With permission from Lavry Engineering:

--

Regarding "Null testing"-

Basically; null tests are useful to determine if two files are identical. If they are different; you hear only the difference.

For those unfamiliar with the term; this test is to determine whether or not two digital recordings (files) are identical. In order to make a meaningful comparison, any potential difference must be addressed so that when the two files are added (mixed) "out-of-phase" they cancel each other, and the difference is silence. This requires the files to contain exactly the same audio content, for the files to be sample-accurate with each other, and for the level to match very exactly.

Typically, the files are placed on two separate tracks and one of the two tracks is moved in the timeline to align the waveforms to a sample-accurate position. The polarity of one track is inverted so the audio content of the two tracks are "out-of-phase" with each other. If there is any difference in the level, the digital level control can be used to minimize this; but unless the difference is zero or exactly equal to the step-size of the digital level control; there will be a small level difference. This would cause two files that are identical in every other way to not cancel out completely.

The usefulness of a null test is extremely limited in determining the difference from the "ideal" as, for example; if either file is "brighter" than the other (has more treble), the difference will be very bright. It is not possible to determine which of the two is brighter by simply listening to the difference; or whether one has more low treble and the other has more high treble and the audible effect is a combination of the two. This applies to ALL differences, including distortion. If one file had only even-order harmonic distortion and the other contained only odd-order harmonic distortion, the null test result would contain both even and odd order harmonic distortion and there would be no way to determine which file contained what distortion simply by listening to the null test result.

So, like many tests; the null test tells you exactly one thing and the results need to be interpreted accordingly. The main reason I used a null test in the past was to determine whether the CD plant had done processing to the Mastered file I sent before manufacturing the CD. In this case; the only important thing was whether or not there was ANY difference. If there was; I knew that something had changed and would contact the plant to arrange to have it re-done. In this case; it did not matter what the difference was (or how it sounded).

The only other application was to determine if the files remained in time with each other. This can be useful if there is a chance that either an edit or a digital "glitch" caused some of the original file to be missing from the copy. In this case; even if there was an audible difference caused by something like AD and DA conversion, it would remain constant unless there was an edit or missing samples. At the point where the change occurred; the "difference" output would suddenly increase in volume because the two waveforms were no longer close enough in content for phase cancellation to "null" most of the signal.

Brad Johnson
Lavry Engineering, Inc.