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.
Topic: Public Multiformat Listening Test @ ~96 kbps [July 2014], AAC, Opus, O (Read 126813 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Public Multiformat Listening Test @ ~96 kbps [July 2014], AAC, Opus, O

Reply #125 –
Nice results on the Opus vs Apple AAC but the FAAC seems like a waste of time. MP3 is always good to include though due to the popularity.
Thanks to everyone who participated and sorry that I couldn't.

Public Multiformat Listening Test @ ~96 kbps [July 2014], AAC, Opus, O

Reply #126 –
Dear all:

This is my first posting on Hydrogenaudio. Nice to meet you all!
It seems that I contributed to the inflation of the scores of every codec examined , but anyway LAME VBR V5 still shows a good quality for me.

Thank you again Kamedo2 and Steve for organizing such a great listening test.
I hope that the statistics will be a guideline to those who are not aware of lossy audio codecs other than MP3.

By the way, the results of OPUS are quite impressive.
I think that upsampling of 44.1kHz original samples to 48 kHz by an OPUS encoder might result in an unexpected distortion of sound.
Since many sources as well as CDs still have sampling rates of 44.1kHz, this feature of OPUS can be a disadvantage.
I wonder about other's opinion for this.

Public Multiformat Listening Test @ ~96 kbps [July 2014], AAC, Opus, O

Reply #127 –
is it so that the numbers of the files represent the position of the files in the abx player? cause if yes, i suppose some cheating is poosible by looking at the file sizes..

Public Multiformat Listening Test @ ~96 kbps [July 2014], AAC, Opus, O

Reply #128 –
The position of the codecs was randomly selected in each sample, easily traceable once you picked out the low anchors.

Public Multiformat Listening Test @ ~96 kbps [July 2014], AAC, Opus, O

Reply #129 –
Quote from: soundman9999 on
is it so that the numbers of the files represent the position of the files in the abx player?...

No, positions are randomly changed every time you open an ABX session.

Public Multiformat Listening Test @ ~96 kbps [July 2014], AAC, Opus, O

Reply #130 –
Quote from: Joungmin Lee on
I think that upsampling of 44.1kHz original samples to 48 kHz by an OPUS encoder might result in an unexpected distortion of sound.

No, Opus has a high quality internal resampler. As You can see Opus comes first. If that was case then the final Opus score would be greatly affected. And that's not the case.

Modern resamplers are transparent today. You can try excessively high quality SoX resampler. It passes a limit of transparency by a large margin.

It's possible to encode by foobar2000 converter this way:
original WAV 44.1 kHz/16 bits -> on-fly SoX resampler  44.1 kHz /16 bits to 48 kHz/32 bits (VHQ: very high quality)-> Opus

SoX plugin for foobar2000 player

Public Multiformat Listening Test @ ~96 kbps [July 2014], AAC, Opus, O

Reply #131 –
Opus uses the Speex resampler plus some tweaks, so you could compare that to Sox if you have some sophisticated tools.  You should find it isn't quite as good as Sox VHQ, but with noise perhaps at -200dB you don't need to panic.  There may be reasons not to transcode your entire lossy music library into Opus just yet, but this isn't one of them.  It is quite likely, depending on your system, that your music is routinely resampled by your computer and you don't give it a second thought.

I'm curious that there are no killer tracks for Opus, but a fairly obvious one for qaac.  At first I thought it was Hotel California which seemed odd, but it looks like that particular blue triangle is "girl", electronic, almost chiptunes.

Public Multiformat Listening Test @ ~96 kbps [July 2014], AAC, Opus, O

Reply #132 –
hy,
i just abx´ed through
sample 24 and 38.

i can clearly find out low and middle anchor, but absolutely no chance to distinguish any of the other clips.

could you tell me what you guys hear? i tried to go for the background noise, but no difference there - for clip 38 its obvious, low and middle cancel some noise, but as soon as i abx the others i cant hear the noise anymore.

the s-sounds are not obvious too, snare drum no difference too, decay of the snare drum sounds the same too me too..

so i cant distinguish them by searching for differences at those points, could you give me a hint for example with sample 38 what distinguishes the encodes?

thx!

Public Multiformat Listening Test @ ~96 kbps [July 2014], AAC, Opus, O

Reply #133 –
Quote
but anyway LAME VBR V5 still shows a good quality for me.

Just a quick reminder that the Lame samples were basically at 128kbps vs 96kbps for the other codecs.  You can of course achieve enough transparency for your needs with any reasonably modern music codec simply by choosing a sufficiently high bitrate.

Public Multiformat Listening Test @ ~96 kbps [July 2014], AAC, Opus, O

Reply #134 –
Quote
so i cant distinguish them by searching for differences at those points, could you give me a hint for example with sample 38 what distinguishes the encodes?

For practice, you could try encoding the original to a lower bitrate, picking some artefacts, then looking for them at 96kbps.  With Opus at 64kbps, I still on this track I still don't find obvious problems, but at 48kbps there is plenty to listen for.  Start with the strings at about 10s.  I used to use female vocals as a benchmark many years ago, but modern codecs seem to have mastered them completely.

Public Multiformat Listening Test @ ~96 kbps [July 2014], AAC, Opus, O

Reply #135 –
Quote from: lithopsian on
Just a quick reminder that the Lame samples were basically at 128kbps vs 96kbps for the other codecs.  You can of course achieve enough transparency for your needs with any reasonably modern music codec simply by choosing a sufficiently high bitrate.


Hi, lithopsian. Thanks for the comment. Maybe my English was not proficient  .

Yep. Definitely, aac, ogg, and opus showed transparency with lower bitrates than mp3 for me. I use qaac for listening music with my smartphone as well.
(as you can see in the test result, I hardly picked out lossy samples, except low and mid-low anchors, from the references, though.)
I wanted to say that I am admiring the advance of the LAME codec in spite of endogenous limitations of mp3 compared to modern formats.
Together with the larger storage devices, that's probably why many people still do not feel the necessity of changing mp3 for the modern codecs.

Public Multiformat Listening Test @ ~96 kbps [July 2014], AAC, Opus, O

Reply #136 –
Quote from: IgorC on
No, Opus has a high quality internal resampler. As You can see Opus comes first. If that was case then the final Opus score would be greatly affected. And that's not the case.

Modern resamplers are transparent today. You can try excessively high quality SoX resampler. It passes a limit of transparency by a large margin.

It's possible to encode by foobar2000 converter this way:
original WAV 44.1 kHz/16 bits -> on-fly SoX resampler  44.1 kHz /16 bits to 48 kHz/32 bits (VHQ: very high quality)-> Opus

SoX plugin for foobar2000 player


Hi, IgorC.

In Windows 8.1, 48kHz is the default sampling rate (maybe in WASAPI? I'm not sure).
I changed the SR to 44.1kHz/16bit. Since the sampling rate of every file I listen is 44.1kHz.
Under this condition, resampling may occur again (by directsound?) during the playback of an OPUS file.
Is it better to change the SR to 48kHz in the Windows control panel for OPUS and resample other 44.1kHz files using the foobar2000 SoX resampler?

Public Multiformat Listening Test @ ~96 kbps [July 2014], AAC, Opus, O

Reply #137 –
Thank you for conducting this test. A huge effort, and well worth it.

Let's be clear: while the quality is remarkable given the bitrate, these codecs are not transparent at this bitrate. They are transparent for some people and/or some content, but for lots of people and lots of content, the difference is audible.

This is quite different from the usual HA recommended settings, where audible differences (due to exceptional people and/or content) are comparatively rare.

The huge number of 4.x results (not too many 5.0s, few very below 3.0 except for anchors) suggest that this was a great test to perform, though maybe the low anchor was a bit too low?

Cheers,
David.

Public Multiformat Listening Test @ ~96 kbps [July 2014], AAC, Opus, O

Reply #138 –
Quote from: Joungmin Lee on
In Windows 8.1, 48kHz is the default sampling rate (maybe in WASAPI? I'm not sure).
I changed the SR to 44.1kHz/16bit. Since the sampling rate of every file I listen is 44.1kHz.
Under this condition, resampling may occur again (by directsound?) during the playback of an OPUS file.
Is it better to change the SR to 48kHz in the Windows control panel for OPUS and resample other 44.1kHz files using the foobar2000 SoX resampler?


Yes, it's better to set one frequency in Windows panel, and the same frequency in foobar2000 SoX resampler (it may be 48 or 96 kHz - according to your sound cadr specification and settings).
🇺🇦 Glory to Ukraine!

Public Multiformat Listening Test @ ~96 kbps [July 2014], AAC, Opus, O

Reply #139 –
Is Sox so much better than the Windows 8.1 resampler?

Public Multiformat Listening Test @ ~96 kbps [July 2014], AAC, Opus, O

Reply #140 –
Quote from: Steve Forte Rio on
Yes, it's better to set one frequency in Windows panel, and the same frequency in foobar2000 SoX resampler (it may be 48 or 96 kHz - according to your sound cadr specification and settings).


Thank you, Steve. Since my soundcard supports up to 192kHz, I changed SR to 96kHz in the Windows panel and enabled the SoX resampler so that 44.1kHz files would be resampled through it.

Now it's time that I enjoy OPUS .

Joungmin

Public Multiformat Listening Test @ ~96 kbps [July 2014], AAC, Opus, O

Reply #141 –
Thank you test organizers and participants!

If the results tell me anything it's that for me, any of these modern codecs is more than adequate at these bitrates. I'm thoroughly impressed with the showing of Opus, and this should more than settle any arguments about which current codec deserves to become the common codec going forward. Not only does Opus eclipse the competition in quality, it does so while being entirely free, open source, and royalty free. The only other codec in this test that can rival that second point is Vorbis which holds it's own in this test, but there's no question that it's younger brother has taken the quality crown at these bitrates.

BTW, you have my nickname in the results as yourload instead of yourlord. I hope I didn't screw up and submit the results with that bad of a typo.

Public Multiformat Listening Test @ ~96 kbps [July 2014], AAC, Opus, O

Reply #142 –
Quote from: lithopsian on
Is Sox so much better than the Windows 8.1 resampler?


Not much, but it's at least configurable. Here is the spectrogram for default quality of WASAPI resampler:



And here's SoX Best + allow AA:




Quote from: Joungmin Lee on
Thank you, Steve. Since my soundcard supports up to 192kHz, I changed SR to 96kHz in the Windows panel and enabled the SoX resampler so that 44.1kHz files would be resampled through it.


Then you can use 192 kHz. Anyway, don't worry about it too much, in 99.9% of cases default WASAPI resampling 44->96 or etc doesn't give any audible distortions. Almost all differences we are talking about are far out of earshot.


P. S. For further discussions on this theme we'd better move to separate thread.
🇺🇦 Glory to Ukraine!

Public Multiformat Listening Test @ ~96 kbps [July 2014], AAC, Opus, O

Reply #144 –
Quote
Then you can use 192 kHz. Anyway, don't worry about it too much, in 99.9% of cases default WASAPI resampling 44->96 or etc doesn't give any audible distortions. Almost all differences we are talking about are far out of earshot.

P. S. For further discussions on this theme we'd better move to separate thread.


Thank you Steve. Let's end discussion on the resampling issue.