I'm going to log my CNC milling attempts and results in here, so I can look back and get a sense of what techniques work the best. Hopefully others will be able to use this info, as well.
Lamp inserts, attempt 1:
Material: 1/8" acrylic
engraving pass:
- 90deg V-bit (sharp)
- 3.6IPM, spindle speed "1"
- 1/16" passes
- result - good, not much fuzzing at all
cutout pass:
- 1/8" spiral bit (sharp)
- 3.6IPM, spindle speed "1"
- 1/16" passes
- result - some tool marking, some melting on the final pass
Lamp inserts, attempt 2:
Material: 1/8" acrylic
engraving pass:
- 90deg V-bit (sharp)
- 9IPM, spindle speed "1"
- 1/16" passes
- result - good, not much fuzzing at all (same as above)
cutout pass:
- 1/8" spiral bit (sharp)
- 12IPM, spindle speed "1"
- 1/16" passes
- result - reduced tool marking, no melting
CNC notes
-
JamesCooper
- Posts: 164
- Joined: Tue Jan 11, 2011 11:46 pm
Re: CNC notes
Tonight, I was milling down solid birch planks. I had 4 planks of 1x5" (actually 3/4 x 4-1/2"). 3 were approximately 48" long, the fourth wasshorter (about 42"). I needed them planed down to 3/8" thickness. Since I couldn't get access to a planer, I attempted to mill them down to size.
I mounted each of them to the table by drilling holes through the ends (9/32" I think) and putting T-bolts up through the holes with washers and wing nuts on top. On our shorter T-bolts (the ones with the solid T end), the bolt extended just up to the top of the wing nuts, which was perfect. With sufficient tightening, I couldn't slide the piece at all by hand.
I purchased a new Freud 75-102 (spiral up cut, 2 flute, 1/4" dia. 1/4" shank, 1" length). It was mounted in our Bosch Colt 1HP palm router on Goliath.
I tried to calculate my ideal speeds using G-Wizard. I had trouble meeting the ideal chipload (0.015 inch-per-revolution according to G-Wizard and validated by manufacturer's suggestions for similar endmills). In almost all cases, I either needed a feedrate of over 400 feet-per-minute, or a spindle speed below 16K RPM. I also had trouble finding pass depths and cut widths that kept deflection below 0.001" and load HP within spec (it kept picking something that resulted in 1.5 HP despite setting a 0.7HP limit). These are the settings I eventually settled on:
The toolpath was generated with Aspire. I created a board of my available length (actually 45" on my 48" boards) and the full width (4.5") and applied a similarly sized rectangle on it. I actually made the rectangle overhang by 1/4" on both sides of the board to make sure it would cut through and not leave corners. I then made a pocket toolpath for this 3/8" deep, using offset mode and conventional cut direction and using the numbers indicated by G-Wizard for cut depth, step over/cut width, feed rate, and plunge rate. I also told it to ramp in the plunge over 5 inches.
I cut this pattern on 3 of the 4 boards (all of the 48" boards). The first I did according to the numbers above. I set the router to a speed of 3, which according to the manual, should be approximately 20k RPM. I noticed that the plunge went quite well, but the initial slot cut was a heavy load and bogged down the router. I slowed it down to about 50% of the feed rate and it coped. After the slot was complete and it started offsetting, the load was much lighter. I gently brought it up to 100% feed rate and it was running well. The motor RPM sounded approximately the same as it did without a load, despite it clearing a lot of wood very quickly. As it rounded a corner, it would slow considerably, but would recover as it went down the straight again. I tried increasing the feed rate to 130% and the RPM didn't appear to change, but it was cutting faster. I did read that the router has a control circuit to maintain RPM under load and it appears that this is doing what it claims, as the RPM didn't drop noticeably as I changed feed rates. On the second pass, the plunge and slot didn't seem to cause quite as much slowdown as it did on the first pass.
As the run was going, I also noticed considerable vibration in the sled holding the router. The path left by the bit also had waves in it in the Y direction. I also noticed that, during the long run down the length of the board, the wood tended to splinter and peel off in chunks, rather than chip. This appeared to be due to cutting with the grain of the wood.
After the run, there were noticeable marks left behind. These were most pronounced in the corners (where it left circular marks) and on the cross-grain cut on the far end (where it left a trail of wood standing up between each offset pass). For some reason, it wasn't as rough in the end towards the operator. In both cases, these cross-grain cuts had a slight lean towards the operator-end of the wood; each pass seemed to cut deeper on that end of the wood and shallower on the far end. Perhaps the Z sled is not quite perpendicular to the table.
On the second pass, I tried playing with the spindle speed and feed rate. I increased
On the second run, I decided to try playing with the spindle speed and feed rate. I started by turning the spindle speed up to about 4.75 and keeping the feed rate at 300 IPM. This caused noticeably less splintering of the wood and more chipping. At this point it was making very nice chips. I also tried increasing the feed rate. I was easily able to raise it to 420 FPM, but I didn't try going further since I wasn't sure what Goliath could handle without losing steps. The increased spindle speed seemed to make the motor slowdown less during the corners. It also appeared that raising feed rate required raising spindle speed to maintain the same level of chipping. Neither seemed to affect the amount of waving on the side of the cut or tool marks on the finished surface.
Between the second and third run, Ryan investigated the chatter that I was seeing. We noticed that several things were loose. The X rails had worn and required tightening. The arm that attaches the lead nut to the Z sled was very loose, so we tightened it. We also realigned the two X screws. Ryan tightened several other things, including a collar on the back side X screw, a collar on the Y screw, and the Y rails.
The third run didn't show any remarkable improvement. Waviness persisted and the tool marks were similar. All we could discern was that the tool appeared to be skipping up and down, causing the Z sled to rotate about the Y axis.
The final board was shorter, so I had to make a new toolpath. Since we noticed that the skipping happened during heavy load in the X direction, I decided to try a pattern that primarily used the Y direction. I decided to use a raster in the Y, but again oversized the pocket so that the turn around and step over that occurs in the X direction would be off of the piece to avoid tool marks. It would also plunge off the table, avoiding that heavy load. Since it would be alternating directions on each pass, we could also see the difference between conventional and climb cutting. For whatever reason, the toolpath was generated with the plunge in the opposite corner from home (in the far, back corner).
The raster cut was considerably different. Again, I started with a feed rate of 300 FPM and spindle speed at 4.75 (possibly around 25k RPM). The initial slot still bogged down, but it was very short, since it only had to pass the 4.5" width of the board. I eventually raised the feed rate to 140%, or 420 FPM. I did notice that it didn't maintain the feed rate well, because it was constantly accelerating and decelerating. Because of this, it took noticeably longer to run. I also noticed a considerable difference in behavior between the conventional cut pass and the climb cut pass. During conventional cut, it threw very nice chips perpendicular to the cut and left a smooth finish on the surface. However, it did tend to splinter the side of the board as the cutter left the side. This splinter grew the lines rastered down the board. On the climb cut movement, there was a very different behavior. In that case, it sprayed a fairly fine powder of saw dust in the opposite direction of the cut. It also left raised wood marks on the surface. The spindle also seemed to bog down more during the climb cut than during the conventional cut. All in all, the conventional cut seemed much better.
On the second pass, something very strange happened. I realized it was going to have to slot, so I hit reset on the feed rate override as it was rapiding back to the plunge for the second pass. For some reason, it then decided that a little over 1/2 way to the back was sufficient and it set down and started rastering back. I had to stop the feed as it approached the front of the piece. This left me with a half-finished piece, but it also gave me an idea for an experiment.
Since I noticed that the conventional pass was quite clean and the climb pass was very rough, I decided to try manually rastering always in one direction to get a constant conventional cut. My pattern was:
As an aside, I did have a problem when I first went to start this. When I rapided back to the unfinished part of the wood, the back X axis stalled and racked the table. I think this may have been due to a considerable buildup of fine sawdust during those climb cut passes, which were spraying directly at the wheels on that end of the gantry. After cleaning it thoroughly and realigning the X axes, it worked properly again.
To get started with my manual pass, I went to a spot just beside the wood near the start of the unfinished section and zeroed X and Y there. I then brought the bit down to the top of the unfinished section, but rather than zero, I clicked on the Z value and manually entered +0.1875 (3/16" above zero). I brought the tool off the piece again and hit Go To Zero, which caused it to line up off the piece, near the unfinished section, at precisely my desired cut depth. From hear, I started manually rastering using the jog keys.
The results were impressive. There were still ridges between each pass, but the roughed up wood on the surface wasn't present -- this was definitely an artifact of the climb direction. In fact, a couple of times I forgot to step back and just returned on the same path that I had cut and this caused the same rough surface issue that the climb cut had made. Perhaps stepping back and running over it in the other direction was responsible for the much smoother final surface. This surface really showed off the fact that the bit wasn't completely flat on the surface, but rather was cutting lower on the operator-side of the table on every pass of the tool. Still, the surface quality was considerably better than the bidirectional raster. Also, it made almost entirely chips and very little sawdust.
For both styles of rastering, long splinters weren't a problem anymore. It seems that cutting along the grain of the wood is what was causing it to split and splinter. Cutting across the grain resulted in much nicer, more consistent chip sizes.
Tomorrow I will post pictures and video, as well as a summary of what I learned during this.
I mounted each of them to the table by drilling holes through the ends (9/32" I think) and putting T-bolts up through the holes with washers and wing nuts on top. On our shorter T-bolts (the ones with the solid T end), the bolt extended just up to the top of the wing nuts, which was perfect. With sufficient tightening, I couldn't slide the piece at all by hand.
I purchased a new Freud 75-102 (spiral up cut, 2 flute, 1/4" dia. 1/4" shank, 1" length). It was mounted in our Bosch Colt 1HP palm router on Goliath.
I tried to calculate my ideal speeds using G-Wizard. I had trouble meeting the ideal chipload (0.015 inch-per-revolution according to G-Wizard and validated by manufacturer's suggestions for similar endmills). In almost all cases, I either needed a feedrate of over 400 feet-per-minute, or a spindle speed below 16K RPM. I also had trouble finding pass depths and cut widths that kept deflection below 0.001" and load HP within spec (it kept picking something that resulted in 1.5 HP despite setting a 0.7HP limit). These are the settings I eventually settled on:
- Cut depth: 0.875" (3/16")
- Cut width: 0.0938" (3/32")
- RPM: 19597
- Feedrate: 300 FPM (my set limit)
- Surface speed: 1283 SFM
- Chipload: 0.0073 IPT
- IPR: 0.0153
- AFPT (with chip-thinning): 0.0088
- MRR: 5.2763 ci
- HP: 0.7914
- Deflection: 0.0006"
- Plunge IPM: 166.5
The toolpath was generated with Aspire. I created a board of my available length (actually 45" on my 48" boards) and the full width (4.5") and applied a similarly sized rectangle on it. I actually made the rectangle overhang by 1/4" on both sides of the board to make sure it would cut through and not leave corners. I then made a pocket toolpath for this 3/8" deep, using offset mode and conventional cut direction and using the numbers indicated by G-Wizard for cut depth, step over/cut width, feed rate, and plunge rate. I also told it to ramp in the plunge over 5 inches.
I cut this pattern on 3 of the 4 boards (all of the 48" boards). The first I did according to the numbers above. I set the router to a speed of 3, which according to the manual, should be approximately 20k RPM. I noticed that the plunge went quite well, but the initial slot cut was a heavy load and bogged down the router. I slowed it down to about 50% of the feed rate and it coped. After the slot was complete and it started offsetting, the load was much lighter. I gently brought it up to 100% feed rate and it was running well. The motor RPM sounded approximately the same as it did without a load, despite it clearing a lot of wood very quickly. As it rounded a corner, it would slow considerably, but would recover as it went down the straight again. I tried increasing the feed rate to 130% and the RPM didn't appear to change, but it was cutting faster. I did read that the router has a control circuit to maintain RPM under load and it appears that this is doing what it claims, as the RPM didn't drop noticeably as I changed feed rates. On the second pass, the plunge and slot didn't seem to cause quite as much slowdown as it did on the first pass.
As the run was going, I also noticed considerable vibration in the sled holding the router. The path left by the bit also had waves in it in the Y direction. I also noticed that, during the long run down the length of the board, the wood tended to splinter and peel off in chunks, rather than chip. This appeared to be due to cutting with the grain of the wood.
After the run, there were noticeable marks left behind. These were most pronounced in the corners (where it left circular marks) and on the cross-grain cut on the far end (where it left a trail of wood standing up between each offset pass). For some reason, it wasn't as rough in the end towards the operator. In both cases, these cross-grain cuts had a slight lean towards the operator-end of the wood; each pass seemed to cut deeper on that end of the wood and shallower on the far end. Perhaps the Z sled is not quite perpendicular to the table.
On the second pass, I tried playing with the spindle speed and feed rate. I increased
On the second run, I decided to try playing with the spindle speed and feed rate. I started by turning the spindle speed up to about 4.75 and keeping the feed rate at 300 IPM. This caused noticeably less splintering of the wood and more chipping. At this point it was making very nice chips. I also tried increasing the feed rate. I was easily able to raise it to 420 FPM, but I didn't try going further since I wasn't sure what Goliath could handle without losing steps. The increased spindle speed seemed to make the motor slowdown less during the corners. It also appeared that raising feed rate required raising spindle speed to maintain the same level of chipping. Neither seemed to affect the amount of waving on the side of the cut or tool marks on the finished surface.
Between the second and third run, Ryan investigated the chatter that I was seeing. We noticed that several things were loose. The X rails had worn and required tightening. The arm that attaches the lead nut to the Z sled was very loose, so we tightened it. We also realigned the two X screws. Ryan tightened several other things, including a collar on the back side X screw, a collar on the Y screw, and the Y rails.
The third run didn't show any remarkable improvement. Waviness persisted and the tool marks were similar. All we could discern was that the tool appeared to be skipping up and down, causing the Z sled to rotate about the Y axis.
The final board was shorter, so I had to make a new toolpath. Since we noticed that the skipping happened during heavy load in the X direction, I decided to try a pattern that primarily used the Y direction. I decided to use a raster in the Y, but again oversized the pocket so that the turn around and step over that occurs in the X direction would be off of the piece to avoid tool marks. It would also plunge off the table, avoiding that heavy load. Since it would be alternating directions on each pass, we could also see the difference between conventional and climb cutting. For whatever reason, the toolpath was generated with the plunge in the opposite corner from home (in the far, back corner).
The raster cut was considerably different. Again, I started with a feed rate of 300 FPM and spindle speed at 4.75 (possibly around 25k RPM). The initial slot still bogged down, but it was very short, since it only had to pass the 4.5" width of the board. I eventually raised the feed rate to 140%, or 420 FPM. I did notice that it didn't maintain the feed rate well, because it was constantly accelerating and decelerating. Because of this, it took noticeably longer to run. I also noticed a considerable difference in behavior between the conventional cut pass and the climb cut pass. During conventional cut, it threw very nice chips perpendicular to the cut and left a smooth finish on the surface. However, it did tend to splinter the side of the board as the cutter left the side. This splinter grew the lines rastered down the board. On the climb cut movement, there was a very different behavior. In that case, it sprayed a fairly fine powder of saw dust in the opposite direction of the cut. It also left raised wood marks on the surface. The spindle also seemed to bog down more during the climb cut than during the conventional cut. All in all, the conventional cut seemed much better.
On the second pass, something very strange happened. I realized it was going to have to slot, so I hit reset on the feed rate override as it was rapiding back to the plunge for the second pass. For some reason, it then decided that a little over 1/2 way to the back was sufficient and it set down and started rastering back. I had to stop the feed as it approached the front of the piece. This left me with a half-finished piece, but it also gave me an idea for an experiment.
Since I noticed that the conventional pass was quite clean and the climb pass was very rough, I decided to try manually rastering always in one direction to get a constant conventional cut. My pattern was:
- Start on front side of wood (Y=0)
- Feed in the Y direction, cutting through the wood. Mach 3 was set to do this at 120 IPM, but I didn't know how to change it, so I left it.
- Take one jog step back in the X (to disengage the bit from the uncut surface)
- Rapid in the -Y to return to the front of the wood
- Jog two steps forward in the X. This would bring us back to the cutting surface, then one more step into it. I cycle my jog step to 0.1", which was approximately my original cut width.
As an aside, I did have a problem when I first went to start this. When I rapided back to the unfinished part of the wood, the back X axis stalled and racked the table. I think this may have been due to a considerable buildup of fine sawdust during those climb cut passes, which were spraying directly at the wheels on that end of the gantry. After cleaning it thoroughly and realigning the X axes, it worked properly again.
To get started with my manual pass, I went to a spot just beside the wood near the start of the unfinished section and zeroed X and Y there. I then brought the bit down to the top of the unfinished section, but rather than zero, I clicked on the Z value and manually entered +0.1875 (3/16" above zero). I brought the tool off the piece again and hit Go To Zero, which caused it to line up off the piece, near the unfinished section, at precisely my desired cut depth. From hear, I started manually rastering using the jog keys.
The results were impressive. There were still ridges between each pass, but the roughed up wood on the surface wasn't present -- this was definitely an artifact of the climb direction. In fact, a couple of times I forgot to step back and just returned on the same path that I had cut and this caused the same rough surface issue that the climb cut had made. Perhaps stepping back and running over it in the other direction was responsible for the much smoother final surface. This surface really showed off the fact that the bit wasn't completely flat on the surface, but rather was cutting lower on the operator-side of the table on every pass of the tool. Still, the surface quality was considerably better than the bidirectional raster. Also, it made almost entirely chips and very little sawdust.
For both styles of rastering, long splinters weren't a problem anymore. It seems that cutting along the grain of the wood is what was causing it to split and splinter. Cutting across the grain resulted in much nicer, more consistent chip sizes.
Tomorrow I will post pictures and video, as well as a summary of what I learned during this.