many people prefer xtreme (and some with settings that result in ~300kbps files). Does this help their ears or does it help their brain?
Go to the top of the page
+Quote Post
mithrandir
post Feb 26 2002, 23:24
Post #28





Group: Members
Posts: 669
Joined: 15-January 02
From: SE Pennsylvania
Member No.: 1032



QUOTE
Does the encode sound better to us, because there is a percievable difference, or simply because of our knowldege that the encode has got to be of better quality.

The trouble is that the two files ARE measurably different. The xtreme file has more data and less quantization error than the standard file. This is something we can verify and prove. We just don't know if the differences are audible by human hearing (and whose ears, since we all hear a little differently). I'm rather convinced that they are appreciable differences if the conditions are right (playback equipment, listening environment, musical material). A ringing sound, like a bell or a strongly-picked guitar string, is an example where the differences between a standard and a tweaked xtreme file can be clear and demonstrable on ABX tests. Not that this is surprising, considering the bitrate difference.
QUOTE
I do believe that with rising bitrate mpc quality scales along, but how much does it take to eliminate an artifact?

Using a high-bitrate setting is not necessarily about eliminating artifacts so much as it is about increasing resolution...revealing that "bloom" around an instrument floating on the soundstage. The standard preset does a commendable job masking audible artifacts, but it doesn't mean that standard is revealing all that is potentially audible. That's just a trade-off of producing a smaller file. It can be a very good trade-off: a 175kbps file with no apparent artifacts COULD be considered "transparent" to a wide array of listeners.
QUOTE
To me it seems this calls for tuning of the psymodel (which is a very difficult, I know) rather than raising the bitrate.

Yes, I believe this is true but from an end-user perspective, we can't change the psymodel...only manipulate the provided settings. Perhaps when SV8 comes out, we will find a few more psymodel enhancements that enhance quality at equal bitrates.
QUOTE
Does this help their ears or does it help their brain?

Undetermined. I'm fully aware of the pitfalls of subjectivism, but I am not entirely convinced that ABX tests should be considered end-alls. I'm not so sure that long-term and short-term listening are exchangable. If you can't tell a difference in a well-chosen 15 second clip, does that mean after you finish listening to a 74 minute album, A and B are going to leave you equally emotionally stirred? Maybe, maybe not.

When big bucks are on the line (a person considering a $10,000 CD player, for instance), then this is a discussion worth fleshing out. However, if going with xtreme settings is only about helping out our brains, the only cost to me is potentially using too much diskspace unnecessarily. That's a cost I'm willing to pay considering I do feel the xtreme settings are worth it and I have experienced an increase in listening enjoyment with them.
Go to the top of the page
+Quote Post
Leto Atreides II
post Feb 27 2002, 10:06
Post #29





Group: Members
Posts: 163
Joined: 13-January 02
From: Eugene, OR
Member No.: 1009



QUOTE
Originally posted by mithrandir

When big bucks are on the line (a person considering a ,000 CD player, for instance), then this is a discussion worth fleshing out. However, if going with xtreme settings is only about helping out our brains, the only cost to me is potentially using too much diskspace unnecessarily. That's a cost I'm willing to pay considering I do feel the xtreme settings are worth it and I have experienced an increase in listening enjoyment with them.

If someone's spending 10 grand on a cd player, then I think they could spend a few hundred bucks more and get bigger hard drives to hold some losslessly compressed audio.
Go to the top of the page
+Quote Post
Dibrom
post Feb 27 2002, 10:36
Post #30


Founder


Group: Admin
Posts: 2958
Joined: 26-August 02
From: Nottingham, UK
Member No.: 1



QUOTE
Originally posted by Garf
My opinion:

in the very few clips where I ever heard MPC fail, they failed as much with -xtreme as they did with -standard.

If the psymodel is failing, most likely it won't allocate enough enough bits with xtreme either.

So, I doubt there is much sense in using -xtreme. It's very arguable it will actually give higher quality, and it's quite a bit larger.


I agree that for the most part, I believe -standard is probably enough. I haven't so far seen a clear case where someone could reliably tell the difference between --standard and --xtreme either.

However, I do happen to use --xtreme --nmt 10. The reason for this is that originally, way back when I was first testing MPC, before I even created the --dm-presets, I actually did find a few difficult autechre (surprise? smile.gif) clips where there was some barely perceptible pre-echo with --standard and --xtreme (I don't remember the clips at this point). I bumped the nmt up to 10 (from 9) when using xtreme and that brought the artifact to a point where I could no longer reliably abx it. Ever since then I've been pretty comfortable with that.

The --nmt 16 --tmn 32 which everyone seems to be using now came from a time when I was discussing another difficult clip, 2nd_vent, which actually caused an audible problem that mostly seemed to go away at that point. Certainly its not necessary for 99% of the other material out there. For that matter, I haven't tested this clip with any of the recent encoder builds, so it may no longer exist even.

Moral of the Story:

Settings significantly higher than --standard, or at least --xtreme, are almost definitely overkill. It's fine to use them "just because", but for someone questioning whether they should be using something much higher than --standard or --xtreme, realize that it's almost definitely not necessary for damn near anything you'd encode.
Go to the top of the page
+Quote Post
Gecko
post Feb 27 2002, 16:06
Post #31





Group: Members
Posts: 934
Joined: 15-December 01
From: Germany
Member No.: 662



I had a listen to the 2nd_vent clip compressed with standard and heard no problem. If I can't even pick it up on a sample where others need such extreme settings to get it to sound decent, then I should be fine with standard, yes?
Go to the top of the page
+Quote Post
Garf
post Feb 27 2002, 16:53
Post #32


Server Admin


Group: Admin
Posts: 4853
Joined: 24-September 01
Member No.: 13



QUOTE
Originally posted by mithrandir

A ringing sound, like a bell or a strongly-picked guitar string, is an example where the differences between a standard and a tweaked xtreme file can be clear and demonstrable on ABX tests.


Clips please...

--
GCP
Go to the top of the page
+Quote Post
superorc
post Feb 27 2002, 19:00
Post #33





Group: Members
Posts: 62
Joined: 10-December 01
Member No.: 621



QUOTE
Originally posted by sony666
>>and I like how much the bitrate display bounces around in WinAMP. I don't know about you all, but watching that bitrate display is hypnotic...I'll get all excited if it breaks 400kbps on a hard music segment. "Ooooh, look. A 400kbps frame. MP3 can't do that!<<

Hmm..
I think some people are so obsessed with this bitrate stuff and geeky encoder switches, they forget to ENJOY their music.

I enjoy my old 128k Bladeenc mp3s because I listen to the music, not to the encoder quality.

I feel sorry for your wasted time looking at that display, mithrandir sad.gif

In 20 years, when I will be around 45, I will give a **** whether my old Metallica album is 160k or 320k. I will probably listen to it once and put it back again.


lol try making a 128kbps with fatboy.wav. in fact i have the whole song just to compare audio codecs. you need either lame -alt-preset insane or FhG fastend from musicmatch @ 320 kbps to get descent quality. AAC, Ogg Vorbis, and MPC at their high quality settings can at least make it transparent
Go to the top of the page
+Quote Post
mithrandir
post Feb 27 2002, 22:24
Post #34





Group: Members
Posts: 669
Joined: 15-January 02
From: SE Pennsylvania
Member No.: 1032



Deprecated.
Go to the top of the page
+Quote Post
mithrandir
post Feb 28 2002, 02:43
Post #35





Group: Members
Posts: 669
Joined: 15-January 02
From: SE Pennsylvania
Member No.: 1032



Here are two samples for standard vs tweaked xtreme comparisons. Note that you aren't looking for artifacts per se, but a difference in resolution, air, palpability, immediacy and other soundstaging/imaging characteristics.

Guitar sample (FLAC)
Guitar sample (MPC 0.90s --standard)
Guitar sample (MPC 0.90s --xtreme --nmt 16 --tmn 32)

Trumpet sample (FLAC)
Trumpet sample (MPC 0.90s --standard)
Trumpet sample (MPC 0.90s --xtreme --nmt 16 --tmn 32)

I'm starting to think that part of the reason why the standard files sound different than the tweaked xtreme files is because the standard files have a more aggressive lowpass. The standard files chop off some of the harmonics in the upper treble bands and this has an audible effect (to a certain degree).
Go to the top of the page
+Quote Post
Dibrom
post Feb 28 2002, 09:44
Post #36


Founder


Group: Admin
Posts: 2958
Joined: 26-August 02
From: Nottingham, UK
Member No.: 1



QUOTE
Originally posted by mithrandir
I'm starting to think that part of the reason why the standard files sound different than the tweaked xtreme files is because the standard files have a more aggressive lowpass.


This difference (which may be a culprit if there is an audible deviation) is based on the more sensitive ath of --xtreme. You can easily modify the --standard preset to use this ath by just using "--standard --ltq fil", which usually only adds a few kbps.
Go to the top of the page
+Quote Post
Garf
post Feb 28 2002, 11:32
Post #37


Server Admin


Group: Admin
Posts: 4853
Joined: 24-September 01
Member No.: 13



QUOTE
Originally posted by mithrandir
Here are two samples for standard vs tweaked xtreme comparisons. Note that you aren't looking for artifacts per se, but a difference in resolution, air, palpability, immediacy and other soundstaging/imaging characteristics.


Thanks for the clips. I'll have a listen and report my findings.

Edit: Oops, due to fatigue I see now that I misread the discussion. I missed the 'tweaked', i.e. I thought you were talking about -standard vs. -xtreme.

--
GCP
Go to the top of the page
+Quote Post
mithrandir
post Feb 28 2002, 19:30
Post #38





Group: Members
Posts: 669
Joined: 15-January 02
From: SE Pennsylvania
Member No.: 1032



What is the difference between "--standard --ltq fil" and "--xtreme"? If xtreme uses this more sensitive ath by default, then I suppose xtreme also uses more sensitive nmt and tmn values?

Perhaps nmt 16 and tmn 32 are a bit overkill. From a sonic perspective, there's nothing wrong with them (unlike -k -ms --lowpass 22.05 in LAME). The challenge is finding the right values that don't waste bits, yet still retain all of those audiophile-oriented sonic qualities I talked about. The built-in presets are supposed to solve this challenge for us, but I have an itching to tweak them.
Go to the top of the page
+Quote Post
mithrandir
post Feb 28 2002, 20:36
Post #39





Group: Members
Posts: 669
Joined: 15-January 02
From: SE Pennsylvania
Member No.: 1032



Should have used the search. Here are the differences between the 3 presets.

-standard (default values):
-ms 2 -cvd 1 -bw 22050 -ltq ank -ltq_gain 0 -minSMR 0 -tmpMask 1 -tmn 18 -nmt 6

-xtreme:
default except -ltq fil -tmn 20 -nmt 8

-insane
default except -ltq fil -minSMR 3 -tmn 24 -nmt 9

Supposedly there are some characteristics intrinsic to each preset than cannot be mimiced via a switch.
Go to the top of the page
+Quote Post
Dibrom
post Feb 28 2002, 20:36
Post #40


Founder


Group: Admin
Posts: 2958
Joined: 26-August 02
From: Nottingham, UK
Member No.: 1



QUOTE
Originally posted by mithrandir
What is the difference between "--standard --ltq fil" and  "--xtreme"? If xtreme uses this more sensitive ath by default, then I suppose xtreme also uses more sensitive nmt and tmn values?


Yes. You can see the differences just by using the verbose switch.

QUOTE
[b]Perhaps nmt 16 and tmn 32 are a bit overkill. From a sonic perspective, there's nothing wrong with them (unlike -k -ms --lowpass 22.05 in LAME). The challenge is finding the right values that don't waste bits, yet still retain all of those audiophile-oriented sonic qualities I talked about. The built-in presets are supposed to solve this challenge for us, but I have an itching to tweak them.


I honestly feel that --xtreme, for almost all music one would encode, achieves this. At the very least, I haven't seen proof of this otherwise.
Go to the top of the page
+Quote Post
Garf
post Mar 1 2002, 00:39
Post #41


Server Admin


Group: Admin
Posts: 4853
Joined: 24-September 01
Member No.: 13



QUOTE
Originally posted by Dibrom

I honestly feel that --xtreme, for almost all music one would encode, achieves this.  At the very least, I haven't seen proof of this otherwise.


My point in this thread was that the above holds for --standard just as much.

I could ABX the clips the original poster provided on standard, so I'm going to try xtreme tomorrow and see what comes out...

--
GCP
Go to the top of the page
+Quote Post
mithrandir
post Mar 1 2002, 16:17
Post #42





Group: Members
Posts: 669
Joined: 15-January 02
From: SE Pennsylvania
Member No.: 1032



QUOTE
Originally posted by Dibrom
I honestly feel that --xtreme, for almost all music one would encode, achieves this.  At the very least, I haven't seen proof of this otherwise.

I've been testing out --insane --minSMR 0 and the results have been promising so far. My reasoning is this: I prefer full-bandwidth encoding, even if we can point to tests claiming it isn't necessary. Insane gives you this with a relatively minor increase in bitrate (depending on the source material). But insane's --minSMR 3 setting seems a little wasteful, so I "disabled" it by setting the value to 0 (like xtreme).

I encoded all of Nirvana's Nevermind album (the Mobile Fidelty UltraDisc II version) and the average bitrate was 212kbps vs. 285kbps for xtreme 16/32.

But I do understand your point. I was listening to those wayitis files that JohnV had up yesterday and I was rather surprised to find out after listening to them that wayitis_772.wav is ~128kbps MPC. It sounded almost like the original to my ears - it still had artifacts, but it was rather listenable...and this was using "downgraded" settings: -radio -ltq_gain 10 -tmn 12 -nmt 4.8. Of course, I made the comparison on a PC with a crappy integrated AC'97 sound chip, so perhaps the differences were mitigated by the hardware. But it was an impressive showing nonetheless, considering I never use 128kbps as a baseline when encoding
Go to the top of the page
+Quote Post
xmixahlx
post Mar 5 2002, 13:14
Post #43





Group: Members
Posts: 1394
Joined: 20-December 01
From: seattle
Member No.: 693



greetings fellow musepackrats [get it? --> a pack rat? as if "some" of us keep too much muse??? ...nevermind]

...especially mithrandir, or shall i say Olorin, Lathspell, Tharkin, Incanus, or simply.... the Grey Fool...

coincidence?

seeing your commandline in your last post i thought i would tell ya that...

i've been experimenting with these first two switches [and threw in --insane as reference]:

[kbps]
+5 ..... --insane --minSMR 0 --nmt 10 --tmn 32 --verbose --forcewrite
+3 ..... --insane --minSMR 0 --nmt 11 --tmn 26 --verbose --forcewrite
+0 ..... --insane

[those last two switches, verbose and forcewrite are just what i use in MPC batch encoder...they don't change the output file..but you know that already and now i look like an ass...]

they both return nearly identically sized files on average...but then the question remains...which is better of the two? [#2?]

the lone --insane switch differed with a difference of @ 4kpbs less than the experimental ones, and as far as i am concerned... i would rather those bits that were being spent by the [in]audibles meandor towards the audible ones...although i was previously o.k. with the --insane switch [edit: in regards to SIZE/kbps], this change just seems make sense to me.

maybe, just maybe, sv8 will come out and change everything...speaking of...did i hear right that KLEMM will not continue the braindead setting in sv8? it seems kinda silly not to if he is just introducing it now...


--------------------
RareWares/Debian :: http://www.rarewares.org/debian.html
Go to the top of the page
+Quote Post
mithrandir
post Mar 5 2002, 17:20
Post #44





Group: Members
Posts: 669
Joined: 15-January 02
From: SE Pennsylvania
Member No.: 1032



QUOTE
Originally posted by xmixahlx
[kbps]
+5 ..... --insane --minSMR 0 --nmt 10 --tmn 32 --verbose --forcewrite
+3 ..... --insane --minSMR 0 --nmt 11 --tmn 26 --verbose --forcewrite
+0 ..... --insane

they both return nearly identically sized files on average...but then the question remains...which is better of the two? [#2?]

In my experience, the second command line is preferred (doesn't mean you should use it, however). I have learned from testing that the noise masking threshold has a greater effect on sound quality than the tone masking threshold. I came across a segment of music two nights ago that demonstrated artifacting even with --nmt 9 (insane's default setting). Perhaps I don't know what to listen for, but I can set the tone masking threshold to a fairly low number and not hear much in the way of artifacts. --tmn 26 is plenty good for almost any situation and --tmn 32 isn't going to have an appreciable advantage. However, I can see situations where setting the noise masking threshold to 11dB instead of 10dB would be advantageous.
QUOTE
[b]i would rather those bits that were being spent by the [in]audibles meandor towards the audible ones

Yes. You do have to ask yourself how to do expect to play back these MPC files. For everything short of archiving, --insane is not necessary. The insane preset, even with the --minSMR 0 switch, is wasteful because it encodes the entire frequency spectrum...and if you aren't making "archival" files, you don't need the full spectrum. I have advocated using insane, but I have not qualified WHEN it makes sense.

In fact, I must admit that I and others have spread around a plethora of command lines without necessarily saying WHY or WHEN you might want to use them.

Let's say you want to make an archival-quality encode. I would suggest you use something like this:

--insane --minSMR 0 --nmt 12 --tmn 26

Make a few encodes and check the bitrate. If you don't mind using more bits, increase the nmt value (to no more than 16), then tmn (to no more than 30-32) and then minSMR (to no more than 3). The insane profile pushes the already very sensitive "fil" ath curve down by 6dB, so I doubt you'll ever need more sensitivity (like braindead's truly whacko -12dB offset).

But I frequently don't have to make archival-quality encodes. Most of the tracks I encode will be played back on my PC, using headphones plugged directly into the sound card. This requires a much different encoding approach.

I spent a lot of time this past weekend really playing with MPC's settings. I discovered that I could save a LOT of bits if I used the right ath profile for my needs and I was very suprised to find out that I should use --ltq ank --ltq_gain 6. This is not a sensitive ath curve at all (in fact it's 1dB above radio's curve!). Does that mean I'm deaf? No, not at all. The reason is because I found out that the droning sounds from my PC's case and CPU fans and dual hard drives really mask a lot of low-level sonic information. If I use --ltq fil, I find that the bitrate at the end of many tracks (say the music slowly fades out) does not dip much until digital silence is reached (when the bitrate reads 3kbps). Well, I thought, if I can't hear anything - thanks to the masking noise from the PC - then why should the bitrate remain at 200kbps!? I have the ath curve set wrong.

So I did some experimenting. To make a long story short, I kept playing with --ltq_gain until the bitrate during end-of-track fades correlated with what I could (and couldn't) hear. So instead of WinAMP reporting a fading-out track's kbps as 214-203-205-192-201-194-197-181-3, it now reports 214-205-208-192-161-124-96-43-5-3. Granted, bitrate is not a statement of volume, but as a track fades into imperceptibility, the bitrate SHOULD decrease as long as I cannot hear artifacting.

When I encode files for the sole purpose of playing them back on my PC, I now use:

--xtreme --ltq ank --ltq_gain 6 --nmt 10 --tmn 22

This produces manageable files (200kbps on average) that display little or no artifacting. If I had a quiet listening environment, I would use a lower --ltq_gain value or even switch to the "fil" profile. But I don't, so this makes sense for me.
QUOTE
maybe, just maybe, sv8 will come out and change everything...speaking of...did i hear right that KLEMM will not continue the braindead setting in sv8?  it seems kinda silly not to if he is just introducing it now...

SV8 won't change everything...or at least not standard 2-channel 44.1KHz encoding. What's there to change? Can we find startling sonic flaws in SV7? Not at all. The only breakthrough that is possible is a dramatic reduction in bitrate...say quality can stay the same while consuming 20% less bits. That would be nice, though I don't think anything near that will happen. Methinks Klemm tossed in braindead as a little joke. If you want to use braindead, but are afraid it will be deprecated, then just use the equivalent switches with xtreme, not that it's necessary.
Go to the top of the page
+Quote Post
Frank Klemm
post Mar 5 2002, 18:20
Post #45


MPC Developer


Group: Developer
Posts: 543
Joined: 15-December 01
From: Germany
Member No.: 659



QUOTE
Originally posted by mithrandir

SV8 won't change everything...or at least not standard 2-channel 44.1KHz encoding. What's there to change? Can we find startling sonic flaws in SV7? Not at all. The only breakthrough that is possible is a dramatic reduction in bitrate...say quality can stay the same while consuming 20% less bits. That would be nice, though I don't think anything near that will happen. Methinks Klemm tossed in braindead as a little joke. If you want to use braindead, but are afraid it will be deprecated, then just use the equivalent switches with xtreme, not that it's necessary.


SV8/standard will be use a little mit more data rate than
SV7/standard and will be from the view of quality between
SV7/standard and SV7/extreme.

Due to better Huffman coding the data rate will be more constant.
Fatboy should be transparent at ~250 kbps, not at ~310 kbps.

The high increase of bitrate for encoding beyond transparency
will be much smaller.

Low quality settings like radio, adsl and thumb will use
Intensity stereo, but in a much more flexible way than MP1,
MP2 or MP3.

But the main advantage will be more stable quality than SV7.
Especially the little bit too much treble problem should be fixed in
SV8. Editing/Fading of files without reencoding is possible. You can
convert SV7 and MP2 to SV8. Streaming is possible (but
not optimized for weak connections). 32 kHz and 48 kHz is
possible. Up to 7 channels are possible. Increases dynamic range
(102 dB => 256 dB). Tagging support. CDdb (not lke cddb
or freedb, but more like imdb).

Still not the focus will be data rates below 110 kbps.
For such datarates I recommend Ogg Vorbis, AAC, MP3Pro,
WMA, RA and Codecs which are designed to be good at this
datarates. The quality of MPC scales very good with the datarate.
For high data rates this is good, for low datarates ...


--------------------
-- Frank Klemm
Go to the top of the page
+Quote Post
xmixahlx
post Mar 5 2002, 22:01
Post #46





Group: Members
Posts: 1394
Joined: 20-December 01
From: seattle
Member No.: 693



hello again...Grey Fool:

the commandlines i referred to were for archival purposes...but then again...i think all of my encodes have been...

my idea is to be able to encode the .mpc's to another format if need be [read "mp3" for a portable player] thus the whole spectrum would be ideal to encode, but to tweak those encodes to achieve a lesser bitrate with greater quality...after using the second command line last night, i found that adjusting the nmt and tmn values even more served my evil needs quite well...

--insane --minSMR 0 --nmt 12 --tmn 32 --verbose --forcewrite

which is very weird considering that the one you offered is so similar, only the higher tmn value on my part...we must think alike...i wonder if that is good or bad...[read: creepy?]

i have just assumed that a tmn ceiling value of 32 was ideal...do i assume wrong? in my experience, 32 is @ 6kbps larger than 26, and well...that ain't much...

the target for me is to never even WANT to encode these files again...even if, in 20 years, SV500 comes out...but it looks like i will, with what MR. KLEMM has just revealed...it is very hard to become upset by these events though, isn't it...

later, mike


--------------------
RareWares/Debian :: http://www.rarewares.org/debian.html
Go to the top of the page
+Quote Post
Gecko
post Mar 6 2002, 01:39
Post #47





Group: Members
Posts: 934
Joined: 15-December 01
From: Germany
Member No.: 662



From my experience you can transcode pretty well from mpc xtreme (perhaps even standard) to mp3. It's not ultimately perfect, but the results are pretty decent. Better than mp3 --> mp3. If you are going to use the portable for unattentive listening, like during train or car rides, I think xtreme will suffice. If you want to use the portable with your home stereo consider more. If you allready need higher quality mpcs to achieve transparency on your regular setup, then you should be fine transcoding from those.

On the other hand I think (unproved!) that it doesn't make much difference, what source you feed mp3. Say an mp3 aps produces a file that is 75% perfect (100%=original) and mpc xtreme produces 90% perfect files. There allready is enough headroom present in mpc xtreme to make good mp3s. NB: I haven't tested this and this could be totally wrong. The %s are just some vague numbers. My ears are still in bad condition from the concert last weekend so don't trust them right now (...and maybe never again sad.gif).
Go to the top of the page
+Quote Post
mithrandir
post Mar 6 2002, 04:12
Post #48





Group: Members
Posts: 669
Joined: 15-January 02
From: SE Pennsylvania
Member No.: 1032



I think --tmn 32 is unnecessary; I can't hear any improvement over, say, --tmn 24. But, as you said, the bitrate increase is fairly benign.

If you may be transcoding to MP3, don't fret about not using insane. MP3 is going to chop all that high-frequency data anyway..."even" xtreme has wider bandwidth than a Lame APS file.

So while I have suggested using --insane --minSMR 0, you can also use --xtreme --ltq_gain [x], as long as [x] is -3 or less.

When you transcode, think of this little example. Assume an original WAV file has a "quality value" of 0, an MPC insane file has a quality value of 5 and a Lame APS file has a quality value of 8. If you want to estimate this hypothetical quality value for a MPC->MP3 transcode, simply take the square root of the sum of the squares: sqrt(5^2+8^2) = 9.43. Suppose you transcoded to a Xing file that had a QV of 18. The resulting files would have a QV of: sqrt(5^2+18^2) = 18.7. Notice how little the QV changed in this case...Xing is so much worse than MPC insane that the transcode has little effect: a slightly worse input will not dramatically change the output of a really crappy encoder. But if you transcode to a high-quality format, then the degradation is greater.

If you never want to reencode again, then go lossless. I guarantee you'll want to reencode your SV7s, not because they are markedly inferior, but it's in "our" nature to be obsessive. I know that once Klemm releases SV8 to the public, I'll be itching to reencode everything, whether I need to or not. Irrational behavior, but it oddly brings me enjoyment to coddle my obsessions.
Go to the top of the page
+Quote Post
layer3maniac
post Mar 6 2002, 09:45
Post #49





Group: Banned
Posts: 529
Joined: 29-September 01
Member No.: 37



QUOTE
If you never want to reencode again, then go lossless.
I think that re-encoding is one of the BEST reasons to go lossless. Rip ONCE, re-encode to your heart's desire!
Go to the top of the page
+Quote Post
lucpes
post Mar 6 2002, 16:56
Post #50





Group: Members
Posts: 517
Joined: 9-October 01
Member No.: 254



QUOTE
I think that re-encoding is one of the BEST reasons to go lossless. Rip ONCE, re-encode to your heart's desire! Rip ONCE, re-encode to your heart's desire!


And my 500 CD's I have now only in mpc would've required how much space???? some 300+GB? no thanks, no lossless for me...

just my 2cents though...
Go to the top of the page
+Quote Post

3 Pages V  < 1 2 3 >
Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 18th April 2014 - 14:56