Hi all,
Recently completed a build. It's a dual opposed sealed enclosure with some budget drivers. I wanted to experience and measure a dual opposed system to get a better idea of the vibration cancellation and the suitability of the design in the context of living furniture that happens to hide a subwoofer in the future (tables, end tables, coffee tables, ottoman, etc as subs in addition).
Dual Opposed Acoustic Suspension Subwoofer
GRS 12SW4HE Drivers x 2 (Series, 8 ohm, drivers in same phase but physically oriented to face opposite)
Sealed 3.7 Ft^3 Gross | Sealed 2.6 Ft^3 Net Internal (15mm Pine Plywood) | 3.07 ft^3 Effective Stuffed Volume
Qtc 0.770 (Simulated) | Qtc 0.919 (Effective Actual, Derived from Measurements)
500w Power Handling
Very best,
Recently completed a build. It's a dual opposed sealed enclosure with some budget drivers. I wanted to experience and measure a dual opposed system to get a better idea of the vibration cancellation and the suitability of the design in the context of living furniture that happens to hide a subwoofer in the future (tables, end tables, coffee tables, ottoman, etc as subs in addition).
Dual Opposed Acoustic Suspension Subwoofer
GRS 12SW4HE Drivers x 2 (Series, 8 ohm, drivers in same phase but physically oriented to face opposite)
Sealed 3.7 Ft^3 Gross | Sealed 2.6 Ft^3 Net Internal (15mm Pine Plywood) | 3.07 ft^3 Effective Stuffed Volume
Qtc 0.770 (Simulated) | Qtc 0.919 (Effective Actual, Derived from Measurements)
500w Power Handling
Very best,
Modeling & Simulation
For modeling I used WinISD and my process was to take the two drivers and put them in a sealed volume and give them their max power handling combined, which is 500w, and then adjust the box volume until I saw the excursion graph down to 10hz happily at xmax. This is merely the starting point. This assumes many things such as linear performance, linear power, etc, but none of that is true so I expect measurements to be different. Still, this is the starting point.
** Note, the model is simulated from published GRS specs. I know full well that the reality won't be the same.
Excursion:
I adjusted enclosure volume until I saw the excursion resting at xmax down to 10hz with my max power handling of 500w.
This resulted in 2.6 ft^3 net internal volume, or half that per driver. They are sharing the same space in this cabinet.
The drivers:
Driver 1 DATS:
Driver 2 DATS:
GRS Publication:
Very best,
For modeling I used WinISD and my process was to take the two drivers and put them in a sealed volume and give them their max power handling combined, which is 500w, and then adjust the box volume until I saw the excursion graph down to 10hz happily at xmax. This is merely the starting point. This assumes many things such as linear performance, linear power, etc, but none of that is true so I expect measurements to be different. Still, this is the starting point.
** Note, the model is simulated from published GRS specs. I know full well that the reality won't be the same.
Excursion:
I adjusted enclosure volume until I saw the excursion resting at xmax down to 10hz with my max power handling of 500w.
This resulted in 2.6 ft^3 net internal volume, or half that per driver. They are sharing the same space in this cabinet.
The drivers:
Driver 1 DATS:
Driver 2 DATS:
GRS Publication:
Very best,
Materials & Initial Cutting
Cost List:
GRS 12SW4HE Drivers x 2 ($59 Each)
4x8 5/8th inch Plywood x 1 ($41)
Titebond II Glue x 1 ($8)
Terminals x 2 sets & wire ($20)
Stains & Poly ($15)
Screws for drivers ($1)
$203 in hardware and materials total including finish
If you want to add an amplifier, you can get a ton of power from a NX1000 in the $250 range or the SPA500DSP at $340 for a 500w amp with integrated good DSP. Total cost for a functioning unit would be $500 roughly, or a bit over with tax, etc, depending on choices. Cheaper, more power, more output, larger, compared to commercial options in the $500 price range.
====================================================
I find I really like lighter, thinner materials. It sort of goes against the overbuilt principals of using thicker, heavier, denser materials but I figure this is another good opportunity to see how material influence and the execution stack up with something like dual opposed where the enclosure experiences less overall vibration but is still expected to have potential flex, so we will brace for flexing but otherwise not worry too much about intricate bracing nor go for heavy materials. I will do a double baffle because I like to flush mount the drivers and to get rigidity there.
I picked up a single sheet of 15mm (5/8th inch) thick 4x8 foot Pine CDX plywood from Ace Hardware for $41. I prefer plywood over MDF for the reasons mentioned above. Yes, there are voids and its thin and flexes. But I've built with thin plywood before with much larger enclosures that handle even higher pressures from multi-drivers and with proper bracing it works totally fine with no artifact. In this situation I will do minimal bracing, because I want to see how dual opposed vibration canceling interacts with this.
Targeting my net internal volume of 2.6 ft^3, I calculated those dimensions and material thickness to get to around 3.7 ft^3 gross volume.
My final dimensions:
Gross external dimensions:
16" Wide
20" Tall
20" Deep
Internal exploded dimensions:
I will double baffle each driver and flush mount each driver in a recessed portion of that baffle.
The internal bracing will also just be a baffle with matching circle size for simplicity. No other bracing planned (cringe and squint, but, let's see how it turns out since cabinet vibrations are supposed to be canceled out, we can test this).
Wen Tracksaw for the initial cuts
Very best,
Cost List:
GRS 12SW4HE Drivers x 2 ($59 Each)
4x8 5/8th inch Plywood x 1 ($41)
Titebond II Glue x 1 ($8)
Terminals x 2 sets & wire ($20)
Stains & Poly ($15)
Screws for drivers ($1)
$203 in hardware and materials total including finish
If you want to add an amplifier, you can get a ton of power from a NX1000 in the $250 range or the SPA500DSP at $340 for a 500w amp with integrated good DSP. Total cost for a functioning unit would be $500 roughly, or a bit over with tax, etc, depending on choices. Cheaper, more power, more output, larger, compared to commercial options in the $500 price range.
====================================================
I find I really like lighter, thinner materials. It sort of goes against the overbuilt principals of using thicker, heavier, denser materials but I figure this is another good opportunity to see how material influence and the execution stack up with something like dual opposed where the enclosure experiences less overall vibration but is still expected to have potential flex, so we will brace for flexing but otherwise not worry too much about intricate bracing nor go for heavy materials. I will do a double baffle because I like to flush mount the drivers and to get rigidity there.
I picked up a single sheet of 15mm (5/8th inch) thick 4x8 foot Pine CDX plywood from Ace Hardware for $41. I prefer plywood over MDF for the reasons mentioned above. Yes, there are voids and its thin and flexes. But I've built with thin plywood before with much larger enclosures that handle even higher pressures from multi-drivers and with proper bracing it works totally fine with no artifact. In this situation I will do minimal bracing, because I want to see how dual opposed vibration canceling interacts with this.
Targeting my net internal volume of 2.6 ft^3, I calculated those dimensions and material thickness to get to around 3.7 ft^3 gross volume.
My final dimensions:
Gross external dimensions:
16" Wide
20" Tall
20" Deep
Internal exploded dimensions:
- Top & Bottom (2x)
- 406mm x 508mm x 15mm
- Sides x 2:
- 478mm x 448mm x 15mm
- Brace x 1:
- 376mm x 478mm x 15mm
- Baffles (front, back, doubled, x 4):
- 406mm x 478mm x 15mm
- Driver circle cutout = 11 & 5/16th inches
- Rabbet flange for flush is 1/2" width, 1/2" depth
I will double baffle each driver and flush mount each driver in a recessed portion of that baffle.
The internal bracing will also just be a baffle with matching circle size for simplicity. No other bracing planned (cringe and squint, but, let's see how it turns out since cabinet vibrations are supposed to be canceled out, we can test this).
Wen Tracksaw for the initial cuts
Very best,
Baffles & Hole Cutting
I combine two baffles for each primary driver front baffle (front and back) so that I can get rigidity but also to flush mount the driver by cutting away a recession into the thicker baffle.
Window brace with equal diameter hole
Very best,
I combine two baffles for each primary driver front baffle (front and back) so that I can get rigidity but also to flush mount the driver by cutting away a recession into the thicker baffle.
Window brace with equal diameter hole
Very best,
Cabinet Assembly
Baffles cut, circles cut, time to assemble the box. I summon a few dozen clamps for aid.
I'm using Titebond II for all glue joints
Test fitting
Very best,
Baffles cut, circles cut, time to assemble the box. I summon a few dozen clamps for aid.
I'm using Titebond II for all glue joints
Test fitting
Very best,
Cabinet Finishing
Glue is cured, so time to finish. I'm going to sand it with a basic Skil orbital with 80 grit, 120 grit and 220 grit. Then, with a hand router and a 1/2 inch round over bit, I round all the edges.
Plus the incredible chore of putting up all the clamps.
Very best,
Glue is cured, so time to finish. I'm going to sand it with a basic Skil orbital with 80 grit, 120 grit and 220 grit. Then, with a hand router and a 1/2 inch round over bit, I round all the edges.
Plus the incredible chore of putting up all the clamps.
Very best,
Wiring, Terminals & Stuffing
I'm wiring the drivers individually internally and doing the rest externally with large terminals. I did this so that I can choose to run the drivers individually with two amps if I wanted to, or wire them in series externally (8ohm) for single amp, or wire them in parallel (2ohm) for single amp (if stable). I like having options and keeping it modular so while the wires and terminals are extra and are not hidden, I like seeing them, and I like being able to tinker with configurations in the future.
I like those big Dayton heavy duty banana terminals.
Time to stuff this box. Before I stuffed it, I did a bunch of measurements without stuffing. I will review those in the last posts. Then I brought it back and took the drivers out and stuffed the box.
I'm using Polyfil. I went for more than the typical suggestion and went with 4lbs of stuffing into 2.6 ft^3 net internal volume, maybe a bit less with driver displacement that I didn't really account or worry over since it was so small and I had lots of space still inside this enclosure. This puts me at 1.538lb per ft^3. We will see what this yields in measurements later.
I came to this above value based on work done by Tom Nousaine.
I'm at about 1.3 ft^3 per driver, though its not that simple of course, but ultimately I used that as a guide to pick a number basically. Based on the measurements in that link, I went with close to 1.5 (1.538) lb per ft^3. We will see how the effective volume (based on behavior) and efficiency changes with this. I expect something in this smaller box loaded like this to at least show up in measurements.
Very best,
I'm wiring the drivers individually internally and doing the rest externally with large terminals. I did this so that I can choose to run the drivers individually with two amps if I wanted to, or wire them in series externally (8ohm) for single amp, or wire them in parallel (2ohm) for single amp (if stable). I like having options and keeping it modular so while the wires and terminals are extra and are not hidden, I like seeing them, and I like being able to tinker with configurations in the future.
I like those big Dayton heavy duty banana terminals.
Time to stuff this box. Before I stuffed it, I did a bunch of measurements without stuffing. I will review those in the last posts. Then I brought it back and took the drivers out and stuffed the box.
I'm using Polyfil. I went for more than the typical suggestion and went with 4lbs of stuffing into 2.6 ft^3 net internal volume, maybe a bit less with driver displacement that I didn't really account or worry over since it was so small and I had lots of space still inside this enclosure. This puts me at 1.538lb per ft^3. We will see what this yields in measurements later.
I came to this above value based on work done by Tom Nousaine.
I'm at about 1.3 ft^3 per driver, though its not that simple of course, but ultimately I used that as a guide to pick a number basically. Based on the measurements in that link, I went with close to 1.5 (1.538) lb per ft^3. We will see how the effective volume (based on behavior) and efficiency changes with this. I expect something in this smaller box loaded like this to at least show up in measurements.
Very best,
The 2 drivers need to be tightly coupled. You need a brace that contacts both magnets the baffle, top, and bottom. As you have pictured in post #1 you are losing much of the benefit of push-push.
dave
dave
Measurements (DATS, Frequency Response Sweeps, Unstuffed vs Stuffed)
I did these measurements unstuffed and stuffed so I could compare them.
Predicting standing waves:
The two internal cavities make the following dimensions:
376mm x 478mm x 205mm
X in meters = 1/2 wavelength
2(X) in meters = full wavelength
Top to bottom:
2(0.478m)=0.956m
343m/s / 0.956m = 358.7hz + harmonics (717hz, 1435hz, etc)
Side to side:
2(0.376m)=0.752
343m/s / 0.752m = 456.1hz + harmonics (912hz, 1824hz, etc)
Front to back:
2(0.205m)=0.41
343m/s / 0.41m = 836.5hz + harmonics (1673hz, 3346hz, etc)
We will look for these potential predicted standing waves in the measurements.
Note, all the predicted standing waves are above my intended bandwidth (10hz to 100hz). So none of these values matter in reality. Just a "good to know" type thing.
DATS:
Unstuffed:
Both drivers wired together in series, 8 ohm (this was my intended design use, out of the options). This is unstuffed. I have two little ripples, around 142hz and 450hz. I noticed two of these before in the single driver measurements, but one went away. The two are not related I don't think. The 3rd was a harmonic of the first ripple and it was diminished with stuffing. So I have a pretty hard resonance at 142~150hz approximately and 430hz~450hz roughly. Neither align with predicted standing waves based on dimensions. So these are from something else maybe. I'm guessing its from the window brace I made perhaps. Or a void in the plywood somewhere.
Stuffed:
Now, stuffed and in series, for 8ohm. This is with 4lbs of stuffing at 1.538lb/ft^3 ratio stuffing to cabinet internal volume. Fs shifted to the left as expected from 47.5hz to 42.6hz. Impedance dropped significantly from 77~78ohm down to 43~44ohm. The two ripples mentioned before at 142hz seems to have shifted to the right a little, closer to greater than 150hz or 160hz range, and the 450hz ripple moved to the left a little, closer to 420~430hz range. Since the ripples didn't go away, they are probably resonance I think and not just standing waves. I can't get rid of them from stuffing, even with heavy stuffing like this. So it's probably structural. That said, will they matter? No, they might as well not be there, they do not result in anything practical in the later measurements since I won't measure over 100hz. Just one of those "the more you know" situations in this context. I won't be running this greater than 100hz (low pass filter) anyways, so these issues are all well over the intended bandwidth of this sub.
Actual T&S and QTc:
My original simulation is based on the published specs. They are not the same as measured. Now that I have measured free air T&S per driver, drivers in the enclosure unfilled and drivers in the enclosure filled, I have measurements to compare them all and derive some information:
From my measurements:
Single driver in free air:
Re = 3.847
Fs = 24.83
Qts = 0.5308
Qes = 0.5896
Qms = 5.317
Vas = 4.147 ft^3
For snippet use:
Fs = 24.83
Qms = 5.317
Qes = 0.5896
Both drivers in the enclosure without filling:
Fct = 47.54hz
Qmct = 5.317
Qect = 1.203
Both drivers in the enclosure with filling:
Fc = 42.66hz
Qmc - 4.824
Qec = 1.136
Using the above to calculate some actual values that differ from published and more specific to what I have going on (differing from simulation, so this is execution):
Qtcto = QmctQect / (Qmct+Qect)
Qtcto = 5.317*1.203 / (5.317+1.203)
Qtcto = 6.396 / 6.52
Qtcto = 0.981
However, if I assume the above for filled:
Qtco = QmcQec / (Qmc+Qec)
Qtco = 4.824 * 1.136 / (4.824 + 1.136)
Qtco = 5.48 / 5.96
Qtco = 0.919
Stuffed final Qtc of the system is 0.919 in reality. The published specs of the GRS are way, way off of course as expected.
Finally, some other changes based on stuffing such as the moving mass ratio after stuffing and the effective volume change after stuffing:
Mac / Mact = FctQec/FcQect
Mac / Mact = 47.54hz*1.136 / 42.66hz*1.203
Mac / Mact = 54.005 / 51.31998
Mac / Mact = 1.0523 (moving mass ratio)
Now, effective volume change:
Vab / Vb = ((FctQect / FsQes) - 1) / ((FcQec / FsQes) - 1)
= ((47.54hz*1.203) - 1) / ((42.66hz*1.136) - 1)
= 56.19062 / 47.46176
= 1.1839
= 18.4% increase in effective volume
Reflection: I probably should have modeled based on actual T&S measurements. The box would have been much, much bigger based on the T&S values. That said, my measured values to not easily validate in a simulation so while I could have done more to possibly model based on measured T&S, it's a lot of extra work to do that and in reality I honestly, nor should anyone, have to do all this just to get reliable data that is supposed to be reliable from the manufacturer and published. Why is it ok for published T&S to be so different? Why bother publishing at all? Sigh. It's not just cheap stuff like GRS. I notice this on every single driver I measure. None of them are close. I have no measured top shelf premium stuff yet though, so maybe it's just that. Still, why is it ok for cheap stuff to not meet spec? Or at least publish spec that actually matches the samples?
==================================================================
Ground Plane Measurements, both 1 meter & 2 meter, Stuffed & Unstuffed
I'm using a Crown XLS1002 amplifier bridged for power. I'm using a Fiio E10 for DAC output to this amp. Running REW on the laptop here. Umik-1 microphone with gain settings modified (in my signature if you are interested) to allow higher SPL readings. Measured at 1 meter and 2 meters. For the ground plane, I setup on a moving blanket out in a flat dirt area under some trees. Not the middle of a field, but good enough. The ground is not perfectly flat, but it will work for these wavelengths. I measured from the side so that my microphone was equally distant from both direct radiators, so from the side-middle and out 1000mm and 2000mm distance measured.
Unstuffed, 1 Meter sweeps. Sweeps were done at REW's 1M resolution, 21 seconds, in 3db increments.
I stopped at 111db at 50~60hz range as this was already plenty loud for me and I found that another +3db resulted in signal clipping (input side, my DAC was limiting things from this laptop). I did notice a weird behavior that I could visually see at 23~24hz but could not explain. It only happened as signal got higher. I'm not sure if it's an enclosure dynamic or an electronic thing. But this was another reason I stopped there. The slope is dropping 12db per octave as expected.
Unstuffed 1 meter vs 2 meter results. I noticed the same 23~24hz dip that I could visually see the drivers hiccup and do there.
Stuffed results this time. I noticed with stuffing, the odd behavior at 23~24hz went away. I lost efficiency above 50hz and it flattened out. But, I gained efficiency in all frequencies under 30hz. More on this in a moment. I stopped at 110db simply because I didn't need to see higher. I noticed what appears to be about 1db of compression in the range from 40~50hz range down to 10hz, you can see the slight difference there. That may or may not be compression, but since it showed up there, I stopped there and call this my stopping point. Still, I'm good with this output for 8ohm and series wiring.
Stuffed 1 meter vs 2 meter results
Finally stuffed vs unstuffed. This is one of those points where I'm always questioning things. I've stuffed boxes in the past and had no effect results that seemed practical. However, this time, with a little more care and after lots of questions and answers, I went for a lot more stuffing, 54% more stuffing than most suggest. This is with 1.538lb per ft^3 stuffing ratio. This is the results where we can see what the stuffing actually did that is practical.
Blue is stuffed
Red (or whatever color that is) is unstuffed
Stuffed loses efficiency above 38hz, it flattened out, this is good to me, this lowers that hump that I would normally EQ away anyways.
Stuffed gains efficiency below 38hz. At first its a little. Not even a db. Then by 27hz and below its 1db and by the time we get to 10hz its a whopping 2db gain.
So I would say stuffing was entirely worth it here and going higher ratio in the smaller box was well worth it.
Very best,
I did these measurements unstuffed and stuffed so I could compare them.
Predicting standing waves:
The two internal cavities make the following dimensions:
376mm x 478mm x 205mm
X in meters = 1/2 wavelength
2(X) in meters = full wavelength
Top to bottom:
2(0.478m)=0.956m
343m/s / 0.956m = 358.7hz + harmonics (717hz, 1435hz, etc)
Side to side:
2(0.376m)=0.752
343m/s / 0.752m = 456.1hz + harmonics (912hz, 1824hz, etc)
Front to back:
2(0.205m)=0.41
343m/s / 0.41m = 836.5hz + harmonics (1673hz, 3346hz, etc)
We will look for these potential predicted standing waves in the measurements.
Note, all the predicted standing waves are above my intended bandwidth (10hz to 100hz). So none of these values matter in reality. Just a "good to know" type thing.
DATS:
Unstuffed:
Both drivers wired together in series, 8 ohm (this was my intended design use, out of the options). This is unstuffed. I have two little ripples, around 142hz and 450hz. I noticed two of these before in the single driver measurements, but one went away. The two are not related I don't think. The 3rd was a harmonic of the first ripple and it was diminished with stuffing. So I have a pretty hard resonance at 142~150hz approximately and 430hz~450hz roughly. Neither align with predicted standing waves based on dimensions. So these are from something else maybe. I'm guessing its from the window brace I made perhaps. Or a void in the plywood somewhere.
Stuffed:
Now, stuffed and in series, for 8ohm. This is with 4lbs of stuffing at 1.538lb/ft^3 ratio stuffing to cabinet internal volume. Fs shifted to the left as expected from 47.5hz to 42.6hz. Impedance dropped significantly from 77~78ohm down to 43~44ohm. The two ripples mentioned before at 142hz seems to have shifted to the right a little, closer to greater than 150hz or 160hz range, and the 450hz ripple moved to the left a little, closer to 420~430hz range. Since the ripples didn't go away, they are probably resonance I think and not just standing waves. I can't get rid of them from stuffing, even with heavy stuffing like this. So it's probably structural. That said, will they matter? No, they might as well not be there, they do not result in anything practical in the later measurements since I won't measure over 100hz. Just one of those "the more you know" situations in this context. I won't be running this greater than 100hz (low pass filter) anyways, so these issues are all well over the intended bandwidth of this sub.
Actual T&S and QTc:
My original simulation is based on the published specs. They are not the same as measured. Now that I have measured free air T&S per driver, drivers in the enclosure unfilled and drivers in the enclosure filled, I have measurements to compare them all and derive some information:
From my measurements:
Single driver in free air:
Re = 3.847
Fs = 24.83
Qts = 0.5308
Qes = 0.5896
Qms = 5.317
Vas = 4.147 ft^3
For snippet use:
Fs = 24.83
Qms = 5.317
Qes = 0.5896
Both drivers in the enclosure without filling:
Fct = 47.54hz
Qmct = 5.317
Qect = 1.203
Both drivers in the enclosure with filling:
Fc = 42.66hz
Qmc - 4.824
Qec = 1.136
Using the above to calculate some actual values that differ from published and more specific to what I have going on (differing from simulation, so this is execution):
Qtcto = QmctQect / (Qmct+Qect)
Qtcto = 5.317*1.203 / (5.317+1.203)
Qtcto = 6.396 / 6.52
Qtcto = 0.981
However, if I assume the above for filled:
Qtco = QmcQec / (Qmc+Qec)
Qtco = 4.824 * 1.136 / (4.824 + 1.136)
Qtco = 5.48 / 5.96
Qtco = 0.919
Stuffed final Qtc of the system is 0.919 in reality. The published specs of the GRS are way, way off of course as expected.
Finally, some other changes based on stuffing such as the moving mass ratio after stuffing and the effective volume change after stuffing:
Mac / Mact = FctQec/FcQect
Mac / Mact = 47.54hz*1.136 / 42.66hz*1.203
Mac / Mact = 54.005 / 51.31998
Mac / Mact = 1.0523 (moving mass ratio)
Now, effective volume change:
Vab / Vb = ((FctQect / FsQes) - 1) / ((FcQec / FsQes) - 1)
= ((47.54hz*1.203) - 1) / ((42.66hz*1.136) - 1)
= 56.19062 / 47.46176
= 1.1839
= 18.4% increase in effective volume
Reflection: I probably should have modeled based on actual T&S measurements. The box would have been much, much bigger based on the T&S values. That said, my measured values to not easily validate in a simulation so while I could have done more to possibly model based on measured T&S, it's a lot of extra work to do that and in reality I honestly, nor should anyone, have to do all this just to get reliable data that is supposed to be reliable from the manufacturer and published. Why is it ok for published T&S to be so different? Why bother publishing at all? Sigh. It's not just cheap stuff like GRS. I notice this on every single driver I measure. None of them are close. I have no measured top shelf premium stuff yet though, so maybe it's just that. Still, why is it ok for cheap stuff to not meet spec? Or at least publish spec that actually matches the samples?
==================================================================
Ground Plane Measurements, both 1 meter & 2 meter, Stuffed & Unstuffed
I'm using a Crown XLS1002 amplifier bridged for power. I'm using a Fiio E10 for DAC output to this amp. Running REW on the laptop here. Umik-1 microphone with gain settings modified (in my signature if you are interested) to allow higher SPL readings. Measured at 1 meter and 2 meters. For the ground plane, I setup on a moving blanket out in a flat dirt area under some trees. Not the middle of a field, but good enough. The ground is not perfectly flat, but it will work for these wavelengths. I measured from the side so that my microphone was equally distant from both direct radiators, so from the side-middle and out 1000mm and 2000mm distance measured.
Unstuffed, 1 Meter sweeps. Sweeps were done at REW's 1M resolution, 21 seconds, in 3db increments.
I stopped at 111db at 50~60hz range as this was already plenty loud for me and I found that another +3db resulted in signal clipping (input side, my DAC was limiting things from this laptop). I did notice a weird behavior that I could visually see at 23~24hz but could not explain. It only happened as signal got higher. I'm not sure if it's an enclosure dynamic or an electronic thing. But this was another reason I stopped there. The slope is dropping 12db per octave as expected.
Unstuffed 1 meter vs 2 meter results. I noticed the same 23~24hz dip that I could visually see the drivers hiccup and do there.
Stuffed results this time. I noticed with stuffing, the odd behavior at 23~24hz went away. I lost efficiency above 50hz and it flattened out. But, I gained efficiency in all frequencies under 30hz. More on this in a moment. I stopped at 110db simply because I didn't need to see higher. I noticed what appears to be about 1db of compression in the range from 40~50hz range down to 10hz, you can see the slight difference there. That may or may not be compression, but since it showed up there, I stopped there and call this my stopping point. Still, I'm good with this output for 8ohm and series wiring.
Stuffed 1 meter vs 2 meter results
Finally stuffed vs unstuffed. This is one of those points where I'm always questioning things. I've stuffed boxes in the past and had no effect results that seemed practical. However, this time, with a little more care and after lots of questions and answers, I went for a lot more stuffing, 54% more stuffing than most suggest. This is with 1.538lb per ft^3 stuffing ratio. This is the results where we can see what the stuffing actually did that is practical.
Blue is stuffed
Red (or whatever color that is) is unstuffed
Stuffed loses efficiency above 38hz, it flattened out, this is good to me, this lowers that hump that I would normally EQ away anyways.
Stuffed gains efficiency below 38hz. At first its a little. Not even a db. Then by 27hz and below its 1db and by the time we get to 10hz its a whopping 2db gain.
So I would say stuffing was entirely worth it here and going higher ratio in the smaller box was well worth it.
Very best,
Measurements (Vibration Tests, Videos)
Now that all the typical stuff is out of the way, I wanted to explore what the dual opposed afforded me in this design. The idea was that it should cancel cabinet vibrations. I wanted to test this with an accelerometer and also just practical tests like a glass of water.
I'm using a WitMotion accelerometer sensor and associated software to gather data on acceleration, angular velocity, etc. I attached it with blu-tac.
First, I measured single driver so that the other driver was merely passive, this means I do not benefit from the cancellation effect and the box becomes a typical enclosure with 1 active driver and effectively 1 passive radiator. So I should see some enclosure dynamics since nothing is being canceled out. I took a 30hz signal at fairly high level and ran it for a while. You can see the obvious vibration pattern (top right in the graphs) of angular velocity following an interesting pattern with an internal walking pattern too.
Now, both drivers active, dual opposed and in phase. The lines are totally flat. Note, I increased power to -3.00 dBFS here.
I graphed the exported values to make it easier to see the angular velocity pattern changes here. This is vibration. It's two patterns really, or more, and repeating. You can see the primary pattern peak descending and the secondary patter peak ascending until they are similar and then they diverge and repeat again. This was interesting to see. This is one driver and a passive one with no cabinet vibration cancellation.
Here's acceleration changes (g's). You can see the obvious pattern here, it was not so easily seen in the above screen shot graphs due to the scale. I reduced the scale heavily. The previous scale was (1g = baseline normal) and scaled to 15gs. Obviously 15 g's is nuts and not practical. I reduced the scale to 1.1g's. Just 10% more and less from baseline 1g. You can see the very minor acceleration changes here, as vibrations, maybe 0.01g change +/- from baseline 1g. You can see the alternating pattern here too, where the peaks are similar, diverge, then one ascends and the other descends. The practical result is you feel vibrations and these patterned vibrations will cause things to move. Anything coupling to these movements may also move.
Next, I wired in series and powered both drivers. Save signal level. The dual opposed (in phase) effect is noted here quite glaringly and very satisfying to see. The angular velocity changes, with the patterns noted above, are completely gone. The measurement is the same scale as the previous measurement in this unit. The angular velocity is just reduced to nothing. Not even a small deviation. Just flat line dead. The box is virtually dead.
Finally I did take a look at acceleration (g). I could still feel a little bit of vibration on the box when I touched it. But I should, there's SPL coming from it. High SPL. I was well over 100db and the excursion was approaching 18~20mm total distance travel. My pants legs were vibrating and moving. Things were coupling in the room and rattling some. The box, however, felt virtually dead, except for a few times I thought I could notice something. Acceleration revealed that despite angular velocity being completely diminished by the dual opposed design, I could still pick up a trace bit of acceleration just like before. Not much. Same tiny scale of 10% more than baseline 1g. The pattern is different and cycles over much more time than the previous pattern. More time between pattern cycle makes it harder to notice and ultimately has less energy related to moving things in this context, so while I can detect minor vibration, it's so minor I had to do this to measure it. As you noted before, the massive vibrations from angular velocity changes was easily noticed and measured significant and was reduced to nothing in this design. Note, this could also be external sourced from my AC, water heater, water pump, wind on my walls, etc, coupling to my subfloor and then transferring to the enclosure. I'd have to run it random for days to see any correlation. So from this, I just go back to angular velocity being dead and call it good.
Videos:
Now, some unscientific tests. Glass of water! The T-Rex test from Jurassic Park.
I recorded it via video, normal time and slow motion, for review.
Very best,
Now that all the typical stuff is out of the way, I wanted to explore what the dual opposed afforded me in this design. The idea was that it should cancel cabinet vibrations. I wanted to test this with an accelerometer and also just practical tests like a glass of water.
I'm using a WitMotion accelerometer sensor and associated software to gather data on acceleration, angular velocity, etc. I attached it with blu-tac.
First, I measured single driver so that the other driver was merely passive, this means I do not benefit from the cancellation effect and the box becomes a typical enclosure with 1 active driver and effectively 1 passive radiator. So I should see some enclosure dynamics since nothing is being canceled out. I took a 30hz signal at fairly high level and ran it for a while. You can see the obvious vibration pattern (top right in the graphs) of angular velocity following an interesting pattern with an internal walking pattern too.
Now, both drivers active, dual opposed and in phase. The lines are totally flat. Note, I increased power to -3.00 dBFS here.
I graphed the exported values to make it easier to see the angular velocity pattern changes here. This is vibration. It's two patterns really, or more, and repeating. You can see the primary pattern peak descending and the secondary patter peak ascending until they are similar and then they diverge and repeat again. This was interesting to see. This is one driver and a passive one with no cabinet vibration cancellation.
Here's acceleration changes (g's). You can see the obvious pattern here, it was not so easily seen in the above screen shot graphs due to the scale. I reduced the scale heavily. The previous scale was (1g = baseline normal) and scaled to 15gs. Obviously 15 g's is nuts and not practical. I reduced the scale to 1.1g's. Just 10% more and less from baseline 1g. You can see the very minor acceleration changes here, as vibrations, maybe 0.01g change +/- from baseline 1g. You can see the alternating pattern here too, where the peaks are similar, diverge, then one ascends and the other descends. The practical result is you feel vibrations and these patterned vibrations will cause things to move. Anything coupling to these movements may also move.
Next, I wired in series and powered both drivers. Save signal level. The dual opposed (in phase) effect is noted here quite glaringly and very satisfying to see. The angular velocity changes, with the patterns noted above, are completely gone. The measurement is the same scale as the previous measurement in this unit. The angular velocity is just reduced to nothing. Not even a small deviation. Just flat line dead. The box is virtually dead.
Finally I did take a look at acceleration (g). I could still feel a little bit of vibration on the box when I touched it. But I should, there's SPL coming from it. High SPL. I was well over 100db and the excursion was approaching 18~20mm total distance travel. My pants legs were vibrating and moving. Things were coupling in the room and rattling some. The box, however, felt virtually dead, except for a few times I thought I could notice something. Acceleration revealed that despite angular velocity being completely diminished by the dual opposed design, I could still pick up a trace bit of acceleration just like before. Not much. Same tiny scale of 10% more than baseline 1g. The pattern is different and cycles over much more time than the previous pattern. More time between pattern cycle makes it harder to notice and ultimately has less energy related to moving things in this context, so while I can detect minor vibration, it's so minor I had to do this to measure it. As you noted before, the massive vibrations from angular velocity changes was easily noticed and measured significant and was reduced to nothing in this design. Note, this could also be external sourced from my AC, water heater, water pump, wind on my walls, etc, coupling to my subfloor and then transferring to the enclosure. I'd have to run it random for days to see any correlation. So from this, I just go back to angular velocity being dead and call it good.
Videos:
Now, some unscientific tests. Glass of water! The T-Rex test from Jurassic Park.
I recorded it via video, normal time and slow motion, for review.
Very best,
Summary & Closing Thoughts
Overall, I'm happy with the results. This is going to be used for 99% music basically and for that it's doing the job. However, I really wanted to build it to measure the phenomena related to dual opposed design to see how the internal dynamics result in something practical and also I wanted to get better measurements to show positive and practical results of stuffing a sealed enclosure.
I was very surprised to see just how dead the box becomes with dual opposed. Truly dead. Water won't ripple or move. No vibrations that are really significant. Totally dead. Yet, the output is in phase and no worries about the face direction of the drivers (below 80~100hz) for phase purposes due to wavelength of the frequency. If I wired them out of phase, the output reduces and its kind of chaotic as the drivers are not aligned nor in phase and so they fight each other a bit and cancel a lot of things out. But when wired in phase, they do really impressive things with internal dynamics where the enclosure vibrations are diminished to nothing yet the radiating pattern of SPL at low frequency remains externally. Super cool to feel it and neat to see it on practical measurements. The measurements and practical "water test" results really show this is excellent design for furniture, or living surface builds where you want to hide a sub in functional furniture, but not have your guests' drinks walk off and spill, or objects jumping and buzzing around. And when the enclosure is isolated on rubber or other material that doesn't transfer vibration, the surroundings no longer couple to the enclosure vibration (though the sound waves still effect the room, it will still rattle things at high SPL and low frequencies).
Time domain stuff is based on the room and since I'm using it near field in my office for music, I don't need enormous SPL at distance, so my time domain is no problem and beyond 300ms the SPL decays to below my ambient noise floor. This has nothing to do with the design however, just how the setup is in the room.
Musically, I really like it. It does what it should do. And I really like how dead the box is. It's wild that I can touch it with my hand or toe and even though I see the driver moving and hear it quite loudly, it's just a dead box when I touch it. If I didn't hear anything I would think it was off as I'm so used to boxes being obvious with vibrations.
The stuffing side quest was also a really satisfying thing to see. Practical results that showed a very good reason for stuffing, to flatten the hump in mid-bass and increase efficiency and gain output below Fs. 1~2db gain is quite a lot at 10~20hz if you think about it. So it was well worth it and really was revealing as a good thing. From my reading, it was measured to be less and less effective like this as true volume increased well beyond 5 ft^3 net internal. But with smaller 1~2 ft^3 enclosures per driver, it helped more. It showed here with the 1.538 lb per ft^3 ratio gaining over 40% effective volume noted as behavior by the driver's output. The equivalent of Qtc dropping from a volume perspective, but a gain in Q from a SPL at Fc perspective.
Sadly, measuring drivers just reveals how much the industry gets away with nearly useless T&S publications which then makes simulation far more of an effort. Also, really annoying if you were to purchase things based on T&S values, only to find out they measure totally different. It's rampant in the industry at all levels. Somehow it's just "ok" and it's just how it is. Thankfully its not so radically bad when it comes to practical results, ie, output and bandwidth.
What would I do differently Now? Now that I've gone through the process, what would I do differently knowing what I know now? Well, the first thing is I think I would select drivers based on wiring potential for pairing them, ie, something that is 8ohm native or via dual voice coils so that I can wire them parallel comfortably around 4ohm. Then, measure via DATS to get the free air T&S values of these drivers. Use those real measurement based T&S values to better guide simulation in software to design the enclosure and shape the response and behavior. Then execute the build. Structurally the only thing I would do differently is I wouldn't bother with the brace window I did, I did it out of curiosity and simplicity but in reality I would rather just do stud posts and I would have doubled them up and contacted all points on all walls and it would have been effortless to cut them to fit with a miter saw. And I think I would add stud ribs to the walls running from the driver to driver plane as overkill bracing. I didn't find any problems with my current bracing, but that's how I would have done it had I gone with my typical method. Ultimately though I think the biggest thing is to use drivers that wire for final loads that are better for "cheap power." There's no cheap power for 8 ohm. The advantages of parallel is acoustic power and if I can get a final load of 4ohm, then its a lot easier to get cheap power to allow for the full power handling of the two drivers. For this reason, circling back, I would instead go for typical drivers that are dual voice coil at 4ohm each so they can be wired in series for 8ohm and then both drivers wired in parallel for 4ohm final load. Or just native 8ohm drivers wired in parallel for 4ohm. Budget drivers that come to mind that would wire better include: Skar SDR-12 or LaVoce SSF122.50L, Pioneer TSW312D4, JBL Stage 122D, Kicker 50CWCD124, MTX 5512-44. Lots of great options that cost more of course, so mainly looking at less expensive options since its double the cost to get two.
For actual use, I will be using a basic DSP (the Dayton DSP-LF, cheap, at least for now just to play aorund; I have the DSP-408 and MiniDSP HD too) to shape its output to a slight house curve to 20hz. Here's my heavy curve influence on the DSP interface and the in-room near field results:
iWoofer app on Android controlling the DSP. No importing of values, the curves are manual. I set these curves as opposite of my in-room response to get a flat result to 20hz and then gave it a gentle hump along the way to get tiny bit of house curve out of it.
Here's my near field, no smoothing, results with the DSP above in my room. I love this. Again, this is for music, but it covers everything so well. And the box is totally dead.
Very best,
Overall, I'm happy with the results. This is going to be used for 99% music basically and for that it's doing the job. However, I really wanted to build it to measure the phenomena related to dual opposed design to see how the internal dynamics result in something practical and also I wanted to get better measurements to show positive and practical results of stuffing a sealed enclosure.
I was very surprised to see just how dead the box becomes with dual opposed. Truly dead. Water won't ripple or move. No vibrations that are really significant. Totally dead. Yet, the output is in phase and no worries about the face direction of the drivers (below 80~100hz) for phase purposes due to wavelength of the frequency. If I wired them out of phase, the output reduces and its kind of chaotic as the drivers are not aligned nor in phase and so they fight each other a bit and cancel a lot of things out. But when wired in phase, they do really impressive things with internal dynamics where the enclosure vibrations are diminished to nothing yet the radiating pattern of SPL at low frequency remains externally. Super cool to feel it and neat to see it on practical measurements. The measurements and practical "water test" results really show this is excellent design for furniture, or living surface builds where you want to hide a sub in functional furniture, but not have your guests' drinks walk off and spill, or objects jumping and buzzing around. And when the enclosure is isolated on rubber or other material that doesn't transfer vibration, the surroundings no longer couple to the enclosure vibration (though the sound waves still effect the room, it will still rattle things at high SPL and low frequencies).
Time domain stuff is based on the room and since I'm using it near field in my office for music, I don't need enormous SPL at distance, so my time domain is no problem and beyond 300ms the SPL decays to below my ambient noise floor. This has nothing to do with the design however, just how the setup is in the room.
Musically, I really like it. It does what it should do. And I really like how dead the box is. It's wild that I can touch it with my hand or toe and even though I see the driver moving and hear it quite loudly, it's just a dead box when I touch it. If I didn't hear anything I would think it was off as I'm so used to boxes being obvious with vibrations.
The stuffing side quest was also a really satisfying thing to see. Practical results that showed a very good reason for stuffing, to flatten the hump in mid-bass and increase efficiency and gain output below Fs. 1~2db gain is quite a lot at 10~20hz if you think about it. So it was well worth it and really was revealing as a good thing. From my reading, it was measured to be less and less effective like this as true volume increased well beyond 5 ft^3 net internal. But with smaller 1~2 ft^3 enclosures per driver, it helped more. It showed here with the 1.538 lb per ft^3 ratio gaining over 40% effective volume noted as behavior by the driver's output. The equivalent of Qtc dropping from a volume perspective, but a gain in Q from a SPL at Fc perspective.
Sadly, measuring drivers just reveals how much the industry gets away with nearly useless T&S publications which then makes simulation far more of an effort. Also, really annoying if you were to purchase things based on T&S values, only to find out they measure totally different. It's rampant in the industry at all levels. Somehow it's just "ok" and it's just how it is. Thankfully its not so radically bad when it comes to practical results, ie, output and bandwidth.
What would I do differently Now? Now that I've gone through the process, what would I do differently knowing what I know now? Well, the first thing is I think I would select drivers based on wiring potential for pairing them, ie, something that is 8ohm native or via dual voice coils so that I can wire them parallel comfortably around 4ohm. Then, measure via DATS to get the free air T&S values of these drivers. Use those real measurement based T&S values to better guide simulation in software to design the enclosure and shape the response and behavior. Then execute the build. Structurally the only thing I would do differently is I wouldn't bother with the brace window I did, I did it out of curiosity and simplicity but in reality I would rather just do stud posts and I would have doubled them up and contacted all points on all walls and it would have been effortless to cut them to fit with a miter saw. And I think I would add stud ribs to the walls running from the driver to driver plane as overkill bracing. I didn't find any problems with my current bracing, but that's how I would have done it had I gone with my typical method. Ultimately though I think the biggest thing is to use drivers that wire for final loads that are better for "cheap power." There's no cheap power for 8 ohm. The advantages of parallel is acoustic power and if I can get a final load of 4ohm, then its a lot easier to get cheap power to allow for the full power handling of the two drivers. For this reason, circling back, I would instead go for typical drivers that are dual voice coil at 4ohm each so they can be wired in series for 8ohm and then both drivers wired in parallel for 4ohm final load. Or just native 8ohm drivers wired in parallel for 4ohm. Budget drivers that come to mind that would wire better include: Skar SDR-12 or LaVoce SSF122.50L, Pioneer TSW312D4, JBL Stage 122D, Kicker 50CWCD124, MTX 5512-44. Lots of great options that cost more of course, so mainly looking at less expensive options since its double the cost to get two.
For actual use, I will be using a basic DSP (the Dayton DSP-LF, cheap, at least for now just to play aorund; I have the DSP-408 and MiniDSP HD too) to shape its output to a slight house curve to 20hz. Here's my heavy curve influence on the DSP interface and the in-room near field results:
iWoofer app on Android controlling the DSP. No importing of values, the curves are manual. I set these curves as opposite of my in-room response to get a flat result to 20hz and then gave it a gentle hump along the way to get tiny bit of house curve out of it.
Here's my near field, no smoothing, results with the DSP above in my room. I love this. Again, this is for music, but it covers everything so well. And the box is totally dead.
Very best,
Thanks; I learned a lot from this build. So far, my vibration measurements of angular velocity and acceleration and my common water test shows that the cancellation effect is very good and that it's not losing "much of the benefit" of dual opposed in phase (push push?).
Very best,
I think perhaps it would be more important, the application you mentioned, in a design with much bigger drivers with much more power handling and excursion capability. I imagine the forces from something that moves 20~25mm one way and handles 1kw per driver or more would be a different vibration animal in a smaller enclosure.
Very best,
First off, congratulations on all the detail put in your write up!
The passive radiator movement would be in phase with the driver above Fb (box tuning frequency), about ~90 degrees out of phase at Fb, and go to 180 degrees below.
At 30Hz, you may be measuring up to double the vibration a single driver would produce.
Shorting the unused driver would give a different result, closer to that of a single driver.
Half the power shared between two drivers.
Nice!
I also similar notice deviations on the blue trace ~16-17Hz, 26-27Hz, and 30Hz.
Wind noise could be responsible for those little "blips".
Less excursion than rated Xmax.
After the suspension is broken in (driven free air or pushed by hand beyond Xmax) and warm, Fs will drop.
Fs will drop when measured at a higher voltage than your test.
Fs will increase when the driver is measured in close proximity to a surface (like your work table).
Fs is normally measured with a warmed up driver suspended solidly with the cone in a vertical orientation.
Other TS parameters will change along with with Fs depending on the test conditions.
Thanks for your posts!
Art
Thanks! Learned a lot!
Here's the measurements when I did the single driver active and the other one passive, acting as a passive radiator:
Could you expand MFB? I'm not familiar with the acronym.
Thanks!
Thanks!
Very best,