diffraction effects from tweeter faceplate

[mazurek]

The effects of diffraction have made my crossover design challenging for my latest speakers because it is hard to tell whether the diffraction is important. My best sounding crossover ignores diffraction bumps and lets the frequency response vary a couple dB on axis. I sold my current system, and will be building a copy, so I have the opportunity to change from the ScanSpeak 6600 to the 6620 which looks like it has diffraction mitigating features. I gave up on applying my own felt, which does improve the graph (unsure about sound improvement), because of aesthetics.

http://www.madisound.com/catalog/PDF/scanspeak/D3004_660000.pdf

http://www.madisound.com/catalog/PDF/D3004_662000.pdf

The main features in the illuminator version of the 6600 Aircirc seem to be a slightly different faceplate that is slightly deeper and has some pattern on it. Response curves look identical, but I think its because they reused their old measurement (my driver to driver consistency is not as good as that).

Any thoughts on whether these features will have an actual effect, or just look cool? If I am convinced, I will post difference curves between the two models.

[dlr]

The effects of diffraction have made my crossover design challenging for my latest speakers because it is hard to tell whether the diffraction is important. My best sounding crossover ignores diffraction bumps and lets the frequency response vary a couple dB on axis. I sold my current system, and will be building a copy, so I have the opportunity to change from the ScanSpeak 6600 to the 6620 which looks like it has diffraction mitigating features. I gave up on applying my own felt, which does improve the graph (unsure about sound improvement), because of aesthetics.

http://www.madisound.com/catalog/PDF/scanspeak/D3004_660000.pdf

http://www.madisound.com/catalog/PDF/D3004_662000.pdf

The main features in the illuminator version of the 6600 Aircirc seem to be a slightly different faceplate that is slightly deeper and has some pattern on it. Response curves look identical, but I think its because they reused their old measurement (my driver to driver consistency is not as good as that).

Any thoughts on whether these features will have an actual effect, or just look cool? If I am convinced, I will post difference curves between the two models.

There will likely be no discernible difference in diffraction between these two drivers. The changes due to the faceplate will not affect the lower frequencies where diffraction occurs. It's not enough of a waveguide in that sense.

The curves you mention are the same ones, reused as you point out. I'm interested in seeing the changes for the new one as well. Were I buying now, unless there is something else significantly different, I'd go with the 6600 because they're cheaper and I doubt that there will be much if any audible change. My guess is that they're going to the new faceplate simply to keep the dome behind the front plate plane and they can claim a thicker, more solid mount.

WRT the diffraction and design, it sounds like you've taken the best route. If you dispense with diffraction control, then the smoothest power response will be that which has some on-axis irregularities. One benefit of good diffraction control is making it easier to design with fewer measurement angles.

Try making a series of measurements, one on-axis and several off-axis, say 5-10-15, left and right, then average the responses (magnitude only), a so-called windowed average. You can use the free version of Praxis for this, it has math functions available (there'll be a learning curve with it). Be sure to not use phase when averaging, as this is unreliable given the mic positioning issues. This will give you a rough idea of the power response, a better indicator of what the perception of the system is in-room.

Consider also using diffraction simulation software such as the Edge and the BDS (FRC Consortium spreadsheet) if you haven't already. The position on the baffle has a significant impact on the diffraction, though you may be well aware of that.

I have a section at my site on windowed average measurements. It's very old, one of my first sections back when I was first learning how to make a site, so I see after moving it a while back hat even the background is missing in spots, but the data (graphs) are there. It was also back in my early days of experimenting with diffraction, before the BDS and other simulation software was available, at least that I knew.

Looks like I should completely re-do that page anyway.

Dave

[mazurek]

Thanks for the suggestions Dave, I will do some reading on the math you suggest, and check into your site. I have done averages in ARTA using their power average function, I'll have to see what type of averaging math you are using. The power average will even out the effects of diffraction, but emphasizes larger values more in the average, so large diffraction effects seem to perturb the average.

I've modeled the response in the Edge, which never gives me exact results, but does help with design decisions. I am keeping a symmetric layout, and will likely push the tweeter closer to the top, into the roundover. After experimenting with some acoustic foam, it appears that the top to bottom dimension is responsible for the most of my problems.

I believe this is the source of the moderate brightness people were talking about at DIYNE. At that time, I had subtly notched the diffraction and level matched the drivers at the xover point (diffraction area), leaving the rest of the tweeter response slightly out of balance. When I got rid of the notch and re-leveled the drivers, they were less bright. I also slightly adjusted the phase on my subwoofers for more punch, so the audience feedback from your event was quite helpful.

[mazurek]

Checked out the praxis manual and the ARTA manual. Leaving windowing aside (I already apply whatever window I can get away with), praxis has 5 types of averages. RMS would be what ARTA does. I think log average may be what I want to figure out the true low end power response, or magnitude.

One thing I did to see what is really happening is just measure the nearfield response, and assume that the tweeter faceplate has no real effects on its power response. Baffle step was fairly intuitive to see in the measurements, so in my view the leftover response difference between farfield - (nearfield - baffle step) is approximately equal to the diffraction effects.

Now that I have my curve of what I feel are the important aspects of the tweeter response, I'll have to see whether there's a better spatial averaging procedure to arrive at the relevent response curve.

[dlr]

Checked out the praxis manual and the ARTA manual. Leaving windowing aside (I already apply whatever window I can get away with), praxis has 5 types of averages. RMS would be what ARTA does. I think log average may be what I want to figure out the true low end power response, or magnitude.

One thing I did to see what is really happening is just measure the nearfield response, and assume that the tweeter faceplate has no real effects on its power response. Baffle step was fairly intuitive to see in the measurements, so in my view the leftover response difference between farfield - (nearfield - baffle step) is approximately equal to the diffraction effects.

Now that I have my curve of what I feel are the important aspects of the tweeter response, I'll have to see whether there's a better spatial averaging procedure to arrive at the relevant response curve.

When I say windowed average, I'm referring to an average of measurements across a listening window, a spatial average, no relation to window as it refers to an MLS measurement. Be sure that the average is magnitude only. There should be an option for that.

The last time I made an average I used LAUD with its built-in function. I've not actually used Praxis for this, I use it for other post-processing. I'm assuming it will average in the same manner, since Bill Waslo authored both packages.

Be careful about assuming an approximation on the diffraction between near- and far-field. Distance can have a large impact.

Dave