-
Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
Constant directivity 2-way à la Geddes
- Thread starter bmc0
- Start date
ctrl
Major Contributor
It depends on your own expectations and how far down you want the "cardioid-like" frequency range to go.
If, for example, the control of the radiation pattern is desired only down to 400-500Hz, then this can be realized directly via the cabinet shape.
Make that 2Pi sensitivity. After baffle step correction what you are left with is about 92dB. Not saying that this is lousy, very few loudspeakers are true 92dB at 1W and 1m distance.
I achieved great results with 2512 in 120 liter ported enclosure with DE250-8 compression driver.
Last edited:
It's $150 per each in the USA. What midbass driver would you suggest in its place?
Crossed over at 12 dB/octave 1500 hZ it sounds fine. I had mine in a tapered 1/4 wave (TQWT) cabinet because they're easy to make.
Perhaps that augmented the low end response.
It still needed a sub.
When decommissioned, I could still reuse most of the plywood.
- Thread Starter
- #25
I haven't made any raw drivers, but I do (manual) room correction with my DSP software, which has full parametric EQ and FIR capabilities. I also designed and built a 4-channel poweramp to use with these speakers (see attached pictures). It uses the LM3886 chip, but is a bit different than most chip amps in that it has balanced inputs. The LM3886 is configured as a difference amp with NE5532 input buffers (basic instrumentation amp topology). Full KiCad design files, including gerbers are available here.
Yes, half space (2pi) is how drive unit sensitivity is usually specified.
I don't know for sure since I haven't seen many good 3rd-party measurements of drivers of this type. One of the Kapplites may be slightly better. B&C has some good offerings too, but they tend to be a bit expensive (in the US, at least).
Here's a raw ground-plane measurement I did of the Deltalite in-box (1/48 oct smoothing):
Attachments
Very tidy amp.
- Thread Starter
- #27
I still haven't done the measurements I intended to do, but I've learned enough about how to use ABEC3 and Gmsh in the last few years to actually simulate the complete loudspeaker. The simulations show that the somewhat odd vertical directivity around 600Hz or so is not a measurement error like I originally thought. Here are the measured curves again—same data as before, but normalized to the intended listening axis (18° horizontal):
And the simulated curves (60Hz-6kHz; same normalization):
The measurement distance in the simulation was 3 meters while the real measurements were done closer (I don't recall the exact distance). Even so, the agreement is quite good above ~400Hz (minimum valid frequency for the measured data).
Measured and simulated horizontal contours:
Vertical contours:
I ran another simulation with the waveguide completely removed and found that the aberrations in the woofer's vertical responses went away. Maybe a freestanding waveguide would cause less disruption?
And the simulated curves (60Hz-6kHz; same normalization):
The measurement distance in the simulation was 3 meters while the real measurements were done closer (I don't recall the exact distance). Even so, the agreement is quite good above ~400Hz (minimum valid frequency for the measured data).
Measured and simulated horizontal contours:
Vertical contours:
I ran another simulation with the waveguide completely removed and found that the aberrations in the woofer's vertical responses went away. Maybe a freestanding waveguide would cause less disruption?
Last edited:
- Thread Starter
- #28
I've been playing with Bempp-cl recently and found that I was able to get slightly more accurate results than with ABEC3, especially off axis. I based my script on the loudspeaker radiation tutorial, but I'm using the LU solver rather than GMRES. It seems to me that LU works much better for Helmholtz problems of this size. GMRES tends to converge very slowly at high frequencies (at least without some clever preconditioning algorithm) and gives spurious solutions at certain frequencies.
I generated two meshes for this simulation—one for 50-5kHz and a higher resolution one for 5kHz-15kHz—then concatenated the results. Measured on the left; simulated on the right:
CTA-2034-A (corrected according to T. Welti):
Horizontal contours:
Vertical contours:
Positive horizontal curves:
Positive vertical curves:
Negative vertical curves:
As before, the minimum valid frequency for the measured data is ~400Hz and the observation distances aren't identical (3 meters for the sim vs 1.5-ish meters for the real measurements, IIRC). Both measured and simulated data are shown normalized to the intended listening axis (18° horizontal). Limitations considered, the agreement seems virtually perfect.
I generated two meshes for this simulation—one for 50-5kHz and a higher resolution one for 5kHz-15kHz—then concatenated the results. Measured on the left; simulated on the right:
CTA-2034-A (corrected according to T. Welti):
Horizontal contours:
Vertical contours:
Positive horizontal curves:
Positive vertical curves:
Negative vertical curves:
As before, the minimum valid frequency for the measured data is ~400Hz and the observation distances aren't identical (3 meters for the sim vs 1.5-ish meters for the real measurements, IIRC). Both measured and simulated data are shown normalized to the intended listening axis (18° horizontal). Limitations considered, the agreement seems virtually perfect.
