In the ongoing debate on other forums over 'who's speakers sound like what', I am still trying to sort out how to interpret the graphs that folks publish.
Since I already know what speakers I like,
its really an intelectual exercise. As part of the exercise I posted a simple question on another forum about what people thought the minimum change in db in a speakers response would noticably colour the sound of a speaker. I was not expecting a lot of responses given that not many really care to know, but I got exactly 0 responses. There are lots of ways to interpret that, but I must say I was dissapointed.
In contrast to the above, I find your response refreshingly strait forward. I love this site!
Hi Fred,
Thanks. In the large body of research conducted by Dr Floyd Toole when he was the resident scientist (psycho-acoustician) at the National Research Council in Ottawa for 25 years (Axiom participated in this research starting back in the early 1980s), Floyd concluded after many years that the deviations that really mattered in terms of listeners making judgments between good and bad speakers all occurred in a fairly narrow band through the upper bass and midrange.
In the hundreds of speakers measured over the years (as an editor of AV magazines that did their measurements and double-blind tests at the NRC, I was part of the listening panels for many years), the speakers that were highly rated and whose measurements correlated with the high ratings, all had very smooth midrange frequency response within about 1.5 dB from the upper bass frequencies to about 4,000 Hz (this nicely supports the statement that "it's all in the midrange" in terms of speaker quality). Larger deviations at higher frequencies --8 kHz to 20 kHz--are not easily detectable or even readily audible with music programming. Of course, we can measure them with pink noise and test signals in anechoic chambers, but we do NOT express our preferences for "good" and "bad" speakers based on these upper octave anomalies when we listen to music.
It's also a function of the fact that human hearing is really peaked in the midrange for evolutionary reasons-- for speech intelligibility, of course, and to detect cracks of twigs etc in the forests and jungles to warn of attacks by predators and enemies. Moreover, contrary to what many audiophiles believe, the bulk of instrumental fundamental tones and important harmonics also occur largely in the midrange. Of course, bass response matters, but the deviations in bass may be more irregular and we won't downgrade the speaker until really gross peaks become audible (fat, boomy peaks). And the nuances of upper treble also contribute, but do not significantly alter our overall listening impressions of a speaker.
To sum up, if a speaker has really smooth, linear response through the midrange--and this only applies to speakers carefully measured in anechoic chambers, not casual or erratic measurements done in rooms--then it's very likely to be heard as an accurate and pleasing music reproducer in average rooms by most listeners with normal hearing.
There is much I've left out, but these conclusions, reached after 25 years of research, are what made Dr. Toole elected to president of the Audio Engineering Society world-wide. It is Canada's unique contribution to the advancement of speaker design and measurement, and how the resulting products correlate with listener perceptions in double-blind tests.
Regards,
Alan
