Here is a link to a site where someone has already solved the problems involved with taking a stock vinyl drag knife and mounting into swivel bearings and holder. His trial and error process is well documented which brought him to his final solution. Clicking on any picture will open a larger view of the picture.http://www.cuttingedgecnc.com/vinyl.htm
My solution is a variation that adds a compression spring.
To make your own DIY Swivel Drag Knife Cutter all you need are:
1. one .25" x .125" rare earth magnet from Amazon.com
2. one 5mm OD, 2mm ID bearing from http://www.bocabearings.com
3. one 5mm OD, 1.5mm ID bearing from http://www.bocabearings.com
4. one 3/8" x 12" brass tube from Hobby Town USA or your local hobby store
5. three 3/16" wheel collars (Dubro or Great Planes collars) from Hobby Town USA or local hobby store
6. one spring that slides easily down the 3/8" brass tube, has an ID > .25" and a compression force of 80-150 grams when compressed 3/16"
7. Roland replacement blades 45 or 60 degree from Ebay.com
8. A copper pipe cutter such as a Rigid Model 104.
9. A tiny amount of thin CA (Cyanoacrolate also known as Super Glue)
10. A tiny amount of Loctite 609
11. Digital calipers (very helpful)
Here is what a package of Carl Goldberg wheel collars looks like.
Here are the collars with bearings pressed in, the magnet and the blade.
Here is a drawing of these parts:
Here are what the key parts look like, laid out as they will be located in the 3/8" brass tube
Here is the assembled result with a shot the pipe cutter I used for crimping.How it works
The bearings provide low friction so the Roland blade can swivel (or caster if you like) keeping the cutting edge of the blade always facing the direction of travel. The rare earth magnet is what keeps the blade sucked up into the top bearing. The magnet is all that holds the knife in place. Only about half of the cone tip extends into the 1.5mm ID bearing's inner race.
It turns out that the 3/16" model airplane wheel collars have a 5mm (.1970") ID exactly. I did have to deburr the inside of the collar with a file where the Allen set screw threads flared towards the inside wall. After doing this the bearings pressed in perfectly. I did not use the set screws that came with the collars. The outside diameter of the wheel collar is .3430". Just the right size to slip into the 3/8" brass tube.
The middle wheel collar which contains the 1.5mm ID bearing maintains a fixed separation from the bottom collar by a crimp in the brass tube. It can, however, slide upward against the spring. Between the middle and top collar is the compression spring. The top wheel collar's fixed position is determined by the tube crimp above it. The location of the crimp is such that the top wheel collar slightly compresses the spring against the middle collar by about 1/16". When upward force is applied to the tip of the blade it pushes against the upper bearing which is locked in the middle collar causing the middle collar to compress the spring. The compressed spring exerts a downward force on the blade by an amount determined by the spring's coefficient.
Most vinyl or stencil material requires about 85-150 grams of downward force on the blade to ensure only the vinyl is cut and not through the paper backing.
For me the hard part was finding the right compression spring with a diameter small enough to slip into the 3/8" brass tube but with an inside diameter slightly larger than 1/4", the diameter of the magnet. It had to have a compression force of 80-150 grams when compressed between 1/16" to 3/16". I found the "perfect" spring in a cheap, old laser pointer that pressed three L44 button batteries together in the battery holder. (I forgot to take pictures of it before I completely assembled the new solution, sorry about that, so the spring shown above is only to represent the idea of where it goes within the assembly of parts. The drawing shows how it really should be. My "perfect" spring's length was .3" uncompressed and about .3" in diameter. Assembly Sequence for Spring Loaded Solution
I started by cutting a clean end with the pipe cutter so the tube would have a slight crimp on that end. This crimp and another just above the first collar will keep the first collar locked in place. The 2mm ID bearing is pressed flush to one end of this wheel collar. Mine pressed in with a snug fit so I did not use Loctite on it. I dropped this wheel collar, bearing side down into the brass tube from the uncrimped, "virgin" end, then crimped the collar into place about .2160" from the end. If the wheel collar was not locked in place snugly I crimped again closer to the top of the wheel collar until it was stationary.
I made a second crimp .5" from the end. This sets the lower limit of the second wheel collar that contains the 1.5mm bearing. The bearing in the second collar is pressed halfway in. I painted around the exposed outer edge of the bearing with a straight pen when then pressed it about 1/16" passed flush so the magnet would be attached to only the collar. The Loctite was allowed to set. The magnet is then centered on the bearing end of the collar and CA'd to the collar using same method of painting a tiny drop of the CA around the outer edge of the magnet with a straight pin where it meets the collar. The CA was allowed to set . Use Loctite and CA sparingly because you don't want this stuff to get in the ball bearings. The second collar is dropped down the tube bearing & magnet side up followed by the spring. Note: The second wheel collar must move freely, with low friction inside the tube. Then the third collar is dropped down the tube so it sits on the spring.
The third crimp point is "approximately" 1.232" from the first collar end of the tube and just above the top of the third collar. This actual crimp point will depend on the spring' length and how far down the third collar has to moved to compress the spring about 1/16". To determine this crimp point, I slide the small tube down the "virgin" end just where it sat on the third collar. I marked a reference line on the smaller tube where it exits the larger tube. I then put another mark on the small tube 1/16" higher and checked to make sure than when I pushed the small tube down to the second mark the spring was being compressed all through the 1/16" of travel. I pulled out the small tube, laid it next to the 3/8" inch tube, lined up the second mark to the "virgin" end of the big tube, then marked on the big tube where the opposite end of the small tube reached. This determines the last crimp location. Using the small tube to push on the third collar and compress the spring all the way, I then made the last crimp as I held the spring in compression.
I used my standard Dremel mounts by making some compression reduction clamps sized to hold the brass tubing. Here are the results.
My first stencil cut was done using my first prototype which did not employ a spring or the third collar. I simply used the accuracy of my "Target Depth" plunge of .003" to pierce the vinyl. For this test I used Top Flite Monocote Self-Adhesive Trim. Total thickness is .010" while the backing is .0075" thick. My blade solution at that time did not incorporate a compression spring but did quite well as you can see below. Work surface flatness must be within .001" in this case.
During this test is when I learned my once flat work surface now had low spots. That's when I decided to use the compression spring since it helps to compensate for minor variations in flatness. The amount of compression force acting down on the blade is adjusted by varying how far Z-axis travel compresses the spring.
Here is the result of cutting thick, Duck brand adhesive shelf vinyl from Lowe's. I used the 45 deg. Roland blade for this with about 170 grams of down force, although other text cuts it worked equally well with about 125 grams of down force. The font is Arial and 1" in height. Total thickness of material is .0125", .0085" is the vinyl and .004" is the backing material. I used a feed-rate of 10 IPM. The most important thing of note here is that the formica work surface is not truly flat with plus or minus .01" variations. This is due to the variance in thickness of the two pieces of 3/4" particle board glued together and covered with formica (it is a piece from an old table top). So, the spring loaded drag knife blade made "flatness" not so critical.
Looking closely at the letters you can see some, not all, of the corners are slightly rounded. This is the result of not applying the .25mm offset of the blade (tool). I've learned that under-compensating for the blade offset results in corner rounding and over-compensating results in corners with "whipped cream" type of peaks.
Ain't this fun!