Elvee... if you email me the files I can put them on Drop-Box and I can reset and rename the poll options to start again.
could attach here as .zip archive - even smaller files with flac 1st, then zip
of course you lose people at every step
http://www.foobar2000.org/encoderpack adds flac to foobar2000
http://www.foobar2000.org/components/view/foo_abx
of course you lose people at every step
http://www.foobar2000.org/encoderpack adds flac to foobar2000
http://www.foobar2000.org/components/view/foo_abx
Last edited:
Most likely the two original files sounded different not because the phase alteration was converted by amplitude distortion to amplitude changes, but because the two files were clipped at slightly different levels.
The survey would be better if there were three options: File1, File2, and Cannot Tell Apart. Otherwise testers who cannot tell them apart cannot do the survey. You may be discarding the majority!
Also, in AB testing, it is well known among experts on such that you should ask poeople to swap back and forth a specific number of times (say 4 times). The human brain is notorious for picking a second example as different, but if the original example is tried again, people sometimes change their mind. Further, if people are allowed to swap back and forth between two options, there is a tendency to think one option is different when it isn't. Good practice in AB testing is to provide 3 or 4 A's, 3 or 4 B's, mix them up, and ask testers to assign each of the 6 or 8 tests to either of two options.
The survey would be better if there were three options: File1, File2, and Cannot Tell Apart. Otherwise testers who cannot tell them apart cannot do the survey. You may be discarding the majority!
Also, in AB testing, it is well known among experts on such that you should ask poeople to swap back and forth a specific number of times (say 4 times). The human brain is notorious for picking a second example as different, but if the original example is tried again, people sometimes change their mind. Further, if people are allowed to swap back and forth between two options, there is a tendency to think one option is different when it isn't. Good practice in AB testing is to provide 3 or 4 A's, 3 or 4 B's, mix them up, and ask testers to assign each of the 6 or 8 tests to either of two options.
Last edited:
The later two files are much more similar to listen to. With the level being at 0db I find I can only listen at the first couple of steps of the volume control (pc) so I guess I'm losing resolution there.
Can't really tell them apart audibly tbh at this stage.
Can't really tell them apart audibly tbh at this stage.
And what do you suppose caused the change in clipping level? I see it differently, I think the clipping level was the same in both cases, but the rearrangement of the harmonic phases changed the peak level of the waveform.
For a given RMS level, the waveform with the random harmonic phases has a higher peak level than the one with the phases set to give crossover distortion. So for the same RMS loudness and the same clipping level, the waveform with random phases gets clipped harder. This harder clipping explains the change in spectral content.
This is an example of what I mean by nonlinearity converting phase distortion into amplitude distortion.
Of course I assume that the OP equalized the RMS levels of the two waveforms, which is necessary for a fair test.
Last edited:
What caused the change in clipping level? I don't know - I don't know why it was clipped at all. I wasn't there when the OP did it.
While it is true that nonlinearity will convert incomming or pre-existing phase distortion into amplitude distortion, here we have the OP saying we added the phase distortion digitally, after the amplitude distortion.
It is doubful that distortion in most tester's equipment is sufficient to do much if any audible conversion, if that's what you meant, unless they deliberately turned up the gain sufficient to totally overload their amplifier.
Meanwhile, inspection of the two spectrums (ie look at the frequwncy domain plots) shows a result typical of what you get from different levels of clipping.
In an ideal world, every aspect of the testing protocol should be strictly controlled and enforced, but this is the internet, and people can do whatever they like, and they will: some will use their favorite player and play randomly the two tracks until they are satisfied, others will use Foobar or similar, and some will also peek at the waveform shape.
Since I have way no way to control any of this, I prefer to leave the test as it is: an informal one, and rely on opinions honestly expressed by the members taking part
I am also unable to differentiate them: when I think I have spotted a difference, I listen one more time and it vanishes.
Let's hope some "golden ears" participate and give their opinion...
There was no change in the level: the clipping was caused by LTspice not taking into account the scaling factor, which caused the ADC to saturate.
Both the levels before clipping and the clipping levels were strictly identical
In fact, it is the opposite: the unaltered waveform sounded harsher with the clipping.
.
Of course
I repeat: there was no change in the clipping level
You didn't change the clipping level relative to the RMS level. But fiddling with the harmonic phases will change the peak-to-RMS ratio of the waveform, so you are changing the peak level relative to the clipping level.
Try synthesizing a square wave that is just below clipping, then flip the phase of the 3rd harmonic 180 degrees. The waveform grows a huge lump and you now have clipping.
When you consider the amount of distortion in speakers and tube amplifiers, I think it is quite possible that significant PM to AM conversion is going on. Imagine a hefty kick drum transient that pushes the woofer cone out, or drives your boutique single ended amp into grid current. Then flip the phase 180 degrees. Now the woofer is getting sucked in, the tube is driven into cutoff, and the distortion products generated are quite different. Are you hearing absolute phase, or just different distortion?
Try synthesizing a square wave that is just below clipping, then flip the phase of the 3rd harmonic 180 degrees. The waveform grows a huge lump and you now have clipping.
When you consider the amount of distortion in speakers and tube amplifiers, I think it is quite possible that significant PM to AM conversion is going on. Imagine a hefty kick drum transient that pushes the woofer cone out, or drives your boutique single ended amp into grid current. Then flip the phase 180 degrees. Now the woofer is getting sucked in, the tube is driven into cutoff, and the distortion products generated are quite different. Are you hearing absolute phase, or just different distortion?
Last edited:
I understand that. But a third option, "cannot tell apart" would likely significantly improve the survey response rate, and give us a better idea of whether any response choosing File 1 or File 2 are real choices or whther they felt foced to pick one when they couldn't be at all sure there is a difference.
Not a real concern. The third harmonic of a square wave is 33% the amplitude of the fundamental, so yes it has a big impact on the waveshape. Here we have a sinewave with about 6% distortion. The third harmonic, which should be the largest amplitude, is thus a bit below approximately 6% of the fundamental amplitude. This is too small to alter the waveshape enough unless the files are played at very high level.
Elvee, I can still alter the poll if you wish and add a third option. Also, its due to close automatically tomorrow anyway. Do you want it extending ?
Altering it to incude a third option, and extending the date, would also provide an opportunity to advise that the tones should be played at a comfortable level not likely to clip, so that Scopeboy is happy.
Yes please, there have been too few answers at the moment to draw any conclusions anyway
I've added the third option and set the close date to the 20th.
It might be an idea for you to alter your first post and add the new zipped files there. You should be able to do that with your "first post edit" as thread starter, if not I can do it if you wish.
It might be an idea for you to alter your first post and add the new zipped files there. You should be able to do that with your "first post edit" as thread starter, if not I can do it if you wish.
- Status
- Not open for further replies.