[Packgrog]

I'm considering switching from my usual LAME 3.97 VBR 0 settings to Ogg Vorbis q7 or q8. This site's test results made this seem particularly intriguing. I encoded a few of my favorite CDs with LAME -V0 --vbr-new, Ogg Vorbis -q7, -q8 (and one album with -q9 and -q10 as well, for curiosity sake), and FLAC 1.1.4 -8. I haven't started doing listening tests yet, but I found the file size results fascination. Nine times out of ten Ogg -q7 was substantially smaller than LAME -V0. Most of the Ogg -q8 was about the same size as the VBR 0 LAME file. I found this particularly interesting given that the target bitrate for -q7 seems to be the same as -V0.

I'm curious to know, though, what everyone else's Worst Case Scenarios might be with these various settings. In what specific songs can you detect differences? What are those differences? Are the lossy, compressed versions preferable (ie: lossless too bright)? And what genres seem to benefit best for you with which codec and compression level? Which ones don't seem to benefit at all from higher quality settings?

I'd really rather this not turn into a flame war between ABXers and non-ABXers. If you think you notice a difference between one setting and another, I want to hear about it, because if I have the CD, I intend to try it myself! My hardware is almost certainly different from yours anyway, so maybe I'll hear the same thing, and maybe I won't. Some starting points would be nice, though. I'm certain that 99% of the time no one will be able to discern any differences between any of these settings. I'm just interested in that last 1%, so I can determine if it would be an issue or not.

Looking forward to reading your experiences!

 

[Packgrog]

Initial impressions with Porcupine Tree - Blackest Eyes (from In Absentia):

Ogg with -q7 and FLAC both seem to have more sparkle in the cymbals than my -V0 --vbr-new LAME 3.97 MP3. All of these were ripped last night. Subtle, but surprising. I'd thought (based on older tests I'd read) that LAME did a better job of representing higher frequencies. Could be placebo effect (I do need to ABX this), but there does seem to be a subtle difference.

Same thing with Trains from the same album. Ogg with even just q7 seems to reproduce subtle details just a little bit better. Cymbal crashes are more extended and lively, and the strumming of the acoustic guitar also seems more lively.

Again, this isn't a blind test, so take this with a grain of salt. That said, my setup is fairly revealing, and is as follows:

iRiver H120 w/Rockbox -> Monica 2 DAC -> Tomahawk -> Atrio m5

Anyone else care to try these examples as well?

 

[majkel]

I'd compare LAME 3.97 V0 or --preset-extreme to something between Q6 and Q7 for ogg. Q8 is even better than mp3 320kb/s because of better soundstage reproduction. Also, Q8~Q9 is a threshold I wouldn't distinguish between ogg Vorbis and lossless. Regarding target bitrates, ogg is the winner. Reggarding battery life - ogg is by 1/3 more power hungry.

 

[Packgrog]

Good to know! Can you give me an example of a song where the difference between -q7 and -q8 is noticeable?

As it stands, VBR 0 was excellent, but -q7 seems just that tiny bit better. And I can't really tell any difference between -q7 and FLAC at this point.

Interesting to know about the difference with Ogg power usage. Hmmm... Will have to keep an eye on that...

 

[majkel]

Not all players and sources let you hear the difference between q7 and q8. The only existing is warmth of the midrange, being full starting from q8 and a bit lacking color at q7 but again, some players sound too "grey" to reproduce such nuances. You can use Katie Melua's voice or Papa Roach - Infest guitars to follow the differences between bitrates. Headphones I know could work: Sennheiser PX100, Sennheiser HD 595...

 

[HFat]

These aren't *my* experiences but you should check the listening tests on hydrogenaudio.org.
In particular, as you seem to be ripping to lossy as opposed to transcoding to lossy, you might want to check guruboolez's latest transcoding test in which Ogg was the best performer (barely) and LAME the worst (by a long shot). If you ever need smaller files (to load on a flash player or something), you'll be happy to have a decent transcoding source.

EDIT: and yeah, a lowpass is often a good thing... Jan Meier (used to) put some kind of lowpass tone control on his amps.

 

[Febs]

Quote:

Originally Posted by Packgrog /img/forum/go_quote.gif Ogg with -q7 and FLAC both seem to have more sparkle in the cymbals than my -V0 --vbr-new LAME 3.97 MP3. All of these were ripped last night. Subtle, but surprising. I'd thought (based on older tests I'd read) that LAME did a better job of representing higher frequencies. Could be placebo effect (I do need to ABX this), but there does seem to be a subtle difference.

At the risk of stating the obvious, you might want to use ReplayGain to ensure that the files are the same volume.

 

[Packgrog]

Quote:

Originally Posted by Febs /img/forum/go_quote.gif At the risk of stating the obvious, you might want to use ReplayGain to ensure that the files are the same volume.

Not sure that's necessary given my current setup. I'm using the optical output of the H120 via Rockbox and comparing the same songs ripped from the same CD. All rips in this case were done last night on the exact same setup with EAC. The only difference should be the codec. And unless either the encoder altered the volume or the Rockbox decoders for MP3 and OGG somehow output at different volumes via the optical output, that should be an issue. Volume levels seem identical to me.

Besides, as far as I know, you folks still haven't implemented the ReplayGain provided by the LAME encoder (something to do with APE tags? Don't remember).

Trying to be careful to compare apples to apples here. And loving the Rockbox & optical output combo. Awwwww yeaaaahhh!

 

[Packgrog]

Well, looks like for the time being Ogg -q7 is the way to go. While running the mix of files (FLAC, Ogg, and MP3) on shuffle, I couldn't really tell the difference. FLAC and Ogg were sometimes a little more detailed in the high end, but only if you're REALLY listening for it.

That alone isn't QUITE good enough a reason to re-encode all of my CDs (and I have enough that this would be tedious), the prospect of saving 20 MB per album while maintaining or even improving the sound quality of my files absolutely *IS* good enough reason. I currently have just over 60GB of audio files. More than half of that is MP3s from my own CDs (and yes, if I like a downloaded album enough, I buy it, as most of my downloads are for trial purposes, and most people rip are lousy bitrates), so that could be a pretty significant space savings. NICE.

Once I get to my handful of classical and jazz albums I'll do some tests with -q8 to see if there's any noticable improvement. Given that most of my music is metal or progressive rock or electronic, and I just can't tell the difference between -q7 and LAME -V0 on most of this stuff, I think I'll stick with -q7 for the most part and save the extra drive space. Awesome.

 

[Packgrog]

Quote:

Originally Posted by majkel /img/forum/go_quote.gif Not all players and sources let you hear the difference between q7 and q8. The only existing is warmth of the midrange, being full starting from q8 and a bit lacking color at q7 but again, some players sound too "grey" to reproduce such nuances. You can use Katie Melua's voice or Papa Roach - Infest guitars to follow the differences between bitrates. Headphones I know could work: Sennheiser PX100, Sennheiser HD 595...

Ugh. Listening to some of the -q7 files I ripped, I was noticing that the vocals of Tori Amos and Sarah McLachlan sounded ever so slightly "digital". So I decided to try -q8...

With even just Tori Amos' Crucify, the cymbals sound more relaxed (not as sparkly, but more natural), and the soundstage seems more natural and spacious. Bwah!? Believe it or not, I'm actually disappointed that I'm hearing a difference, as -q8 files are about the same size as my old -V0 MP3s, so I won't end up saving drive space after all. The sound quality, however, will be nicely improved. I'm not sure I would notice any improvements provided by -q9 or FLAC (and I pray I never do), so I'll run with -q8 now. Meh. And I'd ripped so many CDs to -q7! Now I have to redo them AGAIN!

 

[majkel]

IMHO total SQ including warmth, soundstage and spectrum reproduction goes like this: mp3 V0 =< ogg Q7 < mp3 320kb/s CBR < ogg Q8. So it is still worth spending some bits to have better quality and lower file size than --preset-insane LAME mp3.