What -q value and what build are necessary to reach -aps quality?

1. search the forum. also see this.
2. listen for yourself, nobody can tell you what aps quality is for you but your own ears.

I did. In fact, this may be my first post but I've been reading the forum for over a year now. From my experience with ogg I can tell for sure that it clearly beats mp3 at 128 (didn't realize it untill I got better speakers, though).
I wasn't asking for savant explanations, your or others' personal opinions would be rather welcome instead.

maybe you can post a POLL ...

What Q level equals LAME APS for YOUR hearing?
Q9
Q8
Q7
Q6
Q5
Q4
Q3
Q2
Q1

this is only an indication, not a real fact:

During the testing of autuv, i remember that a regular with very good ears on these boards "approached his limits" at around q6-q7....... however, this was with extreme problem-samples, so something in between q6-q7 is probably similiar to preset extreme.

- Lyx

I've used -q6 for a long time now and swear by it as good enough, well except for my absolute favourite album which gets the rare -q7 encode just because

I've yet to perceive an artifact and cannot ABX -q6 from the original. I could probably get by with less but I'm very comfortable with the bitrate I'm given from -q6.

Good idea. Done.

Nothing much original, since the May 2004 128 kbps multiformat listening test's proved that abundantly.
And nuff said, but you're talking Vorbis, actually. Ogg is just the container.

For me, Vorbis -q3 and --aps are both transparent. Vorbis -q3 is smaller, so I'd choose -q3 over --aps any day.

Are you serious on this? Cause if Vorbis is transparent @ -q3 (110 kbps or so, right?) to you, then I'm guessing LAME --alt-preset medium would be as well (and even -V5 perhaps), to your ears. In which case opposing Vorbis -q3 to LAME --aps doesn't seem like a fair comparison to me.

Let me put it this way. For me, Vorbis -q3 and lossless sound identical. Vorbis -q3 is smaller, so I'd choose -q3 over lossless any day.

K. That's starting to sound a little less biased. Have you actually tried ABX'ing Vorbis -q3 to original PCM? I regret saying that I'm starting to doubt your authority on Vorbis if you claim that -q3 sounds transparent to you.

Edit: left out the quote.
Edit 2: don't mean to sound harsh whatsoever. It just seems so incredible to me to consider -q3 transparent

Yes. I can successfully ABX -q2 *sometimes*, but I fail all the time with -q3.

Why should my ears affect my authority? That sounds very funny.

I realized when I tried to participate in Robertos 128 kbit listening test, that I couldn't distinguish 95% of the samples. Sure, I don't have high/end equipment and I didn't find a totally quiet room, on the other hand, those were all exceptional "problem samples" and I had to concentrate a lot.

The conclusion for me was/is: if you use a modern codec (such as vorbis) everything that is way above 128 kbit is a waste of bits. Consequently, my whole collection is ripped in -q4. Rest assured, for the average joe that is perfectly ok.

PS: what the heck, I used to rip to -q3 before and was fine , so I'm with kjoonlee there

Hmmm, the problem samples statement seems a bit exaggerated to me. They were not intended to serve as problem samples, anyhow.
And BTW, your not being able to distinguish between the various 128 kbps samples themselves is not that surprising. You didn't get to compare them to the original WAVs, did you? If you would've had that opportunity, I'd be really surprised if you wouldn't have been able to discern between them then.

If those were plain casual music samples I stand corrected. Each sample included also a hidden reference, which i failed to identify. So, in fact I compared the original with the compressed version. I don't know if you have tried already, it is more difficult than one might think. Personally I found that test to be an eye-opener.

(needless to say that I didn't mean to identify mp3, aac whatever directly)

You must have heard this a lot of times but I'll repeat it anyway: careful on what hardware (speakers/headphones/souncard) you base your conclusions when you claim uselessness of higher bitrates.
For poor speakers 128 and 256 will most of the time sound the same so this will lead to wrong conclusions.

In view of this I kindly suggest that anyone posting their impressions/experiences in this topic also give a short description of used hardware.

One 'problem' if you can call it that with most listening tests is that they use pieces of music that not everyone knows. The better you know a track the more likely it is that will be able to ABX at a given level. Or conversely trying to ABX between the original and a compressed version of the same track you dont know will be harder than if you knew the piece well.

The problem with this poll is that the version of oggenc is not specified. Various listeners have found verifiable improvements by the 1.1RC1 and megamix2 builds over the 1.0 release. So whereas you may need -q6 or -q7 with 1.0, you might need a lower setting with one of the newer encoders since, for example, they are employ short blocks more readily at lower quality settings.

Good point.
I answered the question with libvorbis 1.1 in mind.

There may not be anything wrong with your ears, physically, but you would probably benefit to train them for critical listening. Hearing comes naturally, of course, but critical listening is actually a learned skill. I can understand why you say you fail all the time with -q3. My sister downloads 128kbps Blade MP3s and never complains about the sound. Her hearing isn't bad, she just doesn't really listen. (Of course, -q3 is much better than 128kbps Blade!)

Although we may all scoff, there is a reason why Microsoft and Sony claims "near CD-quality" for 64kbps WMA and 48kbps ATRAC3 plus, respectively. It's because a suitable part of the population isn't focussed on serious listening...if it sounds somewhat like the original than it must be "near CD-quality".

ff123's site is a good place to start training your ears for serious listening. Again, there is nothing "wrong" with your ears, you simply need some more experience.

I agree with the both of you.
With aoTuV beta 2, I have no problem whatsoever to discern between a -q3 Vorbis file and the originating lossless one. And that's on my cheapo 20 EUR or so PC speakers, without even having to try hard. I'm not bragging or anything, I just consider it very questionable if people claim they can't tell the difference between -q3 Vorbis and the orignal WAV, especially in blind listening test environments.
-q4 is quite a bit more problematic to tell for me. Without headphones, so on my 2.0 PC speakers, aoTuV beta 2 -q4 sounds near-transparent to me.

It is my personal belief that unless you're using *really cheap* speakers or earbuds, you're fine. I tend to view speakers cheaper than approx. 30 EUR or earbuds cheaper than 10 EUR to be not worth using at all.

The speakers I used were EASTERN HF-B34, manufactured at around 1994 in Korea. Consumer price at that time was at around 84,800 KRW. The current rate for 84.8k KRW is around 60 EUR.

The earbuds I used were Leemax (Cresyn) AXE-555 and AXE-599 earbuds, manufactured very recently. They cost around 10 EUR.

edit: The Vorbis encoders used were RC3 and 1.0.1.

I prefer to think that ignorance is bliss when it comes to artifacts and magic tricks. (In other cases, however, I hate that saying.)

edit: And yes, I've tried training before. That's what let me ABX -q3.

Another name for lossy audio compression is psychoacoustic audio compression. Lossy audio coding is also called perceptual coding. I'm sure you are aware of this.

My perception is different from yours, so I use a different setting to encode perceptual audio. What's so questionable about that?

Right, I might have failed to point it out, but the poll is rather addressed to users, so I'd suggest we stick to the encoders recomended in this forum i.e. aoTuV b2, Post 1.0.1 CVS, GT3b2 and as a means of comparison lame -aps.

Ok, I encoded a 26 track album with the following results:

1. mp3 (lame 3.96.1 -aps) 61.3 MB average bitrate 170
2. Vorbis (Post 1.0.1 CVS -q6) 62.6 MB average bitrate 174
3. Vorbis (aoTuV b2 -q6) 64.1 MB average bitrate 178
4. Vorbis (GT3b2 -q6) 71.1 MB average bitrate 197

I do not wish to draw any conclusions; the truth is I was looking to find a setting that would match lame's -aps and consume less space but it turns out I should stick to mp3s...
I will try with other CD's with different types of music.

I actually reached that same conclusion (for the same reason) when deciding what to use for my portable.

why you gave ranges? Q5 is dif than Q6 ... and the max of 10 allows you to ask for 10 options ...

Vorbis Megamix 1 around -q5 is similar to LAME 3.90.3 --preset standard IMO, based exclusively on my ears. Old soundcard and old (but not so bad) amp and speakers here.

If it's file size comparisons you're interested in, try reading this article. Sometimes you can use vorbis at -q 8 and above and still end up with files smaller than lame -aps.

In case you're wondering, I use FLAC exclusively for my music. Not because vorbis at -q 3 or -q 6 aren't good enough; I've listened to quite a bit of music at -q 0 with no complaints. But regardless of how the music sounds, only lossless helps cure paranoia.

i hate to point out the obvious here but that article used lame 3.90.3.
from my own rips, 3.96 encodes around 5-8 megs smaller per cd than 3.90.3 or gt3b2.

Forget to mention this in my earlier post but the oggenc I use is Megamix II. I trust it as I've not really heard any artifacting at -q6. Not to say a problem sample wouldn't do it, but my music just isn't made up of problem samples.

If I hadn't, someone might have complained about not providing ranges... , not everybody uses round values; suppose -q4.67 is the setting you found to be your sweet spot and you can only vote -q4 or -q5. I don't like polls with restrictive options.

I hope everybody will agree with me when I say that the purpose of lossy encoders is to give the best quality possible at the smallest file size possible. Of course I am not talking about archiving, that is the realm of lossless encoders, I am talking about the best quality you can get from your encodes so you can have your music collection handy on your mobile, car or standalone player without having to switch on the computer.

Here are my findings for the second album:

1. mp3 (lame 3.96.1 -aps) 68.7 MB average bitrate 175
2. Vorbis (Post 1.0.1 CVS -q6) 75.8 MB average bitrate 193
3. Vorbis (aoTuV b2 -q6) 78.9 MB average bitrate 201
4. Vorbis (GT3b2 -q6) 90.5 MB average bitrate 231

Interesting, so judging from size only GT3b2 should really sound better than lame -aps...

Um, you can't infer quality from filesize, can you?

altu, have you actually tried ABXing your encodes?

aoTuV b2 at -q4 works for me. I can't distinguish most, if not all samples from the original. I cannot with APS either....so, for this pole I choose Vorbis aoTuV b2 @ -q4. I'd also choose this quality setting over LAME at 128kbps cuz I can distinguish that bitrate from the source.

i have to claim that with some samples ogg vorbis delivers better quality already at -q4 than lame --aps. i have quite a problem sample which is the classic jean michel jarre recording "equinoxe IV", which shows quite strong artefacts even with lame -aps, wheras vorbis at -q4 sounds much closer to the original.

as i was new to ogg and was also interested into buying a minidisc-player i wanted to compare ogg with atrac and mp3, you can find my very amateurish "test" here:

http://www.austinv.com/forum_viewtopic.php?30.3465

my conclusion was that with most material ogg vorbis already sounds transparent to my ears at -q4, with some tunes even at -q3, wheras all other codecs fail at this bitrate. therefore i have to defend kjoonlee if he uses -q3 and claims it to sound good, because it actually does. i have many recordings where -q3 or even -q2 would be enough. it depends on the material. older recordings from the 70s or 80 usually encode easier, because they are not that overprocessed like modern recordings.

my aim was to find a bitrate that sounds good, but also saves space, and only vorbis could fullfill this! later on i also compared mpc at this bitrate, because it was claimed to sound "best", but to my ears vorbis sounds better.

i have a short sample online, one is encoded with vorbis -q4, the other one is original wav. can you tell which is which? (of course without looking at the spectrographs!)

http://homepage.boku.ac.at/h9526506/temp/1.wav

http://homepage.boku.ac.at/h9526506/temp/2.wav

best regards,

o.

Isn't that the truth. I started off with LAME -aps, then moved to LAME -ape for a bit. Then I jumped to MPC, then WavPack at 384 kbps with the -h switch. Eventually I ended up with FLAC, and then Monkey's Audio for the slightly better compression. I really tried to convince myself that WavPack at 384 kbps would be plenty enough for transcoding should I need to, and that based on the tests I've read here by other users, should be audibly lossless (note I didn't say mathematically) for 99.99% of all music. Despite that it takes up a little less than half the space of lossless files, I just couldn't do it. I'm paranoid by nature. It's an annoying personality quirk. I was originally worried about the disk space that lossless would eat up, but so far even with having re-encoded my entire collection of CD's, it's not so bad and I still have tons of free space on my new 200 Gig HD. And being able to transcode has actually be great, not to mention conveniant, for testing different lossy codecs.

FWIW, I recently did do some casual ABX testing on some of the genres I like. I can tell the difference between the original and compressed file below 128k no matter the codec, though personally I thought Nero AAC was pretty decent (artifacts were least annoying). With the little bit of testing I did directly at 128k, I could usually tell the difference between a LAME file and original, but aoTuVb2 at -q4 was very very difficult for me. Since I plan on purchasing a Neuros, and my wife probably won't be able to notice anything amiss (I'm the picky one), aoTuV at -q4 or there abouts is probably what I'll end up transcoding everything to for the portable. For anything that doesn't support Vorbis (for example the DVD player we own which handles MP3 and WMA), LAME -aps will most likely be my choice. I'm kind of glad actually that -q4 sounds good because it means I won't have to worry about the decoding issues that tend to come with higher bitrates when using a portable.

i just thought it was worth pointing out that certain artefacts like pre-echo, and overquantization in the wrong bands is audible on even the cheapest of speakers.

i have some vorbix files that are a staggering 270kbps and yet i could hear artefacts even in spite of an ear infection rendering me almost completely deaf. why? because the artefacts were in the 3-4Khz band, a VERY unwise place to put them.

NOTE: this does not say anything on vorbis. i suspect that whoever encoded these files hadn't a clue what on earth they were doing, and probably "tweaked" the encoder at the command line, making a complete mess. it is for this reason that i eventually bought the album

i haven't voted in this poll because i honestly haven't done more than a casual comparison between vorbis and --aps. I'm a musepack man myself - to my ears subband distortion sounds better than transform distortion (one type adds narrowband distortion, whereas the other type actually removes certain frequencies, leaving "holes"). of course, this isn't true for mpeg layer 2, which is bloody awful at musepack bitrates.

Q4.99

I couldn't ABX that on The Verve's - Bittersweet Symphany, so decided that was good enough for me. Good space savings over MP3's @ -APS, while providing transparent CD quality (to my ears at least). I archive everything to FLAC, and then use OggDropXP to make Vorbis Ogg's.

For me, vorbis 1.1RC1, quality: -q6 is total transparent for music in my headphones, based in my ABX tests.