Data Collection
Frequency response and phase data were collected for each driver individually. All measurements were made outdoors on my driveway with the microphone set up 12 feet from the face of the speaker, laying right on the ground. The speaker was tilted forward so that the microphone, on the ground, was directly in line with the bottom of the tweeter. One-meter test distances work fine for single drivers and provide a common benchmark for comparing one driver to another, but when testing multiple drivers it is best to get as far away from the baffle as practical. I use this ground plane technique to eliminate the effects of the floor bounce on the measurements. This gives me much smoother response data, and definetely better phase data. The floor bounce really screws up the phase data. I prefer to deal with the floor bounce later, when doing actual listening tests. I have found that just putting a few pillows on the floor in front of the speaker eliminates most of the problems assosiated with this reflection. I think we are so used to hearing this way, that eliminating all of it sounds unnatural. Just my thoughts!
The tweeter data is shown first. You can see that the data is not as pretty as the stuff you see in the color brochures. There is a large peak at about 875 Hz, a dip at 2K, and then a steadily rising response up to 20K which is the upper limit of LAud. The phase data wraps around and around and is all but unreadable. Momma didn't warn me about this kind of stuff.
Measurements of the Focal midrange yield similar phase wrap.
It is important to realize that both of these measurements were made with the microphone at exactly the same position. With this in mind we can subtract the time-of-flight from baffle to microphone and unwrap the phase data so we can visualise it better. The result will be the RELATIVE phase data of these two drivers. The actual data will not change at all. Now I already mentioned that the mic was 6 feet from the baffle, so as an approximation we can figure that the sound moves through the air at approximately 1 foot per millisecond. The actual speed is closer to 13.5 inches/millisecond but the 1 foot number is good enough for now. So we can calculate that the time-of-flight from the baffle to the mic will be 6 milliseconds. Using 13.5 inches results in a time of 5.3 seconds. I subtracted 6 milliseconds and didn't get exactly what I wanted (all three phase responses within +/- 180 deg on the plot) so I played around in LAud subtracting various times from both drivers until the phase data from both drivers was unwrapped and between +/-180 degrees. As long as we subtract the same delay from all phase responses, the actual data and the relative phase between the two drivers doesn't change. What this does is allow us to look at the relative responses.
I got the best results by subtracting 5.6 seconds from each response and ended up with the data shown below. The unwrapped phase data is a little ragged due to the fact that an MLS signal was used to obtain the data and there is no option to smooth the phase data like there is with the SPL data. You will need to use an eyeball average on the phase data. When we start optimizing the phase response later, you will see that this is not that hard. The important thing about the phase data at this point is that you can read it in the area of the crossover. So, studying the graph, we can see that the phase is about 150 deg at 2000 Hz and rises to about 360 deg at 6000 hz.
Subtracting the exact same time-of-flight from the midrange data produces a phase angle of 60 deg at 2000 Hz that rises to about 210 deg at 6000 hz.
The phase difference between these two drivers is about 90 degrees without any crossover filter applied. This data looks pretty good, so the next thing I did was to save it in LAud as an ASCII file and export it over to CALSOD. Now the real fun begins.
Just for completeness, I have included the mid-bass Shiva data below. I measured it with the same setup as the other drivers, and then subtracted the same delay from the data. As before, first is shown the raw data in which the phase wraps around and around, and then the time compensated data in which the phase is unwrapped.
With the 5.6 second delay subtracted out of this data you can see that the phase at the expected crossover frequency of 250 Hz is nearly zero. What a coincidence that it came out that way!
You will note that there has been no mention of acoustic center or relative position of the drivers. The drivers are all mounted on a flat baffle and the measurments were taken at a single point in space that is even with the bottom edge of the tweeter. The acoustic center of the midrange is at least an inch farther away from the microphone, and the mid-woofer is another several inches farther than that. None of this matters as we have all of the data relative to the same point in space. This just happens to be the location of my nose when I am seated in my favorite listening chair!
Looking only at the raw data we see that at that point there are significant differences in phase from one driver to another. We will attempt to correct these anomalies with the crossover. As a last resort, we may be forced to offset the drivers or resort to an electronic delay to get the phase response to match up between drivers at the crossover frequency, but I'll bet we won't have to do this. We will explore a number of options to try to get a good frequency response and a good phase response.
Now on to Choosing a Crossover Frequency