Andrew Zerbe - Summer 2019 Blog
(newest entries at the bottom)
SURI Kickoff. Got to see a few of the current BDML projects, at least one of which I will be helping out with this summer. Set up my account/wiki page. Finished with STARS safety training. Alex showed me how to mix and spin-coat some silicone for his stretchable interconnects.
After a joint Skype call, Alex gave me the task of CADing a few different permutations of the meandering interconnect geometries to cut on the UV laser. I read the paper he sent and found five or so more, and then did CAD for interconnects with a few different parameters (learning OnShape in the process). (Onshape file here.) We then cut out two copies of the different options on the UV laser (very cool!) and put them on two spin-coated glass trays - one with the white Moldmax and another with Dragon Skin. Attach:V1Meanders_OneSide
The first thing to do was to mask the pads and the ends of the wires for each prototype interconnect, and then spin-coat another layer on them. I actually found that I needed to spin-coat twice to get a thick enough layer that it covered the interconnects, which werenít perfectly flat due to the hand removal process after lasercutting. I then revised my original plate of meander options to focus on the ones that seemed most promising. Since the laser cutter was occupied, the third thing I did was to experiment with the heat seal TPU we have and see which was the most promising for stretchable sensors, which was part # DS3412. Every other material and thickness would wrinkle and warp upon stretching. Four layers was optimal, so hopefully we can get a sheet of the stuff with twice the thickness (currently each sheet is 50 microns thick). I then tried the same process sandwiching an actual interconnect and we looked at it under the microscope. Attach:TPU_Interconnect_Stretched_V1
June 20, 21, 24
The main thing I've been working on is refinement of the interconnect layout and process. Currently what I do is four layers of the 50 micron TPU. The interconnect is first sandwiched between the cured silicone layer it was cut on and the first layer of TPU, which I iron for 5s at 164 degrees. The TPU is now pretty well stuck to the copper, more than the tackiness of the silicone at any rate, allowing the copper interconnect to be peeled off. Then each subsequent layer of TPU is added with a 2 second ironing time. The longer I iron for, the more bubbles appear, which manifest themselves as a very tangible "pre-stretch" factor. What happens is that the firs time I stretch the TPU, all the bubbles either pop in a direction normal to the sheet, or they join with each other, and the whole thing becomes quite porous and pretty fragile.
The other problem I've been ironing out, so to speak, is fatigue and breakage of the interconnects. It seems that the width of the line has to be comparable to the thickness of the copper, otherwise it will crease at the top of each hemisphere and fail within a few cycles. Lines that are thin enough not to beak from creasing and fatigue have just an unacceptable tensile strength. The solution that I see is to double or triple up each of these lines for each channel. A prototype of this actually got pretty good results embedded in TPU.
I also took the laser safety training and should get UV laser access soon, which will allow me to really refine the interconnects and do a side-by-side comparison of Pyralux and simple copper sheet. I'm especially interested in the 2 mil sheet (~50 microns), which is substantially thicker than the Pyralux stuff. I'm also curious to see how the polyamide backing on the Pyralux influences the bending shape.
I have become quite good with using the microscope to glean meaningful information from my samples. I'll throw those pictures up when I get the chance.
June 25, 26
I did some more background research and modified my CAD for the next iteration of meanders. I also spent some time cataloging a cart of lab e-waste by serial number. I continued to refine my TPU molding process to minimize bubbling. I also prepared for lasercutting, which will take place Thursday morning, by familiarizing myself with the DPSS laser document, choosing my settings, cutting out appropriately sized samples of 1 mil/2mil/pyralux sheet, and flattening them out as best I could. The TAP blue seems insufficiently sticky, and I have a few trepidations about lasercutting the Easy-Tack repositionable adhesive because it says FLAMMABLE in no uncertain terms on the can. The flammable part is probably the aerosol, but I'm not inclined to take a big risk with it. So I spin-coated a layer of Dragon Skin onto a glass plate, and hopefully by the time it cures I will be able to determine if it is tacky enough.
Update: it seems tacky enough. Heat-sealed TPU also does not stick to it at all, so I think this is the ideal system for keeping the copper pieces aligned relative to each other.
I cut the meander patterns this morning on the DPSS laser. It worked pretty much as well as I could have hoped. The laser cut all the way through the Pyralux (even if the cut was a bit excessive), and the Dragon Skin was tacky enough to keep the patterns aligned when the excess material was lifted off--as long as the patterns were lightly pressed down with a pair of tweezers or something and the material was lifted off relatively level.
Heat-sealing preserved the alignment and didn't bond to the Dragon Skin, as hoped (and verified by earlier tests). I psyched myself out a bit on the timing and opted for 10 seconds, which was a bad choice, as many bubbles formed in the process. I should trust the 5 second-1 second rule from my earlier trials.
The final sheet performed very well. Some of the strands broke off at the very end during fabrication (which I think I can modify the design to rectify), but after stretching it quite a bit I don't see or feel any breaks. The breaks that already happened don't completely compromise the interconnects either, which is another benefit of the stranded approach.
The two-strand wires seem adequate, as do the five-strand wires, but I'm not sure which I prefer. The width of the patterns can be reduced, however, without increasing the stress/folding issues.
Maximum 200% elongation before it becomes geometrically impossible - this is true of the CAD and it behaves this way IRL.
Laser settings - 75 mm/s, 10 passes. Overkill but acceptable
Resistance - about 3 ohms over 25mm length ('displacement', not path length)
AutoCAD - save as 2000 DXF otherwise there will be import errors.
Also, it's nice to see the 3D printer actually working. I was definitely too late to the repair process to be useful, but it's fun to dip back into that hobby a bit.
After the main BDML meeting, did a few hours of cleanup, had a doctor's appointment, came back, cleaned some more, and cast a fresh layer of Dragon Skin on a larger glass plate in order to do the actual capacitor set. I also actually fixed (or possibly just fixed the settings for) the 3D printer.
I got the CAD for the actual capacitor plates from Alex, re-did my CAD in SolidWorks in order to use Sketch Blocks, did the CAD for a version of the capacitor plates with interconnects, and cut it on the laser with the fresh layer of Dragon Skin. (Current procedure is sketch in SolidWorks, make SW drawing, export drawing to DXF, delete construction lines and export watermark in AutoCAD.) I also tried the Copper shim stock, but didn't get super great results. I think the polyamide backing on Pyralux makes it stronger and more consistently bendable.
I tried out the new different types of TPU that came in, and settled on 3914 as the new best option. It bubbles far less, and stretches less, which is actually a good thing. Having a bit of stiffness will address a lot of the fatigue and breaking issues I've been seeing with the 3412.
I sealed in the full cut with 3914, and the final product was pretty good. The actual interconnects worked great, but the connections between them and the capacitor pads had a lot of stress concentrations and thin sections, and they all broke pretty much immediately. I got 67% elongation and a very satisfactory amount of tensile strength at that elongation. It curls a little after testing elongation a few times, but the breakage is the main concern, so I made a new CAD that does the connection to the pads right.
I also worked on the 3D printer with Joel. In order to print NinjaFlex, I had to take apart the extruder assembly and clean it out, since the drive gear was gummed up due to previous nozzle clogging. There's an ongoing issue with temperature setting in Gcode vs on the firmware(?) that makes some of the prints fail, and we have to run it at 130% flow just to get a consistent extruded line.
July 3, 5
I cut the new design out and sealed it in TPU, leaving out the actual electrode pads. Before casting the dielectric pillars on to the actual electrodes, we tested the process on a sample of copper and a sample of TPU with/without silicone primer to see what worked best. The non-primed TPU didn't work at all, and even the primed TPU had a lot of pillars that would just shear or fall off. So I decided to go with casting on the copper directly, and did so on the main design. Waiting for it to cure to report on results.
Alex and Jean-Phillippe talked to me about a new idea for ultra-flexible ribbon cable made by progressive layers of conductive fabric shielding, silicone, flat copper wiring, silicone, and sheathing. We cut some preliminary 'wires' out of the 2mil copper shim stock on the UV laser and then tried out the process with Sylgard 184. (Note: use the red double sided tape to stick the fabric to mold-released glass.) Some degassing issues (solvable) and some issues with the fabric having creases and generally not being flat (also solvable but more difficult). Waiting for cure on that too.
July 8, 9
The ribbon cable prototypes turned out reasonably well; well enough to try a full sensor array with them, at least. I cut it out on the UV laser, sealed it in TPU, used wire mesh fabric for one ground plate, cut out another filled-in pattern for the other plate because I donít really trust the fabric, sealed the fabric in TPU, and started casting the pillars on the TPU (w/ primer) for the other side.
Pillars for the other sensor came out rather well. So I forged ahead, cutting out two ground plates from Pyralux, sealing them in with TPU, and soldered jumper wires to all the pads and one of the ground plates.
Week of July 15
(Adapted from an update email) Iíve definitely made some progress. I have a sensor that stretches around 100% without breaking and is sensitive enough on all four pads (I have some video Iíll send along as soon as my technology behaves). I spent far too much of the week dealing with manufacturing issues - getting the laminating process for the stretchable versions/Pyralux down, especially with the stretchier TPU material. I also have to recoat the glass plate with dragon skin after basically every cut, as with the finer interconnects the laser marks basically roughen the surface of the silicone too much for it to be reusable. (I tried easy tack too- boy did that not go well. Iím still suspicious that thereís a more reliable solution, though). Also lots of brass shim experimentation - it just doesnít seem to peel off nicely enough, I think the polyamide layer in the Pyralux keeps it relatively rigid, whereas with the brass, any normal force lifts the traces it right off the silicone. I have some ideas for that too. I want to try doing the stranded wire just on the edges of the meander, and have them join back into a main wire for the part that doesnít define as much. That way, the lift off process will be much less finicky.
We encountered some issues within the active shield, namely that its close proximity to the actual sensor means that touching or otherwise interfering with the shield causes all four channels to respond equally, a telltale sign. Shielding issues wise, Iíve tried some different TPU thicknesses, and I think the next thing to try is actual silicone; as soon as the thickness between the shield and the plates is large enough, it just doesnít stretch very well. So thatís the next thing to try. .6mm or so is what Iím targeting, after some discussion.
I also think that would help with another similar issue, which is that the capacitance increases by a huge amount when the interconnects are stretched. I think this is because the TPU is bonded around the wires rather than really to them, so they end up getting closer/generally shifting when stretched. I think genuine silicone is the thing to try with that too. I think it will actually save time, just because the process has the potential to be more reliable and repeatable.
The first thing I did this week was to start making connects with silicone. I spent far too much time actually figuring out how to do lift-off from the Dragon Skin glass slide onto a separate dragon skin sheet, before realizing that I can permanently attach the bulk Pyralux to the silicone, cut it, and then peel off the excess, just with a bit more force. Iíve since done this with Sil-poxy and with a thin spin-coated layer of dragon skin on top of a larger cured layer, and the latter worked better. It certainly helps to be able to bake the dragon skin (66 degrees C for ~20 min, although this varies as it takes the oven some time to heat up. I just check every 5-10 min).
1. Spin coat at 700RPM
3. Spin coat at 2000RPM (increase?)
4. Add bulk Pyralux
6. Cut out
7. Mask pads / connectors
8. Spin coat at 2000RPM (next time: excluding the GND layer)
10. Add primer to pads and wait to dry, then cast pillars directly on pads
10b. Solder header and jumper wires, cover with sil-poxy
11. Stack up individual layers from bottom to top, using some DS20 made into a thin layer with a razor, and use weights to press down until silicone becomes tacky
12. Then, cure fully in oven.
13. Adhere GND plate to pillars (unsure what the best tactic here is yet).
The most recent prototype works rather well. 84% extension, passes the feather test, little to no response from stretching or tapping on wires. Next steps are to make it thinner / more robust / more stretchable, eliminate hysteresis, and try doing a full-finger version with the jig (as well as wraparound).
I prepared to fabricate V4 next week. I used primer on the bulk Pyralux to bond it to the first layer of silicone, so hopefully the peel off process will go OK but the bonding will be better overall. I definitely plan to use primer on all the other layers too. Iím also going to actually use spin coated TAP blue to bond the pillars to the ground layer this time and hopefully that will solve the weird hysteresis effects. And Iíve bumped it up to 1000RPM for each layer, so hopefully this will finally be thin enough. I also realized that I can keep the shield and the ground on the same layer; I donít really understand why I didnít think that was an option. Hopefully this one will get to 100% (I think the last one only has about 75% or so. Fine for distal but not as good for proximal).
The primer is far too aggressive for peel off but should improve things for the other layers.
Kapton is a horrible mask for silicone because a) itís thin, b) itís far too sticky, and c) the top layer is too smooth so the unwanted silicone doesnít adhere to the tape at all. Sticking (so to speak) to masking tape from now on.
...has been spent fabricating an improved proximal sensor. Thinner silicone, integrated ground and shield, sealed ground and pad layers around the edges (and glue-sticked to flat plate) rather than sil-poxy / glass slide weight (this last one may not be an improvement).
No stretching effects, minimal shield effects, substantial (but not that bad) coupling between the two closer pads especially, some electrical weirdness, and 80% stretch or so with substantial (but, again, not that bad) spring constant. Suspect buckling of TAP blue pillars leads to unfortunate hysteresis.
Started fab on another sensor. Main differences are in the sensor pads rather than the connects, using thicker pillars and a bulk layer where pillars are not needed to try and decouple. Also adding a bulk layer of tap blue to the ground plate to prevent permanent curling when the layer is peeled from the glass slide.
There's yet another sensor that will use super thin cut foam, which Alex proved the concept of today, but otherwise will be identical. (How will we attach foam to copper? Hopefully an incredibly thick TAP blue layer will be sufficient.)
And a third sensor that uses Dragon Skin 10 instead of 20 which should improve stretchability without compromising durability.
A stupid mistake on my part with one of the sensors led to it getting scrapped. Currently there is a DS10 & pillars version in progress and a DS20 & foam version on standby. We are trying out foam attachment methods and foam types on some smaller pad sets without flexible interconnects before Ďwastingí a full set of actual manufactured sensor stock. Spoke to Hojung; it seems that we need to 1) degas 10Ē before and after pouring on sample and 2) have the molded pillars facing down onto the flat side to minimize drip.