Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.

Poll

mp3pro
[ 44 ] (22.8%)
wma8
[ 21 ] (10.9%)
plusv
[ 4 ] (2.1%)
ogg
[ 124 ] (64.2%)

Total Members Voted: 256

Topic: Poll: least-worst-sounding 64 kbit/s format (Read 12456 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Poll: least-worst-sounding 64 kbit/s format

Ok, it's not quite a "proper" listening test, and it's not blind, (and I don't even provide the reference file!) but it should be quite easy to perform except for the downloading time.  Which 64kbit format do people find least annoying?  wma8, mp3pro, plusv, or ogg (post-rc3)?

Listen to the temporary files at the bottom of:

http://ff123.net/samples.html

Poll: least-worst-sounding 64 kbit/s format

Reply #1
OK, this is somewhat unrelated, but...

I think the RealAudio formats sound the best at "streaming dialup modem" bitrates... especially when you get to 24kbps and below... although WMA probably gives it a good run for its money, and maybe with the latest codecs improves on it.

The only problem with RealAudio is the player... uggh...

Poll: least-worst-sounding 64 kbit/s format

Reply #2
I’m dealing with each part of the track separately.  I converted all four mp3 files to wav, and then chopped the wav files into 6 parts, labeling them "# codec" (e.g., 3 ogg).  So I could more easily compare the codecs on each portion of the clip and compare.  Here are my comments, by part of the song.

Part 1 - acoustic guitar
mpv32 has a fake, buzzy sound which seems to be added.  The others all sound quite similar

Part 2 - synth/techno
mpv32 adds a lot of high pitched fuzz, which is quite annoying.  Its replication otherwise seems pretty good.
wma and mp3pro both make the drums come out with some metallic smearing, they’re a little different but I found them equally annoying
ogg is perhaps a touch brighter than wma and mp3pro, but seems to have better clarity

Part 3 - cymbals, drum, bass
mpv32 - terrible echo effects, sounds like multiple cymbal strikes, when it’s clear from the others that there’s only one cymbal strike at a time
ogg enhances/exaggerates the high-pitched range of the cymbals, so that they stand out more than in mp3pro and wma.
mp3pro smears the cymbals a touch, while wma seems to make them sound more metallic than they should be; it gives them too much of a specific tone

Part 4 - spanish guitar
mpv32 the worst, again... fuzzy with lots of echo.
The rest are pretty much equal, at least on the first half (quieter half) of this section, before the drums come in.  wma maybe a touch duller than others, but that may be because mp3pro and ogg exaggerate the highs.
After the drums come in, ogg really makes the drums sound louder by exaggerating the high snare sound.
mp3pro and wma sound similar to each other, mp3pro probably a little better because wma adds more swishing.

Part 5 - percussion
mpv32 the worst, really exaggerates high frequency
others are pretty close in quality

Part 6 - voice (sitting...)
mpv32 the worst, of course
ogg is actually the worst of the remaining 3 - it makes the voice sound overly metallic and artificial - imagine a female Terminator singing
mp3pro probably the best here, because wma is murkier


Overall, mp3pro and wma usually sound very close to each other.  To my ears, mp3pro sounds better than wma when there’s any noticeable difference between the two.
ogg may be the most listenable for the majority of cases, but it does seem to change the sound from the sourcefile more than do wma and mp3pro.  As Monty has said, listenability versus accuracy... but I think that this post-RC3 version of ogg may make too much of a trade-off where "listenability" is concerned.  But it probably will sound more pleasing than mp3pro in most cases, even though mp3pro seems closer to the source wav file.  But I think that RC4 shouldn’t do quite as much artificial brightness as this version does.

I don’t have the original for reference; that would be nice... then I could tell how much ogg brightened some of the clips.

The plus-V technology doesn’t seem to represent itself very well, at least when applied to mp3.  But a lot of this could be due to the fact that the source wav had its sampling rate converted from 44.1 khz to 32 khz, and then back to 44.1 khz.  This could have a large impact on echo and fuzz, I would guess.  For reference, I converted the file hihat.wav, a 2-second clip that many of you are likely familiar with, to 32 khz wav and 22.5 khz wav, and then converted those back to 44.1 khz.  The 22.5 handled it a lot better than the 32, which makes sense because it’s pretty much a 2:1 ratio.  So perhaps, to be fair to the PlusV technology, you should downsample to 22.5 khz if you're going to downsample.

edit: and from what is said on the website, plusV can supposedly be applied to ogg as well as mp3.  Which might produce better quality at a given bitrate, because ogg is better than mp3 at the bitrate before application of PlusV technology.
God kills a kitten every time you encode with CBR 320

Poll: least-worst-sounding 64 kbit/s format

Reply #3
You are perhaps hearing something in plusv that I don't hear, because I found the 22 kHz version much worse than the downsampled version (which is why I downsampled first).

If there were some easy way I could have used plusv with FastEnc, I would have done that.  Using it with lame at such low bitrates isn't showing its capabilities for mp3 to best effect, I think.

In general I found mp3pro to have lots of pre-echo type artifacting, and ogg to have too much of the fluttering type of artifacting it is prone to.  I have been growing increasingly accustomed, and therefore increasingly annoyed at the flutter artifact.  Also, ogg has an annoying harshness, or maybe it's better termed an increase in the sibilance, of the cymbals in the 41_30sec section.  Still, maybe it's the least of the evils at this bitrate.

ff123

edit:

Quote
[Re plusv samplerate conversion]: But a lot of this could be due to the fact that the source wav had its sampling rate converted from 44.1 khz to 32 khz, and then back to 44.1 khz.


It doesn't actually convert back to 44.1 kHz when it plays.  The encoded file played at 32 kHz (really 16 kHz doubled).

Poll: least-worst-sounding 64 kbit/s format

Reply #4
My connection to your site seem hit and miss - is it just me?

I'm getting lots of: "A connection to the server could not be established" errors.

Cheers,
David.

Poll: least-worst-sounding 64 kbit/s format

Reply #5
Quote
My connection to your site seem hit and miss - is it just me? 

I'm getting lots of: "A connection to the server could not be established" errors.


I've been noticing problems with my server since last night.  Hopefully hostrocket takes care of it today.

ff123

Poll: least-worst-sounding 64 kbit/s format

Reply #6
Quote
Originally posted by ff123
It doesn't actually convert back to 44.1 kHz when it plays.  The encoded file played at 32 kHz (really 16 kHz doubled).


Sorry.  Stupid mistake on my part... when I converted them all back to wav, I resampled the mpv back to 44.1  So disregard my above comments on mpv
God kills a kitten every time you encode with CBR 320

Poll: least-worst-sounding 64 kbit/s format

Reply #7
Hmm.  Now you've got me wondering what I actually did when I encoded to lame --alt-preset insane.  I must have decoded plusv to a 32 kHz sampled wav file, and lame must have encoded ok with that.  But I'll have to check.

ff123

Edit:  verified that the lame insane encode is 32 kHz.

Poll: least-worst-sounding 64 kbit/s format

Reply #8
Please note that PlusV seems to be technically on it's early stages and quality isn't comparable to mp3pro or other low-bitrate technologies.

From PlusV Technical faq on http://www.plusv.org/modules.php?op=modloa...ies=Technical#5:

---------

2001-10-18: Is the quality of PlusV going to be better later?

Definitely. This is the very first release of the format, and no _serious_ psycho-acoustic modeling has yet been done. Actually, one of the reasons for putting PlusV out for anyone to use for free is to get wide acceptance and to get the psycho-acoustic gurus on the Internet to play around with the format. We will naturally continue to develop a better psycho-acoustic model, but it is our hope that the Internet community could do this much better if the right people really put their heart to it.

----------

On the other hand, Dibrom & Gabriel Bouvigne were mainly intererested using PlusV for 16+ frequences which are normally difficult to encode with mp3 technology. Bits needed to encode high frequencies properly is the reason for huge bitrates on some songs.

I think that would be a great idea. For example, my hearing capability drops rapidly after 16.3 hz and i hear sounds after that weakly to 17.5 hz. "Shadowing" those frequencies with SBR technology would save a lots of bits and we could use more bits for lower frequencies, without fear that bitrate would "explode" too much. (Edit: "Exploding" of bitrate isn't such a big problem anymore, now that we got Dibrom's excellent settings, but i still think this would be a great idea)

Anyway, i voted for mp3pro because it had less smearing of sounds than other formats and generally sounded "clearer" to me, PlusV was the worst.

Poll: least-worst-sounding 64 kbit/s format

Reply #9
What about adding the regular mp3 to the list?


Of course it will be teh worst sound, but it would allow to mesure the progress of other formats against the mp3.

Poll: least-worst-sounding 64 kbit/s format

Reply #10
Quote
Of course it will be teh worst sound, but it would allow to mesure the progress of other formats against the mp3.


Added it to my samples page.  Is it the worst?  I let people judge for themselves.

BTW, I used lame with --nspsytune, because at the lowpass I wanted (11 kHz), it seemed to do better than FastEnc.

ff123

Poll: least-worst-sounding 64 kbit/s format

Reply #11
Thank you

btw, did you used the same settings with and without +v?

Poll: least-worst-sounding 64 kbit/s format

Reply #12
I tried plusv with the nspsytune settings, but couldn't really tell that there was a significant difference in quality.  Also, it seemed to force the use of abr.  I just used the default settings for plusv (which includes -q4).

ff123

 

Poll: least-worst-sounding 64 kbit/s format

Reply #13
I donno what is plusV, but among Mp3PRO, OGG, WMA, I think the worst would be WMA. 64 kbit/s WMA has a lot of hissing artifacts and heavy metalic sound, and sounds mufled too. OGG is much better than WMA but has a lot of hiss...............
I think Mp3PRO would be the best with it's fake freq. I've tried the sample
"halfsweepinvert.flac"-high freq sweep from ff123. Both WMA 64kbps, Mp3PRO (any bitrate) throw all the high freq and left silents only. OGG is the only one to mantain the high freq of the "halfsweepinvert.flac" even only at 64 kbps.