Thank you for the summary of the discussion at AVSForum. As you may have gathered, we have an issue with the test procedures used, which I brought up before these tests began. At the time, I was assured that these problems were going to be worked through together before any data went public; unfortunately that agreement and the opportunity to increase the accuracy and relevance of the results being tested were ignored after the subwoofer was shipped.
The two main problems with the procedure are the linearity accuracy and the relationship of THD & Noise to real world conditions. First for linearity; the frequency response measurement used by Ilkka, and quite a few others, utilizes a smoothing calculation. The problem with this is that products that are linear simply remain linear (and in fact may even be worsened by smoothing), by pulling down information at the crossover or filter points. The smoothed graph of the EP600, for example, shows a gentle sloping decline or roll-off at the filter points of 15Hz and 100Hz which does not appear in reality with precise measurements. Products that are not linear, however, are made linear by this technique. Linearity is a paramount indicator of the quality of the end performance and should never be modified by using smoothing, or a low number of data points, both of which will hide anomalies in frequency response.
The THD & Noise methodology used in the Finland test is to plot a graph of 7 points from 20 to 80 Hz at a max level of 10% THD. There are a number of problems with this. First, the general problem of noise occurring during the listening experience, such as woofer bottoming, happens below 20 Hz. So in order to relate such measurements as this to the listening experience two things need to be done: one is to establish at what level of THD & Noise the listener experience will be compromised, and two is to then apply this level to all frequencies. We use 100 data points between 150Hz and 10Hz to determine THD & Noise.
Once a level of volume is attained that has audible distortion - at any single frequency - that is your maximum SPL for all frequencies, because the distortion that occurred will intrude on the listening experience each time that frequency is produced. Obviously, listening to movies or music is a dynamic experience, and frequencies change constantly, so that volume is defacto the maximum logical usable SPL output. The method of plotting distortion at various frequencies, and then connecting those points like a response curve, is not a measurement that can be translated into any usable real world condition. For us at Axiom, we have determined through extensive listening tests that this level is when the THD & Noise exceeds the signal frequency. Neither the EP500 nor EP600 will allow the THD & Noise to exceed the signal level at any frequency no matter the volume level setting. If the tester decides that 10% THD & Noise is the level at which audible distortion will occur in the listening experience (our tests have never shown this to be the case), then this, at the very least, needs to be applied to all frequencies.
As you can see, this is a fascinating subject that has spawned numerous Audio Engineering Society papers and decades of research at the National Research Council (NRC). The real fun begins when you start to prove out each measured number utilizing scientifically controlled listening tests. I can hardly believe I have been doing this research for 25 years now but I look forward to continuing to discover more over the next 25 years.
