The midrange and tweeter measurements

This pair needs to be measured carefully to build an accurate crossover. Not only do we need to address the cone breakup challenges of the midrange, we also need to have a carefully phase-coherent crossover point between these two drivers. Set up my Dayton mic and Speaker Workshop and took some measurements. This is what I got.

Here is the published and the measured SPL curve for the tweeter. The published curve is screen-grabbed from the Scanspeak Discovery datasheet.

And here is the published and the measured SPL for the midrange (the Dayton RS150).

It's nice to see the measured responses are similar to the published curves. Some of the differences between them will of course be due to my measurements being on the actual enclosure while the published figures are of drivers fitted on an IEC baffle.

Both these measurements were taken with the mic in the exact same position and with the various gain settings unchanged, therefore I am quite certain that I am getting accurate relative amplitude and the data is phase-aligned for crossover design purposes. No time alignment to correct here.

Nearfield measurements of midrange and woofer

The woofer measurement need to be near-field. I don't need to worry much about its phase coherence (at frequencies as low as 150Hz, phase is a pretty gross issue) and its amplitude will be corrected by parametric equalisers in my crossover anyway, for room correction purposes.

I put the mic a few millimetres from the cone of the driver, adjusted gains at various points, set the number of repetitions in Speaker Workshop measurement settings to 20, and fired away. I realised how sensitive the gain settings are for nearfield. A tiny turn on the mixer's gain knob makes the VU meter reading in SW jump from 18000 to max. Got to tweak it to keep it somewhere in the 20,000s.

I took three pulse readings of each driver, moving the mic around a bit, keeping the mic a few mm away from the cone, and somewhere off-centre. I then took a fourth reading placing the mic exactly a few mm in front of the tip of the phase plug. The textbooks say that nearfield measurements need to be taken with the mic in front of the dust cap, but then the dust cap moves with the cone, but these drivers have phase plugs, and phase plugs don't move. Therefore, I thought I'll take a few readings from just in front of the cone surface and one from in front of the phase plug.

While taking the woofer readings, I got lazy and did not move the mic around much for the first three. The fourth, with the mic in front of the phase plug, is done by moving the mic to the new position.

From the pulse measurements, I computed the frequency response curves, ignoring gating and taking into account the mic's calibration curve. I used the Blackman algorithm for the transformation. (Does the algorithm make a big difference?)

This is what I got for the midrange.

And this is what I got for the woofer.

All the curves have X-axis from 10 Hz to 2000 Hz. All curves have been smoothed to 1/16-octave. The raw unsmoothed curves were very "hairy". I don't usually smooth my measurements, but these were uniformly too hairy to look at. My gated farfield measurements have no smoothing.

I have shown all four readings in each case. There are repeatable peak-to-peak spikes and troughs of 10-12dB in both drivers. What causes these?

Try as I might, I could not get clean and realistic-looking graphs. I tried with both the midrange and woofer, at various mic distances, various sampling rates and sample sizes, at various gain settings. Nothing worked. So I called up Gooroo Angshoo, who is my go-to gooroo when all else fails. He suggested various things, but one of his suggestions struck me: he said that he never takes a pulse response for nearfield measurements. He takes a direct nearfield frequency measurement by telling Speaker Workshop to do the job. (Internally, the software must be going by the pulse route, but somewhere there is a difference.) So I tried it and immediately got believable curves:

I am showing two curves here, taken at different distances from the cone surface, to show that the curves were quite consistent. The absolute dB level of the SPL is inconsequential, because each movement of the mic position changes the absolute dB levels sharply with nearfield measurements. All these curves have been smoothed 1/16 octave.

For the midrange, I got this:

Once again, this has been smoothed 1/16 octave.

From these graphs, it was clear that I was getting acoustic rolloff of the mid at about 100 Hz, and for the woofer at about 50 Hz. I did not need the precise Q of the rolloff, because the final sound I will experience will be impacted much more by room modes than the Q of the speaker.

So, now I could think of designing a crossover.