Ask Hackaday: Building nano scale antennas

As an RF engineering student, [Camerin] is usually tasked with pointless yet educational endeavors by his advisor and professors. Most of the time (we hope) he sees the task through and ends up pulling something out of his hat, but a few days ago a professor dropped a bombshell on him. After reading this article on nano scale antenna fabrication, a professor asked [Camerin] if it was possible to build a 3D inkjet printer with a ludicrous amount of accuracy and precision.

The full article, Conformal Printing of Electrically Small Antennas on Three-Dimensional Surfaces, was recently published in Advanced Materials and is available via Google Scholar. The jist of the article is that three-dimensional antennas printed on a sphere approach the physical limits of how good an antenna can be. To test out these small, spherical antennas, the authors of the paper built an extremely high-precision 3D inkjet printer that draws antenna traces on a glass sphere with conductive ink.

The positional accuracy of this printer is 50 nanometers, or about half the size of an HIV virus. The conductive silver ink is delivered by a nozzle with a diameter of 100 to 30 µm and prints onto a glass sphere about 6 mm in diameter. This is a level of precision that companies and research institutions pay top dollar for, so we’re left wondering how the authors built this thing.

We’re turning this question over to the astute readers of Hackaday: how exactly would you build a 3D inkjet printer with this much accuracy and precision? Would it even need to be that precise? Post your answer in the comments.

69 thoughts on “Ask Hackaday: Building nano scale antennas

  1. If I were to be tasked with something this precise, I would probably go looking up some Piezoelectric positioning tables.

    I’ve used some from Nanomotion before, but there are numerous manufacturers that offer this type of equipment. Prices generally start at a few hundred dollars plus drives, but in theory you could just build your own if you know how the physics work.

    1. ^^^ This, or the head positioning drive. I don’t know the resolution they have but a CD/DVD/BD track can’t be very wide.

  2. In the past I have thought that having some stepper motors ( somehow ) use a fulcrum with the pivot very close to the work area. That might translate larger inaccurate movements into very small very accurate movements.

    1. Ex-diemaker here, I’ve built stuff near this kind of precision now and again, master gauges to check the production gauges on fussy things,and the like. Newport’s rotary platforms are too imprecise, ±5 degrees frankly sucks. Their CNC Stewart Platforms are nice though. Probably doable with one of those.

      I’d go with a pair Moore Rotary tables, ±0.1 second of arc, instead and a stationary inkjet. The glass marble is 6mm, around 1/4 inch, not that small. He is getting a angular resolution around 1.7 arcseconds to get a arclength of 50 nm.

      Delmar’s idea using harmonic drives is a good one too, I think my old wire EDM’s used a pair for stepping the X-Y axis, accurate to a micron, the U-V axis used standard NEMA steppers though.

      1. I goofed. 1.7 x 10 ^-5 radians, around 3.4 arcseconds. You get what you paid for ;)
        Actually, I gave the inscribed angle instead of the plane angle, which also happens to be 1.7″

        A 3.4 arcsecond resolution is definitely doable.
        Outside of telescopes and surveying instruments you don’t see high precision angular measurements very often.

        If anyone is curious how a toolmaker gets precise angles look up the wikipedia article on sine bars

        The smaller bottom sine bar in the picture there is the better design. On the larger top one you have only one chance to get the distance right between the cylinders, and on the scales that I used to work at, metal moves and flows over time. 5 years and the larger sine bar would be garbage, while the smaller one you could grind off a few millionths off of the right hand side to move the cylinders closer together or grind off the left one to move them apart. The distance between the cylinders is the hypotenuse, and you use gauge blocks to get your side opposite. The Master I studied under had me ponder what was wrong with that design without telling me, I’ve given you the crib notes. It’s a clever and elegant tool.

  3. As one who is building crude CNC machines, I will personally attest that belts and screws are out (I’d also argue against levers as there is too much room for error). My first thought was voice coil / solenoid actuators (like on hard drives). The optical lens positioning hardware on DVD drives would be a another good start. If speed wasn’t a concern, would thermal expansion/flex of bi-metallic strips be an idea? A laser bounced off a mirror to form the feedback loop of a servo type controller would also be critical.

    1. Belts have been shown to repeat at least to 5 microns when tested against laser interferometers. I never bothered to go beyond this, but I suspect it could improve. You do need to calibrate to some standard and do a linear interpolation if you want accuracy as well as precision.

  4. any reason you can’t build it like any other 3-d printer and just use an insane gear ratio?

    This printer doesn’t have to be precise over a huge area. it’s only printing in a space 10mm x 10mm x 10mm.

    1. Unfortunately, it takes different design techniques. Not only do you need to design for really small motion, you also have to design for really small error.

      The rules of thumb for really high precision are to make your error stack as short as possible, to get your slop as close to zero as possible, and to keep your binding forces as close to zero as possible.

      Gear trains don’t do any of those particularly well. They’re primarily force multipliers, with increased or decreased motion being a side effect of the change in force. For a reducing train, the big problem is that your position errors multiply in the wrong direction.

      Say you have a gearbox with a 1000:1 reduction ratio. Now let’s say the last gear has .001″ of backlash error. That multiplies up to a full inch of error at the input gear, assuming the rest of the train is perfect. If you drive the train with one gram-centimeter of torque (again, ignoring errors), you end up with a kilogram-centimeter of force at the output end.. far more than you need, and possibly enough to make things flex, adding even more error.

      The general process for getting high precision is to start with a mechanism that’s large and has good precision, then jigger it around to work incredibly well over a small fraction of its range. In this case, the stage they used is a monster with air bearings that float on granite rails.

      You can also buy precision by trading away speed. Run closed loop, measure like mad, and move slowly enough that your errors will settle and there’s no chance for oscillation.

  5. This is the way I would do it….

    If you could build the 3D printer specifically to print on a sphere, using a longitude and latitude coordinate system, and made the print head ridiculously stable and impervious to vibration; it would be possible to print small details….

    Think of it this way: draw two concentric circles, one much larger than the other. Then draw a straight line from the centre of the circles to the outside circle. Where the line crosses the circles move a quarter of the way around, you’ll have moved moved further along the large circle than the small circle. The concept should also apply in 3D

  6. Microscope Stages (x/y) and focus mechanisms (z) have nanometer resolution, but aren’t cheap. They may have had access to an automated microscope.

  7. It might not be a general purpose device, but one built for this exact purpose and dimensions. You could have it so a curved track moves the ink jet nozzle tip just above the surface, moving it from the bottom of the sphere to the top. Then you would just need to spin that along its axis.

  8. i would rotate the 6 mm sphere on a very big wheel, the bigger the outer diameter of the wheel the more precision you will go. You can go very precise with a low precision motor at the outer side of the wheel. a Two meter wheel would be cool
    (similar to a dynamo on a bycicle wheel)

    1. as motion in the other two directions is one way only you could possibly attempt traditional motion control geared down (backlash not being an issue)

    1. The virus is not being used as a unit of measurement, anymore than the human hair, or a gnat’s ass where ever used as a units of measurement

  9. Not trying to accept a challenge. but i think it could even be done by hand… i’ve soldered up a 64BGA chip to a breakout board with a large mag glass. thats what 10mm/10mm and when you think about those small pads on those things why not with a hypodermic needle and a steady hand…

    1. Perhaps you don’t understand the scale. The claimed accuracy is 0.00005mm or 0.00000196850394 inches. That is orders of magnitude more than DIAMONDS flex at such a localized scale. I am not saying their claim is true, I believe it is not in practice, however the idea that a human could possibly do anything at those scales is insane.

      Merely having a heartbeat makes work anywhere near ‘normalized’ CNC scales impossible. This process is 1000 times smaller.

      Using a hair width tool to get this process done would be directly equivalent to using the moon to dig a swimming pool in your back yard without the neighbors noticing.

      1. It’s quite likely that they can achieve the resolution they are claiming. It sounds like they’re using a piezo drive which – it you’re careful – can achieve 10s of picometer resolution. STM images can regularly get that sort of resolution. It require cryogenic temperatures and a vacuum generally but is doable.

        Now, the question is whether that level of resolution is actually required for making a functional antenna. My guess would be no.

      2. I understand what their claim is…But thay also claim that the dome is 6mm in diameter.. look at that close up… that antenna printed on there is at least 2.5 mm wide. or am i missing something??

    1. They are reporting 4 antennas: 1.7 GHz, 700 MHz, 1.69 GHz and 3.5 GHz. Of course the efficiency is reduced at the 700 MHz band. The reason were looking at the paper is for implantable sensor communications. The sensors would be on the mm scale… and producing electronics on glass at that scale isn’t easy…

  10. I wonder how precise a linear motor would be, not to be confused with a linear actuator. The company I work for uses a big ass linear motor to accelerate 1500lbs at 1G and is ridiculously accurate. Speaking of accurate the company also has something called an accurate robot which is a kuka titan robotic arm with extra encoders and compensation algorithms based on FEA models all the way down into the ground under the robot and laser tracking data. It’s accurate to within 3 thou over its entire range of motion. Oh well, I don’t know about positioning in the nanometer scale. The long axis on the machine I’m working on is 100 feet.

  11. I know nothing about RF engineering, and no more about 3D printing than I learned from reading HaD articles. However, I wonder if maybe there are better ways to create the antennas (antennae?) than constructing a very precise printer. Off the top of my head:

    1) what about printing on a flat substrate that could be vacuum-formed onto the sphere? Uniformity might be an issue, I guess

    2) What about laser-etching the sphere, and depositing a conductive substance into the resulting groove and drying/baking it? I would expect precise laser etching wouldn’t be that hard, and the sphere could be polished afterwards to make sure the conductive material was only left in the grooves

    I’m not saying that these ideas are better, but at the very least I would expect both of these would be better for producing more than one at a time.

      1. This is somewhat how photomasks are applied to silicone to create 20 nanometer width trace circuits.

        Photo lithography – project an image and then mask and etch it.

  12. Would there be any way to reasonably apply typical semiconductor manufacturing methods to a spherical substrate? I admit I don’t know enough about how Intel et al produce their chips to guess.

  13. Someone got the units wrong. 50nm accuracy is rediculous. 50 micro m is pretty good. And it seems in line with the video. Half a human hair is the usual comparison. The syringe in the video is a standard syringe. They are colour coded. I happen to have the slightly bigger green ones here. A 100 micro m (tenth of a mm) output channel sounds reasonable.

  14. Someone got the units wrong. 50nm accuracy is ridiculous. 50 micro m is pretty good. And it seems in line with the video. Half a human hair is the usual comparison. The syringe in the video is a standard syringe. They are colour coded. I happen to have the slightly bigger green ones here. A 100 micro m (tenth of a mm) output channel sounds reasonable.

    1. I think you are right.

      Is the photo and it shows a centimeter for scale comparison.

      1 cm = 10,000,000 nanometers.

      A centimeter is 10 millimeters. A millimeter is 1000 microns. A micron is 1000 nanometers. Typical human hair can range from 200 to 50 or so microns. 3D printer resolution is limited to roughly 15 microns on high end industrial SLA and polyjet machines.

      This is micron range, not nanometer range. BIG (heh) difference.

    2. Roger, the Aerotech table they are using should enable them to get superb resolution (1nm) and accuracy (10nm at least, I would think) if they couple it with a laser intereferometer. I think this is way overkill for the example they are showing in the photo.

  15. If the printing is to be done only on a sphere of glass, then it would be better to keep the printing nozzle/needle still and rotate the sphere. I think it would result in better accuracy.

    1. I’ve seen one used to print tiny strain gauges using a liquid carbon nano-tube solution. It was so small it looked like a little spec by the naked eye, but with a microscope it looked just like a regular strain gauge.

  16. Of course, we well know that gear trains, leavers and such will leave room for backlash / slack. What if one were to build that same gear reduction setup with a heavy spring pretensioner? Say a table gear reducted down with x y drive, and a ‘big ole’ spring on the -x y corner. Really a silly workatound, butWouldn’t it takeup the slack and reintroduce precision back into the system?

  17. I have a few issues with what the photo seems to imply. The dispense tip appears (based on the yellow hub and fuzzy profile!) to be a standard EFD/Nordson 32 gage tip (#5132-1/4-B)with a 100 micron (0.004″) ID. That makes the sphere in the photo roughly 3/4″ in radius. Each of the “zigs” in the photo would then be roughly 0.04″ or 1mm up and my guess is that each line in the photo is 0.01″ wide – maybe a hair less. The OD of the EFD needle is 0.01″

    Perhaps this was a larger case for illustrative purposes or something like that. Perhaps I am wrong about the tip.

    I confess I did not read the original article. I found abstracts through Google Scholar, but not the whole article.

    Also, nano (10^-9) is quite small with normal positioning systems being rated in the micro (10^-6) end of things. At this scale, accuracy vs precision (think of precision as the ability to repeat to a position every time versus accuracy which is the ability to hit an actual, NIST traceable coordinate) becomes a pretty relevant argument and getting nano scale repeatability is easier to assess than accuracy. An HP 5530 laser interferomoter system would tell you if you were both accurate and precise (1 nm resolution for $30-40K). Actually Areotech also manufactures their own laser interferometer as I recall.

    As far as how to get there, microstepping does a good job for low cost. Pair up an 15 tooth MXL pulley (1.2″ pitch circ) to a 1/256 microstep drive (Like a PacSci 6410) and you jump down to 23 nanometers per step. Pairing the motor with something like copley’s stepnet would enable an interferometer to provide positional feedback. Do a linear calibration to a NIST traceable standard and then interpolate between points.

    My other concern is with the conductive silver material. Getting such a material to extrude through a tip that small is not without its challenges. Most of the conductive silver coatings out there (I think) use flake from Spraylat and much of that is 30 um copper flake with electroplated silver. The smallest workable needle for our work was a 27 gage but clearly the material in the photo is a different beast.

    I guess I was just taking note of this because we did stuff with conductive silver in 2D back in 2004 (400 micron dots which is 4x larger) and 250 micron lines in 2005. We did 50 micron dots of epoxy back in 1999 or so.

    The other thing thats bothering me (as I now realize) is that the blurb talks about inkjet (which could get down to the dot sizes discussed) but a traditional extrusion type tip is shown in the photo.

    The gentleman who suggested that this could be done by hand would be in for a severe challenge. Even aided with a stereo microscope I have trouble (and more so now that I am older) aligning things to 0.005″ (127 microns) by hand. The robot is faster too with likely speeds on the order of 30 to 120 inches a minute depending on conductive silver rheology.

    At this scale the temperature of materials is crucial. All laser interferometers need an air temp sensor to obtain accurate measurements.

    Don’t mind me – I am putting off my real work!

    1. I regularly lined up dieblocks by eye within 0.0001″
      Occasionally I would line one up to where I couldn’t see any change on my indicator, I used a small Starett Last Word for lining up things in the WEDM’s,(but used a tenths Interapid for real measurements) and the indicator points had a spring loaded ratchet which I would check to see if it stuck. When that didn’t seem to be the case I would use the touch sensor on the EDM. Every once in a while (about twice a year, to tell the truth) I would get blocks lined up less than 0.001mm over a 300mm long block(near the limit of travel on my machines) so it is humanly possible.

      There is also a trick involved, the blocks I would put in the machine were ground, and the T-slots in the WireEDM machines table were also ground and very accurate, its pretty easy to look along the bottom ground edge of the steel I was cutting and line it up with the T-slot. It becomes pretty easy with some practice. New wireEDMs generally have a rail along the outside edges which should be parallel and perpendicular to the X and Y axises , but these old AGIEs I used had pedestals. Now, setting up the machine so the Z axis was actually perpendicular to the XY plane was a challenge.

      I’m with you on the failing eyesight, getting old is a PITA.

  18. A couple of 1000 to 1 harmonic drives coupled together would give you a 1,000,000 reduction, or three would give you a 1,000,000,000 to 1 reduction.
    You could drive a pretty accurate stages.

  19. I was thinking voice coil motors like we used to use on laser head assembly of the ESI Model 44 laser trimmers back in the day. We had roughly 1/10 of a mil (1/1000 of an inch) accuracy once the X-Y table stopped and settled on each piece we were trimming. I bet some math and predictive algorithms could increase that resolution

  20. My job has me aligning optical fibers to ~0.1 microns in the x/y and 1 micron in the z (don’t need to be nearly as accurate in z for what I do), by hand (ok, we use levers and such) but I can put a fiber anywhere within a 5x5mm area at least (probably a bit bigger, but I don’t want to chance bending anything). Add some stepper motors and a longer lever and that could get down to 50nm easily. (I apologize, but I can’t talk about the actual jig design since it is proprietary and designed in house, NDA and all that jazz).

  21. I would just use dip pen lithography and a custom AFM, should be good to a few nm of accuracy.

    Alternatively I could sputter coat the object and use a FIB to just cut the metal back off to line widths…

  22. I think you guys are going about this all wrong. Why print at such an insane scale when you can print at a reasonable size then shrink it down. I’m sure you know about shrink tubing. Get some of that stuff and print your antenna on it. Heat it precicely and shrink it down to the needed size. Trial and error plus a machine to measure what frequency the antenna best receives might work.

  23. Kinda hoping to see comments speculating the need for such precision. I really hadn’t studied the 1948paper to understand it well. All antennas AFAIK are, and have been “3D” so I’m not sure why that term is important, as in practical use any antenna constructed this way is going to face problems that any antenna faces.

  24. Sounds feasible,
    Currently working on using HV850 to drive piezos, as this chip can be set to alternate output +/- 60V.
    Just make sure to use output limiting as piezos can generate a lot of voltage if flexed by accident ie from a workpiece impact.

  25. Don’t need one, have one in the lab. We use it to print resonant circuits on polymers for localizing them later by RF

  26. Say, you could make tiny antennae like that.

    Could you conceivably make a Geordi LaForge Visor Or at least something like an camera for the radio spectrum?

  27. I did it with a balloon and got BETTER results. He goes all high tech and gets highlighted on Hack-a-day. Stupid society and their love affair with high tech solutions.

    For those interested I coated a BB with a thin layer of platinum silicon. Over-inflated. lightly applied dots in the pattern I wanted by hand. Then deflated to size.

    I had to add/remove/change additives to get the silicone to maintain enough surface tension to ensure a uniform thickness, yet stay super stretchy. Slightly changing pressure allowed for trimming

  28. @Steve, really? Thats pretty neat.
    Like RFID, or just passive resonant circuits..

    something like this with a 125 kHz printed antenna would make a very small RFID chip indeed, with the active semiconductor inside the sphere.
    Then print on the outer coating with the same uv cured Epoxy, etc.

    Next step, put LEDs in them and make the world’s first “active” tattoo :-) that would be a whole new class of badassery.

  29. Use a photo-mask which you get an external company to produce to make a mold. This mold will than be used to make a stamp which applies photo-resist to the sphere, the antenna is then deposited onto the sphere and the photo-resit washed off.

    For a better explanation George Whitesides paper soft lithography.

    Yay micro-fluidic module in third year mechanical engineering.

  30. My mind like there television, radio samthimes off and samthimes on. Let me companset. To day our techinology response and stimule natural body with out affect geneticaly.

    Class good by mind off and brake hack{
    Public string void main(string args[]){
    String.out.printIn(“good by mind off and brake hack!”);
    Class i myself. My world. Computant world. My universty and the place where work are secular. International rule of low{
    Public string void main(string args[]){
    String.out.printIn(“i myself. My world. Computant world. My universty and work area are secular. International rule of low;”);

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s