Oh dear i have no clue how to work with python stuff
It is a bit of a rabbit hole, that is true. For me, it is what I know since my day job is as a programmer. Using parametric programming as CAD tools was more familiar and less daunting to me than trying to learn a normal 3d modeling tool like Fusion 360. Also didn't hurt that the parametric modeling tools, when I looked, cost a lot of money and the python framework was free 😆
Thats really nice !! thank you so much ! ill take a look into it, else i might take up on you offer. and i would be happy to donate something for it ! i can click print but i suck at 3d work 🙂
No donations needed, I do this for fun and not for profit. And your help has already been invaluable to me. Without your videos and already published documentation of how to build a planar, I would not have attempted to build mine. So I would gladly modify the code for the corrugator according to your desired dimensions and compile and publish the stl files.
I think the current design would scale to as big as you want, although some minor adjustments will probably have to be made.
Here is a 500 mm wide version as an example:
I bumped the edge thicknesses to 18 mm. The left and right edges would still work well on the 3d printer but the top and bottom will probably be best to route in plywood on your CNC since plywood is a lot stiffer and won't have to be split in two parts. The corrugator wheels themselves can be split in the middle and printed in 2 parts 250 mm tall each, since the XY precision of the 3d printers are amazing and there is a steel pipe in the middle that should give enough stiffness. And filament peg pins are part of cog wheels which would ensure that both parts are aligned perfectly.
The current design has an inner steel pipe axle with an outer diameter of 12 mm. But if we want to be on the safe side with a very wide corrugator we could widen that to say 18 mm or even more if we go crazy wide like 1000 mm.
All the parts are parametric so it would not be much more complicated than setting a desired pipe outer diameter, find ball bearings with at least 8 mm larger inner diameter than the pipe to make enough room the pins that lock all the plastic parts together.
Last edited:
Thanks ! im glad i did somethign to inspire others, while i was inspired by other members and im sure many others are at you work right now !!
love the diy community !
I might need to think what might be a nice size, before i ask you to poop out something 🙂 i bet this corrugates real nice with a motor on it as well. i noticed that if you ever stop corrugation or change the speed you see it in the end result. (not a huge thing , but in my unfinished corrugator a always occurring theme haha) i bet the handle already improves that allot!!
love the diy community !
I might need to think what might be a nice size, before i ask you to poop out something 🙂 i bet this corrugates real nice with a motor on it as well. i noticed that if you ever stop corrugation or change the speed you see it in the end result. (not a huge thing , but in my unfinished corrugator a always occurring theme haha) i bet the handle already improves that allot!!
Some minor experimentation with the goal of solving the dips here. But mostly the first dip since the second is far up in the frequency that I don't care as much:
As I experimented with previously with the PT2522 in this thread, I wanted to try to try phase plug and waveguides.
Here is the phase plug or mid vane, this one is 15 mm but I tried 8 mm also.
And as you can see in the measurements, it doesn't do much of anything.
Which is good to know, thus I can discard that idea. The reason I wanted to try was that the mid vane works great for the PT2522, but that is probably because the main problem with the PT2522 is that it has a dipole peak @ 7.5 khz where, if you use a mid vane on the rear, it mitigates this by a lot.
My diy planar does not have a problem with a dipole peak since all 4 rows are radiating equally, so this fix isn't needed. But a negative result is also a result so I am glad I tried it.
Next I wanted to try a real waveguide, since my gut feeling is that it could be beneficial to control the upper directivity somewhat and mitigate the dip in that way. Or at least make it easier to EQ by bumping the 0 degree response compared to the 15-45 degree responses.
The 90 degree response looks messed up but that is because the null is now shifted a bit forward, but can be fixed if I make the waveguide symmetric.
To be honest it worked quite well. In fact it probably worked too well since now the 9 khz + response falls off a bit too much. But the 2 - 9 khz region looks good. The directivity now has a smoother transition than the naked variant where it blooms a bit in the 7 khz region and has a bit too much off axis output.
I think I will investigate this further, but with a less aggressive throat angle. This waveguide is 90 degrees but say if I went with 110 - 120 degrees, that might work better.
I also tried to add fabric to the inside of the magnets, like how GRS does for their PT2522 and I think BG also did it with the Neo3:
The fabric does seem to help a small bit, at 8 - 9 khz the 30 degree line is a bit more separated from the 0-15 degrees which is nice. While not a huge change, it does not seem to hurt, if anything the opposite. It also does not seem to hurt efficiency, and will act as a barrier preventing stuff from falling into the drivers I will probably use such fabric in my final driver.
As I experimented with previously with the PT2522 in this thread, I wanted to try to try phase plug and waveguides.
Here is the phase plug or mid vane, this one is 15 mm but I tried 8 mm also.
And as you can see in the measurements, it doesn't do much of anything.
Which is good to know, thus I can discard that idea. The reason I wanted to try was that the mid vane works great for the PT2522, but that is probably because the main problem with the PT2522 is that it has a dipole peak @ 7.5 khz where, if you use a mid vane on the rear, it mitigates this by a lot.
My diy planar does not have a problem with a dipole peak since all 4 rows are radiating equally, so this fix isn't needed. But a negative result is also a result so I am glad I tried it.
Next I wanted to try a real waveguide, since my gut feeling is that it could be beneficial to control the upper directivity somewhat and mitigate the dip in that way. Or at least make it easier to EQ by bumping the 0 degree response compared to the 15-45 degree responses.
The 90 degree response looks messed up but that is because the null is now shifted a bit forward, but can be fixed if I make the waveguide symmetric.
To be honest it worked quite well. In fact it probably worked too well since now the 9 khz + response falls off a bit too much. But the 2 - 9 khz region looks good. The directivity now has a smoother transition than the naked variant where it blooms a bit in the 7 khz region and has a bit too much off axis output.
I think I will investigate this further, but with a less aggressive throat angle. This waveguide is 90 degrees but say if I went with 110 - 120 degrees, that might work better.
I also tried to add fabric to the inside of the magnets, like how GRS does for their PT2522 and I think BG also did it with the Neo3:
The fabric does seem to help a small bit, at 8 - 9 khz the 30 degree line is a bit more separated from the 0-15 degrees which is nice. While not a huge change, it does not seem to hurt, if anything the opposite. It also does not seem to hurt efficiency, and will act as a barrier preventing stuff from falling into the drivers I will probably use such fabric in my final driver.
hmmm thats weird it should hurt efficiency and top end.. we get really different results. what is the top end look like ? i see it drop of pre mature at 10 khz its not looking to good, a magnepan can reach higher then that ?
There probably is a drop, although the felt is very thin so might not be that much.hmmm thats weird it should hurt efficiency and top end.. we get really different results. what is the top end look like ? i see it drop of pre mature at 10 khz its not looking to good, a magnepan can reach higher then that ?
I have printed some 120 degree waveguides to play around with during the weekend. But also, I've been looking at my measurements some more and I am considering back-tracking to the square open cell foam. The EPDM foam performs better on the low end and while a 300 hz crossover would be nice, 400 hz is good enough to satisfy my requirements so it might not be worth sacrificing the top end to achieve it.
In preparation of potentially backtracking to the plain square white foam as suspension, I came up with a solution to make consistent 6 mm wide strips (I can only find 9 mm wide, and cutting by eye does not have satisfactory consistency)
Place the foam strip into the jig and then push it until it bottoms out, currently at 1.5 mm thickness.
Notice that one side is longer
The shorter piece is shorter by the width of my scissors such that when I cut them at the edge, the end result is a straight cut
And from below
Now I release the parts and I have a nice and straight edged foam strip, cut at exactly 6 mm
This jig length works great for my smaller parts, but I probably have to print a longer one such that I can cut my longer parts in one go.
Place the foam strip into the jig and then push it until it bottoms out, currently at 1.5 mm thickness.
Notice that one side is longer
The shorter piece is shorter by the width of my scissors such that when I cut them at the edge, the end result is a straight cut
And from below
Now I release the parts and I have a nice and straight edged foam strip, cut at exactly 6 mm
This jig length works great for my smaller parts, but I probably have to print a longer one such that I can cut my longer parts in one go.
And while on the topic of improving the top end response, there is one thing I have thought about on and off that could be interesting to try and see how it performs.
First to give some background:
Dimentionally, my diy planar is very similar to the BG Neo3W, of which the GR Neo3 is a clone of. Therefore, it makes sense that in the best case, my planar would have comparable top end dispersion to that driver. And that driver looks like this:
It is OK, but nothing to brag about.
The BG Neo3PDR, of which the GRS PT2522 is a clone of, however, does not radiate equally from all 4 rows, it radiates more from the middle 2 rows since it only has 3 rows of magnets instead of 5 like the BG Neo3W. This results in a better off axis response as long as we ignore the huge dipole peak @ 7.5 khz. But such a huge dip should not be ignored, hence why the 5 magnet rows Neo3W is preferable for dipoles. A common workaround for this is to turn the PT2522 sideways, which eliminates the dipole peak but has even worse off axis response.
But what if we could have both? My idea: a coaxial membrane where the outer rows are highpassed where the cutoff frequency is above the dipole peak, say I start with a conservative 12 dB LR2 @ 8 khz.
It would probably be easiest to test it with an active setup, one channel for the inner and another for the outer coil. I would not need to test distortion or efficiency, I would just need enough output to create some off axis measurements.
First to give some background:
Dimentionally, my diy planar is very similar to the BG Neo3W, of which the GR Neo3 is a clone of. Therefore, it makes sense that in the best case, my planar would have comparable top end dispersion to that driver. And that driver looks like this:
It is OK, but nothing to brag about.
The BG Neo3PDR, of which the GRS PT2522 is a clone of, however, does not radiate equally from all 4 rows, it radiates more from the middle 2 rows since it only has 3 rows of magnets instead of 5 like the BG Neo3W. This results in a better off axis response as long as we ignore the huge dipole peak @ 7.5 khz. But such a huge dip should not be ignored, hence why the 5 magnet rows Neo3W is preferable for dipoles. A common workaround for this is to turn the PT2522 sideways, which eliminates the dipole peak but has even worse off axis response.
But what if we could have both? My idea: a coaxial membrane where the outer rows are highpassed where the cutoff frequency is above the dipole peak, say I start with a conservative 12 dB LR2 @ 8 khz.
It would probably be easiest to test it with an active setup, one channel for the inner and another for the outer coil. I would not need to test distortion or efficiency, I would just need enough output to create some off axis measurements.
Last edited:
This is of course, wrong! I meant to write lowpassed.outer rows are highpassed
Other than the intended improvement of the dispersion, it will be really interesting how the co-linear membrane will behave throughout the spectrum.
The inner tweeter section might spill over to the outer midrange sections and thus nullify the dispersion improvement, have you considered to have any dampening in those outer openings?
The inner tweeter section might spill over to the outer midrange sections and thus nullify the dispersion improvement, have you considered to have any dampening in those outer openings?
Other than the intended improvement of the dispersion, it will be really interesting how the co-linear membrane will behave throughout the spectrum.
The inner tweeter section might spill over to the outer midrange sections and thus nullify the dispersion improvement, have you considered to have any dampening in those outer openings?
Exactly.
Hmm, now that you say it it would be more simple to just add damping. I did try that previously with outer damping material, but now that I have the fabric on one of the drivers I could add some wool yarn in the holes.
Would be a lot more simple than fiddling with a coaxial membrane.
Another rabbit hole is acoustic lenses. I don't know how they should be constructed for a linear source though.
I have had some more fun today:
First I wanted to see if I could use the EPDM rubber but with the top end smoothness of the foam. A solid membrane is too stiff and has the dips I want to avoid, but a gapped membrane where there are gaps around the traces is not stiff enough along the corrugations which causes a huge distortion spike at 300 hz, which is no good.
So my thought was, what if I, instead of having a solid fill, use a ladder shaped fill + EPDM rubber?
But it it any good? Well no, it was really bad so now I know another setup to avoid! 😆
Next up is adding wool yarn in the outer holes (EPDM rubber as suspension):
Compared to without the yarn from earlier:
So the yarn definently helps. In fact the first dip is eliminated completely and the second is less severe.
But now it is time for the crazy coaxial membrane, I had lots of fun with that one. Note that this driver does not have the fabric and yarn, it is a normal naked driver:
EDIT: For the coaxial driver I used the plain white foam, not the EPDM rubber! The reason for using the foam is that here I want the best top end response possible.
The first ones are with a flat +- dB gain of the outer coil relative to the inner coil, so no lowpasses or shelfs yet.
And now the more fancy parts, with a lowpass on the outer coil instead of a flat gain.
Conclusion:
Coaxially driving the membrane works beautifully. For maximum dispersion a target - gain of -6 to -12 dB sounds good. And when done as a lowpass it is interesting that the first order lowpass performs a lot better than the second order, maybe problems with the larger phase difference of the steeper filter? In fact the 1st order lowpass also reduces the off axis response blooming at 6-7 khz which is even more nice. And since there are diminishing returns when the attenuation is more than 9-12 dB it might more sense to make it a low shelf instead of a lowpass filter.
It would also be interesting to combine the fabric and yarn with the coaxial setup and see if it performs even better. If I go the coaxial route, I would want a passive solution so a single capacitor in parallel with the outer coil would be a nice and simple solution.
One negative but managable effect of doing a coaxial membrane, is that if I also want to have shading, which I do, then I have to have to run the coils on both sides of the membrane.
First I wanted to see if I could use the EPDM rubber but with the top end smoothness of the foam. A solid membrane is too stiff and has the dips I want to avoid, but a gapped membrane where there are gaps around the traces is not stiff enough along the corrugations which causes a huge distortion spike at 300 hz, which is no good.
So my thought was, what if I, instead of having a solid fill, use a ladder shaped fill + EPDM rubber?
But it it any good? Well no, it was really bad so now I know another setup to avoid! 😆
Next up is adding wool yarn in the outer holes (EPDM rubber as suspension):
Compared to without the yarn from earlier:
So the yarn definently helps. In fact the first dip is eliminated completely and the second is less severe.
But now it is time for the crazy coaxial membrane, I had lots of fun with that one. Note that this driver does not have the fabric and yarn, it is a normal naked driver:
EDIT: For the coaxial driver I used the plain white foam, not the EPDM rubber! The reason for using the foam is that here I want the best top end response possible.
The first ones are with a flat +- dB gain of the outer coil relative to the inner coil, so no lowpasses or shelfs yet.
And now the more fancy parts, with a lowpass on the outer coil instead of a flat gain.
Conclusion:
Coaxially driving the membrane works beautifully. For maximum dispersion a target - gain of -6 to -12 dB sounds good. And when done as a lowpass it is interesting that the first order lowpass performs a lot better than the second order, maybe problems with the larger phase difference of the steeper filter? In fact the 1st order lowpass also reduces the off axis response blooming at 6-7 khz which is even more nice. And since there are diminishing returns when the attenuation is more than 9-12 dB it might more sense to make it a low shelf instead of a lowpass filter.
It would also be interesting to combine the fabric and yarn with the coaxial setup and see if it performs even better. If I go the coaxial route, I would want a passive solution so a single capacitor in parallel with the outer coil would be a nice and simple solution.
One negative but managable effect of doing a coaxial membrane, is that if I also want to have shading, which I do, then I have to have to run the coils on both sides of the membrane.
Last edited:
I got an idea on the topic of the low passed coaxial variants. If the off axis problems at 12 khz and upwards are due to phase differences of the two coils then I think I have a potential solution.
Assuming the 12 dB / oct LR2 lowpass on the outer coil, I think if I also add a 1st order all-pass to the inner coil then both coils should, in theory, be in phase. Thus the lowpassed response should more closely match the flat gain variants.
Assuming the 12 dB / oct LR2 lowpass on the outer coil, I think if I also add a 1st order all-pass to the inner coil then both coils should, in theory, be in phase. Thus the lowpassed response should more closely match the flat gain variants.
Have you to tested the resilience of this foam ? one hour, one week ?white foam as suspension
The main disadvantage with foam is that it does not compress uniformally:
The active side wins over the passive.
EPDM does compress uniformally:
EPDM does compress uniformally:
Yes, but EPDM seems not to support the same response in the high frequency... I wonder why...
//
//
Yes, in OllBoll's case it's a conundrum.
I've found it hard to do SPL measurements with sweeps.
Using pink periodic noise as input and moving the microphone while doing RTA measurements captures the true SPL better.
I guess that a steady microphone doesn't capture the outputs from the slots very well.
Here's an RTA measurement with EPDM as surround where the sweep showed a lot more HF roll off:
If the straight line at 50dB is the EQ:ed level, HF only drops 5dB frow 13kHz to 20kHz.
One might even desire that.
I've found it hard to do SPL measurements with sweeps.
Using pink periodic noise as input and moving the microphone while doing RTA measurements captures the true SPL better.
I guess that a steady microphone doesn't capture the outputs from the slots very well.
Here's an RTA measurement with EPDM as surround where the sweep showed a lot more HF roll off:
If the straight line at 50dB is the EQ:ed level, HF only drops 5dB frow 13kHz to 20kHz.
One might even desire that.
With these types of loudspeakers, I've found that the microphone's position impacts the results a great deal.
To mitigate that, the Moving Microphone Method is good way for the SPL measurements.
For distortion measurements you need to move the microphone to a "good" spot.
To mitigate that, the Moving Microphone Method is good way for the SPL measurements.
For distortion measurements you need to move the microphone to a "good" spot.
- Home
- Loudspeakers
- Planars & Exotics
- DIY midtweeter planar, physically curved and shaded to be used in a dipole CBT