I’m a little bemused by proponents of electrostatic panel speakers, open baffle or dipole speakers, or anything without a proper enclosure, conveniently describing conventional speakers as suffering from “boxy” sound as compared to their wonderful “airy” marvels. As far as I can tell, most panel speakers exist purely to reduce material, construction and logistics costs. 😆 Is it all just silly talk from box-phobic ninnies?
What is it anyway? There’s a whole lot of talk about it, but it seems a bit vacuous. Is it some kind of distortion? Is it poor off-axis response? So with a good calibrated Earthworks microphone I tried to measure “boxiness” of my own DIY speakers. They don’t come much squarer than these!
The active crossover already provides basic baffle step compensation to remove a 6dB midrange rise, and a phase shift network to align the tweeters with the midrange. Apart from seeing what happened off-axis, I also wanted to see if there was any of the documented (and simulated) 2 or so dB peaking response at around 1kHz and rippling caused by baffle edge diffraction above that.
Many people seem to think that baffle step is tweeter-related. It isn’t. A simulation prior to building the speakers generated this on-axis baffle step response of the midrange driver positioned (as-built) on its baffle:
I set up an Earthworks DM-23 microphone on its stand at tweeter height:
Then drew a kind of protractor:
Carefully (more or less 😀 ) aligned each time using the tape measure:
Ensured Doggie was out of the room and accurately measured baffle to microphone tip each time and swept the test signals. Close-range bass measurements don’t show anything relevant for the seated listening position, so I switched the woofer plate amp off and only measured from 100Hz up (crossover filter is 150Hz 4th order).
The drivers are offset from the centreline of the baffle. The measurements were done to the “outside”. If taken from the “inside”, no doubt there would have been some differences.
Here are the traces with 1/48 octave smoothing:
These graphs can be clicked for higher resolution if wanted. With 1/48 octave smoothing it still shows reflections (“grass”). Here it is again with 1/6 octave smoothing and with each trace labelled:
So basically these show SPL as measured at the same microphone height when moved around a 1m horizontal arc from directly in front of the speaker to 60° off-axis to one side. It shows quite tame progression from 2π to 4π energy dispersion with increasing frequency until about 6kHz where the tweeter’s beaming becomes more obvious. As mentioned, the bass amplifiers were switched off – hence the very steep roll-off. The pre and post impulse windows were set to 10 and 50 milliseconds respectively. This is of course a compromise, but it provides a response resolution of 16.67Hz and ample extension to cover the midrange and tweeter frequency bands. It eliminates some room reflections, but not many.
Dec 2017 Addendum (after re-reading Floyd Toole):
Toole uses 1/20 smoothing and an anechoic chamber “spinorama” which is beyond my capabilities, but nonetheless I can present here something more along his guidelines using 1/24 smoothing and an IR gating of 3ms to eliminate the early reflections in the room:
It seems that the 15° off-axis response is the best – especially flat across and above the 2.2kHz crossover region. I subsequently adjusted the toe-in of the speakers to take some advantage of this.
And the off-axis traces are a fairly nice replica of the on-axis trace in the critical band so the nice timbre of the speakers in the room seems to be pretty well reflected by the measurements. These need to be re-done at Toole’s recommended 2m microphone distance and once done I’ll add them to the “wings” page.
I suppose you could say there’s a bit of a 1kHz blip there in the red rectangle with ripples like the simulation, but it doesn’t really stand out like a sore thumb – say compared to the one at 500Hz which might be a room effect at this fixed radius *. Maybe a pair of curved baffle extensions could be added to the left and right of each midrange driver if it’s a problem? Perhaps apply a simple EQ filter – say 1020Hz/-2dB/8Q, but who cares? Actually, if they were near-field monitors (which they’re not) you might address it, but way back at 5 metres or so where the seat is, the room swamps it.
On-axis distortion traces:
And at 60°:
THD is about 10dB below the 31dB noise floor of the room across the entire range. And these traces include distortion of the sound card and microphone!
Here are some more 1m on-axis graphs:
Group Delay and Excess Group Delay:
I don’t have one really, or maybe I do. I do still shake my head in bewilderment though over the ever-growing “open baffle brigade” stomping around at the online audio forums. Comparisons between the above plots (which are real in-room measurements of DIY sealed boxes) and those published by the “others” including the so-called experts might be telling – and I don’t mean their hand drawings, simulations or anything veiled behind ⅓ octave smoothing! Hey there’s a reason I use sealed rigid and internally damped boxes. They contain and suppress the unwanted antiphase! Dipole speakers have net zero power response and are mere velocity sources. Don’t even get me started on the OB subwoofer stupidity! These are under-damped proximity effect transuces requiring very near-field listing to half work. They cannot pressurise a room and therefore are unintersting and unly to be used by STUPID people!
Addendum April 2015:
I added sealed box subwoofers for supreme “LF boxiness” and you can see what that looks like here.
It’s basically this:
Something that made a superb improvement to the soundstage and imaging of these speakers is now here (but they’re still boxes 😀 ).