IPB

Welcome Guest ( Log In | Register )

 
Reply to this topicStart new topic
Any hard data for high bitrate fetish?
How many (killer) samples can you personally ABX?
You cannot see the results of the poll until you have voted. Please login and cast your vote to see the results of this poll.
Total Votes: 62
Guests cannot vote 
rpp3po
post Mar 30 2009, 21:17
Post #1





Group: Developer
Posts: 1126
Joined: 11-February 03
From: Germany
Member No.: 4961



Yes there are very few killer samples. That's the reason some people encode whole collections at very high data rates. I wonder if this is necessary. And the probability for artifacts generally decreases while bitrate increases. But only up to a certain point. Most "killer" samples I know are codec specific, then they fail even for the highest bitrates. So increasing bitrate above a certain point does not decrease the probability for artifacts any more.

Now variable bitrate encoding modes suck up more bandwidth when needed, but maintain very reasonable averages over a collection. So I allege: there is NO reason for going above a certain constant-quality setting, even for killer samples. Either the codec fails at any bitrate or it is transparent for anything above a certain setting.

In my experience the sweet spot are the ~192kbit/s average AAC settings (Q127 and q .5-55) and probably V2 for MP3 (haven't tested the latter myself).

Is there any contradicting evidence?

If you vote other than '0' please include a reference to the sample in question.

EDIT: Editing finished. Accidently I touched the enter key before the poll setup was complete. Please vote again.

This post has been edited by rpp3po: Mar 30 2009, 21:58
Go to the top of the page
+Quote Post
ojdo
post Mar 30 2009, 21:25
Post #2





Group: Members
Posts: 894
Joined: 18-June 06
From: Germany
Member No.: 31980



I am missing the option "never tried to ABX killer samples so far" to participate wink.gif


--------------------
http://freemusi.cc/
Go to the top of the page
+Quote Post
DVDdoug
post Mar 30 2009, 23:07
Post #3





Group: Members
Posts: 2537
Joined: 24-August 07
From: Silicon Valley
Member No.: 46454



QUOTE
That's the reason some people encode whole collections at very high data rates. I wonder if this is necessary.
It's not "necessary" for me to use a lower bitrate... I've got plenty of disk space*, so I just use V0. I don't worry about using the "ideal" bitrate, and I've never bothered with ABX testing.


*Well.... I have plenty of disk space for my MP3s, but not enough for FLAC.

This post has been edited by DVDdoug: Mar 30 2009, 23:08
Go to the top of the page
+Quote Post
rpp3po
post Mar 31 2009, 10:24
Post #4





Group: Developer
Posts: 1126
Joined: 11-February 03
From: Germany
Member No.: 4961



C'mon, is that '1' vote just defiance or where is the reference sample?
Go to the top of the page
+Quote Post
insane_alien
post Mar 31 2009, 10:49
Post #5





Group: Members
Posts: 64
Joined: 26-July 08
Member No.: 56345



i just use flacs, i've got oodles of space available and currently nothing better to do with it. although i do keep a 256kbps ogg mirror for my laptop and a 128kbps ogg mirror for my S9.

i suppose i could get by with the 128kbps oggs(i can't ABX all but 2 tracks between the 128 and 256 tracks)
Go to the top of the page
+Quote Post
odigg
post Apr 1 2009, 21:54
Post #6





Group: Members
Posts: 629
Joined: 25-July 08
From: USA
Member No.: 56264



QUOTE (DVDdoug @ Mar 30 2009, 18:07) *
QUOTE
That's the reason some people encode whole collections at very high data rates. I wonder if this is necessary.
It's not "necessary" for me to use a lower bitrate... I've got plenty of disk space*, so I just use V0. I don't worry about using the "ideal" bitrate, and I've never bothered with ABX testing.


I have ABXed MP3s and V0 (or even V2) is beyond my ABX capability. However, I still encode everything at V0. Like you, I've got a so much free drive space it hardly makes a difference. V0 also offers me a bit of psychological security. After all, there is a chance (however small) that I will be bitten by a radioactive bat and I will get super hearing.

What's that? Batman calling?

This post has been edited by odigg: Apr 1 2009, 21:55
Go to the top of the page
+Quote Post
singaiya
post Apr 2 2009, 04:56
Post #7





Group: Members
Posts: 365
Joined: 21-November 02
Member No.: 3830



I could ABX two killer samples at -V 2 (castanets, kraftwerk). Might be able to do more if I tried more samples, but I'm not so interested in high bitrate mp3 at the moment. I haven't done any high bitrate AAC tests. I'm interested in testing low bitrate AAC, around 100 kbps, since usually that's my sweet spot and so far everything has been transparent for me there. I'm also interested in transcoding tests from lossywav or high bitrate AAC, since I'm not willing to go lossless on the whole collection.
Go to the top of the page
+Quote Post
rpp3po
post Apr 2 2009, 13:05
Post #8





Group: Developer
Posts: 1126
Joined: 11-February 03
From: Germany
Member No.: 4961



QUOTE (singaiya @ Apr 2 2009, 05:56) *
I could ABX two killer samples at -V 2 (castanets, kraftwerk)...


1. That there are samples which you can ABX is not the point of this thread. I'm only interested wether you can ABX V2 (Q127, q .5) if you cannot ABX the same sample with the same encoder at a higher setting like V0 (256kbit/s+, q .6). For the former there are already endless threads.

2. Boo to the cowards, who vote other than '0' but don't present references! Pushing a button is the wrong tool for strengthening your self conception as being different from the mass. Just try it a 1000 times and you will see: it's still the old Joe Average sitting in front of his computer.

To the mods: What sense does it make to hide voter lists from polls? This is an objective community and there shouldn't be the need to hide (already anonymous) names in polls. It will not improve a poll's data quality, rather the opposite.

This post has been edited by rpp3po: Apr 2 2009, 13:11
Go to the top of the page
+Quote Post
Slipstreem
post Apr 2 2009, 13:42
Post #9





Group: Members
Posts: 966
Joined: 7-July 06
Member No.: 32660



I think the poll feature may be broken. I haven't voted but the information at the bottom of the poll says that I have. All I did was click on the "view poll results" button. blink.gif

Cheers, Slipstreem. cool.gif
Go to the top of the page
+Quote Post
Remedial Sound
post Apr 2 2009, 13:57
Post #10





Group: Members
Posts: 504
Joined: 5-January 06
From: Dublin
Member No.: 26898



To the OP, I didn't vote in the poll on account of the fact that I haven't tried to ABX anything in years - back then I found that I couldn't ABX lame V4 or V5, and since then I've been pretty much living in 'ignorant' bliss. I keep my flac backups on an external HD, mirrored on my laptop (lame V4) for computer and portable listening.

Casting aside problem samples (which as you point out are in most cases a limitation of the format and not the bitrate/encoding level), I'm inclined to agree with your theory and believe there are probably very few, if any, exceptional ears out there that can ABX a V2 encode but not a V0. If you're looking for a more scientific answer you're probably better off conducting a public listening test than a poll - in which say the participants attempted to ABX V0, V2, V4, V6 vs an original across a range of "normal" music styles.
Go to the top of the page
+Quote Post
halb27
post Apr 2 2009, 14:11
Post #11





Group: Members
Posts: 2424
Joined: 9-October 05
From: Dormagen, Germany
Member No.: 25015



QUOTE (rpp3po @ Mar 30 2009, 21:17) *
... Either the codec fails at any bitrate or it is transparent for anything above a certain setting.

In my experience the sweet spot are the ~192kbit/s average AAC settings (Q127 and q .5-55) and probably V2 for MP3 (haven't tested the latter myself). ...

Yes, for me too the sweet spot of a good transform encoder is close to but a bit lower than 200 kbps.
However if storage space isn't an issue why not use more like -V0 in case of Lame? There are real life samples which make a subtle though audible difference.

I am a bit low on storage space with my new DAP. That's why I encode nearly everything at -V1.5. For great tracks (musically speaking) I use -V0 or lossyWAV, for simple instrument music or singing I downsample to 32 kHz by using --lowpass 15.2.

This post has been edited by halb27: Apr 2 2009, 14:14


--------------------
lame3100m -V1 --insane-factor 0.75
Go to the top of the page
+Quote Post
rpp3po
post Apr 2 2009, 14:18
Post #12





Group: Developer
Posts: 1126
Joined: 11-February 03
From: Germany
Member No.: 4961



QUOTE (Remedial Sound @ Apr 2 2009, 14:57) *
If you're looking for a more scientific answer you're probably better off conducting a public listening test than a poll - in which say the participants attempted to ABX V0, V2, V4, V6 vs an original across a range of "normal" music styles.


I consider the the non-ABXability of well over 97% of music at ~150kbit/s (AAC) a fact. This thread is about wether there are any reasons at all (even isolated tracks) to go over a certain variable bitrate setting.

QUOTE (halb27)
There are real life samples which make a subtle though audible difference.


Which would that be for LAME? I know that according to all public information, here at HA and on Google, the non '0' votes for Quicktime must be either a lie or unknown facts (I am not tending towards the latter). But maybe it's different for LAME.

This post has been edited by rpp3po: Apr 2 2009, 14:51
Go to the top of the page
+Quote Post
Yirkha
post Apr 2 2009, 14:37
Post #13





Group: FB2K Moderator
Posts: 2359
Joined: 30-November 07
Member No.: 49158



QUOTE (Slipstreem @ Apr 2 2009, 14:42) *
I think the poll feature may be broken. I haven't voted but the information at the bottom of the poll says that I have. All I did was click on the "view poll results" button.
That button is called "View Results (Null vote)" and has a tooltip which says "View results, but forfeit your vote in this poll".


--------------------
Full-quoting makes you scroll past the same junk over and over.
Go to the top of the page
+Quote Post
halb27
post Apr 2 2009, 15:28
Post #14





Group: Members
Posts: 2424
Joined: 9-October 05
From: Dormagen, Germany
Member No.: 25015



QUOTE (rpp3po @ Apr 2 2009, 14:18) *
... There are real life samples which make a subtle though audible difference. ...
Which would that be for LAME?

Let me start with that sample which made me conscious of potentially restricted mp3 quality though it did not come from real life. That was 'trumpet' which I ran upon when I came to HA during the alpha development phase of Lame 3.97 when I did listening tests as did so many HA members. It was so easily abxable at -V0 that I was shocked and it was not one of those pre-echo samples prone to mp3 restrictions. The problem was not improved substantially with 3.97 final, and it was there with previous Lame versions. The problem was increased by using VBR - with very high bitrate ABR or CBR the results were at least a lot better. 3.98 improved significantly on trumpet and other non-pre-echo problems like those of the 'sandpaper noise' kind, or the musically very simple track 'herding_calls'. 3.98 also improved on pre-echo samples like the terrific problem sample 'eig' (aka 'Abfahrt Hinwihl') which BTW maybe one of the easiest sample to be ABXed when using -V0 (or any other very high bitrate setting of any mp3 encoder - Lame 3.98 even does a better job than many other encoders).
I'm into non-pre-echo problems because I'm not sensitive to pre-echo. Harpsichord music provides special problems to mp3 even at highest bitrate.
Anyway Lame 3.98.2 is doing fine, and -V1.5 yields non-annoying errors to me in case it's not transparent. -V0 does a better job on the problems mentioned though isn't totally transparent in any case.

Unfortunately it's also about problems that did arise in my real world encodings. My personal specific experience of subtle (!) Lame errors even at -V0 was with french woman singers. I was able to abx it and a sample I published was confirmed by AlexB. It was 'Là Où Je Suis Née' from Camille's album 'Le sac des filles'.
The problems are very subtle though with -V0 and also -V1.5, and I simply encode emotionally important music with lossyWAV or Lame -V0, and the vast majority of music with Lame -V1.5. If I weren't short on storage space with my new DAP I'd use -V0 throughout when not using lossyWAV. It's just a little bit more care-free, and takes 227 kbps on average for typical music of mine when using a 17.5 kHz lowpass. Not extremely much IMO.

BTW I was one of those who gave a '0' problems to the AAC encoders. I don't have real experience with them but was obliged to do a choice.

This post has been edited by halb27: Apr 2 2009, 18:24


--------------------
lame3100m -V1 --insane-factor 0.75
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: 29th July 2014 - 03:52