<.
I conclude that more constraint is needed to keep bit on-axis:
* Keep direct drive, but increase stiffness by using multiple orings, or polyurethane driver wheel.
* Separate bearings so that potential well is steeper. Keep direct drive.
* Surround the bit with metal berings, and drive bearings (or keep direct drive).
###Next steps.
* Design using smaller bearings. Three 10mm bearings fit around bit at 120Degrees in one plane.
* Why not stick the bit inside a 1/8" ID bearing and drive it directly? Can't be any worse!
*
##Week of April 8 - 15.
### Iteration b: Two bearings.
Mount two bearings (1/8"ID x 3/8"OD x 1/8" height) to hold bit, and drive in cneter with tilted o-ring driver.
The driver is now rotated by the DC brushed motor that came along with the Genmitsu CNC machine. This motor has an E14 collet, which accepts 1/8" shafts. I designed driver pulleys to fit.
* Ripple on DC input to motor is at 10x the rotation speed (ascertained by using Hall sensor).
* 7.2V corresponds to ~ 40 RPS = 18kRPM for the 1/8" bit. No load, 0.45A. Load when driving 0.8 - 1.4A. All cutting is done at this speed.
* This is convenient because, with my power supply in voltage limit mode, speed stays fairly constant as the load varies.
<img src="b-assembled-1.jpg" width="30%"/>
<img src="b-assembled-2.jpg" width="30%"/>
<img src="b-detail-1.jpg" width="30%"/>
Results:
* Very quiet, very low runout. Test pattern cuts down to 0.001" trace.
* Only works for the 1/64 bits that I have, which I measure as 0.0002" less than 0.125". Other bits, which measure -0.0001" or +0.0003" do not fit.
* Need a DIY centerless grinder.
### Iteration c: Six-bearings, drive on outside of bearings.
I printed a clamping arrangement, using 623 size bearings (3mmx10mmx4mm). Initially intended to mount 2x3 bearings, but thichness was too great to allow driving between bearings. Instead, I used only one, and put bearings on either side. I decided to drive on the ouside of the bearings with th o-ring driver. The z-constraint for the bit was an 1/8" steel ball inserted into the wood bracket.
<img src="c-assembled-1.jpg" width="40%"/>
<img src="c-d-detail-1.jpg" width="40%"/>
<img src="c-d-detail-2.jpg" width="40%"/>
Results:
* Driving on the outside is quiet.
* With no cutting load, the 'screwiness' of the setup is enough to make the bit rise to the top when driven in the cutting direction. When used to cut a trace, the bit dives into the work.
* Need a larger positive-z force if driving on the outside. If I had a way to exert such an upward force, this would be a good arrangement, since it would not require such fine adjustment of the driver wheel.
### Iteration d: Six bearings: drive above bearings.
While I had this assembly, I moved the driver up to drive above the bearings. This did not last very long. I don't recall precisely what happened, but I moved on quickly from here.
<img src="d-assembled-01.jpg" width="40%"/>
### Iteration e: 2x3 bearings: drive bit in center.
I redesigned bearing holders to allow room for a tilted driver between the bearings. Tilt is about 1/40. This required that I thin down the PLA to 6mm. Turned out not to be enough material to keep from distorting under the force needed to constrain the bit.
<img src="e-assembled-1.jpg" width="40%"/>
<img src="e-bed-cut.jpg" width="40%"/>
<img src="e-detail-1.jpg" width="40%"/>
<img src="e-detail-3.jpg" width="40%"/>
<img src="e-line-test.jpg" width="40%"/>
<img src="3-bearing-plate.png" width="40%"/>
Results:
* Nice traces, with low runout - down to 0.001" lines with small scallops <0.001".
* Cut out board, using 1/32" mill with no problem.
* Clamping force decreases - PLA is deforming.
* Began to cut wood sac layer with single flute 1/8" bit when z-stop gave way.
* Bearings tilt - need better support.
* Some bearing material (on the OD) is wearing away as harder bit abrades tilted bearing. See black stuff accumulating.
###Next steps?
* Full constraint of the bit with bearings seems to work.
* Need more material, or stiffer material (Al or Phenolic composite) to keep bearings squeezing appropriately.
* Driving on the OD of bearings is advantageous, because pressure on oring is smaller, tilt is not needed.
* Could add a tilted idler as the sixth bearing. The idler would have an oring to provide upward force through friction against the bit, in the same way that the driver did in the last iteration.
* One issue is that the little cheap conical bits have cylindrical portion only about 12 mm long.
<img src="sketch.jpg" width="70%"/>
### Iteration f: 3 bearings in two separate cages. Idler in top. Drive OD of bearings.
April 18-21. Designed and built a sandwich of bearing cages with a tilted idler in one. Drive by contacting OD of two bearings with the usual o-ring driver.
<img src="f-idler01.jpg" width="40%"/>
<img src="f-idler02.jpg" width="40%"/>
<img src="f-top_cage.jpg" width="40%"/>
<img src="f-assembled01.jpg" width="40%"/>
<img src="f-assemble02.jpg" width="40%"/>
<img src="f-assembled04.jpg" width="40%"/>
Videos below show what happens when the assembly is driven with endmill shaft sicking out the top.
First one: top only - bit goes in correct direction. Second video: asssembled. Bit moves in wrong direction.
<video width="320" height="240" controls>
<source src="f-top-spin.mp4" type="video/mp4">
</video>
<video width="320" height="240" controls>
<source src="f-assembled-spin.mp4" type="video/mp4">
</video>
###Results:
* Idler wheel works as expected when only the upper cage is driven.
* When the other bearing is addded, the overall bias drives the bit back down when turning the correct way to cut.
* This would not be too bad if the bit stayed in place. Still quite a bit of friction and inconsistent friction from the idler. The idler and driver need redisign to be more robust and precise and effectively stiffer.
* It may be difficult to drive from OD of bearings, since the bearing-bit friction is less than the bit-idler friction.
###Conclusions and possible steps....
* This design is getting complex, and has too many parts.
* Could return to design a, keep it simple, and use other means to increase the stability.
* Better capture geometry.
* Redesign with smaller-width o-ring, aluminum driver.
* This design does not work for providing net upward force.
* Could work on greater precision for bearing holders, using aluminum or Bakelite.
Another thought: Can I grind off a few .0001" in a controlled way to use bits with 0.125" ID bearings? (My dial micrometer has ~0.0001 precision.) Answer: maybe. Below, I'm using a wet sharpening wheel to abrade the diameter of two cylinders. In the second picture, the 1/8" pin has had 0.0002-0.001 removed in a not-so-nice way. It now fits into bearings. The 1/8" (0.1250 on my dial micrometer) endmill on the right is harder - after 15 minutes of grinding, it was about 0.0001" smaller, and almost fit into the 1/8" bearing. Might be fun to continue on this track, just to have a lab-made spindle that works.
<img src="centerless.jpg" width="40%"/>
<img src="ground-parts.jpg" width="40%"/>
### Iteration g. Bearing cage (3x2) with built-in axial bias. Drive on OD of bearings.
April 21 conversation with Jake. He suggests that rather than use a rubber idler,let the six constraining bearings provide the axial bias. I learned above that small distortions in the bearing arrangements lead to this kind of screwiness. I can do this on purpose, and drive from the outside.
I tried this, modifying the top cage of the previous iteration so that top holes are displaced clockwise (looking down) by about 0.5 mm. This gives a noticable distortion. Pictured below.
<img src="g-screwy-top.jpg" width="40%"/>
And a video:
</video>
<video width="640" height="480" controls>
<source src="g-screwy-top.mp4" type="video/mp4">
</video>
Looks like this works. One way to effect this in a simpler design is to use three screws to support all six bearings, and give a twist to the supports. Will need to figure out how to insert the bit!
<img src="g-sketch.jpg" width="40%"/>
###April 28
* made a new top with holes displaced by 0.25 mm. Appears to be lots of upwards force - using 1/8 in ball as stop, requires ~2A to run. Pushes insert out of top part. Need to make less angled part.
* Recreated the top bearing holder with 0.1mm rotation. Now, the movement is about 1mm per 10 revolutions.
* Still using ball for thrust bearing. Sac layer cutting OK, but bit got loose. Reinserted, tightened, and cut traces. Bit rises as cuts progress. Where's it going?
* On third try of test pattern, bit stayed down, and traces ran OK. At end, everything was loose. Current down to 0.5A, and driver was slipping when tested by hand. Phone app dB meter says 55 dB at 0.3 m distance.
* Tighten up everything, and run traces for a stepper driver. Still a problem with z stability. Now, the cutting depth gets larger by ~0.5 mm, making depp trenches. Time to redesign, and perhaps use a thrust bearing.
### May 05 updates. Iteration h.
#### Making a new six-bearing assembly
Made a new six-bearing stack, driven by the same o-ring driver on the OD of the bearings. To create the positive z force, the bearing assembly is given a twist by inserting spacers next to the tabs on the bearing support discs. Results:
* Bottom symmetrically inserted 0.010", Top has 0.020" on left: rises 0.08mm per rev. (angle 0.008)
* Bottom symmetrically inserted 0.010", Top has 0.015" on left: rises 0.032 mm per rev.
* With the top configuration, It seemed that the bit travelled upwards as the cap distorted. The lower angle gives stable results. I'll know more from cutting with 1/8 bit later.
<img src="h-sketch.jpg" width="40%"/>
Below are the parts. The cap holds a 3.125 mm steel ball as the thrust bearing. That cap can be modified to hold a lower friction manufacuted thrust bearing if needed.
<img src="h-screenshot01.png" width="40%"/>
<img src="h-screenshot02.png" width="40%"/>
<img src="h-screenshot03.png" width="40%"/>
<img src="h-assembled.jpg" width="40%"/>
#### Clamp is loosened to allow insertion of tool, and tightened to clamp tool in bearings.
[link to insertion video](https://youtu.be/ZU1q-gGhicY)
[link to step file](./Clamp_v9.step)
[link to Fusion 360 file](./Clamp_v9.f3d)
####Below are test patterns. Runout on smallest traces is not visible.
<img src="h-test_pattern.jpg" width="40%"/>
<img src="h-traces.jpg" width="40%"/>
#### Cam actuator for z motion.
Built prototype of a cam actuator for z-motion. Stepper motor is attached to a large bearing with an offset cam. Motion goes as r*sin(angle) where r is the offset distance, in this case 1 mm. With 200 steps per rev, this gives 0.01 mm resolution (more or less) with single-stepping.
Next step is to build a simple flexing support that allows +/- 1 mm travel in z.Can run this motor with Clank, and for testing, incorporate into the Clank stages to give x-y motion. Clank's z motor can give a coarse z-adjustment.
Then, all I need is sufficient x-y travel to make a PCB machine. This could be a flex-stage ala Brussels Fablab's Urumbu.
[link to cam video](https://youtu.be/0QTuHu_hz7Y)
<img src="zcam01.jpg" width="40%"/>
<img src="zcam02.jpg" width="40%"/>
### May 06, 2021. Cutting a trace with the "cam" actuated z-axis.
Below is a sketch and picture of the assembled spindle with z-axis actuator.
<img src="h-cam-sketch.jpg" width="40%"/>
<img src="h-cam-assembled.jpg" width="40%"/>
Traces cut with this spindle arrangement. To do this, I attached clank z-motor leads to stepper on cam rotor. Using the same firmware, I found a setting in which depth of "0.05 mm" and jog height of "0.5 mm" worked well enough when I used mods to generate the usual nc file. Corners good, vibration in the x-y plane is not noticable. Seems to be some problem with consistency in depth.
<img src="h-cam-traces.jpg" width="40%"/>
[link to video of trace cutting](https://youtu.be/yPuyHpz9I1c)
###Driving. Stepper motors.
[specs for Nema 11 motor.](./nema11.pdf) Double-step, looks like 6000 pps possible. This is ~1800 RPM.
[specs for Nema 8 motor.] (./nema8.pdf) Double-step, looks like 10k pulses possible. (That means 10k full steps per second??). For 20k RPM endmill, need about 2500 RPM, or 8k pulses per second.
May 11: Ran Nema 8 motor at down to 60 microseconds per step. Maybe run at half that speed. Torque may be adequate? Run at 24V with current limit at 0.85A. Worth a try.
]]>