Pretty straight forward test idea here. I have two copies of two songs but one of the copies of each song has been run through an unneeded crossover filter and summed back together. Lets keep this one blind. If you feel like it later we can do ABXing with two clearly marked examples.
Example 1 A
zSHARE - Lou Reed - Walk on the Wild Side A.flac
Example 1 B
zSHARE - Lou Reed - Walk on the Wild Side B.flac
Example 2 A
zSHARE - Yusef Lateef - _I Dont Stand_ A Ghost Of A Chance With You A.flac
Example 2 B
zSHARE - Yusef Lateef - _I Dont Stand_ A Ghost Of A Chance With You B.flac
Untouched Phase vs LR24 @ 1.5kHz round 1
vs
Example 1 A
zSHARE - Lou Reed - Walk on the Wild Side A.flac
Example 1 B
zSHARE - Lou Reed - Walk on the Wild Side B.flac
Example 2 A
zSHARE - Yusef Lateef - _I Dont Stand_ A Ghost Of A Chance With You A.flac
Example 2 B
zSHARE - Yusef Lateef - _I Dont Stand_ A Ghost Of A Chance With You B.flac
Untouched Phase vs LR24 @ 1.5kHz round 1
An externally hosted image should be here but it was not working when we last tested it.
vs
An externally hosted image should be here but it was not working when we last tested it.
Er those huge things were thumbnails when I posted. Anyway if you really can hear a difference between the files a good way to prove it to yourself and others is to use Foobar2000 and a PC to ABX and generate a report.
Do you mean its same one (A or B) of both tracks you have modified? Or is it random (could still be the same letter) ?
It's blind and random. I know which one is which. You can actually verify the filter I used independently but not sure I am telling how right now. But in both examples there is one (either A or B) that has an all pass LR24 applied to it. This is significantly more distortion than the average crossover.
Highlight the two tracks> Right click> utilities ABX.
If you don't have it you have to reinstall I think. The option is under the installation customizations.
Tell us what you hear as well.
And of course if you can't hear a difference there is no need to really start the ABX. Sometimes A/Bing can make it easier to hear a difference that is present though Guess my point is if you can't hear the difference still let us know.
If you don't have it you have to reinstall I think. The option is under the installation customizations.
Tell us what you hear as well.
And of course if you can't hear a difference there is no need to really start the ABX. Sometimes A/Bing can make it easier to hear a difference that is present though Guess my point is if you can't hear the difference still let us know.
Last edited:
Here are some links to the use of all-pass (phase rotator) filters in broadcasting:
The W3AM Phase Rotating Asymmetry Eliminator!
Allpass paper by W4ENE
http://tinyurl.com/hv2ef
When you listen to broadcast music , the odds are high that it's been passed through a phase rotator.
The following are mono voice WAV sound files. Two very different voices.
One of the channels (L or R) has gone through a 4 pole phase rotator.
The other channel is unmodified.
http://tinyurl.com/mqq2wy
http://tinyurl.com/mo536h
(Note that the phase rotator causes a pseudo-stereo effect when listening on
stereo headphones even though the audio source is monophonic)
There is a possible hitch to ABX testing on these types of files. A phase rotator diminishes the peak to average levels of the audio signal. In casual ABX testing of the above files, a couple of people could distinguish L from R. They both reported that the phase rotated signal was subjectively heard as louder.
Just thought I would throw this information out there for your consideration.
Cheers.
ZAP
The W3AM Phase Rotating Asymmetry Eliminator!
Allpass paper by W4ENE
http://tinyurl.com/hv2ef
When you listen to broadcast music , the odds are high that it's been passed through a phase rotator.
The following are mono voice WAV sound files. Two very different voices.
One of the channels (L or R) has gone through a 4 pole phase rotator.
The other channel is unmodified.
http://tinyurl.com/mqq2wy
http://tinyurl.com/mo536h
(Note that the phase rotator causes a pseudo-stereo effect when listening on
stereo headphones even though the audio source is monophonic)
There is a possible hitch to ABX testing on these types of files. A phase rotator diminishes the peak to average levels of the audio signal. In casual ABX testing of the above files, a couple of people could distinguish L from R. They both reported that the phase rotated signal was subjectively heard as louder.
Just thought I would throw this information out there for your consideration.
Cheers.
ZAP
Key,
Your second 1.5khz graph indicates MUCH more phase shift than a LR24 crossover would have. I see quite a few wraps when there should only be one. Confusing.
Cheers,
Dave.
Your second 1.5khz graph indicates MUCH more phase shift than a LR24 crossover would have. I see quite a few wraps when there should only be one. Confusing.
Cheers,
Dave.
Well A & B sound different so far (I think)... but couldn't say which I think is the modified one after a couple of listens.
Do you have a release date/time in mind key? It would be nice to have in order to make a decision before the cats out of the bag....
Do you have a release date/time in mind key? It would be nice to have in order to make a decision before the cats out of the bag....
24th Order. I was going to do 64th but I thought that would be too heavy handed 🙂
I dunno should I give it about 2 weeks? More? People can always ABX after I tell people or I can make some new examples with different order or something. Any requests?
I dunno should I give it about 2 weeks? More? People can always ABX after I tell people or I can make some new examples with different order or something. Any requests?
Actually, I like the idea of the logged ABX. As soon as I get back home this weekend, I'm going to try a 10 or 12 trial run, and would encourage others to do the same.
If the upload is free for you, you might make a nice test by randomly coding ten files (flip a coin before each one, it might be 5 and 5, might be 6 and 4), and having them available for download. The user then has to sort or label each of the ten files as A or B. That improves the 50% probability from guessing on just two files, and with ten coded files, you could then reveal in advance which of the original files was filtered.
If the upload is free for you, you might make a nice test by randomly coding ten files (flip a coin before each one, it might be 5 and 5, might be 6 and 4), and having them available for download. The user then has to sort or label each of the ten files as A or B. That improves the 50% probability from guessing on just two files, and with ten coded files, you could then reveal in advance which of the original files was filtered.
Let me be the first to say, that i dont hear a difference between example 1A and B. I know people who are quite good at finding mp3 encodes in abx tests, which i pretty much fail at with higher bitrates, so maybe one can train his ears to find phase errors, but i have never done so.
I let my friend who is also good at mp3 finding listen to it and he immediatly did an abx with the foobar2000 plugin and got 14 of 14 right. He could hear a very small difference in the background, some kind of ringing and some added noise in 1A. I on the other hand totally fail at finding the noise he describes.
Edit: I have to add that the points in this track where he could spot a difference was with long steady notes like between 3:54 and 3:56. However, he says that neither version sounds inferior to him, just very slightly different.
Edit: I have to add that the points in this track where he could spot a difference was with long steady notes like between 3:54 and 3:56. However, he says that neither version sounds inferior to him, just very slightly different.
Last edited:
I spend some more time with example 2 and i got 10/10 right. Version 2B seems to have some extra hissing on some notes. Since this was my first real try at ABX testing (previously i only clicked around in the different tracks), maybe one should disregard my first post. Its like solving a puzzle, the missing piece is there somewhere and its hard to find, but once you got it, the difference is clear.
That's the thing, if you don't know what you are listening for, it's hard to hear. That's where training or at least having it pointed out to you helps.
On my first go thru, the difference was not very obvious to me.
On my first go thru, the difference was not very obvious to me.
Hello,
I'm a poor audiophile.. I can not hear any difference between 1A and 1B
I was using headphones.
I did check the time domain waveforms and there is a low frequency delay of about 4ms in the other sample. At high freqs the delay is nonexistent.
Possibly song 1 does not have high energy content at 1.5kHz to bring the effect into hearing.
- Elias
I'm a poor audiophile.. I can not hear any difference between 1A and 1B I was using headphones.
I did check the time domain waveforms and there is a low frequency delay of about 4ms in the other sample. At high freqs the delay is nonexistent.
Possibly song 1 does not have high energy content at 1.5kHz to bring the effect into hearing.
- Elias
- Status
- Not open for further replies.