The heatsink on the 12π doesn't get very hot. It dissipates heat as fast as it sinks it. It feels warm to the touch, but at full power it's only 130º. Without the heat exchanger, motor temperature rises to 195º under the same conditions. At that temperature, the pole piece re-radiates heat back into the voice coil, literally baking it. The glue burns and becomes brittle, and fails to hold the voice coil. When the voice coil comes undone, the speaker begins to buzz as the winding vibrates against the pole piece in the gap. Eventually, the voice coil breaks and the speaker fails entirely. The use of a heat exchanger prevents this. Power levels can be substantially increased, and the motor stays nice and cool so the voice coil glue never fails.

I don't know if you've seen the tests I did on the heat exchanger, but I did a series of temperature measurements. I measured the pole piece at various power levels, with and without the heat exchanger installed. So that kind of lets you know what to expect in terms of longevity.

• LAB12 destructive test
  
• LAB12 test with heat exchanger installed

As for compression due to voice coil heating, the acoustic measurements at the Prosound Shootout shed some light. The sweeps were actually pretty long duration, not gated MLS signals. So they probably heated the voice coil pretty well at high power levels. Compression can be seen as less than 3dB increases between measurements where power was doubled. Each speaker rises 3dB between measurements until it reaches a certain point, where SPL doesn't increase much anymore because compression has set in.

We had planned to do another series of tests that included a lengthy heat soak period, but we simply ran out of time. That would tell us what the speakers acted like after the magnets were very hot. The heat exchanger is really of most benefit in that situation, becaue it prevents the magnet from getting so hot that it re-radiates heat and bakes the voice coil.

Why us Zmin used and not some sort of average Z over used bandwidth?
Which is more accurate?

We could use any value as long as it was consistent for all speakers tested and we all agreed. But it's nice to choose a figure that is easy to repeat in other tests and appropriate for comparison with other measured datasets. We didn't have time to accurately determine average impedance using the calculated area under the curve but we could have examined each chart and estimated an average over the 20-100Hz range. We considered using (min+max/2), but this would have given an artificially inflated figure. And we considered using Zmin, which is the most conservative method of all.

In the end, we chose Zmin. It was the most consistent and least subject to interpretation. Of course, using Zmin as the value to determine RMS voltage for test gave the lowest power and SPL. If we had used some sort of average impedance value, the levels would have obviously been higher. It appears to me that using Zmin resulted in choosing voltage levels that correlated very well with expected SPL values. Zmin was David Lee's suggestion, and I think I agree with him that it is probably the best value to use when testing and when considering subwoofer setups, amplifier choices, etc.

Thanks.

James

It's an amazing box. Sound guys look askance at it until I fire it up. The sound quality from a $40 woofer and $10 worth of piezos is insane. Along with the volume from only a couple hundred watts. Not an easy build.