The following is a description of my test conditions. While I don't claim that these are NRC quality, the value is in the fact that all of the drivers are tested under identical conditions. So while the results may not be absolute, they are relatively correct.
Test conditions are included in the filename.
The frequency response measurements are done at 1 and 2 meters using an MLS signal. The nearest boundary is 65" from the driver, so for farfield measurements the gate is 7ms for 1 meter, and 5.5ms for the 2 meter measurement. For a quick primer, the gate length determines the low frequency cutoff AND the frequency resolution. For 7ms the cutoff is 143hz. This means the first valid data point is 143hz. The frequency resolution is also 143hz, meaning the second valid data point is 286hz, and so on. The graph has a shaded area under the cutoff so you know where it is. Personally I consider anything under the second data point to be invalid. All of the responses are UNSMOOTHED. The SPL level is calibrated to be within a 1dB. Due to the soundcard cutoff you will see a spike in the response above 20khz that is not real. This also causes a ridge in the CSD.
The nearfield measurement is done with the mic on the driver center axis and intersecting the baffle plane. Gate for these is 500ms. The SPL level is NOT calibrated. 1/3 octave smoothing is applied. Under 30hz or so the response is totally affected by the rumble of the environment so don't consider that accurate. Now I use a calibration file for the mic (ECM8000), but I have read that this not because it is inaccurate, but because it is an omni mic, which causes a high frequency rise on axis. With the mic this close to the woofer it's possible I should not be using the cal file. The high frequencies may have more rise in reality than is shown in the nearfield plot when I use the cal file. I've included a pic of the cal file response, use this to help you determine how the response would look if I was not using it. If anyone has some advice here, I'd appreciate it. I can always go back and run the response with out the cal file later if I want.
I use a large baffle that was optimized to provide the smoothest diffraction down to the cutoff frequency. Attached is a pic of the sim. It measures 65" W by 88" H and the driver is centered at (29,40). The driver back is open, no enclosure is used.
Impedance is pretty straightforward, although I will point out that I have a little wiggle between 10-20khz. Despite my best efforts and different soundcards it's always been there. 500ms gate is used.
These are actually quite tricky to do well and consistently, but I'm now comfortable with me method. The driver is mounted vertically on the test baffle. I use the added mass method which can be tricky, but I have it working pretty well now. I settle the suspension with a 25hz sine before each measurement, and all drivers are well broke in. I measure the diameter using about 1/3 of the surround. Soundeasy uses a curve fitting method for impedance that is quite accurate if you spend some time to optimize it. The one thing that all of these softwares have a problem with are drivers with copper sleeves over the pole. Some of these types of drivers may have a measured Le slightly higher than what it really is. Obviously the exact value isn't that important on these drivers, as they tend to be very good already. An "x" is put in any parameter that I haven't measured or the manufacturer doesn't have reliable info for. I'm currently not measuring the 15" drivers as the large cone mass is just to difficult to deal with.
Le(x) is probably the most overlooked parameter by the average diy'er. It's not hard to do even crudely, just hold the cone at roughly +/- xmax while you take an impedance measurement. While even a novice diy'er has the ability to do impedance and frequency response measurements, these are really static or small signal measurements, leaving the diy'er without any knowledge what a driver does under real world use, i.e. dynamically. Le(X) is a static test that can give you a window into how the driver behaves dynamically. By definition it is the change in inductance versus cone excursion. This can tell you many things. If the inductance changes-as seen in the resulting change in impedance-the frequency response will also have to change. This is then magnified by the passive crossover, which has values based on the "at rest" impedance. Then you have the effect on non-linear distortion. Changes in inductance have a huge effect on IMD and cause some of the most audible distortions. The reason for shorting rings is to combat this very issue. As the inductance rise is the area of interest, don't mind the lower frequencies, they will not look good due to the cone be held in place. I've decided to do a standard excursion for a given family of drivers, instead of at the rated xmax of each individual driver. This better fits the idea of xmax being a spec based on performance and not simply the VC versus gap height.This also allows us to compare drivers on an equal footing, since in the end one of the things we may want to know is whether driver A is going to affect the passive crossover operation more or less than driver B when both are moving say, 4mm.
Power compression testing is done by applying ten minutes of pink noise at 105dB @ 1m, unless otherwise posted.
Distortion measurements consist of a harmonic distortion sweep in the usual passband of the driver. Standard is 100dB at about 30 inches, and the mic at 15 inches. Additional plots may be done and the measurement conditions will be in the file name. Not all drivers will have the same setup, scale, or range, but similar drivers will have the same setup. There also several multitone plots with the second and third tones being +/- 10% of the fundamental and 6dB down, and a two tone plot with equal amplitude tones. These are to get an idea of IMD. The Xmax testing is also a good indicator of IMD.
I had intended to use Klippel's Performance Based method for determining xmax as outlined here: http://www.klippel.de/pubs/Klippel%20papers/Assessment_of_Voice_coil_peak_displacement_XMAX_02.pdf But I ran into two roadblocks, the soundcard clips at around 19 volts, and the room totally dominates the harmonic distortion in the bass and inflates the figure. Basically the Klippel method is a two step test. First step is a simple harmonic distortion test by driving the speaker at its Fs to a given excursion and recording the THD, and stopping when THD equals 10%. The second step is to again drive the speaker at Fs then apply a second tone at 8.5 x Fs. This creates intermodular distortion, and again raise excursion until IMD equals 10%. Whichever step reached 10% at the lowest excursion defines the xmax. Now my hope was that most drivers would be limited by IMD which isn't to affected by room. As it turns out many are THD limited. But because of room effects the HD portion of the test is not absolutely accurate, only relatively accurate to the other drivers. Obviously we need an absolute number to since we are defining xmax from it. But the tests do still have value as a relative comparison so I did them all anyway and posted them in the Xmax section. There will be a text document that records the % distortion I recorded on Soundeasy at the given excursion, and some pics of the spectrum analyzer. Because of the 19v clipping issue I did not go out to the same excursion from every driver. For example the B&C 15NW76 needs a hell of a lot of power to drive it to large excursions. I could only get 3mm before the soundcard clipped.
EDIT 8/24/09: I've had to change the testing protocol. The preceding method will be referred to as the original method, and will be noted in the small group of 12" and 15" woofers for which it is was done. The new method will use a standard tone of 40hz instead of driving at the respective drivers' Fs. The second tone will still be at 8.5Fs or 340hz, and will still be 12dB down. Since it is often impossible to either directly measure some cones with my Linkwitz excursion widget (like the Accuton) nor is it always possible to do TS measurements, I'll be using the manufacturers TS to predict the excursion for a given input voltage. This obviously introduces some error into the method that was not there for the original testgroup for which I directly measured excursion. But there isn't any alternative. Also I will only be listing the IMD as you can figure out the THD by looking at the HD sweeps.
The results of these drivers are comparable to the ones in the older measurement group, except for harmonic distortion and Le(x). In the old group the Le(x) was as crude as can be, totally eyeballed. Also I changed the test baffle to 3/4" birch ply from the old MDF, I'm using more room damping for distortion testing, and last year I added some 1/2" insulation to my garage door. So while the distortion results should be close to the old method, I can't with confidence say they that they are. So be aware.