Flipped phase cancel doesn't always mean equality?

[MikeMitchell]

Just read over 'adding a master fader makes it sound less open...' thread. I've always heard in passing and specifically from a few people whose opinion I highly regard that using a phase cancel is not an accurate way to judge a difference between two files. I'm not sonically schooled enough to understand why, or if this is even true. I can't logically on my own come up with a reason why they wouldn't cancel. Anyone disagree or care to attempt an explanation?
Thanks, Mike

[Lagerfeldt]

Originally Posted by MikeMitchell

that using a phase cancel is not an accurate way to judge a difference between two files.

Well, you should start thinking less of those people then.

If you take two identical files and line them up in a DAW and flip the phase, there will be no (none/zip/zero/nill) sound.

If the files are not identical there will be some sound.

This is referred to as a "null test".

It's basic physics really.

[narcoman]

Originally Posted by Lagerfeldt

Well, you should start thinking less of those people then.

indeed so!!! How could anyone argue otherwise - unless they actually didnt know anything. Ignorance can be an unfortunately powerful thing.....tutt

[Kiwiburger]

I don't think any sane person would argue about the validity of a phase inversion null test. If two files null, they are identical. If they don't, they aren't.

There can be times where it's impossible to use a null test for technical reasons. One example might be where an effect using random or non-synced modulation might be involved. For example, you can process the exact same sample with the exact same reverb and create two files, and they won't null because of random variations between them. The conclusion that the two files are different would be correct, but the conclusion that the sample or reverb were different would be wrong.

I believe what the people were saying in the case you mentioned was that the null test of bounced files wasn't necessarily proving anything, because the A and B files created were going to be the same anyway.

If you believe you hear a difference with the master fader engaged (visible) or disengaged (not displayed) then it's probably an audible difference. IF you recorded the streaming output I would expect you to end up with two files that don't null.

I suspect in this case that the resources required to render the master fader, and calculate the meter display etc, were affecting the audio output. People have commented on this effect before: drawing the graphics on screen takes up resources, and if your system is struggling, perhaps swapping virtual memory pages to disk, then it probably will screw up the rendering of your audio.

But - if you bounce the mix to disk non-realtime, that memory bottleneck will be a non-issue. So with or without the master fader shouldn't make a difference.

I'm speculating wildly here, but that's a possible reason.

[arf]

If a DAW system produces an audible difference in playing the same data but in two different modes, then the system is likely in need of better clocking, either by means of a DAC that removes incoming timing errors - Lavry, Mytek, Benchmark, Weiss, or by means of an external clock. I don't think there is any substitute for jitter-immune DACs in a DAW system.

[Kees de Visser]

Originally Posted by MikeMitchell

...using a phase cancel is not an accurate way to judge a difference between two files.

If I'm reading this statement correctly, you are talking about judging differences that are not zero. A null test can be used to check if audio data of two files (or streams) are identical. If they are, there should be no audible difference (unless the monitoring chain is broken). If the difference is not zero, it can give a clue about where to look (listen) for the difference, but it can't be used to judge its audibility since it doesn't take into account the masking effects. In other words: if the difference signal is audible, that doesn't imply that the two audio files (data) are audibly different.
An ABX test is the best option to judge audibility of differences (>0).
Could that be the reservation that was made about the "phase cancel" test ?

[Plec]

I really love people who argue against null tests...
I don't know how many times I've heard and read "I don't need to do null-tests, I trust my ears" when faced with hard to take-in facts...... just want to take a stick and freakin' ram it up their *******