I cut the woofer holes & rebates and set the donuts on top for a few more photos. A little spooky looking, even without the drivers!! The donuts will be removable, held in place with four 1/4-20 bolts, washers, and tee nuts. I want to be able to remove the donuts for final sanding, painting, and finishing.
I gave the vertical tower edges a 3/4" radius round-over, but stopped 6 inches from the bottom so that I could flush mount 4 angled base boards later. These angled base boards will be designed to tip the entire speaker back by about 5 degrees to compensate for the forward, top heavy weight of the big donuts.
I plan to mount the crossover externally on top of the tower in the empty space directly behind the donut. This will look kind of ugly from the side, so I am going to cover the xover with a half dome piece of 8" diameter concrete forming tube.
Hey Bill- If you have the option, I would try the Tweemid in the donuts as you have them, and rotated 180 degrees. One driver may react differently to the position than the other, and possibly for the better.
This will be an accidental ML-TL alignment. So I built a small "trap" to go directly below the woofer. This trap moves the woofer approximately 10 inches down the line. The trap also extends the line length from 31.5 to 47 inches. The area on both sides of the trap and the area behind the woofer will be stuffed completely with about 16 ounces of denim. The area below the trap will be left open as a resonating chamber and lined with 1/4" thick denim insulation.
Hey Bill- If you have the option, I would try the Tweemid in the donuts as you have them, and rotated 180 degrees. One driver may react differently to the position than the other, and possibly for the better.
Good idea. When I run up the FRD/ZMA's in the new week or so, I will take measurements with the Tweemid's rotated 180 degrees. Be interesting to see which configuration measures better on and off axis.
I gave the vertical tower edges a 3/4" radius round-over, but stopped 6 inches from the bottom so that I could flush mount 4 angled base boards later. These angled base boards will be designed to tip the entire speaker back by about 5 degrees to compensate for the forward, top heavy weight of the big donuts.
I enjoy visiting this thread to see what you have done next. You've put in a metric $hit-ton of work for sure, not to mention thinking outside the box. Maybe I missed it, will these be coming to a meet?
Your dedication to try the road less traveled always amazes me Bill. These certainly have the MWAF unique design category nailed!
Thanks, Ed. Once again, I am attempting to boost my MWAF originality score. The originality score is more than appearance. I also need to come up with some good scientific reasoning behind my design. I'm not a very good public speaker, so when it comes my turn to talk, I will have to practice coming up with a quick way to "splain" how diffraction ripple works.
I enjoy visiting this thread to see what you have done next. You've put in a metric $hit-ton of work for sure, not to mention thinking outside the box. Maybe I missed it, will these be coming to a meet?
Thanks, Kornbread. Yes, Dr. Diff & I will be attending a number of meets this year. My current plan is InDIYana in April, MWAF in July, and IowaDIY in October.
Port & stuffing progress: I clamped the back on temporarily for some NF cone & port measurements. They are really looking good so far. The measured near field half space F3/6/10 is 29/25/22, better than what I was expecting based on my initial WinISD models. My WinISD standard model showed an Fb of 35, with a somewhat higher F3/6/10 of 32/29/27.
With the same length port (about 5 inches) the final ML-TL tuning dropped to 31 Hz. Should dig fairly deep and play moderately loud before crashing.
To improve the alignment, I replaced the upper cabinet denim stuffing shown in the picture with a smaller amount of Acousta-Stuf, which helped to boost the main resonance by about 2dB. With the heavy denim, I was losing too much output at resonance due to overstuffing. The resulting line cancellation/reinforcement modes are still well controlled, even with the smaller amount of stuffing used. Hopefully it will sound as good as it measures.
Alignment question: The latest version of WinISD (0.7.0.950) has a check box for "Use Transmission line model for port simulation" on the advanced tab. In my alignment, the difference is a tuning of 31Hz (TL model) verses 35Hz (std model). The TL model shows the same low frequency roll off as the standard model, but my actual summed NF measurements show a greater low frequency extension of about 3 to 5Hz for the TL model. I've browsed the forums for more info on the WinISD TL check box, but came up empty. My guess is that WinISD, like Hornresp, does not include stuffing in the model and therefore cannot give an accurate F3 prediction for elongated, partially stuffed cabinets (ML-TLs) The Leonard Audio TL program, which I have used quite a bit, includes stuffing in the model, but the stuffing entry is a guessing game and needs to be boosted by approximately 2X. As I understand it from reading many threads, only the MJK TL spreadsheets can model stuffing accurately (I have never used the MJK TL program). http://quarter-wave.com/ Does that sound reasonable?
To keep things precisely positioned & square, I made a 1.5" thick template to drill the 4 donut mounting holes. I cut the template to the exact size of the donut bottoms. I used my drill press to do this to prevent vertical hole drift (or run out).
I then clamped the template to the top of the towers and drilled the 4 holes with a 5/16" drill bit. Since I needed to use my hand held drill to do this, the extra thickness of the template helped me to keep the bit perfectly square as I drilled.
I then installed four 1/4-20 T-nuts from the underside (inside of cab).
I then flipped the donuts upside down and sandwiched them between the two towers. I clamped the entire assembly securely in place with two long bar clamps. This made it possible for me to drill from a very stable standing position.
I then clamped the mounting hole template to the bottom, squared it up, and drilled the 4 holes through the donut bases. The thick template kept the bit perfectly square as it went through the thick 2.6" thick donut bases.
Finally, I installed the four 1/4-20 by 3.5" long bolts/washers. Both speakers produced a perfect fit, with no binding or alignment problems.
Horn Response does include stuffing in the TL model. IMO the model differences are due to port end corrections differences. Are you adjusting the port spl by the sd method?
I adjusted the port SPL with the equation used by Joe D'Appolito in his book "Testing Loudspeakers". (ie., 20 times the base 10 log of the ratio of the two diameters, which yields an adjustment of -6.8dB.) Then I applied a time delay of 2.8ms to account for the time it takes for the port output to travel around the cabinet and up to the woofer cone.
I did some more browsing around in Hornresp and cannot find where to do stuffing in an ML-TL. It only seems to allow adjustment to the lining of the rear chamber, which is not used in an ML-TL schematic. Mass loaded transmission line sections and ports are, as far as I can tell, modeled only as flared horn segments, and, these horn segments cannot be stuffed or lined. When I run the Acoustic Power FR graph they always have the big peaks and dips of an unlined TL. I'm using version 50.50 of Hornresp. I must be missing something.
Port update: I don't like the thin-walled plastic flares and ports that are generally used. Seems to me that these thin-walled plastic pipes and flares could easily vibrate or resonate badly at some mid or high frequency compared to the thicker PVC. So, I made mine out of schedule 40 PVC plumbing pipes and added 3/4" radius flares to both ends. Then I boosted the overall length from 5 to 6 inches to compensate for the flares. Tests show that the FB is still right at 31Hz, so I am good to go.
I adjusted the port SPL with the equation used by Joe D'Appolito in his book "Testing Loudspeakers". (ie., 20 times the base 10 log of the ratio of the two diameters, which yields an adjustment of -6.8dB.) Then I applied a time delay of 2.8ms to account for the time it takes for the port output to travel around the cabinet and up to the woofer cone.
Port and driver output should overlap at low frequencies, and converge at 0Hz. It looks like your port output shown above is about 5dB too high.
Port and driver output should overlap at low frequencies, and converge at 0Hz. It looks like your port output shown above is about 5dB too high.
Thanks for the tip. I was not aware of this. I must be entering the port/cone ratio incorrectly when I plug it into the adjustment formula. I will make the correction and re-post my graph.
No prob. The formula of 20log (port diameter / cone diameter) is assuming both measurements are near field and you haven't adjusted any volume / gain knobs in between measurements. The visual check is easier IMO and avoids doing icky math.
I adjusted the port SPL with the equation used by Joe D'Appolito in his book "Testing Loudspeakers". (ie., 20 times the base 10 log of the ratio of the two diameters, which yields an adjustment of -6.8dB.) Then I applied a time delay of 2.8ms to account for the time it takes for the port output to travel around the cabinet and up to the woofer cone.
Port and driver output should overlap at low frequencies, and converge at 0Hz. It looks like your port output shown above is about 5dB too high.
Here are the revised graphs. Thanks for steering me in the right direction on this. Turns out, my original NF cone measurement was in error; I must have bumped the amp volume control setting between measurements and/or placed the mic in the wrong NF measuring position (or something like that).
I therefore re-did both measurements. As luck would have it, my new summation comes out even better looking than before. The peaky looking response at 40Hz is now much flatter, down about 3dB, in the new summation. And, as you can see, the individual cone and port curves now line up at the very low frequencies, as they should. The port corrective adjustment is still -6.8dB (20log(5.25/2.4). The new measurements use flared 6" ports instead of the original 5" non-flared ports. Also, I revised the internal line stuffing a little bit to improve port resonances in the 100-800Hz region.
I should also mention one other thing that may have contributed to the measurement error. My experience has been that the SPL level of near field measurements is very sensitive to exact microphone tip placement. Very small changes in mic distance to woofer cone or port terminus plane can have a significant impact. Moving the mic as little as 1/8" closer or further away can bump the SPL up or down 1dB or so!! You have to be very careful. With a port flare, I place the mic tip half way into the flare. With a non-flared port, I place the tip in the center of the exact plane of the port terminus. With a woofer cone, I place the tip 1/4" from the center of the cone.
A little more progress (slow, but sure, I'm getting there):
With the box fully lined and the port tuning done, I could now glue the rear panel in place. So I sanded the perimeter with 40 grit to rough up the pre-laminated particle board surface.
I clamped several tweeter cutout slugs along the perimeter as alignment gauges, and then applied a thick bead of glue along the perimeter.
I dropped the main cabinet into the alignment gauges and watched the glue slowly squeeze out onto the wax paper below. As this happened, I went around the box checking the alignment, carefully adjusting it with a small hammer until it was perfect. In addition, I waited for about 5 minutes or so before applying the bar clamps because I wanted to give the glue a little extra time to soak in.
After first round of XSim modeling: 12 parts! 900 & 3.7kHz; 2nd order electrical; tank on woofer to tame the 5kHz peaking. Good phase tracking, however, I found it necessary to flip the tweeter phase. XSim mod delay offsets are -0.03 on the tweeter and + 2 inches on the woofer. BSC: 4 to 5dB; Sensitivity: 85dB.
Comments
Some progress:
I cut the woofer holes & rebates and set the donuts on top for a few more photos. A little spooky looking, even without the drivers!! The donuts will be removable, held in place with four 1/4-20 bolts, washers, and tee nuts. I want to be able to remove the donuts for final sanding, painting, and finishing.
I gave the vertical tower edges a 3/4" radius round-over, but stopped 6 inches from the bottom so that I could flush mount 4 angled base boards later. These angled base boards will be designed to tip the entire speaker back by about 5 degrees to compensate for the forward, top heavy weight of the big donuts.
I plan to mount the crossover externally on top of the tower in the empty space directly behind the donut. This will look kind of ugly from the side, so I am going to cover the xover with a half dome piece of 8" diameter concrete forming tube.
InDIYana Event Website
This will be an accidental ML-TL alignment. So I built a small "trap" to go directly below the woofer. This trap moves the woofer approximately 10 inches down the line. The trap also extends the line length from 31.5 to 47 inches. The area on both sides of the trap and the area behind the woofer will be stuffed completely with about 16 ounces of denim. The area below the trap will be left open as a resonating chamber and lined with 1/4" thick denim insulation.
Good idea. When I run up the FRD/ZMA's in the new week or so, I will take measurements with the Tweemid's rotated 180 degrees. Be interesting to see which configuration measures better on and off axis.
Thanks, Ed. Once again, I am attempting to boost my MWAF originality score. The originality score is more than appearance. I also need to come up with some good scientific reasoning behind my design. I'm not a very good public speaker, so when it comes my turn to talk, I will have to practice coming up with a quick way to "splain" how diffraction ripple works.
Port & stuffing progress: I clamped the back on temporarily for some NF cone & port measurements. They are really looking good so far. The measured near field half space F3/6/10 is 29/25/22, better than what I was expecting based on my initial WinISD models. My WinISD standard model showed an Fb of 35, with a somewhat higher F3/6/10 of 32/29/27.
With the same length port (about 5 inches) the final ML-TL tuning dropped to 31 Hz. Should dig fairly deep and play moderately loud before crashing.
To improve the alignment, I replaced the upper cabinet denim stuffing shown in the picture with a smaller amount of Acousta-Stuf, which helped to boost the main resonance by about 2dB. With the heavy denim, I was losing too much output at resonance due to overstuffing. The resulting line cancellation/reinforcement modes are still well controlled, even with the smaller amount of stuffing used. Hopefully it will sound as good as it measures.
Alignment question: The latest version of WinISD (0.7.0.950) has a check box for "Use Transmission line model for port simulation" on the advanced tab. In my alignment, the difference is a tuning of 31Hz (TL model) verses 35Hz (std model). The TL model shows the same low frequency roll off as the standard model, but my actual summed NF measurements show a greater low frequency extension of about 3 to 5Hz for the TL model. I've browsed the forums for more info on the WinISD TL check box, but came up empty. My guess is that WinISD, like Hornresp, does not include stuffing in the model and therefore cannot give an accurate F3 prediction for elongated, partially stuffed cabinets (ML-TLs) The Leonard Audio TL program, which I have used quite a bit, includes stuffing in the model, but the stuffing entry is a guessing game and needs to be boosted by approximately 2X. As I understand it from reading many threads, only the MJK TL spreadsheets can model stuffing accurately (I have never used the MJK TL program). http://quarter-wave.com/ Does that sound reasonable?
Some more progress:
To keep things precisely positioned & square, I made a 1.5" thick template to drill the 4 donut mounting holes. I cut the template to the exact size of the donut bottoms. I used my drill press to do this to prevent vertical hole drift (or run out).
I then clamped the template to the top of the towers and drilled the 4 holes with a 5/16" drill bit. Since I needed to use my hand held drill to do this, the extra thickness of the template helped me to keep the bit perfectly square as I drilled.
I then installed four 1/4-20 T-nuts from the underside (inside of cab).
I then flipped the donuts upside down and sandwiched them between the two towers. I clamped the entire assembly securely in place with two long bar clamps. This made it possible for me to drill from a very stable standing position.
I then clamped the mounting hole template to the bottom, squared it up, and drilled the 4 holes through the donut bases. The thick template kept the bit perfectly square as it went through the thick 2.6" thick donut bases.
Finally, I installed the four 1/4-20 by 3.5" long bolts/washers. Both speakers produced a perfect fit, with no binding or alignment problems.
John,
I did some more browsing around in Hornresp and cannot find where to do stuffing in an ML-TL. It only seems to allow adjustment to the lining of the rear chamber, which is not used in an ML-TL schematic. Mass loaded transmission line sections and ports are, as far as I can tell, modeled only as flared horn segments, and, these horn segments cannot be stuffed or lined. When I run the Acoustic Power FR graph they always have the big peaks and dips of an unlined TL. I'm using version 50.50 of Hornresp. I must be missing something.
Bill
Port update: I don't like the thin-walled plastic flares and ports that are generally used. Seems to me that these thin-walled plastic pipes and flares could easily vibrate or resonate badly at some mid or high frequency compared to the thicker PVC. So, I made mine out of schedule 40 PVC plumbing pipes and added 3/4" radius flares to both ends. Then I boosted the overall length from 5 to 6 inches to compensate for the flares. Tests show that the FB is still right at 31Hz, so I am good to go.
Bill use the Tools Loudspeaker Wizard
Thanks, John. Found it.
Just goofing around in photoshop. Space alien green over particle board finish:
danger, danger will robinson...
I'll have to re-name my speakers the "bubble headed boobie" (Dr. Smith's name for the lost in space robot)
Here are the revised graphs. Thanks for steering me in the right direction on this. Turns out, my original NF cone measurement was in error; I must have bumped the amp volume control setting between measurements and/or placed the mic in the wrong NF measuring position (or something like that).
I therefore re-did both measurements. As luck would have it, my new summation comes out even better looking than before. The peaky looking response at 40Hz is now much flatter, down about 3dB, in the new summation. And, as you can see, the individual cone and port curves now line up at the very low frequencies, as they should. The port corrective adjustment is still -6.8dB (20log(5.25/2.4). The new measurements use flared 6" ports instead of the original 5" non-flared ports. Also, I revised the internal line stuffing a little bit to improve port resonances in the 100-800Hz region.
I should also mention one other thing that may have contributed to the measurement error. My experience has been that the SPL level of near field measurements is very sensitive to exact microphone tip placement. Very small changes in mic distance to woofer cone or port terminus plane can have a significant impact. Moving the mic as little as 1/8" closer or further away can bump the SPL up or down 1dB or so!! You have to be very careful. With a port flare, I place the mic tip half way into the flare. With a non-flared port, I place the tip in the center of the exact plane of the port terminus. With a woofer cone, I place the tip 1/4" from the center of the cone.
A little more progress (slow, but sure, I'm getting there):
With the box fully lined and the port tuning done, I could now glue the rear panel in place. So I sanded the perimeter with 40 grit to rough up the pre-laminated particle board surface.
I clamped several tweeter cutout slugs along the perimeter as alignment gauges, and then applied a thick bead of glue along the perimeter.
I dropped the main cabinet into the alignment gauges and watched the glue slowly squeeze out onto the wax paper below. As this happened, I went around the box checking the alignment, carefully adjusting it with a small hammer until it was perfect. In addition, I waited for about 5 minutes or so before applying the bar clamps because I wanted to give the glue a little extra time to soak in.
After first round of XSim modeling: 12 parts! 900 & 3.7kHz; 2nd order electrical; tank on woofer to tame the 5kHz peaking. Good phase tracking, however, I found it necessary to flip the tweeter phase. XSim mod delay offsets are -0.03 on the tweeter and + 2 inches on the woofer. BSC: 4 to 5dB; Sensitivity: 85dB.
Here is the on-axis XSim 3 pane model: