Thanks Sean. And now back to the discussion. Here is an excerpt from something I wrote recently:
"A speaker’s frequency-response (FR) plot can tell you a lot about its sound. The most obvious example is how FR defines tonal balance. If there’s a rise in the midband (100-1000Hz), then voices will typically sound more forward in the soundstage and have greater presence. If the very high frequencies (above 10kHz) are rolled off, the sound will lack air and the overall presentation will be somewhat dull. If there’s a hump in the midbass (60-100Hz) but little in the way of true low bass (below 40Hz), the sound will be weighty and hard-hitting, but will lack the foundation that the deepest bass provides. The examples are endless. The key is to understand how what you see in the graph correlates with what you hear at your seat. The sound you hear will always be a combination of your loudspeakers’ output and the room’s contribution to and shaping of that output. How these combine at the listening position is what you hear.
"But as anyone knows who’s had any experience with in-room measurements, while a speaker’s FR plot can tell you a lot about what you’re hearing, it can’t tell you the whole story. There is a long list of other sound characteristics that, in my experience, have no correlation with in-room frequency response: resolution, transparency, soundstaging, imaging, etc. Although a set of anechoic measurements can get to the bottom of some technical elements, such as distortion, I’m not aware of any that can tell you exactly how a soundstage will develop in your room."
I never can wrap my head around why someone is either all measurements or all subjective listening. Most experienced audiophiles know that both combine to tell the larger, more complete story that defines the product.
I agree with your first paragraph.
The 2nd paragraph needs more explanation. As I have shown in my paper about predicting loudspeaker preference, in-room measurements above 300-400 Hz lack important information about the frequency response of the speaker that are evident in anechoic measurements but absent in in-room measurements: you cannot easily separate the direct, early and late reflected sounds in an in-room measurement, hence they don't predict sound quality as well as anechoic measurements.
The directivity of the loudspeaker based on its anechoic measurements will tell you a lot about some perceived spatial properties as shown in the work of Klippel and others: Spaciousness, envelopment, apparent source width are all enhanced by wider dispersion loudspeakers. As long as your room acoustics/setup promote strong lateral reflections, then you will benefit from these loudspeaker characteristics. Highly directional speakers (dipoles, narrow horns, horizontal line arrays) will sound less spacious and give you higher ratios of direct/reflected sound - all things being equal. On the other hand directional speakers give your more pin-point imaging, which some people like.
When you move from stereo to surround sound, the directivity of the speakers may matter less since the cues in the surround channels may dominate your impressions of spaciousness and envelopment. Spatial qualities in stereo reproduction are highly dictated by the microphones techniques used in the recording. Spaced omni's produce much more uncorrelated signals at the ears than coincident microphone techniques, and hence will sound more spacious. Listen to the Telarc recordings: most of them are made using spaced omnis (which BTW typically have more bass than a directional microphone).
Don't discount how important the frequency response of the loudspeaker influences spatial attributes: our perception of distance has a spectral element to it. Further sound sources are less bright than closer ones, due to room and air absorption at HF).
Finally, there are binaural measurement techniques for measuring and predicting the location of a sound source, the sense of envelopment and apparent source width. If you look up the work of Professor Jens Blauert, and Dr. Wolfgang Hess (a former PhD student of Blauert) who works at Harman, you can see how this works. This method could be applied to loudspeaker measurements in the future with some more research.
Nonlinear distortion is the biggest research problem to solve. I am working on it right now, so hopefully there will be something to report in the future. For now, nonlinear distortion is a very small factor in the performance of well designed loudspeakers - at least until you start pushing them beyond their excursion limits. Masking helps out a lot here.
