[Distroyed]

Quote:

Originally posted by blessingx I'm curious, why? Understandably for extremely quite music, VBR schemes without reasonably high minimums can have problems, but with LAME's presets this isn't really an issue in most cases, and certainly fine for most genres (with less rock being the exception than classical for the opposite). In fact most newer codecs seem directed towards VBR. Anyway curious why you feel that way?

I was trying to say that VBR is fine for what he's using it for. I only recommend the avoidance of vbr when it comes to, as you said, low volume music, which is generally only encountered in the less active electronic genres and some soundtracks and the like.

 

[Stephonovich]

If you just up the minimal bitrate up to 128 or 192 or something, you'd be fine for even the most demanding music. Well, no. You'd be better off using MPC, or just FLAC and be done with it

(-:Stephonovich

 

[tigre]

Stephonovic:
Seems like your results are quite interesting. You say you've ABXed.

Does this mean you have performed ABX tests successfully ('Probability you are guessing' <= 0.01 / 1%) in every single case you mentioned?

If yes - what did you ABX?
- Original vs. aps
- Original vs. aps modified (-m s, -k) and/or
- aps vs. aps modified

Another question: What's your setup:
What's used for decoding (replaygain used to prevent clipping, resampling used) and for playback (Operating System, ABX software and its output settings, Soundcard, Amp, Headphones/Speakers, anything else?)

 

[Stephonovich]

Less than 1%? Uh, no. That takes, what 7 runs, or something like that? Most I did was 3 runs. I figured if I could ID it in different parts of the song 3 consecutive times, there was a difference. Also, I figured the people here wouldn't mind as much as HA.org members. Geez, you better have at least 5 ABX runs and everything perfectly documented if you don't want to get flamed...

Anyway, first time, I did --aps vs. --apfs. After determining that --aps was the better of the two (albeit slightly), I did --aps -m s vs. --aps, which defaults to joint stereo. Again, I took the winner, (--aps -m s) and did that vs. --aps -m s -k. My final conclusion was that --aps -m s -k gave the best results, with about a .5-1MB size increase.

As for setup, I replaygained all the files to 89dB before testing, and no, I didn't resample. For one, I've got an SB Live! which can only go to 48KHz, and two, I honestly can't see how interpolating for data that was never there is going to imrpove sound quality. If I ever upgrade my soundcard, I'll try it, but for now, call me an unbeliever. I'm running WinXP Corporate (Professional for all purposes, except with no activation required), and I used WinABX for the testing, set to ABXY, it's default mode. My soundcard, as mentioned, is a SB Live! X-Gamer, with the latest KX Audio drivers, and headphones (Sennheiser HD 280 Pros) plugged into the rear out. I have a CMoy amp, but due to it being in pieces, it wasn't used

If it matters, I used EAC on secure mode to rip all music used, and encoded into FLAC with --verify turned on.

As I've mentioned, I don't expect someone to take my results as a Bible, but it may help someone who's making a casual decision. If you want nitty gritty, go to HA. I lurk there, as there's a wealth of knowledge available.

(-:Stephonovich

 

[tigre]

Stephonovich, thanks for taking the time to answer.
Quote:

Originally posted by Stephonovich Geez, you better have at least 5 ABX runs and everything perfectly documented if you don't want to get flamed... <snip> If you want nitty gritty, go to HA.

I hope you didn't get the impression I want to flame you - not my intention. Let me explain why I asked:

Your results (about adding -m s and -k) are different to my own results from tests I performed some time ago and to what I've read from others. Anyway, I believe it's possible that there are differences like the ones you describe, therefore I was interested in repeating your tests myself. To decide if it's worth the time/effort for me I wanted to know these details.

Quote:

For one, I've got an SB Live! which can only go to 48KHz, and two, I honestly can't see how interpolating for data that was never there is going to imrpove sound quality.

Most Creative cards like SB Live! resample internally all 44.1kHz input to 48kHz before D/A conversion in a poor way. The resulting distorsion could have influenced your results, so decoding to (= resampling to) 48kHz using high quality software like SSRC or foobar2000's resampler+diskwriter would result in better quality and more reliable tests.

 

[Stephonovich]

No, I didn't take it as a flame. Maybe a bit towards HA... I was just pointing out the differences. I've read threads where someone posts something similar to what I did and they get brutally attacked with "your test results are useless", and told to go read the TOS before they get banned.

Really? What results did you have, or others? Any threads/sites you can give me? I'd be interested in reading about that. To me, though, it makes sense that anything that has the possibility of reducing quality (a lowpass filter or joint stereo) would conceivably do so at some point in the file. I mean, even VBR messes up every now and then for really complex passages.

In reading, KX Drivers can supposedly bypass the resampling. I found something called AC3 Passthru, where you can select SPDIF 1, 2, or 3 as the output. I would presume this might work for speakers hooked into the digital output of the soundcard, but as I use headphones (which aren't hooked into that), it's useless.

I tried using SSRC, BTW, but ssrc_hp just crashes (in WinXP), and ssrc claims to have made a file, but it ends up being just a few KB in size, hence, not containing any music.

I suppose I should have used Foobar, as it has ABX... never occured to me. How much can the resampling effect the sound quality? Perhaps I'll get Foobar and do a couple quick tests with resampling on and off to see if I can hear anything...

(-:Stephonovich