Skip to main content

Topic: safe, theoretical best archive quality with VBR (Read 4402 times) previous topic - next topic

0 Members and 1 Guest are viewing this topic.
  • user
  • [*][*][*][*][*]
safe, theoretical best archive quality with VBR
Hi all,

just want tell that of reasons of convinience the combination of EAC with lame is really great.

After some trials I even managed to include ID3 tags in one step ripping, encoding...

In EAC choose queued encoding. That saves time (after first track has been extracted to wav, Lame gets started to convert to MP3 and at the same time EAC extracts next track....) and gives Lame/EAC the possibility to include ID3 in mp3 files.

After some time of getting used to all of the settings now I changed my mind and so now I prefer clearly the short form --dm-preset insane ....  (one reason is: EAC accepts not so long command lines...)

--dm-preset standard -X3 -b224 -F --scale 0.98

--dm-preset insane -b224 -F --scale 0.98
www.High-Quality.ch.vu -- High Quality Audio Archiving Tutorials

  • CiTay
  • [*][*][*][*][*]
  • Administrator
safe, theoretical best archive quality with VBR
Reply #1
Hi user,

IMO, this line performs better:

--nspsytune -V0 -mj -h --lowpass 20 -b192 --nssafejoint -Z --athtype 2 --ns-sfb21 -1 --scale 0.98


I also tested -b224, but it didn't work exactly as good as -b192. I don't know why that is, might be that there's less room left to "breathe" with -b224.

  • JohnV
  • [*][*][*][*][*]
  • Developer
safe, theoretical best archive quality with VBR
Reply #2
I would try something like --dm-preset insane -X3
or variants with -X3. Seems that X3 noise calculations handle extreme bass pulse clips like Fatboy pretty well.
Juha Laaksonheimo

  • Dibrom
  • [*][*][*][*][*]
  • Administrator
safe, theoretical best archive quality with VBR
Reply #3
Quote
Originally posted by JohnV
I would try something like --dm-preset insane -X3
or variants with -X3. Seems that X3 noise calculations handle extreme bass pulse clips like Fatboy pretty well.


I agree with JohnV.  Instead of just bumping the bitrate or using a higher -V to achieve higher quality, one really needs to look at different techniques altogether.  One of the things that I am starting to work on for my latest tweaks is a different noise measuring function (which is what the -X switches handle).  I think I have an idea for a new one which may provide much improved results.

-X3 btw basically provides for maximum noise measuring AND total noise measuring while -X1 (which is what nspsytune by default uses, except in --r3mix which turns it off in favor of -X0 - average noise) just measures maximum noise.

Something like --dm-preset standard -X3 (without -Z even) will probably provide higher quality than any of these lines listed.

  • ears
  • [*][*][*]
safe, theoretical best archive quality with VBR
Reply #4
dibrom,

Just for clarity's sake, let me see if I've got this right.  You'd say that one will get better quality from:

--dm-preset standard -X3

than

--dm-preset insane -X3

Excluding all criteria other than quality?  Or are we talking size vs. quality again?

BTW, I'm using the compile included in the win32LAMEv1.7 bundle.


-ears

  • Dibrom
  • [*][*][*][*][*]
  • Administrator
safe, theoretical best archive quality with VBR
Reply #5
Quote
Originally posted by ears
dibrom,

Just for clarity's sake, let me see if I've got this right.  You'd say that one will get better quality from:

--dm-preset standard -X3

than

--dm-preset insane -X3

Excluding all criteria other than quality?  Or are we talking size vs. quality again?


Well I'm not 100% sure, but I would personally go with the standard line over the insane line.  Basically, I wouldn't use Insane currently as it is more experimental than anything else, and I'm not convinced that the methods for achieving higher quality that are used within it are the best way to go.  In my new compile coming up I am going to be revamping all of the presets to provide a more linear quality gain and I will be applying a few things I have learned about which approaches may be more effective than others in attaining higher quality.

With --dm-standard -X3, you should get pretty damn good quality from LAME I believe.  I wouldn't even really feel the need to add -Z to that line personally.

  • user
  • [*][*][*][*][*]
safe, theoretical best archive quality with VBR
Reply #6
Hi,

so I understand that dm standard is quite safe for AQ, and xtreme and insane are still experimental and not really proven to result in better quality than standard.

That would be analogue to r3mix AQ test, too, and to  opinions of some experienced listeners.

Do the actual settings get integrated into dm presets of actual alpha versions of lame ?
www.High-Quality.ch.vu -- High Quality Audio Archiving Tutorials

  • Dibrom
  • [*][*][*][*][*]
  • Administrator
safe, theoretical best archive quality with VBR
Reply #7
Quote
Originally posted by user
so I understand that dm standard is quite safe for AQ, and xtreme and insane are still experimental and not really proven to result in better quality than standard.


Yes.  When I was doing my original testing, I do believe I did generally hear a "cleaner" sound with xtreme, due to a lower ath curve, but it is very slight.  In further practice though standard has been proven to be quite high quality, and I don't believe there has been any clear case where the other presets have been proven to be a big step up.  What is most likely going to happen with the newer presets is that the higher the setting you go, the better they will perform on critical samples.  Stuff like fatboy and drone.  On normal music, it is hard to get much higher quality within LAME currently.

Quote
Do the actual settings get integrated into dm presets of actual alpha versions of lame ?


Yes, updates to the presets and whatever I find to perform better, will be included in the newer versions of LAME eventually.

  • RD
  • [*][*][*]
safe, theoretical best archive quality with VBR
Reply #8
Dibrom,

Now speaking *theoretically*, I wonder why you would not recommend --dm-preset xtreme -X3 as the current highest quality setting with Lame 3.90 alpha7 August 17th build for people who do not care about filesize.

I understand that your new compile will be better because you are altering many things (e.g., something like all short blocks with be 320 kbps stereo, etc.), but FOR NOW would the above commandline be better than --dm-preset standard -X3?

Since xtreme and standard are identical except for xtreme's lower ath.. why wouldn't you recommend that?

I can only come up with a few reasons:

(1) standard has been tested more and performs more consistently than xtreme for some as of yet undetermined reason.
(argumentum via experience.)

(2) nspsytune may not work as well (bugs, incompatibility?) with ath2 as with ath4 (analogue: nspsytune and -q0 are not friends but -q2 works well...)

(3) The V2 (vbr setting 2) of xtreme is slightly more inefficient when combined with the lower ath curve (doubtful, because it should scale, right?)

(4) -X3 and ath2 are not as compatible for some undetermined reason...

I would like your insights on both the theory and what you have found out in practice, because I don't care about filesize and I am curious as to why I should choose:
(A) --dm-preset standard -X3 OVER (B) --dm-preset xtreme -X3

(B) CANNOT be worse, can it?

Thanks
RD

  • Dibrom
  • [*][*][*][*][*]
  • Administrator
safe, theoretical best archive quality with VBR
Reply #9
Quote
Originally posted by RD
(1) standard has been tested more and performs more consistently than xtreme for some as of yet undetermined reason.
(argumentum via experience.)


I believe it would be this one.  Now, that is not to say xtreme performs in an unstable manner, but on at least 1 sample (hihat.wav) I found that standard actually sounded better as did JohnV and a few others.  I have a theory on this though.  I think the reason this happened is because, neither sound like the original a whole lot, but standard seems to sound better, even though it is less accurate on the ath.  What I think is happening is that xtreme is trying to encode more above 16khz, but in this case it cannot do so well enough to prevent encoding that content without more artifacts.  Standard on the other hand isn't really affected because it never tried to encode that content in the first place.

It's like this basically.  Wouldn't you rather have a lowpass at 14khz than go up to 16khz and have that last bit be heavily artifacted?  That isn't exactly what happened here, but it is an example of why I think standard sounded better in that case.  Hope that makes sense.

  • RD
  • [*][*][*]
safe, theoretical best archive quality with VBR
Reply #10
Thanks,

your comments make a lot of sense, and now I think i understand...

Anxiously awaiting the new compile...
RD

PS for now you recommend --dm-preset standard -X9
not: --dm-preset standard -X3?

  • Dibrom
  • [*][*][*][*][*]
  • Administrator
safe, theoretical best archive quality with VBR
Reply #11
-X9 is the new mode that I will be addding in my compile, it currently doesn't do anything in the normal LAME out there.

  • user
  • [*][*][*][*][*]
safe, theoretical best archive quality with VBR
Reply #12
Hi,

CiTay wrote that he experienced that -b192 resulted better. How can that be possible ? He suggested of bigger reservoir.
Could it be true ?  I thought that the reservoir is still there, even bigger if -b224 instead of -b192 is chosen. In ABR it is clear that -b must be about maximum of 75 % of average bitrate, because ABR works with reservoir. But VBR decides on frame-to-frame-basis, which bitrate is chosen, that I thought until I read CiTays remark. 

My general question is: If Lame decides in one of the VBR presets to use eg. 128 kbit for a frame, (eg. because of silent or not so loud passage) but eg. -b192 or -b224 is chosen, what is the result ?
Is the "overkill"-bitrate used for furthermore information on sound quality or is it completely wasted in zero information ?
www.High-Quality.ch.vu -- High Quality Audio Archiving Tutorials

  • HomeK
  • [*]
safe, theoretical best archive quality with VBR
Reply #13
Another question: will this decrease quality for LAME 3.90a7?

--dm-preset standard -X3  --vbr-mtrh

  • Dibrom
  • [*][*][*][*][*]
  • Administrator
safe, theoretical best archive quality with VBR
Reply #14
Quote
Originally posted by HomeK
Another question: will this decrease quality for LAME 3.90a7? 

--dm-preset standard -X3  --vbr-mtrh


Yes, vbr-mtrh currently doesn't work correctly with the -X modes.  The functionality to allow this is being added, but is not in place yet.

  • user
  • [*][*][*][*][*]
safe, theoretical best archive quality with VBR
Reply #15
--dm-preset standard -X3 -b224 -F --scale 0.98

Added the current best command line, right ?
www.High-Quality.ch.vu -- High Quality Audio Archiving Tutorials

  • Dibrom
  • [*][*][*][*][*]
  • Administrator
safe, theoretical best archive quality with VBR
Reply #16
Quote
Originally posted by user
--dm-preset standard -X3 -b224 -F --scale 0.98

Added the current best command line, right ?


There is no need for the -b224 or -F, and --scale is a bad idea when you can use mp3gain.

  • user
  • [*][*][*][*][*]
safe, theoretical best archive quality with VBR
Reply #17
Hi,

what's that ?

I thought a little bit --scale 0.98 would reduce most of the clipping. There has been an example in the other board by somebody who tried scale 0,98 down to 0,9 in steps and he found out that 0,98 avoids a lot clipping, 0,95 most clipping and something around 0,9 nearly all of the clippimg ?
www.High-Quality.ch.vu -- High Quality Audio Archiving Tutorials

  • Dibrom
  • [*][*][*][*][*]
  • Administrator
safe, theoretical best archive quality with VBR
Reply #18
http://www.geocities.com/mp3gain/

Mp3Gain is a utility which allows you to losslessly adjust gain and remove clipping.. in contrast, --scale is NOT a lossless process.  With mp3gain you can reverse the adjustments also.

For that matter however, .98 isn't going to solve that much clipping overall.  For music which is not very heavily clipped it might, but for quite a bit of music these days you will need at least around .95 or even .9 to really stop clipping.  Since mp3gain can do this losslessly, there really isn't much of a point to using --scale.

  • user
  • [*][*][*][*][*]
safe, theoretical best archive quality with VBR
Reply #19
Would a minimum bitrate of 224 cause problems, decrease quality (eg. of reduced bitrate reservoir ?) ?

Or would the bitrate reservoir be the same, independent if -b224 is used or not, so that -b224 would just be a kind of overkill, but would not cause trouble ?
www.High-Quality.ch.vu -- High Quality Audio Archiving Tutorials

  • Dibrom
  • [*][*][*][*][*]
  • Administrator
safe, theoretical best archive quality with VBR
Reply #20
Quote
Originally posted by user
Would a minimum bitrate of 224 cause problems, decrease quality (eg. of reduced bitrate reservoir ?) ?

Or would the bitrate reservoir be the same, independent if -b224 is used or not, so that -b224 would just be a kind of overkill, but would not cause trouble ?


I'm not really sure either way.  However, if you haven't actually heard a difference which you can abx, then I don't see the point in using the switch, especially if it could somehow cause unexpected behavior with vbr.  These days, when I am working on improving quality, I only change something if I can hear a difference.  If I can't, then I don't consider it to be a problem.  It could be that on certain clips the -b224 increases quality (on clips other than on something like fatboy) but so far I don't believe I have found one where there really is a clear difference. If you have though, let me know about it.