riverlethe
100+ Head-Fier
- Joined
- Aug 10, 2007
- Posts
- 484
- Likes
- 48
I'm always puzzled when someone uses a single average (in this case THD) to summarize a complex transducer. The particular average chosen embodies assumptions about what's discernible that bias mathematical and measurement convenience over the (still poorly understood) subtleties of human perception. I learned this many times over when I worked on speech processing at Bell Labs next to scientists and engineers developing perceptual audio coding (what eventually became AAC). Actually, perceptual audio coding teaches us lessons two ways: on the one hand, perceptual masking effects allow surprisingly large mangling of the signal without discernible effects; on the other, some changes can be surprisingly discernible, at least to trained listeners. Which teaches that human hearing and its higher-level cortical components are way more complex and variable than any small set of averages (THD, bit rates, whatever).
Oh, definitely. Most manufacturers' "specs" are worthless. I just brought it up in response to Jason's comment about 0.0005% vs 0.0007% THD. Are you familiar with the GedLee metric? I believe it was partly inspired by the research you mention, but I won't pretend to understand how it's calculated. Can you give some examples of surprisingly detectable distortion? I assume these were established through blind ABX.