Speaker Measurements 101 Page 5
A Pretty Good Way: In-Room, Averaged, Smoothed
Pros: Easy, cheap, captures what a speaker does in a real room
Cons: Doesn't allow much technical analysis of a speaker
About 15 years ago, an off-hand comment from Scott Bagby of Paradigm changed my life-or at least the part of my life where I'm measuring speakers. Hearing of the difficulties I'd had in scaling and summing bass responses, then finding the right place to splice them to quasi-anechoic response curves, he said, "Doing an average of five or six in-room measurements from different mic positions can tell you a lot about a speaker."
Boy, was he right! Since then, if I ever feel my bass measurements might be off, I'll do an averaged in-room measurement to confirm. Stereophile does an averaged in-room measurement along with quasi-anechoic speaker measurements, to give a "real world" portrayal of how a speaker performs in an actual room.
Of course, to do this, you need real-time analyzer software that can perform magnitude averages. TrueRTA, an affordable and popular PC app, can do this easily. For more on DIY measurement gear and techniques, read my article on the subject.
But where do you put the microphone? What different positions should you use? I tend to put the microphone at my seated ear height, take one measurement with the mic where my head would normally be, then move the mic about 2 feet to either side of my seat then 2 feet forward and back of my seat, staggering all the mic positions slightly so no two are ever in a line parallel to the room's walls. Stereophile's John Atkinson places the microphone in each corner of a vertical rectangular grid 36 inches wide by 18 inches high, centered on the positions of his ears. I've seen manufacturers and installers place a microphone in each of a room's seating positions, with the mics all at or near seated ear height.
I'd hoped Devantier would give me some inviolable golden rule of in-room measurement mic positions, but when I asked which technique is best, he simply said, "We don't know." But he quickly added, "It doesn't make that much difference, as long as you're moving the mic a couple of feet in each direction."
For the averaged in-room measurements we did with HATS in my living room, Devantier brought the same exact system he used to calibrate the Staples Center: a multichannel mic preamp, a computer interface, and numerous B&K measurement microphones. This way, we could do 12 measurements of one F208 all at once, with the mics placed in various locations in and around my listening chair. You can see the result in the graph with the blue trace. (Yes, I know I should have followed my own advice and smoothed the result more; this one's smoothed to only 1/12-octave, which is what I use for most of my quasi-anechoic measurements.)
Once you have your five or six or eight or 12 response curves taken at different mic positions, you can average them all together to get your speaker's averaged room response. Then apply smoothing. Start with 1/12-octave, but if that's too hashy, try 1/6-octave.
According to Devantier, there's no firm standard yet for what a typical good averaged in-room response should look like. But the frequency response plot should look fairly smooth, and should show a mild downward tilt in the tonal balance, with the peak bass output perhaps +6 dB greater than the response at 20 kHz. Incidentally, Harman has developed a way to convert its "Spin-O-Rama" 360-degree horizontal/vertical into a predicted in-room response curve; you can see the result in the graph at right with the red trace.
Even after all this effort, involving three complete measurement setups in three locations, I still somehow didn't feel quite satisfied. After all, we'd measured one of the world's best-measuring speakers. In order to get a complete picture of what our averaged in-room method can accomplish, we had to test a bad-measuring speaker.
Fortunately, I had just such a speaker handy: a homemade job I built using a 6.5-inch full-range Audio Nirvana driver. On-axis or off, this speaker measures pretty rough. Would this show up in an in-room measurement? Well, there's the averaged response at right, in the graph with the green trace. Judge for yourself.
While you probably can't use this measurement to do in-depth analysis of a speaker, such as the effects of its crossover, it will, as Bagby suggested, give you a quick and accurate picture of what you'll actually hear from a certain speaker in your listening room. And I can promise you, if you ever tell a speaker engineer that you're measuring your system this way, he'll approve.