IPB

Welcome Guest ( Log In | Register )

 
Reply to this topicStart new topic
Ogg Vorbis 500 Kbps vs lossless
mickywicky
post Jan 10 2005, 00:48
Post #1





Group: Members
Posts: 67
Joined: 10-January 05
From: London, UK
Member No.: 18958



Hi all,

I currently own an iRiver H120 and use Linux on my desktop. At the moment I am looking for an alternative to a lossless format, as the iRiver does not support any (well, it does support .WAV...).
This would be for encoding certain songs only, probably a total of 200 out of 2800 in my collection. Right now everything is ripped to Ogg Vorbis at 192 Kbps (old rips), 224 Kbps (most of them) and 256 Kbps for classical music or operas. So that's quality 6 to 8.
Ideally I would encode my few favourites to FLAC, but then it's not going to happen on the iRiver. So I am looking at the maximum quality Ogg Vorbis can offer, so a VBR 500 Kbps.
Myself, beyond 256 Kbps I cannot tell much difference, but it may have to do with my equipment of course.
WMA is not an option since playing those on Linux is a hassle.
So did anyone compare Vorbis 500 Kbps with a lossless codec before? I could not find any such comparison. I might end up doing a spectrum analysis myself, as soon as I find some tools that can do it (any recommendation, Linux or Windows?)

Thanxx for any input,

Micky
Go to the top of the page
+Quote Post
Jens Rex
post Jan 10 2005, 01:47
Post #2





Group: Super Moderator
Posts: 605
Joined: 18-December 01
Member No.: 680



QUOTE (mickywicky @ Jan 10 2005, 12:48 AM)
Ideally I would encode my few favourites to FLAC, but then it's not going to happen on the iRiver. So I am looking at the maximum quality Ogg Vorbis can offer, so a VBR 500 Kbps.
*

You could do that... but that's insane. You should rather do some ABX testing to find out at what levels Vorbis is transparent to you. I believe it's around 5 or 6 for most people, so if you can't be bothered to ABX, that's a safe choice.

QUOTE (mickywicky @ Jan 10 2005, 12:48 AM)
Myself, beyond 256 Kbps I cannot tell much difference, but it may have to do with my equipment of course.
*

Probably not. You don't really need godlike equipment to spot flaws in codecs. A decent pair of headphones should do it.

QUOTE (mickywicky @ Jan 10 2005, 12:48 AM)
So did anyone compare Vorbis 500 Kbps with a lossless codec before? I could not find any such comparison.
*

Probably because encoding Vorbis at those levels is complete overkill.

QUOTE (mickywicky @ Jan 10 2005, 12:48 AM)
I might end up doing a spectrum analysis myself, as soon as I find some tools that can do it (any recommendation, Linux or Windows?)
*

A spectrum analysis will tell you absolutely nothing about the quality of a codec. This has been discussed to death here, so I suggest you search the forum for more info on that. The same goes for ABX testing and recommended Vorbis settings. There's also a FAQ somewhere, that you should probably have a look at.

Edit:

CiTay said it well when he said:
<CiTay> using that high bitrate is like driving a Skoda, removing the floor, and pushing the gas pedal down to the street... yes it's getting more gas, but it's not going faster than before

This post has been edited by JensRex: Jan 10 2005, 01:52
Go to the top of the page
+Quote Post
Triza
post Jan 10 2005, 02:09
Post #3





Group: Members
Posts: 367
Joined: 16-November 03
Member No.: 9867



Using Vorbis -q 10 is not an overkill in certain scenarios. If you download from an online store that does not provide lossless, but it has Ogg I would choose ogg and choose -q 10. (Probably ogg is lossless codec with the highest max rate.) I expect (no tests so far sorry) that if I have very high bitrate then TRASCODING is possible from this source without noticable artifacts.

Or even for archiving. I find that FLAC works very well with classical music most of which has no clipping. You normally get 500-600 kbps FLAC. But modern pop/rock especially heavily clipped compressed can only go down to 1000-1100 kbps. You might want to arcive with ogg vorbis at -q 10 instead and transcode into the format of your choice from this source.

Again these are just untested ideas, but I think it is expected that -q 10 is very TRANSCODABLE rate. If I remember correctly somebody did some comparisons in this forum and showed that -q 10 is usually do not deviate more than 1% from the lossless signal. So it is almost identical.

Just my 2 pence

Triza
Go to the top of the page
+Quote Post
Canar
post Jan 10 2005, 02:22
Post #4





Group: Super Moderator
Posts: 3373
Joined: 26-July 02
From: To:
Member No.: 2796



QUOTE (Trizamano @ Jan 9 2005, 05:09 PM)
Again these are just untested ideas
*


Your ideas are both untested and invalid. Please refrain from elaborating upon such untested ideas in the future. If you've done a test, then by all means contribute what you've found its results to be. Don't attempt to pass hypothesis off as fact though.

QUOTE (Trizamano @ Jan 9 2005, 05:09 PM)
If I remember correctly somebody did some comparisons in this forum and showed that -q 10 is usually do not deviate more than 1% from the lossless signal. So it is almost identical.
*


This statistic means nothing to the way the audio sounds. Vorbis is not suitable for music archiving like FLAC is. If you'd read the thread, you'd realize that Vorbis has not been aggressively tuned for high bitrates. I'd suspect it would do well in its current incarnation, but there's no testing, and any gains you get from going from Q5 or Q6 up to Q10 is extremely negligible. 500kbps is completely overkill and unrecommended for any perceptual lossy codec.

This post has been edited by Canar: Jan 10 2005, 02:24


--------------------
You cannot ABX the rustling of jimmies.
No mouse? No problem.
Go to the top of the page
+Quote Post
Jens Rex
post Jan 10 2005, 02:23
Post #5





Group: Super Moderator
Posts: 605
Joined: 18-December 01
Member No.: 680



QUOTE (Triza @ Jan 10 2005, 02:09 AM)
Using Vorbis -q 10 is not an overkill in certain scenarios. If you download from an online store that does not provide lossless, but it has Ogg I would choose ogg and choose -q 10. (Probably ogg is lossless codec with the highest max rate.) I expect (no tests so far sorry) that if I have very high bitrate then TRASCODING is possible from this source without noticable artifacts.
*

If we ignore for a moment that transcoding is evil and wrong and eats kittens, encoding Vorbis at Q10 is still insane. You're still wasting space. If you were to transcode, it would probably be for portable use at something like Q2. I dare you to ABX a Q2 transcode from Q10 and Q6.

QUOTE (Triza @ Jan 10 2005, 02:09 AM)
Or even for archiving. I find that FLAC works very well with classical music most of which has no clipping. You normally get 500-600 kbps FLAC. But modern pop/rock especially heavily clipped compressed can only go down to 1000-1100 kbps. You might want to arcive with ogg vorbis at -q 10 instead and transcode into the format of your choice from this source.
*

No one ever said anything about lossless and transcoding. Archiving to lossless with later transcoding in mind is perfectly acceptable (if you have the space), but using Vorbis as a "lossless" codec is retarded.

QUOTE (Triza @ Jan 10 2005, 02:09 AM)
Again these are just untested ideas,
*

You're kidding...

QUOTE (Triza @ Jan 10 2005, 02:09 AM)
but I think it is expected that -q 10 is very TRANSCODABLE rate.
*

Probably not any more than Q6. See comment about transcoding above, though.
Go to the top of the page
+Quote Post
shadowking
post Jan 10 2005, 06:17
Post #6





Group: Members
Posts: 1529
Joined: 31-January 04
Member No.: 11664



Contrary to what most people here believe, you can have a bridge between lossy and lossless. Using ogg, mp3, mpc isn't ideal as they use digital filtering, but using wavpack lossy or optimfrog DS at around 450kbits can have real benefits. You achieve 55% compression over real lossless for all except classical / ambient music. You will have transparent music that transcodes to anything without artifacts. I say this with lots of abx experience.

Anyway using high bitrate encoding in portables is ideal if the same files will be listened to on a PC. Transcoding lossless to high bitrate lossy for portables is a bit silly IMO. For one, these portables no matter how good they sound are still marketed and tuned for 128k audio. I am having a hard time hearing defects with Lame -V5 on a panasonic discman and now a creative zen touch - even using good gear like senheiser px-100 cans. My PC setup: TB santacruz + grado MS1 is more sensitive to artifacts.

To the OP, I would use vorbis Q6 if your HP120 supports it or consider hybrid codecs like optimfrog DS for transcoding.


--------------------
Wavpack -b450s0.7
Go to the top of the page
+Quote Post
DreamTactix291
post Jan 10 2005, 06:19
Post #7





Group: Members (Donating)
Posts: 552
Joined: 9-June 04
From: A place long since forgotten...
Member No.: 14572



iHP-120 supports full range of Vorbis without problems in my experience, but with a hit on battery. With that said I use -q 5 Vorbis for mine because it is both transparent to me and around the size of --preset medium usually.


--------------------
Nero AAC 1.5.1.0: -q0.45
Go to the top of the page
+Quote Post
mickywicky
post Jan 10 2005, 11:54
Post #8





Group: Members
Posts: 67
Joined: 10-January 05
From: London, UK
Member No.: 18958



Thanks for all the replies so far.
I suspected Vorbis was not tuned to high bitrates, above 256 Kbps.
For the record, I run rsync between the iRiver and my desktop, which is connected to a hi-fi system, so the files are identical all around; which is why lossless or 256 Kbps Vorbis makes sense. Also, space is not much of an issue as my entire collection takes 17GB at the moment, and only grows slowly.

>For one, these portables no matter how good they sound are still marketed and
>tuned for 128k audio. I am having a hard time hearing defects with Lame -V5 on
>a panasonic discman and now a creative zen touch - even using good gear like
>senheiser px-100 cans

Though with Etymotic ER6's, q6 and q7 do sound different, hence the original question.

Guess I'll try to find out whether I can tell the difference between q8, q9 and q10 and stick to the lowest one that sounds identical to the FLAC or CD version. Bet it'll be q8 or q9, but from then on, it's more to do with my ears than anything else.

Thanxx

Micky
Go to the top of the page
+Quote Post
Faelix
post Jan 10 2005, 14:31
Post #9





Group: Members
Posts: 54
Joined: 6-July 02
Member No.: 2496



QUOTE (shadowking @ Jan 10 2005, 02:17 AM)
For one, these portables no matter how good they sound are still marketed and tuned for 128k audio. I am having a hard time hearing defects with Lame -V5 on a panasonic discman and now a creative zen touch - even using good gear like senheiser px-100 cans.
*


I understand perfectly the "marketing" part, but what does the "tuned" thing mean? It is indeed good to hear that Lame 3.96.1 provides very listenable music at 128kbps (and now we can finally have all the capacity advertised on portables), but I don't get what you mean, unless it is that more bitrate offer few reasonable improvement of quality on these portables.
Go to the top of the page
+Quote Post
shadowking
post Jan 10 2005, 15:43
Post #10





Group: Members
Posts: 1529
Joined: 31-January 04
Member No.: 11664



They don't have enough fidelity to make artifacts sound bad. I know quite a few problems on V5 that get softer with V4 (on the PC) - I don't think I will ever hear them on V4 with a portable. On portables V5 will sound good and I have doubts about improvements past V4 for normal music. So I am saying that around 160k a trained listener can abx small problems here and there, untrained will be nearly impossible - but do these apply to portable non-abx listening ? I doubt it.


--------------------
Wavpack -b450s0.7
Go to the top of the page
+Quote Post
Gambit
post Jan 10 2005, 22:19
Post #11


Burrrn developer


Group: Developer
Posts: 917
Joined: 25-November 01
From: Bratislava, Slovakia
Member No.: 534



Using Q8/Q9 for portables is just crazy. Actually, anything above Q6/mp3-aps is.
And I'm quite sure you can't ABX Q8/Q9 not even on your home equipment...

This post has been edited by Gambit: Jan 10 2005, 22:20


--------------------
Burrrn - http://www.burrrn.net/
MPEG Audio Collection - http://mac.sourceforge.net/
Go to the top of the page
+Quote Post
Supernaut
post Jan 10 2005, 22:48
Post #12





Group: Members
Posts: 136
Joined: 9-April 04
Member No.: 13359



QUOTE (Gambit @ Jan 10 2005, 10:19 PM)
Using Q8/Q9 for portables is just crazy. Actually, anything above Q6/mp3-aps is.
*

Indeed.

In fact, I find Vorbis to have the least (if at all) annoying artifacts at lower(ish) bitrates, and [if I had a Vorbis-capable portable] I'd be pretty comfortable using quality 3 or 4 for portable use (especially given the extra battery life).

QUOTE (mickywicky @ Jan 10 2005, 11:54 AM)
Guess I'll try to find out whether I can tell the difference between q8, q9 and q10 and stick to the lowest one that sounds identical to the FLAC or CD version. Bet it'll be q8 or q9, but from then on, it's more to do with my ears than anything else.
*

Please do try the lower quality levels as well. If you haven't done any ABX tests up until now, do try it (as other members have also recommended), you might end up at q5 or lower smile.gif Note that you'd not only be saving disk space (which you claim as unimportant), but also battery life for your portable.
Go to the top of the page
+Quote Post
Triza
post Jan 11 2005, 01:09
Post #13





Group: Members
Posts: 367
Joined: 16-November 03
Member No.: 9867



JenRex,

QUOTE
If we ignore for a moment that transcoding is evil and wrong and eats kittens, encoding Vorbis at Q10 is still insane. You're still wasting space. If you were to transcode, it would probably be for portable use at something like Q2. I dare you to ABX a Q2 transcode from Q10 and Q6.


I am misunderstood. I know that transcoding is generally bad and should be avoided. (I am not new to this game you know.)

I did not mean transcoding for portable use! You assumed that.

I give you 2 examples.

1) Online download. Your choices MP3 or Ogg. I choose Ogg at -q 10 because based on common sense I expect that I get something that is closer to the original then MP3 at 320 kbs CBR. So I can achive that and transcode down to -q 6 or even -q 5 for my hifi needs now or to some other format with probably unnoticable artifact if you are sort of space in your PC. Test it. The point is that I cannot predict the future and I think -q 10 is more transcodable than -q 6 especially if your target is -q 6 kinda rate.

2) If you are fed up that modern music is hardly go below 1000 kbs in lossless. I know because 90% of my music (if not more) is in FLAC and even the other lossless are just marginally better based on some lossless comparisons I cannot recall at the moment. So you might experiment with the transodability of Ogg -q 10 and if it is successful you archive your CD-s with -q 10. I think you stand a good chance with this strategy.

This -q 10 due to these reasons made me fancy Ogg even more. Although there were a lot of other more important reasons.

Canar,

QUOTE
Your ideas are both untested and invalid. Please refrain from elaborating upon such untested ideas in the future. If you've done a test, then by all means contribute what you've found its results to be. Don't attempt to pass hypothesis off as fact though.


I have 3 points to you.

1) I think I have the right to muse on these things whenever I want. Especially because these little ideas may sound useful for some and they may want to experiment with them and we may find some quantitative results in this regard from which we all benefit. It may be thought provoking and adds value to this forum.

2) I clearly stated that I have no proof only expectations based on common sense. I did not violate TOS. I did not recommend anything and I certainly this not present these ideas as FACTs and I stressed that several times. Please read it again. You will find that I did not edit that post (after you) so rest assured what you referred to is what is there now.

3) You DID violate TOS by saying that my ideas are "invalid" because you did not provide proof that supported that they are "invalid". No ABX nothing.

'Nuff said

Triza
Go to the top of the page
+Quote Post
music_man_mpc
post Jan 11 2005, 01:31
Post #14





Group: Members (Donating)
Posts: 707
Joined: 20-July 03
From: Canada
Member No.: 7895



QUOTE (Triza @ Jan 10 2005, 04:09 PM)
1) Online download. Your choices MP3 or Ogg. I choose Ogg at -q 10 because based on common sense I expect that I get something that is closer to the original then MP3 at 320 kbs CBR. So I can achive that and transcode down to -q 6 or even -q 5 for my hifi needs now or to some other format with probably unnoticable artifact if you are sort of space in your PC. Test it. The point is that I cannot predict the future and I think -q 10 is more transcodable than -q 6 especially if your target is -q 6 kinda rate.

blink.gif Why not use -q 6 and not transcode if you plan to use -q 6?
QUOTE (Triza @ Jan 10 2005, 04:09 PM)
2) If you are fed up that modern music is hardly go below 1000 kbs in lossless. I know because 90% of my music (if not more) is in FLAC and even the other lossless are just marginally better based on some lossless comparisons I cannot recall at the moment. So you might experiment with the transodability of Ogg -q 10 and if it is successful you archive your CD-s with -q 10. I think you stand a good chance with this strategy.

Fair enough, but personally I would prefer either using twice as much space and recieving the certainty of lossless or using less then have as much space and knowing that the files will still sound really good.

QUOTE (Triza @ Jan 10 2005, 04:09 PM)
1) I think I have the right to muse on these things whenever I want. Especially because these little ideas may sound useful for some and they may want to experiment with them and we may find some quantitative results in this regard from which we all benefit. It may be thought provoking and adds value to this forum.

Fine but you are still expected to test these "little ideas" before posting. HA.org wouldn't be very authoritative if half the threads contained "little ideas" like this and most of them turned out to be wrong.

QUOTE (Triza @ Jan 10 2005, 04:09 PM)
2) I clearly stated that I have no proof only expectations based on common sense. I did not violate TOS. I did not recommend anything and I certainly this not present these ideas as FACTs and I stressed that several times. Please read it again. You will find that I did not edit that post (after you) so rest assured what you referred to is what is there now.

Perhaps not, however one does get the impression that you are recommending the use of -q 10, if you aren't then you should think about rephrasing your stance on it. If you are then this does border on a TOS violation.

QUOTE (Triza @ Jan 10 2005, 04:09 PM)
3) You DID violate TOS by saying that my ideas are "invalid" because you did not provide proof that supported that they are "invalid". No ABX nothing.

You made a claim. Canar said that your claim was invalid, which means that your claim cannot be seen as being authoritative. Invalid doesn't mean the same thing as wrong. Thus Canar certainly didn't violate TOS #8, he didn't even make a claim about subjective audio quality. You did, Triza, thus the burden of proof is on you.


--------------------
gentoo ~amd64 + layman | ncmpcpp/mpd | wavpack + vorbis + lame
Go to the top of the page
+Quote Post
den
post Jan 11 2005, 06:04
Post #15





Group: Members
Posts: 315
Joined: 18-February 03
From: Perth, Western Australia
Member No.: 5050



I actually posted some ABX results on HA a couple of years ago, where I transcoded Vorbis q10, Lame api, mpc q10 and Wavpack lossy 320 kbit into ATRAC (for minidisc).

Vorbis q10 was clearly discernable in the sample tested, which was the introduction to Blue Monday by New Order if I recall correctly. (Wavpack was the closest to transparent to these ears.)

A search for Wavpack, ABX, Vorbis and ATRAC will probably find it somewhere.

This probably doesn't help the original poster's question, but I just want to make the point that assuming an extreme bitrate will save you from transcoding artifacts is a mistake, as this is not the case in my experience.

Later

Den.


--------------------
Den
My blog - http://www.iinet.net.au/~den
Go to the top of the page
+Quote Post
Cygnus X1
post Jan 11 2005, 07:06
Post #16





Group: Members (Donating)
Posts: 676
Joined: 5-June 02
From: New York
Member No.: 2224



QUOTE (den @ Jan 11 2005, 12:04 AM)
This probably doesn't help the original poster's question, but I just want to make the point that assuming an extreme bitrate will save you from transcoding artifacts is a mistake, as this is not the case in my experience.

*


Same here. Way back when, in my MD days, I basically came to the same conclusion: hybrid codecs that simply do quantization (like wavpack lossy) seemed to transcode with the least amount of transcoding artifacts, followed by MPC, which was followed by Ivan's PsyTel AAC encoder ("transcoding" setting). MP3 and Vorbis did the worst, which is no wonder considering the format-inherent problems of each that aren't mitigated by simply throwing in more bits (pre-echo and "thickness" for MP3 and coarseness/HF boost for Vorbis).

Unfortunately, I don't have the original ABX results from this, but the observations above provide a summary of what I found.
Go to the top of the page
+Quote Post
mickywicky
post Jan 11 2005, 15:10
Post #17





Group: Members
Posts: 67
Joined: 10-January 05
From: London, UK
Member No.: 18958



Supernaut:

>Please do try the lower quality levels as well. If you haven't done any ABX tests
>up until now, do try it (as other members have also recommended), you might
>end up at q5 or lower Note that you'd not only be saving disk space (which you c
>laim as unimportant), but also battery life for your portable.

I did a few weeks ago, though I had no clue this was called an ABX test tongue.gif
On portable q5 was extremely close to transparent (I could only hear a difference with certain tracks), until I started using the Etymotic earphones, and that went up to q6.
I see your point with the battery life though. It went from 16 hours to 13 or so. But I am lazy, and on my home system, q5 definitely sounds different. Since I mirror both repositories...

den:
>I actually posted some ABX results on HA a couple of years ago, where I
>transcoded Vorbis q10, Lame api, mpc q10 and Wavpack lossy 320 kbit into
>ATRAC (for minidisc).

Excellent, I'll check these out. I tried a few searches before posting, but without knowing what ABX meant I wasn't going to look for it, was I tongue.gif

Very good forum, I'm glad I found this.

M.
Go to the top of the page
+Quote Post
xmixahlx
post Jan 11 2005, 19:16
Post #18





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



playing wma on linux is easy (but not recommended smile.gif )

there is an xmms plugin and a lamip plugin (based on the xmms one) that i know of


later


--------------------
RareWares/Debian :: http://www.rarewares.org/debian.html
Go to the top of the page
+Quote Post
Pio2001
post Jan 13 2005, 22:29
Post #19


Moderator


Group: Super Moderator
Posts: 3936
Joined: 29-September 01
Member No.: 73



QUOTE (Triza @ Jan 11 2005, 02:09 AM)
I choose Ogg at -q 10 because based on common sense I expect that I get something that is closer to the original then MP3 at 320 kbs CBR.
*


Yes, but being closer to the original is not necessarily an advantage. We are talking psychoacoustic compression here, and a file closer to the original can very well sound worse. If the Vorbis codec used had got a flaw, for example, it would be present at any quality level, and would end up worse than MP3, even at q10.
Go to the top of the page
+Quote Post

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: 27th December 2014 - 08:49