It also depends on how you use the multiple smaller woofers. In the XSD, they are opposed slot loaded drive with the magnet on one facing the cone on the other. The motion cancels suspension induced even order distortion to a large extent. You can get really clean response from qnty 8 x cheap 6.5in woofers. These 8 woofers cost about $100 and have the equivalent cone area of more than two 12in woofers but with a much smaller frontal cross section due to SLOB mounting. This is for moderate drive level of 2.0Vrms (1W into 4ohms).
Or just don't understand what (or even why) you're measuring.
In some cases not having enough or the right amount of knowledge or skills.
But I guess that kinda means the same as what you were saying.
It also always sounds that "measuring" itself is a well defined proces, almost black and white, that can be learned from a textbook or so.
In practice it very rarely is.
It's a very iterative proces, based partially on skill and intuition, mixed with a good chuck of theoretical and practical knowledge.
At the risk of sounding like an academic, which I'm not although I was indoctrinated by a school full of them, and at further risk from joining this discussion late, I'd like to add my two cents:
time domain (impulse response) and frequency domain are alternate, mathematically equivalent ways of looking at the same thing.
these mathematical analyses depend upon the system being linear, which for speakers is generally only true for small signals. thus ,an argument for designing with lots of headroom. Even if the speaker remains in its linear range, the room will have its way with the response. Thus, the need/desire for (quasi) anechoic measurements.
fourier analysis applies to periodic waveforms and therefore yields steady state frequency response, but music isn't steady state so clearly a need to look at more than just frequency response.
we look at impulse response, step response, frequency response, distortion graphs, CSD, DI, polar maps, ... all of which highlight different aspects of the speaker's performance. We need that highlighting because its extraordinarily difficult to see all that there is to see in an impulse response. Its a critical question of what is a sufficient set of measurements or graphs to check before declaring the job done. I don't have an answer for that; I just know that the one you leave out is where you will get bitten.
time domain (impulse response) and frequency domain are alternate, mathematically equivalent ways of looking at the same thing.
these mathematical analyses depend upon the system being linear, which for speakers is generally only true for small signals. thus ,an argument for designing with lots of headroom. Even if the speaker remains in its linear range, the room will have its way with the response. Thus, the need/desire for (quasi) anechoic measurements.
fourier analysis applies to periodic waveforms and therefore yields steady state frequency response, but music isn't steady state so clearly a need to look at more than just frequency response.
we look at impulse response, step response, frequency response, distortion graphs, CSD, DI, polar maps, ... all of which highlight different aspects of the speaker's performance. We need that highlighting because its extraordinarily difficult to see all that there is to see in an impulse response. Its a critical question of what is a sufficient set of measurements or graphs to check before declaring the job done. I don't have an answer for that; I just know that the one you leave out is where you will get bitten.
We can agree on the validity of the pulse reproduction of the DUT and following FFT analysis like KEF used I presume. My take is that sweeps, MLS and periodic noise have proven to be very useful replacements with higher S/N. And the classic sweep or multi tone tests bring enough on nonlinear distortion for even the discerning home user.
The better being the enemy of the good, of course and likely there are better ways. But I think in normal listening environments including elaborate home theatre settings above ways are more than adequate. And I think backing results up with ABX testing is mandatory.
Last edited: