Help - Search - Members - Calendar
Full Version: newbie: my personal quick ogg observations
Hydrogenaudio Forums > Lossy Audio Compression > Ogg Vorbis > Ogg Vorbis - General
oluv
i am quite new to ogg vorbis, to tell the truth i tried it yesterday for the first time, and i was fascinated by the quality it already achieves even at 4 or lower. before i was using lame at 192 to 256kbps.

i tried to compare some of the different ogg builts to hear what the differences are between them.

the first thing i found is, that the original 1.0.1 and the GT3b have a higher frequency response compared to the others, but also have a strange high-frequency amplification. at -q4 they reach around 21khz when looking at the spectrograph.
now aoTuv and megamix2 which i also tried, do not have this HF amplification, but there seems to be some low-pass filter, as at the frequency-response reaches only 19khz at -q4, besides the files-sizes of megamix2 are bigger than all the others at the same quality rate. i don't know why in the latest builts the highest frequencies were cut off, compared to the original ogg.

i have no idea what changes and improvements have been made between the codecs, but i would love to know what the things are now. which codec includes the features of the others etc. and which is up-to-date?
is there some "official" release that is gonna come, or do i have to select between all the different ones?

i have read some things about pre-echo and other problems. which codecs suffer from this problem? i couldn't hear this with my own music-samples, maybe someone could give me a hint which particular sounds and frequencies are affected or give me a link to sample-files.

my goal would be to use the lowest possible setting and get still decent quality. until now i noticed that 128kbps is already very good, and with some material i could even go down to 96kbps, which i wouldn't do though. i could use 8 or higher, but my concern is the file-size. i do not want 1:1 cd-quality. if i want CD i listen to CD. i only want to archive most of my cds on a portable rig to listen to them on the run at "near" cd-quality.

besides i have also a question: i use CDex for encoding and until now i was dependent on the dll-libraries, as i couldn't get an external encoder to work within CDex. if i link to oggenc.exe i get an error: Cannot send data to external encoder.
maybe i am missing some setting, but i tried nearly everything. i would appreciate any help.

thanks for any response and clearing some things up for me smile.gif

best regards,

o.
QuantumKnot
QUOTE(oluv @ Aug 12 2004, 08:07 PM)
i am quite new to ogg vorbis, to tell the truth i tried it yesterday for the first time, and i was fascinated by the quality it already achieves even at 4 or lower. before i was using lame at 192 to 256kbps.

i tried to compare some of the different ogg builts to hear what the differences are between them.

the first thing i found is, that the original 1.0.1 and the GT3b have a higher frequency response compared to the others, but also have a strange high-frequency amplification. at -q4 they reach around 21khz when looking at the spectrograph.
now aoTuv and megamix2 which i also tried, do not have this HF amplification, but there seems to be some low-pass filter, as at the frequency-response reaches only 19khz at -q4, besides the files-sizes of megamix2 are bigger than all the others at the same quality rate. i don't know why in the latest builts the highest frequencies were cut off, compared to the original ogg.


The original 1.0.1 and GT3b1/2 lowpass filter at 20 kHz while this value was changed in aoTuV beta 2 (which is the base of megamix 2, 1.1RC1) down to 18.9 kHz. Whether this decrease of 1 kHz is audible is a matter for blind listening testers smile.gif

Just a note that any claims of quality need to backed up by blind listening tests in order to be treated seriously here. It's the only accurate way of removing the placebo effect. smile.gif

QUOTE
i have no idea what changes and improvements have been made between the codecs, but i would love to know what the things are now. which codec includes the features of the others etc. and which is up-to-date?
is there some "official" release that is gonna come, or do i have to select between all the different ones?


The original 1.0.1 and GT3b1 had some issues with HF hiss/boost on sharp attacks as well as some coarse grainy sound heard in classical music. GT3b1/2 has less pre-echo at q 5 and above, though this is at the expense of 'wilder' bitrate jumps.

aoTuV beta 2 attempted to reduce these HF problems which have been verified in some recent listening tests. aoTuV beta 2 does not include the aggressive pre-echo tunings of GT3b1/2, hence we have Vorbis Megamix which combined aoTuV beta 2, QKTune beta 3.2 (pre-echo improvement at q 2 to 4.99), and GTune beta 2 (pre-echo improvement at q 5 to 10). However, while I was doing the merging, I left out something from GT3b2 which didn't bring out the pre-echo tunings affecting microattack samples. So Vorbis Megamix 2 came about, based on the new 1.1 RC1 from Xiph.Org (which was based on aoTuV beta 2 + bug fixes and tweaks), which includes better microattack handling.

The other versions include the stock 1.1 RC1 encoder plus some additional advanced switches which allow you to "do your own pre-echo tuning". This gives you more choice and control over the compromise between bitrate and quality. Do a search of the Ogg Vorbis-Tech forum to find this encoder with the two special switches.

QUOTE
i have read some things about pre-echo and other problems. which codecs suffer from this problem? i couldn't hear this with my own music-samples, maybe someone could give me a hint which particular sounds and frequencies are affected or give me a link to sample-files.


For pre-echo, look for a sample file called castanets.

http://lame.sourceforge.net/download/samples/castanets.wav

In order to train your ears to this particular artifact, encode castanets at q 0, and try to do an ABX listening test and see if you can recognise the noisy smearing. Then you should be able to pick it out in the higher q's.

As for what versions deal with pre-echo, GT3b1/2, Megamix, Megamix 2, and 1.1RC1 with the special switches. In light of HF hiss problems at medium to low q's, I suggest considering the last two. smile.gif

QUOTE
my goal would be to use the lowest possible setting and get still decent quality. until now i noticed that 128kbps is already very good, and with some material i could even go down to 96kbps, which i wouldn't do though. i could use 8 or higher, but my concern is the file-size. i do not want 1:1 cd-quality. if i want CD i listen to CD. i only want to archive most of my cds on a portable rig to listen to them on the run at "near" cd-quality.


According to the recent 128 kbps listening tests, which compared Ogg Vorbis with lame MP3, iTunes AAC, Musepack, and WMA, Ogg Vorbis tied with Musepack as the winner. So one would think q 4.35 or even q 4 would be a nice place to start encoding. smile.gif

QUOTE
besides i have also a question: i use CDex for encoding and until now i was dependent on the dll-libraries, as i couldn't get an external encoder to work within CDex. if i link to oggenc.exe i get an error: Cannot send data to external encoder.
maybe i am missing some setting, but i tried nearly everything. i would appreciate any help.


To rip 'on-the-fly' using an external encoder in CDex, the encoder needs to read from standard input. So for oggenc, use '-' for your input

CODE
oggenc -q 4 - -o %2


You can also add in the other switches to set tag info like artist, album, year, track num, etc.
oluv
many many thanks!

i was already afraid that i would get an answer like: search the forum smile.gif

well i did some research, but it is quite hard to follow as there are so many different builts around. i was missing something like your summary that explains what have been done here and there, and what is the state by now.

QUOTE
The original 1.0.1 and GT3b1 had some issues with HF hiss/boost on sharp attacks as well as some coarse grainy sound heard in classical music.

well i hear it more like a continuous emphasis in the higher frequencies, as if someone would have played with an EQ and about 12 or 14 khz. maybe we are talking about different effects.

i will take megamix2 into consideration. but the bigger file sizes at the same quality-setting compared to the other codecs put me off a bit. as i remember megamix2 produced the biggest files.

also thanks for the castanets link i will try this out. i think i know which effect is meant.
before i also played around with atrac a lot because i primarily planned to buy a HI-MD recorder. but atrac3 132kbps seems to have serious problems with attacks as they are smeared. maybe it is not what is considered as "pre-echo" but i think i know what i should be looking for.

i will try your cdex settings for oggenc.

thanks a lot and best regards,

o.
Lyx
don't forget the blind-listening part. You will only betray yourself if you just compare by simple listening - besides most ABX apps out there also offer a nice UI to make comparisions easier and more comfortable than when doing it manually.

- Lyx
oluv
now i made some ABX tests with the "famous" castanets-sample. i tried it only at -q4 and there is really a difference in the attack compared to the original. at -q4 megamix2 seemed to have the sharpest attack and performed best, although it was still distinguishable from the original, but it also had the biggest file-size as already mentioned. i didn' try it on higher rates.

but i have the impression that with most of music the compression artefacts are not so obvious. i did some tests with "normal" music at -q4 and it was really hard to hear any difference. i normally concentrate on the music as whole, not on a 4 second sample, that i repeat again and again. i am not very nitpicking. the music as a whole is important for me.

of course for you it is different, as you are working on the codec and want to improve it.
i am only looking for a tool that is able to reproduce my music in the best way, by using the smallest size possible. and i came to the conclusion that with some piecies even q3 could be enough if. i will continue to try it out on further tracks but until now it looks as if i would stick to -q4.

i only wanted to mention that even -q0 at 64kbps ogg is very "listenable". i will probably use this setting to encode some music for my girlfriend that she can listen to it on her smartphone. she didn't notice much difference to the original wink.gif
and what i like about ogg is that at 64kbps it still tries to stick to the original regarding frequency-range. all the other codecs wma, mp3, atrac etc. block the higher frequencies much more at this bitrate and even at higher rates.

i wanted to congratulate all of you for developing this wonderful tool and having brought it so far. i love ogg and i love you all laugh.gif no, just kidding! im not jacko, although i love to use his tracks for comparing the quality. quincy jones and bruce swedien had also done a great job on his recordings!

best regards!
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.
Invision Power Board © 2001-2008 Invision Power Services, Inc.