PDA

View Full Version : .OGG VBR Quality 10 ~400kbit/s lossyness analysis


spontabmark
2005-03-27, 07:07 PM
could be me being stupid and noobish but:

i was messing around with convert a lossless file into different lossy formats and back and analysing them in EAC and i did this with .OGG VBR Quality 10 ~400kbit/s and at first it looked identical to the original file in the spectrum and freq, i did a couple of analysis' of lower bitrates to see what to look for and went back to the 400 and just noticed a very slight variable trim at the top of the right channel on the spectrum but the freq was normal, i then did both channels in AnalFreq and saw the dip at end on the right channel. is that all there is to spot .ogg compresion at this level cus i wouldn't spot that ordinarily or are these programs not good enough etc. thanks :)

Five
2005-03-27, 07:14 PM
EAC isn't the greatest for spotting lossy sources. check the mp3 detection link in my signature. I'd love to discuss this further, could you please post some screenshots for us?

spontabmark
2005-03-27, 07:15 PM
the EAC spectrum i did, i just zoomed onto 2 secs and you can see it a bit more clearly than these so not too hard to spot i guess. but goes to show should always zoom on 2 secs :lol

spontabmark
2005-03-27, 07:19 PM
i did this first so i was really confused lol

spontabmark
2005-03-27, 07:25 PM
on zoom can see blocks so you can tell in EAC

spontabmark
2005-03-27, 07:31 PM
and in analfreq i normally only do the left channel as default so wouldn't have noticed

thats all of them Five :D

Five
2005-03-28, 08:36 AM
I just took a look at all this stuff.

One thing worth trying is setting the FFT on the EAC FA to maximum. Then you might see some of what analfreq is showing, altho I think analfreq will probably provide a better analysis anyways.

The SA for the OGG VBR Quality 10 is pretty hard to spot... it looks a little grainier and there's that little dropoff in the right channel as you pointed out.

It might be worth trying some of the other progs like Audacity and Audition (audition can still do SA after the demo has expired). Lossy at the highest quality settings is pretty difficult to spot. The artifacts are the most subtle. When you're using SA it's worth it to do a little zooming in and out to see if the blocks come into focus.

Again, watch out for EAC when it comes to analysis... check this thread (follow it to pg2).
http://www.thetradersden.org/forums/showthread.php?t=1331

The show looked clean in EAC but then when checked with other progs it was very obvious it was lossy-sourced.

There's some more links here...
http://www.thetradersden.org/forums/showthread.php?t=4288

I would love to find a reliable way to spot lossy sources encoded at the highest quality settings. Right now it's really hard to tell, even with the before and after analysis right in front of me.

spontabmark
2005-03-28, 10:39 AM
does audacity only do spectrum? i'm not seeing any difference between ogg and original with it at the moment on spectral

Five
2005-03-28, 11:02 AM
make sure you go to flie > preferences > spectrograms and set the maximum frequency to 22000 Hz. For some reason, the default setting doesn't show the highs and we really need to see those.

Five
2005-03-28, 11:03 AM
make sure you open a decompressed WAV. It will load up a FLAC but the sound isn't right.

here's the button where you can select to view SA:

Five
2005-03-28, 11:07 AM
for fa with audacity, select up to 21.8 seconds (or if you select more, it will just use the first 21.8 seconds), then go to view > plot spectrum.

Five
2005-03-28, 11:12 AM
here's the sa for my example file... this is wma audio from a wmv file encoded with "cd quality". Audacity uses white for no activity, unlike CEP which uses black so it takes a bit of getting used to.

spontabmark
2005-03-28, 11:29 AM
ah thanks for the help, yeah the results are the basically the same as EAC, you can see the dip on the FA but i wouldn't notice it when having no source to compare to and the blocks in the spectrum are the same but a bit easier to see

Five
2005-03-28, 11:38 AM
maybe you should get the trial version of Audition and try that. for my example, I found it hard to see the blocks, but in Audactiy they're there. These are the kind of blocks I'm talking about:

spontabmark
2005-03-28, 12:38 PM
got audition and came up with:

Five
2005-03-28, 01:01 PM
There you have it. You can see the blocks way up top, especially on the right channel just like analfreq showed. Blocks like this would never appear on a pure lossless source.

Because the encoder was set to 10, very little information is lost from the original audio. It doesn't look like it's even touched anything below 20kHz.

The reddish parts are where all the music is. the purplish "fog" from about 6kHz up to 22kHz is 99% noise.