IPB

Welcome Guest ( Log In | Register )

2 Pages V  < 1 2  
Reply to this topicStart new topic
-V2 gives way too high bitrate!?!
[JAZ]
post Mar 31 2010, 13:16
Post #26





Group: Members
Posts: 1706
Joined: 24-June 02
From: Catalunya(Spain)
Member No.: 2383



Thanks robert. I've used your explanations as a way improve the article.

Btw... do you still remember the updated docs i did waaaaaay ago?

http://www.hydrogenaudio.org/forums/index....showtopic=65756

I've checked the last changes to the documentation, and probably they will need to be updated too, but at least remember that those exist.



@ shadowking : I was going to say that the switch is not experimental, but two sources from LAME claim the opposite (it is not listed in the switches.html, and it is explicitely shown in the experimental part in lame --longhelp)

But i have to disagree partially with the reasoning behind not using -Y on -V2. -Y is not the only parameter that changes between -V3 and -V2, and -Y does not necesarily affect quality.
If we were to recommend a setting, it is clear that -V2 -Y is not something to recommend.
If we were to recommend a setting for the current case of this topic, -V2 -Y is a recommendation, if the user wants to keep using -V2.

We would need to find a sample where -Y had an impact, and -V2 vs -V3 was ABXable to determine how much -Y affects audibly, if any, in that case.



Ooops! I forgot to put the link:

http://wiki.hydrogenaudio.org/index.php?title=LAME_Y_SWITH

This post has been edited by [JAZ]: Mar 31 2010, 13:47
Go to the top of the page
+Quote Post
Ajax
post Mar 31 2010, 14:16
Post #27





Group: Members
Posts: 13
Joined: 11-October 06
From: California, USA
Member No.: 36216



The original poster "masterridley" (a new member) clearly stated in his first sentence that he is using dBpoweramp from Illustrate software to conduct his LAME MP3 encodings. AFAIK, the dBpoweramp GUI does not allow for -Y switches or -Y parameter changes in the LAME MP3 configuration choice. It does however have a user selectable choice for "high or normal quality" MP3 encodings but I do not think this is the same as a -Y switch (however I could be wrong about this). So "masterridley" if you are still reading this topic, you should know that the -Y switch discussion in your topic may not directly affect your usage of dBpoweramp to encode MP3 files.

Also, a message for Winamp developer Benski who posted before me in this topic.

Benski, it would be nice if there was a way for Winamp users to make -Y switches as well as more detailed LAME MP3 parameter changes in the Winamp GUI. For example, I believe the LAME MP3 configuration GUI in Winamp's current version 5.55 is getting a little outdated and should be updated. Winamp still uses the older "-- alt preset fast standard" LAME MP3 VBR quality selection format which I believe is outdated (but still the same quality as the newer -V selection methods).
Go to the top of the page
+Quote Post
robert
post Mar 31 2010, 14:25
Post #28


LAME developer


Group: Developer
Posts: 783
Joined: 22-September 01
Member No.: 5



global gain: is the global quantization step size (value range 0 - 255)
scale factor per band: amount how much to reduce the global quantization step size for this band, (value range different per bands, ie sfb0: 0-15)

CODE
    ifqstep = (scalefac_scale == 0) ? 2 : 4;
    for (sfb = 0; sfb < psymax; ++sfb)
        quantization_stepsize = global_gain - ((scalefac[sfb] + (preflag ? pretab[sfb] : 0)) * ifqstep) + subblock_gain[window[sfb]] * 8);


This post has been edited by robert: Mar 31 2010, 14:35
Go to the top of the page
+Quote Post
pdq
post Mar 31 2010, 14:47
Post #29





Group: Members
Posts: 3304
Joined: 1-September 05
From: SE Pennsylvania
Member No.: 24233



@[JAZ]

You might also point out that for older folks like myself the use of --lowpass less than 16 kHz makes -Y a moot point, since sfb21 will now be empty.
Go to the top of the page
+Quote Post
shadowking
post Mar 31 2010, 14:53
Post #30





Group: Members
Posts: 1523
Joined: 31-January 04
Member No.: 11664



We should not be recommending or documenting -Y unless there is a big issue. I think since v3.94 and the -V scale, there is no need for -Y, -ms etc . IMO if a few album come out @ 270k its not a big deal. That should be an exception rather than the rule. Should we recommend -Y if the rest of the OP's collection is around 200k ?

If V2 is too high use -V2.5, -V3 or --lowpass 17.5, 17 . -Y encourages tinkering like the old days.


Basically, I'm saying that in the old 3.90 days -Y was mentioned because there was only APS / APE and everything outside wasn't tuned.. Today there is a tuned V scale for precise control, So there is no need for things like -Y and probably even --lowpass.

This post has been edited by shadowking: Mar 31 2010, 15:10


--------------------
Wavpack -b450x1
Go to the top of the page
+Quote Post
db1989
post Mar 31 2010, 15:04
Post #31





Group: Super Moderator
Posts: 5141
Joined: 23-June 06
Member No.: 32180



QUOTE (Knowledgebase)
LAME Y SWITH

Surely I needn't say it, but perhaps there's a better title?
Go to the top of the page
+Quote Post
[JAZ]
post Mar 31 2010, 20:45
Post #32





Group: Members
Posts: 1706
Joined: 24-June 02
From: Catalunya(Spain)
Member No.: 2383



DOH!

I saw something strange in that, but i didn't realize the typo. Fixed!

http://wiki.hydrogenaudio.org/index.php?title=LAME_Y_SWITCH

If you really meant a completely different title... then i'm open to opinions.



As for documenting, explaining and/or recommending the -Y switch, i've done so to have a place to point to, when that question arises again.

Since it only applies to the settings -V2 , -V1 and -V0, because everything else uses it, i cannot see how it can be seen as a detrimental or damaging setting.
It surely makes the resulting file lower quality than not using it, but if the problem manifests, it can be seen as offering *more* quality than what it targets. (Not because of bitrate increment, but because of why it increases).

This post has been edited by [JAZ]: Mar 31 2010, 20:54
Go to the top of the page
+Quote Post
db1989
post Mar 31 2010, 21:09
Post #33





Group: Super Moderator
Posts: 5141
Joined: 23-June 06
Member No.: 32180



The spelling was one thing. Also, it may not matter for reasons other than aesthetics (people are unlikely to directly search for the article title, redirects could be set up, and I think wikis 'normalise' case), but must it be ALL CAPS? I thought brick wall limiting was frowned upon here. biggrin.gif

Personally, I would use "LAME -Y switch" or such--but as I said, it doesn't much matter, as redirects could catch most users trying other titles.


Edit: grammerz, etc.

This post has been edited by dv1989: Mar 31 2010, 21:35
Go to the top of the page
+Quote Post
Aleron Ives
post Mar 31 2010, 23:58
Post #34





Group: Members
Posts: 167
Joined: 22-March 10
From: California
Member No.: 79208



I would like to thank robert for explaining the correlation between -Y in VBR and the default CBR/ABR behaviour. My question has finally been answered.

QUOTE (shadowking @ Mar 31 2010, 03:52) *
-Y is still 'experimental'. It was never supposed to be used by non-developers. This is what Dibrom once said. I think these matters are too technical and will never be fully understood.

While it's true that the Y switch is quite technical and its purpose is difficult to explain, I don't think that means that trying to explain it is automatically a bad idea. I came into this topic having very little understanding of the way MP3 works and having no idea what "sfb21" was, and now I have at least a rudimentary comprehension of what sfb21 is, why it's different from the other sfb's, and why it can cause excessive bitrate in -V 0 - 2. I don't claim to fully understand everything, but I think this topic has taught me enough to make a moderately well-informed decision on whether I want to use -Y. I think that everyone should at least have the chance to receive the same information, so they too can make informed decisions on how they use LAME. No, not everybody will understand -Y after reading the explanation, but that's no reason to deliberately withhold the information, IMO. Isn't the purpose of this site to give people a venue to discuss digital audio, and hopefully learn about it in the process?

As for the "experimental" nature of -Y...

QUOTE ([JAZ] @ Mar 31 2010, 05:16) *
I was going to say that the switch is not experimental, but two sources from LAME claim the opposite (it is not listed in the switches.html, and it is explicitely shown in the experimental part in lame --longhelp)

It seems that -Y has moved into a netherrealm between "experimental" and "standard". While its existence is hidden from the user, -Y is actually probably being employed by many people who use LAME. It's enabled by default from -V 3 all the way to -V 9, and since the -V n presets have been fine-tuned for various quality levels, it seems to me that the developers have already decided that -Y isn't really "experimental"-- it must have a clear benefit that doesn't cause disproportionate decreases in quality, or they wouldn't be including it in the VBR presets.

This post has been edited by Aleron Ives: Apr 1 2010, 00:00
Go to the top of the page
+Quote Post
shadowking
post Apr 1 2010, 10:49
Post #35





Group: Members
Posts: 1523
Joined: 31-January 04
Member No.: 11664



This is still going back to my point. Users don't need to make an informed decision unless the current V system is broken. This was the case 7 yrs ago when the was no V system and preset medium couldn't be 'trusted' . Today from what I read people are having a hard time with -V5 - not -V3 or even -V4.

All mentions of -Y in the wiki look confusing and not professional. In the last 128k tests -V5.7 was used to obtain desired bitrate and NOT some random collection of switches. This worked put very well.

It would be wiser to mention *possible* bloating on some material - rock / metal due to mp3 limitations and that encoder is still doing the right job (its VBR after all). To maintain reasonable quality at an economic bitrate one should consider -V2.7 or -V3. It should be said those settings are still quality tuned but encode HF to lesser precision.


--------------------
Wavpack -b450x1
Go to the top of the page
+Quote Post
Aleron Ives
post Apr 1 2010, 11:18
Post #36





Group: Members
Posts: 167
Joined: 22-March 10
From: California
Member No.: 79208



QUOTE (shadowking @ Apr 1 2010, 02:49) *
This is still going back to my point. Users don't need to make an informed decision unless the current V system is broken. This was the case 7 yrs ago when the was no V system and preset medium couldn't be 'trusted' . Today from what I read people are having a hard time with -V5 - not -V3 or even -V4.

All mentions of -Y in the wiki look confusing and not professional. In the last 128k tests -V5.7 was used to obtain desired bitrate and NOT some random collection of switches. This worked put very well.

Maybe it's just me, but I think deliberately trying to keep people in the dark is "unprofessional". It seems like you're saying, "Most people are too stupid to understand any of this, so let's not even bother trying to explain things in the first place."

There's nothing wrong with having a very simple, well-tuned set of presets for the novice user to get good results with LAME, but I also don't think there's anything wrong with giving more in-depth explanations of how things work. LAME has a ton of configuration options that aren't all very well-explained in the longhelp, so having another resource to turn to for information is beneficial. If the purpose of LAME was to just provide a very simple and static set of MP3 encoding options limited to preset configurations, then all the switches wouldn't exist. The fact that the developers haven't removed them suggests to me that they figure somebody somewhere might have use of them. Consequently, it's nice to explain to people what they do should they wish to read such information. It's not like we're forcing people to read any of the in-depth wiki information. They can easly just see "use -V 2" on the main LAME page or in the LAME recommended usage message if they want the super-simple answer. The fact that someone is reading the wiki page on LAME suggests that they do in fact want more information.

While I can see the merit of trying to separate the in-depth information from the basic information, I think more than just the Y switch is "confusing" in that regard. The entire technical table of V settings ought to be moved to an advanced LAME settings page, as under the assumption that "people don't need to make informed decisions", knowing things like the lowpass for each setting is superfluous.
Go to the top of the page
+Quote Post
[JAZ]
post Apr 1 2010, 11:22
Post #37





Group: Members
Posts: 1706
Joined: 24-June 02
From: Catalunya(Spain)
Member No.: 2383



Robert, can you comment on this line of the article?

QUOTE
If global gain is zero, resolution will need to be increased (and quantization lowered) on every other scale factor band.


This is mostly extracted from the mp3-tech.org explanation of the sfb21 defect, but from your explanation of the relation between the global gain and the scale factors, isn't it the opposite? Like this:

QUOTE
Since the range of the scalefactors is small, if global gain is reduced and any of the scale factors cannot be incremented to compensate this reduction, the resolution of that band is increased innecessarily.



I've put up some updates to the article. (orthographic corrections and a couple of further explanations/formatting)

This post has been edited by [JAZ]: Apr 1 2010, 11:23
Go to the top of the page
+Quote Post
robert
post Apr 1 2010, 11:37
Post #38


LAME developer


Group: Developer
Posts: 783
Joined: 22-September 01
Member No.: 5



QUOTE
If global gain is zero, resolution will need to be increased (and quantization lowered) on every other scale factor band.

The sentence from the wiki makes no sense. Though the corresponding section at mp3-tech.org does explain the sfb21 problematic very well.

QUOTE
Scale factor refers to how much quantization (loss of precision) is applied to each band, where higher quantization causes greater compression, and consequently less variation between the minimum and maximum values (resolution).

- scale factors allow to reduce the quantization step size per sfb, within a limited range, as a deviation from the global step size (global gain).
- quantization is a process. larger quantization step sizes allow greater compression, this results in more quantization noise.

QUOTE
Global gain is an extra quantizer that affects all bands simultaneously.

This is plain wrong.

QUOTE
The only way to increase the resolution on sfb21 is therefore to reduce the global gain quantization, since global gain applies to all bands.

Needs correction.

QUOTE
If global gain is zero, resolution will need to be increased (and quantization lowered) on every other scale factor band.

If global gain is zero, then there is nothing you can do anymore.

QUOTE
LAME implements the -Y switch as a way to activate the alternate logic that CBR uses in respect to quantization noise in the sfb21 band.

No, it's an off switch. It turns off code that's used in higher quality VBR settings.

QUOTE
Since the sfb21 does not have quantization, its quantization noise is not evaluated.

Makes no sense. The content of sfb21 will surely get quantized.

QUOTE
In MPEG2 (16,22,24Khz), it is sfb12.

This is wrong. Long blocks have 22 sfbs (sfb0 - sfb21), short blocks have 13 sfbs (sfb0-12). The short block sfb12 has no scale factor of its own, like the long block sfb21.

This post has been edited by robert: Apr 1 2010, 13:44
Go to the top of the page
+Quote Post
[JAZ]
post Apr 1 2010, 13:39
Post #39





Group: Members
Posts: 1706
Joined: 24-June 02
From: Catalunya(Spain)
Member No.: 2383



shock1.gif wacko.gif blink.gif crying.gif blush.gif

Talk about failing to read... rolleyes.gif


I didn't fully understood how could the scalefactor be decreased if the gain was decreased too, and that's why I "invented" the other definition. It is much more clear now that I know that what is stored is not the scale factor itself, but a relative factor.


I've amended the article.

This post has been edited by [JAZ]: Apr 1 2010, 13:39
Go to the top of the page
+Quote Post
greynol
post Apr 1 2010, 18:28
Post #40





Group: Super Moderator
Posts: 10000
Joined: 1-April 04
From: San Francisco
Member No.: 13167



QUOTE (shadowking @ Apr 1 2010, 02:49) *
Users don't need to make an informed decision unless the current V system is broken.

I see great benefit in any wiki article that prevents continually recurring discussions/misconceptions about technical matters.

QUOTE (shadowking @ Apr 1 2010, 02:49) *
All mentions of -Y in the wiki look confusing and not professional.

So what?!? They don't necessarily have to be, do they?


--------------------
Your eyes cannot hear.
Go to the top of the page
+Quote Post

2 Pages V  < 1 2
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: 16th April 2014 - 14:58