Using the "industrial" sharpies, fine and ultrafine point, I made two attempts at an AVR RFID card with integrated PCB antenna, with mixed results.
I designed all but the antenna in Eagle and exported to EPS, then imported into Make-the-Cut. The spiral antenna I created using Inkscape spiral tool, then cut-and-paste into Make-the-cut, and just rotate and move so the traces connect.
I draw a rectangular bounding box around the circuit and put into a separate layer, which I print first onto a piece of paper taped on the cutting mat. This shows me exactly where to place the blank circuit board.
Next I tape the circuit board to the paper using masking tape at the very edges only. I snipped off one of the grey friction rollers so it doesn't roll back and forth over the design and ruin it while the ink is still wet.
I remove the Expression keyboard to make room for the pen, and wrap a long thin strip of duct tape around the pen close to the tip to make it big enough for the tool holder to grip. I "cut" (draw) from Make-the-Cut at Extreme speed, low pressure, and etch in the usual way.
The experiment was very successful in that I was able to get pretty reliable traces in two sizes, roughly 0.5 and 0.8 mm, plenty for a sporting shot at surface mount stuff.
Unfortunately, while a magnet-wire coil produces upwards of 10v peak-to-peak, I can't seem to get better than 850mV or so across the pcb coil leads, regardless of my choice of tuning capacitor, so for now running the Tiny85 is still out of the question.
The two antennas have about 25 and 45 turns respectively, but so far I've been totally unable to get them tuned to produce enough voltage to run the AVR.
I've been probing the tag coil across the capacitor leads while the tag is being stimulated at 125khz with my homemade reader.
The second antenna is pushing the resolution limits in the spiral because the "jaggies" from the machine reduce clearance between lines. I had to do a little cleanup work with a razor blade, separating adjacent lines that had shorted in a very few places.
I spent a few hours reading about 125k antenna formulas and designs to try and better understand the relationship between desired frequency, loops of wire, wire gauge, and the value of the tuning capacitor.
I worked through some of the formulas in this datasheet, plugging in 100 turns, 66mm, 30 gauge to see what it would come up with for inductance, ideal # of turns, and capacitor value, to see if they would match the 1nf & 100 turn values specified in scanlime's blog entry. I wrote a simple java program to calculate the values for me after not finding what I thought I was looking for on various online antenna calculators (they seemed to be all too complicated).
I found this interesting note along the way which I had not seen elsewhere: "... For copper wire, the loss is approximated by the DC resistance of the coil, if the wire radius is greater than cm. At 125 kHz, the critical radius is 0.019 cm. This is equivalent to #26 gauge wire. Therefore, for minimal loss, wire gauge numbers of greater than #26 should be avoided if coil Q is to be maximized....".
So, #26 gauge wire next time!
Unfortunately my program isn't putting out believeable numbers yet, so I have more work to do.
In the meantime, I also discovered this other datasheet for a neat little 0.32 cent RFID chip (the Philips Hi Tag-S) that has a good description of the details of the encoding methods used starting in Section 7.3 (including our manchester, for future reference)
And here's another good discussion of the antenna design and coupling (different frequencies, same concepts and formulas).
I played with the open source 4NEC2 Antenna Modeler & Analyzer for a while trying the helical generator, but it kept blowing up at simulation time with my RFID design, complaining that the antenna was connected to ground at the end (duh, this is what the tutorial shows, I don't get it!).
While further developing my printed-circuit-board method, I inadvertently figured out how to do filled shapes with the pen using the free version of "Eagle" PCB design software.
1) You'll need to have Eagle (free version is fine), Inkscape, and Make-the-Cut all installed and available. I use Gimp for editing and converting file formats.
2) Open the desired image in Gimp (or other image editor) and save it out as a BMP format image, with as few colors as possible. A 1-bit monochrome with no dithering is best, like line art. (you set this via Image->Mode->Indexed... in Gimp)
3) Start up Eagle, and do a New->Board.
4) Do File->Run, and run the script "import-bmp.ulp". This will be located in the ulp subfolder of your Eagle installation directory. It will prompt you to browse to your BMP file and open it. Keep your bitmap size modest, 300 DPI or less, more just slows everything down. If you have a black-on-white drawing, select White only and click OK. Click the DPI radio button under Format, and set your DPI value; it will set the Scale factor for you automatically. Click OK.
5. It will think and popup window saying "Accept Script?", click "Run script".
6. Your bitmap should now appear in blue in the Eagle Board window. Save your project, then Run File->CAM Processor.
7. Select Output Device HPGL. Deselect all layers except #200 "200bmp". You'll have to scroll down to find it. Click "File" and browse to your desired output file location and name. Be sure to give the file a ".plt" extension which indicates the HPGL file type. Set the Pen Diameter to the actual width of the line your pen produces. 1 mil = 1/1000 inch. A ballpoint pen tip ranges from about 20-50 mils. Click "Process Job". This will render your .plt file.
8. Start up Inkscape, do File->Import... and import your .plt file. This can take a little time. If the application freezes, give it some time.
9. Eventually, your image will appear in Inkscape, already selected. Click "Edit->Copy".
10. Switch to Make-the-Cut, now do "Paste-in-Place". Voila!
Note, depending on how large the filled areas are, and the diameter of the pen, it might take quite a while to calculate when you click the "Cut Project With..." button. (the example image has 65000 points). Fortunately, since the fills are done using mostly horizontal and vertical lines, the machine can run pretty fast.
Here are another couple screenshots to show the pattern closeup. If you wanted a grid you would tell Eagle that the pen is wider than actual.
This method should work really well with engraving tips..
MTC reported this image was about 11000 cutting points which really seems pretty modest to me. I haven't tried actually printing one this big yet. If you (in MTC), right click on the imported image, and do Shape Magic->Advanced->View Path Detail... and scroll down through the path segments with the cursor key, you'll see the path taken is pretty sensible and doesn't waste a lot of time scrolling wildly all over the place.
I created a circuit in Eagle to test the resolution and fills using a fine tip pen in the Cricut.
The pen itself is a 0.01 inch diameter permanent marker, shortened to fit the machine. The test image is made of three identical copies of a diagram. The diagram contains some sample SMD and through-hole pads, as well as traces in 0.01, 0.03, 0.05, and 0.07 inch widths.
The legend reads "10mil pen, 20mil pen" and "30 mil pen" to indicate the pen size as given to the Eagle CAM processor function, but the entire drawing is rendered using the 10 mil pen in order to see which circuit will have proper fills with no white lines inside.
The results prove to me that the best fills do indeed result when the pen size is accurately set in the Eagle CAM processor. When the pen in smaller than Eagle thinks, it doesn't lay down enough lines to completely fill the solid traces. The TQFP and SMD resistor pads look nearly usable, even the horizontal jitter is almost nonexistent near the middle of the page. (Maybe the duct tape pen mount absorbs vibration better than my custom metal pen holder?) The worst flaw just seems to be incomplete coverage in the diagonals of the largest (0.07 inch) traces, which should be easily retouchable by hand.
Here's the workflow:
1) Create the board in Eagle
2) Run the CAM processor, select output device HPGL, layers Vias, Pads, Top or Bottom, and Holes.
3) Start a new drawing in Inkscape. Import the HPGL file into Inkscape. Select the imported circuit diagram, and click "Copy".
4) Mount the pen in the Cricut tool holder.
5) Start Make-the-Cut, and do "Paste in Place" (ctrl-shift-V) to paste without resizing. Arrange as needed and then cut to the machine.
Last night at the Hackerspace I tried a new way (to me) of doing PCBs on the Cricut. So far it looks like the most promising yet.
This time instead of trying to scratch off an etch resist, I'm directly drawing it on using a plain old mini Sharpie pen like a plotter would. I understand the Stadtler Lumicolor pen is also recommended.
I just now discovered this link that shows exactly what I need to try next as far as the pen goes: PCB Plotting
-Then, in Eagle, run File->CAM Processor.
-Select Output Device EPS
-Click File and select your output file path.
-Don't worry about the offset and page size.
-Select the "Pads", "Bottom" or "Top", and "Vias" (it will complain if "Vias" is not selected).
-Click "Mirror" if you are doing the bottom layer.
-Click "Process Job," this will write the output file.
-Install a copy of Ghostscript & GSView, and run "ps2pdf [options] input.[e]ps output.pdf" to convert the EPS file from Eagle to a vector PDF.
-Fire up Make-the-Cut, and do "File->Import->Vector PDF File", leave "Import Strokes and Fills" selected, select your PDF file and click "Open".
-Select the imported image and click "Ctrl-B" to Break the circuit up into its pieces.
-Deselect all, then click on each of the four border lines and delete.
-Select all, and click "Ctrl-J' to Join the circuit back up into a single piece.
-Position the circuit on the cutting pad as needed.
-Load up your Sharpie in the tool holder and print a test piece on paper to verify positioning.
-Load up your copper in the machine. If you're running anything thicker than 0.01 you may need to raise the pen in the holder. I use double-stick tape or at least a fresh spritz of spray-tack.
-Print your design. I don't know how many coats are necessary, but I am doing two coats, one after the other. Don't use Multicut for this!!! It does each line multiple times immediately instead of doing the whole pattern completely and repeating it: This causes the pen to dissolve the previous coat and move it around a little.
-(optional) Put the board into the toaster oven just briefly to make sure the ink is fully dry.
I tested a few different etch-resists over the weekend with good luck, but the process itself still needs improvement before the boards will be usable.
The three resists I tried were Johnson "One-Step No Buff" floor wax, Krylon purple spray paint, and black Lacquer spray paint.
I made three test pieces of PCB material, one using each coating, then scratched each one manually with a nail, and etched. The wax was applied by pouring a thin stream over the top of the metal and propping it up at an angle on a paper towel to let the excess run off.
After etching, the floor wax was a clear failure, the coating was just too thin and failed over about half the surface. The purple Krylon seems like it worked well, as did the black Lacquer. Clear paint looks awesome but its hard to tell when you've sprayed enough on.
Also, one thicker coat seems to work better than two thin coats like you might normally use.
I still have a couple of problems to solve before the process will be more useful however: I need a wider scratch mark, and there's an annoying jitter in the tool during the first quarter or third inch of motion on the X-axis.
After making the test pieces, I scratched a full-size test piece using the matte black Lacquer.
Edit: Unfortunately, the etched scratches still do not completely separate the areas of copper, resulting in 100% shortage across the board surface.
Clearly I need to find a better resist and tool usage combination. The fine parallel lines still resolve clearly, but are also not fully clean. They are also not close enough to merge when etched. I think a softer resist might be desirable, so now I'm thinking .. what about melting a very thin layer of candle wax onto the surface with a hair dryer and then scratch wax but not metal with something pointy but not sharp, like a tooth pick?
Also, much hotter etchant may work better as well (this most recent board was done at room temperature and appears to be incompletely etched before the resist started to fade. Previously I used a double ziplock bag with a few TSP of etchant in a hot water bath with good results.
I've created this writeup on my current Cricut PCB process, and posted to the hackerspace Wiki.
I make PCBs on the Cricut by the following process:
1. Design a one-sided board in Eagle. Export the trace and pad layers only to a monochrome PNG file. 150 dpi seems to work well so far.
2. Using GIMP, add a fine outline around all traces.
- select the entire background color (black) with the "Select by Color Tool". - Do a "Select->Shrink" and shrink the selection by 1 pixel. - Fill the entire selection with the trace color (white) using the Bucket Fill tool. Make sure you have "Fill whole selection" set. - "Select->Shrink" by 1 pixel again. - Fill the entire selection with the background color (black). - Finally, save out your image as a PNG file.
3. Import the GIMP image into whatever software you are using to drive the Cricut (such as Make-the-Cut). In Make-the-Cut, I use the following settings for the highest accuracy tracing: Threshold 180, Resample x5, Smoothing 0, Optimize 0. If you use a PNG image it should import to the correct size. I have had trouble with other image formats becoming larger or smaller when importing.
4. I highly recommend using a fine-point pen tool to draw a preview test of your design on paper to check for sizing and other issues before potentially ruining expensive copper.
5. Prepare your PCB material. I am currently using super thin 0.015 single sided FR4, but the machine should be able to handle thicker materials, as long as your tool holder and tool can still clear the work piece. For an etch resist, I put on a single thick coating of spray paint as my etch resist, but plan to try Future floor wax next.
Information about using Future floor wax as etch resist.
5. Load some kind of "scratching" tool in the Cricut. My current preference is a deck screw or a sharpened nail. (The scribing tool is a bit too fine and can result in bridging). You can use a custom tool holder or just masking or packing or other "hard" tape. Pay careful attention and set the installed height of the tool high enough to make sure it won't drag across your work piece while "up".
6. Finally, use the Cricut to scratch your design into the prepared PCB, and etch. I have been experimenting with different settings for pressure and multicut, but currently I like "high" pressure, multicut x2.
One of the first things it seems like every interested hacker asks me about the Cricut is, "how can we get it to do PCBs?" Well, I've finally done it!
I created a single-sided circuit design in Eagle, using 50 mil traces, then exported just the pads and traces as a monochrome PNG file.
Next, I sprayed a coat of clear spraypaint on some single-sided 0.01" PCB material and let it dry well. (Next time I will use two coats)
Finally, I imported my design into MTC using pixel trace and used my scribing tool to scratch the design onto the PCB (multicut 2, pressure high). This removed the spray paint around the edges of my traces. After it was done, I used a toothbrush to brush the removed-paint-bits off the board.
Finally, I etched with a few tablespoons of ferric chloride in a double-bagged-ziplock immersed in hot water. The whole etching process took less than ten minutes. Aside from some unwanted specks where my one coat of spray paint was a bit light, the result looks perfect!
Obviously I still have to do more work to make this board functional, I'll report back when I have more to show.
EDIT 3/13: I tried drilling the board and soldering up one of the oscillators, and it turned out that there were a couple of shorts in my etched diagonal lines where some "jaggies" were really close together. Also I discovered that the "donuts" need to be larger in diameter to leave more solder pad remaining after drilling the hole, especially the ones for the 555 chip.