My 3d printer has been ressurected! And so I immediately ordered it to start printing!
I have de greased the steel plates and sprayed a light coat of black paint on the outer sides, the sides I will not glue magnets to. The curve on the smaller parts is hard to spot so the idea behind the paint is to make it easy to avoid gluing the magnets to the wrong side.
I plan to start gluing magnets tomorrow. (the grey plastic parts are my glue jigs to help fit magnets at every odd hole. Then I have smaller jigs to help fit magnets in between the then glued magnets.
And meanwhile, I have test printed an edge part to test the fit. And it fits nice and snug. With a small dash of superglue this should fit together perfectly. I just need enough glue to hold it together before assembly. When the final assembly is done then the 3 screws will hold it toghether and provide the strength.
In my previous iteration I used small M3x3 mm threaded inserts but if I tightened the screw too much they came loose inside the plastic part. That would then be a recipe for disaster since it would probably rattle. This iteration with a little glue and then every other hole with a plastic dowel pin to hold it in place is a much easier and more reliable solution.
If it continues to progress as nicely and fast as this, then I might have the two test drivers with 4 and 6 rows of membrane traces respectively done by next weekend and ready for measurements.
I have de greased the steel plates and sprayed a light coat of black paint on the outer sides, the sides I will not glue magnets to. The curve on the smaller parts is hard to spot so the idea behind the paint is to make it easy to avoid gluing the magnets to the wrong side.
I plan to start gluing magnets tomorrow. (the grey plastic parts are my glue jigs to help fit magnets at every odd hole. Then I have smaller jigs to help fit magnets in between the then glued magnets.
And meanwhile, I have test printed an edge part to test the fit. And it fits nice and snug. With a small dash of superglue this should fit together perfectly. I just need enough glue to hold it together before assembly. When the final assembly is done then the 3 screws will hold it toghether and provide the strength.
In my previous iteration I used small M3x3 mm threaded inserts but if I tightened the screw too much they came loose inside the plastic part. That would then be a recipe for disaster since it would probably rattle. This iteration with a little glue and then every other hole with a plastic dowel pin to hold it in place is a much easier and more reliable solution.
If it continues to progress as nicely and fast as this, then I might have the two test drivers with 4 and 6 rows of membrane traces respectively done by next weekend and ready for measurements.
Last edited:
Some progress:
Still have to glue the magnets for the front plates & for the 4-trace-row variant. But many small steps.
My schedule has changed though so will most likely not have measurements for the 2 variants this weekend, but definently by the weekend thereafter. And in that next week, my weekend starts at thursday 🙂
Still have to glue the magnets for the front plates & for the 4-trace-row variant. But many small steps.
My schedule has changed though so will most likely not have measurements for the 2 variants this weekend, but definently by the weekend thereafter. And in that next week, my weekend starts at thursday 🙂
Small adjustment.
I didn't like how abrutply the edge foam ended when the magnets ended, so I removed some of the plastic. Since there is less plastic, this means I will have to make the top pieces on the edges out of steel, but that is easy enough to do when ordering next time. Until then I will make them out of plastic but much thicker to ensure they don't bow under the pressure.
I also didn't like that the white foam was not as evenly applied as I would like, but it was easy enough to remove and reapply. And using the full foam width without cutting it and fastening it with the self adhesive glue is very easy. I pray to the gods of audio that the 6 row setup will work great, that the coaxial magic works as well here such that I can get as good top end dispersion on this wider driver as the 4 row setup, because the foam is much easier to apply on the wider version.
I didn't like how abrutply the edge foam ended when the magnets ended, so I removed some of the plastic. Since there is less plastic, this means I will have to make the top pieces on the edges out of steel, but that is easy enough to do when ordering next time. Until then I will make them out of plastic but much thicker to ensure they don't bow under the pressure.
I also didn't like that the white foam was not as evenly applied as I would like, but it was easy enough to remove and reapply. And using the full foam width without cutting it and fastening it with the self adhesive glue is very easy. I pray to the gods of audio that the 6 row setup will work great, that the coaxial magic works as well here such that I can get as good top end dispersion on this wider driver as the 4 row setup, because the foam is much easier to apply on the wider version.
More progress! All magnets for the 4 row & 6 row experiments are now glued!
And sandpapering the magnets individually took a lot of time, so I devised a method to speed it up which turned out great! I cut up an A4 sized plate of 2mm steel. Then I placed a layer of thick paper on top of the steel, and then flat glue jigs from earlier which are as high as the magnets. Then I dropped the magnets into the jigs to make them stick to the steel plate.
Then I sanded (80 grit sandpaper) the magnets by hand over the jigs such that only the top surface is sanded. In the third image I have sandpapered the top right 5 columns. It worked perfectly and afterwards I vacuumed the magnets and the jigs to remove particles.
Afterwards I removed the plastic jigs and combined the magnets. When sanding I made sure that the top row had one polarity and the bottom row another, and then when reassembling into a long stack I had some colored magnets in the middle to help me know which is which.
And sandpapering the magnets individually took a lot of time, so I devised a method to speed it up which turned out great! I cut up an A4 sized plate of 2mm steel. Then I placed a layer of thick paper on top of the steel, and then flat glue jigs from earlier which are as high as the magnets. Then I dropped the magnets into the jigs to make them stick to the steel plate.
Then I sanded (80 grit sandpaper) the magnets by hand over the jigs such that only the top surface is sanded. In the third image I have sandpapered the top right 5 columns. It worked perfectly and afterwards I vacuumed the magnets and the jigs to remove particles.
Afterwards I removed the plastic jigs and combined the magnets. When sanding I made sure that the top row had one polarity and the bottom row another, and then when reassembling into a long stack I had some colored magnets in the middle to help me know which is which.
I assume you mean like them chomping together to crush say a hand? If so then that is the main reason I only have 5 lengthwise rows of magnets per piece on the top.
And no more progress on the test drivers today, that will have to wait until thursday.
But I did start to dissasemble some of my old drivers. I want to try to see if I can recycle the magnets. If I can recycle say 80% of the magnets then that would be a huge saving.
And with the CA glued magnets without any sandpapering on the surfaces, they are very easy to knock off because the glue is very brittle. I put the small particle board piece against the magnets and then lightly tap it with the wood mallet. This is enough to break the glue and let me remove them. I start with removing the 2 outer rows and then collect the magnets, then repeat untill I have all the rows and then go to the next piece.
When I was at it I decided to not only try to recycle the magnets from my last full scale driver but also my earlier test drivers. And it worked great to remove the magnets except for 1 piece. At first I didn't understand why the magnets didn't get unstuck as easily but with more force I got some magnets off. But while most of the magnet got off I believe the copper-nickel alloy protecting the magnet from oxidation was still stuck to the steel plate...
It was about then I remembered that yes, for this plate I didn't use CA glue but rather epoxy. And it turns out that epoxy is way way stronger and less brittle than CA glue.
This made me want to repeat the strength test but with sanded magnets and steel plates. If the CA glue is still this brittle even with properly sanded surfaces then I am considering switching glue to epoxy when scaling up to full size. It will take longer to glue everything, but I do not want the magnets to get loose in the finished driver and if epoxy is what gives that then epoxy is what I will use. Especially since the finished driver will only have the rear steel solid, so might flex a bit until it is permanently mounted in an enclosure. Epoxy might have a lot less risk of cracking during this potential flexing than CA glue.
No further work will be done until thursday when I will continue with gluing the 3d printed pieces to the 4 trace row plates & adding the foam suspensions. But most importantly I will start working on the membranes. The plan is to produce 2 membranes of the best config I found, which was 23 um mylar + 30 um alu with 30 um acrylic self adhesive. Whole membrane baked at 150 degrees C in between two layers of 40 um aluminum which after the baking process are discarded.
I have already prepared the corrugation roller to support the 6 trace row membrane which will be 82 mm wide as opposed to the 4 trace row variant which is 55 mm wide.
Some minor progress, just finished the membrane models for 4 & 6 rows respectively.
It will be interesting to see if coaxial driving is magic enough to go with 6 rows of traces without significantly impacting the top end off axis response. And then secondly, I have 6 pads on the 6 row variant because I want to see if I need to triaxially drive it or if I can get away with coaxial. Coaxial would be much easier because then I could cut the membrane on a single side, on top of needing less passive components for the lowpass network.
It will be interesting to see if coaxial driving is magic enough to go with 6 rows of traces without significantly impacting the top end off axis response. And then secondly, I have 6 pads on the 6 row variant because I want to see if I need to triaxially drive it or if I can get away with coaxial. Coaxial would be much easier because then I could cut the membrane on a single side, on top of needing less passive components for the lowpass network.
Some more minor progress:
In addition to cutting a 4-row & 6-row membrane, I'm not sure exactly how much shorter the corrugated membrane will be. So based on previous photos I calculated +20% but I also cut a membrane with +15% length.
The short pieces with a "B" and "T" are stiffening supports that I will add to the underside of the membrane. The idea is to make it much stiffer on the ends where I mount it to the enclosure. It also adds extra protection during soldering since it negates the risk of burning through / ripping the mylar.
Next up is to mount it on some 23 um mylar, cut the surrounding alu sheets for stiffness during baking and then after baking remove them and solder the wires. Then it is time for assembly!
If everything goes well I should have measurements on saturday.
In addition to cutting a 4-row & 6-row membrane, I'm not sure exactly how much shorter the corrugated membrane will be. So based on previous photos I calculated +20% but I also cut a membrane with +15% length.
The short pieces with a "B" and "T" are stiffening supports that I will add to the underside of the membrane. The idea is to make it much stiffer on the ends where I mount it to the enclosure. It also adds extra protection during soldering since it negates the risk of burning through / ripping the mylar.
Next up is to mount it on some 23 um mylar, cut the surrounding alu sheets for stiffness during baking and then after baking remove them and solder the wires. Then it is time for assembly!
If everything goes well I should have measurements on saturday.
I have also realized a plausible reason of why my and @solhaga s distortion measurements differ so much. Mine has too much distortion @ 300 hz and needs to be crossed at 400 hz where his can handle 300 hz without too much distortion.
The reason is most likely that my previous experiments were only 26 cm tall (in the middle section with magnets), where his was 40 cm tall. My new experiments are 31 cm tall so will probably handle the low end a bit better, but the most important part is that this implies a relationship between distortion @ say 300 hz to membrane length that is not linear.
Or in short, consider the two following setups.
If this is true, this is great news for me because it will let me cross lower in my full scale driver than I previously assumed, even if I go with 4 rows traces instead of the wider 6 rows of traces). 300-400 hz is probably doable, instead of 500 hz which I was afraid I would have to settle with.
The reason is most likely that my previous experiments were only 26 cm tall (in the middle section with magnets), where his was 40 cm tall. My new experiments are 31 cm tall so will probably handle the low end a bit better, but the most important part is that this implies a relationship between distortion @ say 300 hz to membrane length that is not linear.
Or in short, consider the two following setups.
- A single 40 cm tall membrane outputting 90 dB @ 300 hz
- Two 20 cm tall membranes together outputting 90 dB @ 300 hz
If this is true, this is great news for me because it will let me cross lower in my full scale driver than I previously assumed, even if I go with 4 rows traces instead of the wider 6 rows of traces). 300-400 hz is probably doable, instead of 500 hz which I was afraid I would have to settle with.
Some steps forward, some steps backward:
I corrugated all 4 membranes and then left them an hour because I needed to eat, but when I got back the 3/4 of the corrugated membranes had un-corrugated themselves!
I forgot to take a picture, but basically the mylar was strong enough to push away the top and bottom 40 um alu support foils. The point of the support foils is to encapsulate the inner membrane and help keep it's shape during baking.
EDIT: here is a rough picture of one of the membranes in the image on the top as seen from the side, after the mylar un corrugated it:
Hence an important lession!
After corrugating the membrane with the support foils, bake them immediately! Do not wait like I did!
So when I make the full size membranes, because of this it is probably a good idea to corrugate and bake one membrane at a time so there is less time for things to go wrong.
And another smaller detail:
With the wider 6 trace rows membrane, the corrugations are less deep than with 4 trace rows variant which has resulted in my corrugation ratio being messed upp. Basically, even my +15% membrane is too long as you can see in this photo:
This is not a huge problem, however, since I can make new holes and just let the top of the membrane extend a bit outside the driver. I even think I have some extra self adhesive rear supports I can add to make it stiffer and to ensure that the new holes are aligned properly. The above membrane skews to the bot on the top end where the ruler is, the holes are just slightly off from where they should be.
I corrugated all 4 membranes and then left them an hour because I needed to eat, but when I got back the 3/4 of the corrugated membranes had un-corrugated themselves!
I forgot to take a picture, but basically the mylar was strong enough to push away the top and bottom 40 um alu support foils. The point of the support foils is to encapsulate the inner membrane and help keep it's shape during baking.
EDIT: here is a rough picture of one of the membranes in the image on the top as seen from the side, after the mylar un corrugated it:
Hence an important lession!
After corrugating the membrane with the support foils, bake them immediately! Do not wait like I did!
So when I make the full size membranes, because of this it is probably a good idea to corrugate and bake one membrane at a time so there is less time for things to go wrong.
And another smaller detail:
With the wider 6 trace rows membrane, the corrugations are less deep than with 4 trace rows variant which has resulted in my corrugation ratio being messed upp. Basically, even my +15% membrane is too long as you can see in this photo:
This is not a huge problem, however, since I can make new holes and just let the top of the membrane extend a bit outside the driver. I even think I have some extra self adhesive rear supports I can add to make it stiffer and to ensure that the new holes are aligned properly. The above membrane skews to the bot on the top end where the ruler is, the holes are just slightly off from where they should be.
Last edited:
I forgot to take a picture, but basically the mylar was strong enough to push away the top and bottom 40 um alu support foils. The point of the support foils is to encapsulate the inner membrane and help keep it's shape during baking.
EDIT: here is a rough picture of one of the membranes in the image on the top as seen from the side, after the mylar un corrugated it:
Here is an photo of what happened:
And to solve the membrane being too long, I added another underside support piece with holes:
I also have measurements! But something is wrong...
Either somethign is rattling or the membrane is not centered, since there is a lot of distortion when pushing xmax.
I was about to dissassemble the driver when I stopped myself and came to the realization that while the sub 1k response sucks, what I need to test isn't the low end response but rather the top end. And the top end looks fine even with the problem.
So I hooked it up coaxially such that the 2 outer traces on each side (4 in total) were low-passed with LR2 @ 5 khz. The 2 innermost traces were run full range except that they were first order all passed @ 5 khz to match the phase response of the low-passed traces.
And it works! But we have definently reached the limit of adding more rows and magically solving the top end off axis response by low passing the outer traces.
Here is the 4 trace rowed variant with same filters but at 8 khz:
The 4 row variant is much more clean on the top end. In fact the top end is so much better that I can get away with a more simple filter, just having a first order low pass on the outer traces and no all pass on the inner. And such a filter would be very easy to realize with passive components, just a single capacitor in parallel with the 2 outer trace rows.
I still plan to assemble the new version of the 4 trace row variant:
But if it measures as expected, which is like the previous iteration, then that is what I will go forward with. Crossing lower would be nice, but I would rather cross at 400-500 hz and have a great 10-20 khz response than crossing at 250 hz and have a bad response at 10-20 khz.
The results are in, and the 4 trace variant does have the distortion problem that the 6 trace variant has. Or more likely the problem that the current assembly of that 6 trace variant has.
But most importantly, the 4 trace variant measures pretty much exactly as expected, which is what I hoped for. Here are 8 khz first order low pass vs second order @ 50 cm:
This variant is a little bit longer (31.25 cm long where the previous was 26.25 cm), so the off axis blooming above 5 khz might be because interferance due to the driver height since I don't have any shading here. So I also measured at 25 cm mic distance:
It is still there, but it is less. It will most likely be reduced when going full length with shading. But I also plan to experiment with small waveguides which have worked well in other experiments. A waveguide like this, which is from a previous planar iteration:
In conclusion, because the 4 trace variant measures exactly as expected, I will go with the 4 trace variant for my full size driver. Hence the next step is to finalize the models and order steel for a full size 4 trace variant of the new mechanical setup. Having the rear steel plate in one piece worked great at ensuring that the membrane is perfectly aligned, which is what I hoped for.
Also, here is a photo of my timing reference speaker. It is two 32 ohm headphone drivers wired in series hooked up to the same sound card that forwards the SPDIF signal to my DSP amplifiers.
But most importantly, the 4 trace variant measures pretty much exactly as expected, which is what I hoped for. Here are 8 khz first order low pass vs second order @ 50 cm:
This variant is a little bit longer (31.25 cm long where the previous was 26.25 cm), so the off axis blooming above 5 khz might be because interferance due to the driver height since I don't have any shading here. So I also measured at 25 cm mic distance:
It is still there, but it is less. It will most likely be reduced when going full length with shading. But I also plan to experiment with small waveguides which have worked well in other experiments. A waveguide like this, which is from a previous planar iteration:
In conclusion, because the 4 trace variant measures exactly as expected, I will go with the 4 trace variant for my full size driver. Hence the next step is to finalize the models and order steel for a full size 4 trace variant of the new mechanical setup. Having the rear steel plate in one piece worked great at ensuring that the membrane is perfectly aligned, which is what I hoped for.
Also, here is a photo of my timing reference speaker. It is two 32 ohm headphone drivers wired in series hooked up to the same sound card that forwards the SPDIF signal to my DSP amplifiers.
Is it really wise to have these ref speakers at 90 degrees angle vs the DUT? I would tink having them as close to the DUT as possible would be much better?
//
//
Having it below seems to work just fine since the mic is omnidirectional. If I have understood correctly what the ref speaker does, then it acts as a reference for the time windows and thus ensures that wall reflections aren't interpreted as the actual signal. Which in my case is important since I measure the 90 degree dipole null, the wall reflection is way way stronger, so without the ref speaker I might not be able to measure the null correctly.
If so, then my interpretation is that it should not matter where the reference speaker is, as long as the ref timing sweep is picked up by the measurement microphone. Although with the asterisk that if the timing ref speaker is at a different distance than the DUT, then the window times might need adjusting.