Dayton CX150-8 Coaxial Project

[4thtry]

Thanks again for the files, John.

After reviewing your model and blended frds, I think I have found two big mistakes that I have made, one of which may be a contributing factor to the matchiing problem you are experiencing (sorry about that). When I initally started taking my on-axis measurements, I put the loudspeaker up on my horizontal polar response turntable, which raised the center of the woofer 25.25 inches above the floor. At an OmniMic distance of 39 inches and height of 55.375 inches (on tweeter axis) this created a large floor bounce reflection at 133Hz. So I removed the turntable and set the speaker on the floor for all on-axis measurements (mic 37.6 inches above floor at a distance of 39 inches). This helped to eliminate the floor bounce.

But putting the speaker on the floor was a big mistake on my part. With the speaker on the floor, the center of the woofer is only 7.5 inches above the floor. This near floor placement creates early reflections that completely invalidate the "W alone.frd" file that I posted in #131 above. And the T+W.frd, M+W.frd, and T+M+W.frd files that I posted in #132 are invalid for the same reason. This was probably a contributing factor to why you could not get a good match!

This next week I am going to place the speaker up on a 25 inch high stand that has an open frame below it. I will then re-measure all of the on-axis FRD's, using the bass removed track for the tweeter instead of the cap, and post the results. On this stand, the center of the woofer will be 32.5" above the floor. The tweeter/mic to floor height will then be 62.6", which is OK because the ceiling in this room is sloped and is 144 inches above the floor at this point.

The other big mistake that I am making is that I am trying to do my design using the raw OmniMic FF files alone, without using NF data, the blender, or minimum phase data. In comparing our files, it is obvious that my model is lacking significant BSC compared to yours. One of my problems is that I do not have excel and it is difficult for me to run the blender in open office. But, since my last failed attempt to use the blender, I have now developed some "workaround" procedures to extract the diffraction models from the blender. After adding the files together, I can blend them in OmniMic and extract minimum phase in XSim.

So it is back to the drawing board. Two steps backwards. One step forward. But a very good learning experience!!!! Thanks for all the excellent feedback!!!

[4thtry]

Here is a new set of FRDs. They were all measured without a protection capacitor, using the full bandwidth, track 2, sine wave sweep. I decided that it was easier to simply test the tweeter full range, instead of using the "bass removed" track. To be safe, I kept the testing duration as short as possible. After the tests, I played some music through the tweeter to make sure it was still OK. Subjectively, it sounded about the same as the other tweeter. No rattling or buzzing of any kind. I'm probably good to go.

TESTING SPECS:
Mic: 1 meter on tweeter axis at a height of 62.5 inches.
Speaker: Placed on top of a 25" high stand; woofer 32.2" off floor; tweeter: 62.5".
Settings: Track 2, sine sweep, blended mode, 5ms gate, 1/48th octave smoothing.
Protection cap: None
Ambient noise: Furnace turned off for all measurements
Other: Mic not moved or bumped for all measurements. Amp at 32dB volume for all.


All of these FRD's still have the raw phase data, so you will need to replace this with minimum phase data.

T alone 2.frd
M alone 2.frd
W alone 2.frd
T+M 2.frd
T+W 2.frd

[4thtry]

Here are the rest of the replacement frd's:

M+W 2.frd
T+M+W 2.frd


And the NF woofer cone measurement will also need to be replaced, as the one I posted earlier was made with the woofer near the floor. This near floor placement caused a 500-1200Hz bounce that messes up the file. So I placed the speaker on top of the 25" stand to make a new NF measurement. The bounce disappears on this new measurement.

NF woofer cone 2.frd

[4thtry]

Here are the minimum phase "get file" matches in XSim. Minimum phase was generated using XSim's derived Hilbert transform function. I attempted to use tail frequencies and slopes to provide the best overall "get file" matches. Please note that I needed to change the XSim "mod delay" values for the minimum phase verses raw phase "driver sets" as follows. In this way, I could easily flip back and forth from raw to minimum phase by simply swapping out the driver icons.


Tweeter Raw Phase mod delay: -0.88 (in)
Tweeter Min Phase mod delay: -0.03 (in)
Midrange Raw Phase mod delay: 0.00 (in)
Midrange Min Phase mod delay: 0.00 (in)
Woofer Raw Phase mod delay: +10.18 (in)
Woofer Min Phase mod delay: +9.89 (in)


The matches using the raw phase data are a little bit better than the matches using minimum phase data. The minimum phase match is a little off in the 5-10K range for the T+M and T+M+W matches. The minimum phase T+W match does not show this problem, which indicates that this is due to the tweeter and midrange phase interacting is some way. Perhaps this was due to some type of measurement error, but I do not think so. The raw phase data matches do not show this problem. I was very careful not to touch the microphone or amplifier level as I swaped wires around during the tests.

[jhollander]

Bill those are almost the same offsets I got and the same miss-match on the T+M. Only thing I can think is there is some narrow range peaking that is obscured by Omni mic smoothing.

Edit, where the smoothing makes the derived phase not accurate.

[4thtry]

I blended the NF and FF files for the woofer and midrange using Jeff B's FRD blender worksheet, open office 4.1.3, OmniMic 5.03, and XSim 1.1. Below are my output files with minimum phase replacing the raw phase data. I used a multi-step workaround process to create these files because I do not have Excel and I cannot get the blender worksheet to work properly in open office.

M blend min ph.frd
W blend min ph.frd

Edit: Correction: Charlie Laub's blender program. Sorry for error.

[Wolf]

Just for clarity- the Blender is not one of Jeff's programs. Charlie Laub created that one.

Wolf

[4thtry]

Here is my latest XSim 3-pane using blender adjusted minimum phase files. Reverse null shows the new crossovers to be 510Hz and 3500Hz. System impedance is now above 5 ohms at all frequencies. Impedance phase angle looks good. This model has the tweeter and midrange significantly padded down compared to the model I posted earlier.

Thanks, John, for posting your frd files. They helped immensely. After I merged & replaced the phase, I found that it was necessary to boost the woofer inductor all the way up to 5.6mH, just like you did, to flatten out the baffle step. I think I am finally figuring out how to do this stuff!

[4thtry]

I put together a prototype crossover with alligator clips and measured the horizontal polar response from 0-90 degrees. I am by no means an expert at interpreting this type of information, so feel free to critique. The on-axis is fairly flat from 400Hz to 20kHz, but as you move off-axis the response dips down quite a bit from 4 to 5kHz, depending on measurement angle. At 15 degrees off-axis, there is a somewhat sharp notch centered at 8kHz. This notch disappears at angles other than 15 degrees.

[rpb]

A dip at 8k is strictly the tweeter section of the speaker. The mid response is too far down to have any effect at 8k. In other words, there's nothing you can, or need to do about it.

The dip at 5k is likely the tweeter as well. I'd suggest measuring the off axis of the mid and tweeter separately, and see which one is dipping.

[ugly woofer]

The question Bill, is how does it sound? I look forward to hearing these in april. It should be fun to compare the relatively inexpensive dayton coax with Ben's much pricier coax, and for that matter my homemade junk, as well as others.

[Wolf]

I hope I get it done in time....
Wolf

[4thtry]

A dip at 8k is strictly the tweeter section of the speaker. The mid response is too far down to have any effect at 8k. In other words, there's nothing you can, or need to do about it.

The dip at 5k is likely the tweeter as well. I'd suggest measuring the off axis of the mid and tweeter separately, and see which one is dipping.

Thanks for the tip! Will do this and report back.

[4thtry]

The question Bill, is how does it sound? I look forward to hearing these in april. It should be fun to compare the relatively inexpensive dayton coax with Ben's much pricier coax, and for that matter my homemade junk, as well as others.

It is starting to sound the way I like it. I put L-pads on the tweeter and midrange so that I can shift the response up and down relative to the woofer. Problem is, if I adjust the baffle step down too much to get it to sound about right at home, it always sound too "thin" in the mid-bass when I get to one of the larger event spaces.

[jhollander]

Before I trusted the blender I would take the speaker outside and try to get a 1 meter gated response down to 150 -200 hz to check the woofer match.