Kevin Haskins wrote:
What is the best way to set levels for distortion testing? Obviously if your working with a 6.5" driver and trying to normalize it somewhere for unit to unit distortion test you have to calibrate it to a level. If you pick an individual frequency you may end up on a peak or dip of an individual driver. If you pick an average over a range you have to choose the range & that might be fraught with peril if your unit has a large dip or peak in that range. Correct on all counts - it's difficult to do. I tend to set a nominal level as averaged (1/12th octave spacings) over the center 2 octaves (or most important range; for a 6.5" woofer, that would be typically 250 Hz to 1 kHz).
The other way is to show a true relative distortion level to the output; graphs that show the fundamental (frequency response) along with the harmonics are good, but you must interpret the harmonics relative to the fundamental level. You can't just look at the harmonics and say "lower is better!" with those types of graphs.
For example, at
the 6.5" woofer HD test, look at the Extremis and the AA6.5 (often compared against each other).
Down in the deep frequencies - right at 20 Hz, we see the following for the first two harmonics:
Extremis: -52 dB/2nd -44 dB/3rd (0.25% and 0.63%)
AA6.5: -56 dB/2nd -56 dB/3rd (0.16% and 0.16%)
MANY have looked at that and said "see! The AA6.5 has less THD than the Extremis - XBL² is junk, it doesn't work!"
BUT, look at the output levels:
Extremis: -2 dB
AA6.5: -16 dB
So the Extremis was generating 14 dB MORE output at those distortion levels! Now, you typically cannot just add the output offsets to the distortion levels (it's typically highly conservative; THD in a driver does not increase linearly with excursion), but for the quickest shot, we'll do it:
Extremis: -50 dB/2nd -42 dB/3rd (0.31% and 0.79%)
AA6.5: -40 dB/2nd -40 dB/3rd (1% and 1%)
Pretty different now, eh? That big advantage is completely gone once we consider output level. This is what you would expect - worse "conservative" case - if the output levels were equalized.
Fundamentally, THD testing is a tricky thing. IMHO you need to always consider the following:
1. Measurement setup - am I in the proper acoustical setting (near/far field), is the ambient noise low enough to accurately measure the levels I want?
2. Am I setting the relative SPLs in a significant/relevant way?
3. Is the data as presented easy to understand/simple to reference to each other, for the target audience?