Originally Posted by arnyk
I did a little research and came up with this:
It's actually worse than that. This polar plot is normalized to the above graph and shows how much louder or quieter the speaker is at various frequency and horizontal angle combinations than it is on-axis.
People hear timbre as a combination of the spectra believed to be the direct sound and its reflections, especially those off the side-walls which come from far (like 60+ degrees when speakers are toed in to face the listener) off the speaker's design axis. Continuously decreasing high frequency content in the reflections sounds natural to us (perhaps because things in our environment like plants diffuse and absorb more at high frequencies) although reversals in that trend make those frequencies sound louder and the sound less natural.
Cone and dome 2-ways with traditional cross-over points and nothing done to physically prevent the tweeter's nearly uniform radiation into half space have an output drop moving through the midrange followed by an abrupt broadening crossing to the tweeter which yields reflections that contain disproportionally more more energy in the 2-4KHz range than elsewhere with surplus energy in that frequency range associated with "harshness"
Note the peak from 3-4Khz.
People like to say that you can't tell how a speaker sounds from measurements,
We can do better than merely telling how speakers sound.
We can actually predict preference based on measurements thanks to Sean Olive's 2004 formula built on numbers like flatness and smoothness of on and off-axis response plus bass extension.
He works for the Harman research group where Harman has a vested interest in making speakers that people prefer to buy.
Compare that with this speaker:
This is the classic NHT Superzero. Note the generally smooth response and absence of peaks. A little less bass, but look at the comparison in physical size: 5.5" W by 9" H by 5" D. External volume 247.5 cubic inches. Compare the SZ to the Totem: 14" H by 6.8" W by 9" D. External volume 630 cubic inches.
The NHT polar response (normalized) looks like this
Note the 3-4KHz off-axis peak which is why most speakers sound like speakers and I recommend a less conventional design that avoids the problem (tweeter wave guide matching the mid-range directivity at the cross-over, acoustically small point source approximation, acoustically small dipole, Danley Synergy Horn or Unity Horn, for limited output levels a small full-range driver on a back-horn terminating at the floor designed to account for baffle step, there are _lots_ of engineering options with different trade-offs), a beefier tweeter crossed over lower, and/or a 3 or more way with a smaller midrange.
These are my favorite small speakers (Linkwitz Pluto, as in designed by Siegfried Linkwitz of Linkwitz-Riley cross-over fame with an additional thirty years of speaker design experience/experimentation and Audio Engineering Society paper digestion under his belt plus the Stereophile Loudspeaker of the Year award for the lowest-cost version of the flagship model from his first commercial venture) that I use in my bedroom system. I've heard $40,000 (with Pass Labs electronics) and $80,000 speakers which didn't do as well overall (although the $80K pair played a lot louder cleanly and had good bass through the first octave).
The premature roll-off is an uncorrected artifact from the measurement microphone; they actually extend to 15Khz in this version. 6" wide at the top, 4.5" for most of their height, with a 12x8" footprint on the floor. 42" high, but you want to get the tweeters at or a little past ear level and would be using stands to do so with "book shelf" speakers.
Measurements are of the original version; the current 2.1 version is flat to 40Hz (-3dB) not 60Hz and compensates for the high frequency roll-off (one great thing about active speakers is that you can boost output without compromising sensitivity at other frequencies).
Nothing stands out about the sound which is excellent - you want to be engaged by the recordings and not be distracted by the speaker's editorial contributions (a little brighter here, a little boom there, a little dull elsewhere). Music sounds like music played by musicians and is very natural as long as you respect their output limits (no worse than other speakers with 5-6" mid-bass drivers) just as you might predict from the measurements (I bet $800 on that building them without hearing them first. I was right).
They're a $1000 (inflation stinks) DIY project (with one afternoon of soldering (a lot more than in a passive speaker - you're building two power supplies, six amplifier channels, and time-alignment circuits plus fourth order filters and the usual accommodations for drivers) and one putting everything together) and $3000 commercial offering including built-in amplifiers which puts it within hiscore's budget if he sells his speakers too.
I pair them with the Pluto+ sub-woofers to get more mid-bass headroom and have the first octave.
When you position a Pluto, one Pluto+ sub-woofer, and microphone next to the ground and drive them off Siegfried's cross-over designed for the purpose they measure like this (if I used the system for critical listening I'd probably increase the system high-pass pole frequencies to avoid room gain):