Headache

[esaj]

I'll write about this a more detailed post later (probably over the weekend or so), but for now:

It's still very much a work-in-progress (I'm actually using a separate Arduino Uno + CNC-shield to test this), but getting there. So far, it seems of at least semi-decent quality, although the spindle-holder and the sleds under the table are 3d-printed and not aluminum. Also, no limit-switches, I was playing around driving it manually with bCNC here, and held the arrow key down... well, as I release it, the motion won't stop!  I was just about to cut power near the end of the video when it finally stops

Only problem so far has been that the M6x10mm bolts that are meant to fix the sleds and the part through which the trapezoid screw goes under the table (which isn't attached on the video, ie. the Y-axis) are slightly too short. They barely catch on the threads of the square nuts that go in the rails, so it looks like I'll have to cut longer bolts for it (I looked around the house and garage, but only could find M6-threaded rod, no actual bolts ). In longer run, I think this can be improved further (open-collector hall-sensors + neodymium-magnets for limit switches, have to check the sled-dimension, if they can be replaced with something like SBR10UU's or such etc). Also, at some point I'll try if my higher-current Nema17's I bought separately give better torque/speed/accuracy (once I even get to try how good or bad the accuracy really is), and also have some "elastic" couplers on the way, that should prevent backlash.

Edited September 14, 2016 by esaj

[esaj]

Got my first test board milled (there are two traces that join together near the "zig-zag"-pattern, but that was my own fault ):

This is without auto-leveling, no contours and no drilling files, just plain simple SOP8 + pins for the legs. The last picture shows that the legs go right into correct spots (that's a SOP16-chip actually, but the pin pitch is the same ). I must admit that I'm pleased, especially considering that the machine cost something like 240€ (128€ for the machine + little under 60€ freight + 30ish € VAT and 24.80€ for DHL clearing).

I'm still learning to use the software tools, the first cut was too deep (the very corner of the lower left pin in the "worse" cut), then I took it too high up (not cutting deep enough), and then managed to move the entire shape out of position (so I couldn't try a third run on top of it, couldn't find undo in bCNC ). So just moved it aside a bit and made another run with better Z-axis setup (and still managed to partially f*** it up, one of the pins is cut off when the mill was moving into position too low )... The tool path was Fritzing -> FlatCAM -> bCNC, but I likely switch to Kicad instead of Fritzing for "real" projects.

 

 

 

Edited September 16, 2016 by esaj

[esaj]

Testing full tool-cycle, isolation routing + drilling + cutout:

The routing and drilling went pretty much perfectly (SOP16 -> two 2x8 pin headers, 1mm drill holes), things are connected to where they're supposed to and not connected to anywhere else. The cutout-part didn't go that well, first I had the bit running too deep, which caused the controller to shutdown (probably due to overheating of motor drivers), and I had to reset the damn thing (losing position). After getting the X/Y -axes closer to the correct position, I decided to leave more room on the Z-axis. Of course it was too much and it didn't cut all the way through...  But this is just practice (although actually, I do have use for that adapter board in testing some H-bridge drivers).

Also, this thing's hungry for blood: I've cut my finger on the V-bit twice before (minor scrapes), but trying to get it out to replace it with the drill bit, I was pulling it downwards (later on, I found out that I had forgotten to loosen the second screw holding the bit in place), but fingers slipped, banged the copper board and bounced up... end result: the bit went THROUGH one finger tip and cut the other (in different hands). Those things are sharp...

Edited September 17, 2016 by esaj

[esaj]

Ok, I've now tried through the entire toolchain from Kicad to Flatcam to bCNC with autoprobing, milling, drilling and cutouts.

This one went pretty well (except the drill holes went a bit to the side as I screwed around with the offsetting and lost position ):

Forgot to add a ruler to see the scale, but the entire copper-clad board is 70mm wide, the largest traces are 1mm and smallest are 0.25mm, the drill holes are 0.7mm (which is too small for a TO-220 power mosfet that was supposed to come to the right-side), I skipped the "tool change -stops" -options in FlatCAM... And destroyed one bit, as you can see from the lower failed attempt

Didn't do the cutout on this one, as I don't think I'm able to solder all the legts to the smalles pads without causing short circuits  But at least I can say that the accuracy is easily enough for my use, just need to design the pads and traces larger.

[Rehab1]

OMG esaj.....I need to stop by this post more often!!! Sorry! Your CNC machine is amazing! Interesting...ships get christened with champagne? your new CNC with bloody dna  ?  Exceptional job!?

 

[esaj]

I promised I'd write up more details on the machine over the weekend, and this at least partially repeats stuff from above, but here goes:

I'd been eyeing on a CNC-machine for PCB milling for several months. At first, I was thinking something like CNC 3040 or similar with all the bells and whistles (ball screws, strong steppers, limit switches etc). But those machines cost quite a lot (like 800-1000€ and up with all the costs), and I didn't want to spend that much. Then, when once again browsing the machines on Aliexpress, I saw a smaller, but much cheaper machine called CNC 2418. Despite being smaller, the work area is still enough for me (240x180x30mm), enough for even larger PCBs and milling out front panels for plastic boxes I use as enclosures (front panel-size up to about 210x125mm). Some sellers claimed the accuracy of below 0.1mm could be achieved, although I was sceptical about that, seeing that the screws were trapezoidal and some parts were 3D-printed plastic instead of aluminum or other metal. It could also be fitted with engraving laser, but I ordered one without the laser-head, as I have no use for it.

I found one on sale at around 128€ + 59.xx € for shipping (there were some sellers selling the machine itself for even slightly cheaper, around 110€, but the shipping would have been more expensive, making the entire package cost more). 

After the package was shipped (via DHL), it took less than 4 days before it arrived. And one of those days it was just sitting waiting for customs clearance. Really fast shipping. DHL handled the clearing, as they've changed policy here so that you either pay 40€ to get the documents to do the customs clearing yourself, or 5% of the taxes & customs duty, or a minimum of 24.80€ if the 5% amount would be under that. So for me, the cheaper option was letting them handle it, as the customs duty percentage was only 2.7% for the product, leaving the total amount below 10€, so no customs duties were charged, "only" 24% VAT of the machine price (about 30.70€). So in total it cost me about 240€.

The machine was packaged well, two boxes inside each other, with extra paddings inside and the parts were separately bagged and rolled into foam mats and bubble wraps to protect the more sensitive parts.

 

 

Everything needed to get the machine working was in the package, although I used pieces cut from M6 threaded rod instead of the M6 * 10mm bolts that were in the package, as those bolts seemed slightly too short for their intended use (holding the sleds and the part through which the trapezoidal screw goes under the milling bed), and could barely reach through the T-slot to the T-slot nuts (they were only holding for a couple of rounds of threads, and I suspected they might loosen or strip the threads in the long run). I contacted the seller about this, and suggested packing M6 * 15mm or such instead with the machines, and apparently they're going to do that in the future. Also, for most simple parts (nuts, screws, T-slot nuts, etc) there are some spares with the package (in case you lose some or strip the threads or whatever).

 

 

Putting the machine together wasn't hard, although it still took me about 12 hours in total. There's a small (8cm) DVD with the package that contains assembly instructions (that aren't very step-by-step but easy enough to follow), as well as some software and the Windows-driver for the CH340G USB-to-TTL -converter used by the controller (the controller is based on Arduino and runs the open source GRBL-software for controlling the motors). For Linux, the driver isn't needed, it should work out-of-the-box. 

 

When assembling the machine, take care to make sure the frame comes out "true", ie. the aluminum "squares" are really square, and not diamond-shaped etc, it might require some tightening, loosening and repositioning to get it all good. Also when putting in the rails, move the bed and the spindle-head all the way to the ends when lining up the axes, it might take a couple of rounds of loosening and tightening the parts that hold the linear guide rails (under the milling bed and at the ends connecting to the frame) before you get everything straight. Getting them to run smoothly and in straight line is essential for good machining accuracy and quality.

 

 

The motor couplers are rigid, I have some spiral flexible couplers, but haven't tried them yet. The flexibility of the coupler helps the motors, as they don't have to work as hard, because the trapezoidal screws and linear guides themselves might not be 100% straight (were talking like for example 1mm of "bend" across the entire axle, but tight fitting sleds and the guide-part cause this to transmit to the motor coupling), causing sideways forces during linear movement that further stresses the motor. The downside is that the flexible coupling could potentially have (minor) effect in the accuracy (or not, depending on many factors ;)).

After getting the machine assembled and learning the basics of the software, I ran a few tests, some of which didn't end that well:

 

After figuring out the basics of the software tools, I managed to mill a small test for a SOP8-chip and pads for pin headers (without drill holes):

 

As you can see, the pins are exactly where they're supposed to. The lead-spacing of SOP-chip is 0.65mm. This was a pleasant surpise, the accuracy of the machine seems very good.

Further testing, a SOP16-to-pin-headers -adapter board (I have such boards as factory made, but this was a good test too):

 

Again, no problems with the accuracy, although I probably cut a bit too deep, as the traces are very thin (but still all connected).

At this point, I also built the wires for auto-probing the board surface. bCNC supports autoprobing (or just probing, or depth probing or whatever you call it), a process where the software can ask the controller to check when the milling bit hits the board surface. The GRBL-controller supports this through the A5-pin, which is set up as input with pull-up; all you need to do is make two wires (I used basic Dupont-headers and alligator clips), one going to the A5 and one to the board ground. Then you connect one wire to the milling bit, and the other to the board. As the board surface is copper and the milling bit is metal (tungsten carbide or something), they both conduct electricity, and once they touch each other, the input is pulled down and the controller informs the Z-depth where the "collision" happened. You can select the amount of X and Y points (and of course the area) that is scanned through, and then bCNC will build a "height map" of the entire milling area and adjust the milling depth on-the-fly during the process. This gets you very high accuracy in the milling, as you can see here:

 

This is a 555-based PWM-controller designed first in KiCAD, then exported as Gerber, the Gerber-files are then processed to G-code through FlatCAM, and finally bCNC is used to autoprobe and mill the board. As you can see, the accuracy and detail of the board is outstanding (well, at least for a machine of this price-range!). Actually, it's more accurate than what I really need: I can't solder that small pads by hand!

After drilling, I messed up the positioning and offset myself (stupid mistake :D), but as I already knew I'd need to re-design the board with larger pads and traces anyway, I just went with it (you can't learn without making mistakes and all that ;)):

 

A small problem that I had run into a couple of times was that sometimes with longer use period / longer job running, the controller would go into some form of "alarm"-state, where it stops milling and needs to be reset (usually I also lost the position at this point). It turned out that the A4988-drivers where overheating and entering thermal shutdown to protect themselves. At first I used a separate fan blowing into the board, that seemed to help (it no longer did that), but it was balanced on top of plastic boxes, and powered by LiPO-battery with alligator clips (kinda dangerous setup, should the fan fall and cause the clips to short). 

So, todays job was to make an encasing for the controller board with the fan built-in (the controller-board itself has 12V output). I took a basic 12V fan, "suitably" sized plastic board, made some holes and installed the control board:

 

 

I had to snip of small pieces of the corners of the controller board to fit it there, but those parts aren't necessary anyway, actually, the parts that are nothing but green can be snapped off if you like, there's already cut holes there so you could do that just by hand. But those larger holes in the corners happen to be lined up so that they hit the aluminum profiles of the frame with T-slot nuts, so I wanted to keep them and line up the box with them.

After getting the controller encased, it was time to add the fan. I had drilled the larger holes, but needed to cut the large hole for the fan to suck air through (or blow out through, depending which way around it's installed). If I only had a machine that could do it... oh, right:

 

 

 

No more heating problems!

There's still ways to go for me to really learn all the ins-and-outs of the milling processes (feedrates and milling depths for different tools like 1/2/4 fluted end mills, fishtail mills, different degreed V-tips etc), learn to avoid the pitfalls when building the G-code with FlatCAM (that thing seems to have a ton of options), as well as avoiding problems already at the design phase etc. But I can quite confidently say that the machine or it's accuracy or such won't be the problem (at least for through-hole and larger SMD-part boards).

Edited September 18, 2016 by esaj

[esaj]

After a few more days and learning to better use the software, I've now got the workflow pretty much down (of course, I'll likely alter it as I progress and find better ways to do things). Hobby16 asked about the downsides of the machine, and of course there are some. Nothing too bad, and it mostly comes down to why the machine is so cheap (plus, you can fairly easily make it better by exchanging some parts, but quality parts can become quite costly):

- The "collet/chuck" is just a cheap motor coupler, I think it's brass, like this:

Getting bits aligned properly can sometimes be hassle (although the other end is supposedly 3.175mm, ie. the "correct" size for the typical bits) and those small nuts will probably wear out fast with repeated tool changes. Most of the time the bits go straight without problem, but a few times during tool changes I've noticed a slight wobble in the bit, so I've had to release and re-attach the bit to get it straight.

A new one like that costs something like 1€, and the machine came with 2 pieces (+ 1 extra nut, in case you lose/strip the threads on one), and I'm ordering some spares + at least one "real" ER11 -collet/chuck (but those things are more pricey, the chuck itself is something like 8-10€ at cheapest and the collets are a few euros a piece):

 

 

- The motor is low powered (something like 130W or so at best, maybe not even that) R775, so no fast feedrates (especially when doing the cutout) and no metal milling (the copper layer on PCB surface is so thin that it won't be a problem though). Don't know the RPMs, the seller says 7000 @ 24V (it came with a 24V / 6A power adapter, that says "Made for Lenovo" on top of it ), but trying the spindle speed-control in BNC, it seems to be running at full speed already with the setting somewhere above 1000. Of course it could be that the setting in bCNC doesn't match the real motor speed. The spindle wattage isn't that important in isolation milling, but when doing cutouts or such, you need more power (torque), especially with any higher feedrates, as otherwise the bit can get stuck because the motor isn't powerful enough to keep it spinning while doing "heavier" cutting. The RPM is important for isolation milling, as the V-bit needs to spin fast to cut the copper (the faster the RPM, the higher feedrate can be used). Industry-level PCB milling is said to use RPMs in the 30000 to 100000 (30k-100k) -range.

Also, the motor is probably brushed, and I doubt it has very long lifetime.

- Everything else is in the package, but you need to do get/your own holders (there are 4 extra T-slot nuts to use, I ended up cutting pieces from M6 threaded rod, then drilling holes to pieces of 4x40mm aluminum flat and filing the edges and attaching them with wing nuts, you can see the holders in the picture where I've cut the fan-hole for the encasing in the above post)

- The A4988 -drivers had a tendency to overheat during longer stress (cutout, trying to carve wood too deeply in one go), even though they have small heatsinks (they also came with the machine, but you need to put them on yourself)

I encased the controller with a fan that seems to help (although even since then, it has once stopped the job during faster cutout probably due to a driver going into thermal shutdown)

- There are no limit switches, control buttons or probe-wires with it, but since the board is based on Arduino / grbl, it's easy to add them:

 

The pins are exposed (the header rows at the lower left), so you can add limit switches, stop/pause/continue-buttons and the probes yourself. I have already made the control- & limit-switches, last night was cutting small about 1x1.5cm boards to hold the microswitches & connectors, I'll probably just use double-sided tape to attach them. The control buttons still need a housing though, but I should get some suitable small plastic encasings this week or the next.

You can see that there's some form of built-in fan at the top of the motor, but since the holder is plastic, it probably does very good job at THERMALLY INSULATING the motor :

Although the fan should still blow through it... I've never tried if the motor feels hot.

Don't know how long that thing's going to last, maybe I should already order a replacement (as the Aliexpress orders can easily take 4 weeks to arrive ). Cheap 15k RPM (@24V) 775's cost about 10-15€ a piece (with free shipping). 

Just thought to add these here too from the private messages, in case someone's thinking about getting one of these machines.

 

Edited September 21, 2016 by esaj

[esaj]

A board milled for a friends' synthesizer-project, largest board I've milled so far in one go:

Almost perfect, but not really... The vias were meant to be drilled with smaller drill-bit, but I didn't have the correct size, so I went with 1mm everywhere. The isolation went fine otherwise, until I had depth-problems near the top corners(where the holders were), and had to re-mill those parts. Which of course then became a bit too deep   But I guess it can still be used. The larger holes used in the vias cut the traces at points, which is far from ideal, but can be worked around     I guess he used vias instead of actual holes/pads where the wires are attached to...

Probably I should start trying using multiple passes to get wider isolation around the pads, those can get tricky to solder by hand on the smaller ones.

Edited September 26, 2016 by esaj

[Lefteris]

it looks much prettier than the last one 

you are getting the hang of it, great !!! 

[esaj]

I (of course) ended up ordering higher torque servos, a gripper and brackets after that one sleepless night when I built the makeshift-robot arm... This is mostly just testing some servos, I still need to build a "proper" structure, and am still waiting for some more servos and other stuff. Also, sorry for the camera pointing to wherever at times, I was more concentrated on controlling the arm...

Edited September 26, 2016 by esaj

[Rehab1]

50 minutes ago, esaj said:

I (of course) ended up ordering higher torque servos, a gripper and brackets after that one sleepless night when I built the makeshift-robot arm... This is mostly just testing some servos, I still need to build a "proper" structure, and am still waiting for some more servos and other stuff. Also, sorry for the camera pointing to wherever at times, I was more concentrated on controlling the arm...

Oh that is so awesome! Great work! So what do you even consider a sleepless night?

[esaj]

19 minutes ago, Rehab1 said:

Oh that is so awesome! Great work! So what do you even consider a sleepless night?

The kind where I actually try to sleep, but don't manage...  On a partially related note, late last week I actually applied for a job at a company that produces software (and mechanical design, and electronics) for (industrial) robotics projects and such. Haven't heard back from them yet, and they might not even respond, as my application was a bit rushed and not really polished, plus they seem to have pretty high bar (on average, their employees have 10+ years of experience, and I think most are embedded developers), and also, even if I managed to get an interview, and get the job, it would probably mean that I have to try and get back to "normal people" schedule, like 8AM-4PM job, plus have to commute (like I've said numerous times, I get to work from home almost all the time currently). Gotten a bit too used to sleeping in late and working at "weird" hours...

Edited September 26, 2016 by esaj

[Rehab1]

I cannot imagine with your knowledge, skill level and motivation the company not taking advantage of an opportunity to hire you! Yes, I am sure you love your current hours and being able to work from home but this job opening must have really caught your eye!

Regarding your servo gripper, my company use to fabricate myoelectric prosthetic devices for both upper and lower extremity patients. Currently there is a great deal of governmental R&D funding for myoelectrics especially with all of the IED's exploding and taking off US soldier's extremities in the middle east Unfortunately when I downsized the company 7 years ago I sold that part of the practice. 

Well I'm off to watch the 2 candidates box it out on stage in the first presidential debate. Always great to communicate with you! Cheers!

 

Edited September 26, 2016 by Rehab1

[esaj]

On 27.9.2016 at 2:07 AM, Rehab1 said:

I cannot imagine with your knowledge, skill level and motivation the company not taking advantage of an opportunity to hire you! Yes, I am sure you love your current hours and being able to work from home but this job opening must have really caught your eye!

Didn't hear back from them this week... Maybe they already found what they're looking for, or are just otherwise busy. I don't stress over it though, as I don't really need a new job

 

Quote

Regarding your servo gripper, my company use to fabricate myoelectric prosthetic devices for both upper and lower extremity patients. Currently there is a great deal of governmental R&D funding for myoelectrics especially with all of the IED's exploding and taking off US soldier's extremities in the middle east Unfortunately when I downsized the company 7 years ago I sold that part of the practice. 

I've seen some of videos of those "robot arms" that read the muscle/nerve signals from the remaining limb and people actually seem to learn to use them (over time) almost like they're a natural part of the body. Really cool stuff, didn't know the term "myoelectric" though before, thanks for that

 

Quote

Well I'm off to watch the 2 candidates box it out on stage in the first presidential debate. Always great to communicate with you! Cheers!

You think Trump could actually win? Weirder stuff has happened...

 

I just finished putting together a 15-channel servo controller, it came out somewhat big, as there's lots of traces to run the PWMs through and the chip is quite large (28-leg DIP). Pretty basic stuff (as usual), and you can actually buy ready-made controllers (using the same chip as mine, LTC5940, except the ready-made ones use SMDs and two-sided boards and are thus much smaller ):

Board milled and cleaned up, I made 2 isolation passes to get wider cuts. Some of the cuts are wider than the others, at first I though that the machine repeatibility was playing part there (ie. how accurately it returns to a position when moved around and told to go back to exactly the same spot as before), but actually looking at my KiCAD-files, I've instructed the GND (Ground) -network to use wider isolation than other parts, thus the wider cuts facing the ground plane (although it looks like the area between the PWMs should be part of the ground-plane, it's actually isolated from two points, if you look close, that's why most of the PWM-lines coming from the chip don't have as wide isolation). 

Piss poor solder job, especially next to the lower right corner of the text block (those are two ceramic disc capacitors that had very thin legs in comparison to the 0.9mm holes, and I tried to get the solder to fill the holes, but it just spread all over the place), but hey, it works   Also tried to get the solder to spread better around the thermal reliefs in the ground-connections, but that didn't go too well either... Maybe in the future I'll just leave solid copper around the ground-contacts. Solder mask would be nice, but kinda hard to do at home  And at least there are no short circuits anywhere.

Component-side, nothing much special here. There's one resistor under the socket, as I had some space issues, but otherwise I'm pretty happy with the layout. I "only" used 15 channels of the 16 of the chip, as I doubt I'll never need more than 10 anyway, plus using the last channel would have either cut the ground contact from the chip (requiring a jump wire or two-sided board) or I'd have had to lay it out differently, so I just left the last channel out.

Chip and jumpers in place, plus some markings so I don't have to figure out what pin does what, haven't really tested it much yet. The jumpers are used to select between using the LM7805 (5V regulator) to regulate from higher voltage to 5V for the chip, or direct +5V line from outside the board, and the other one selects between using the same power source as the chip for the servos, or using a separate power source for them.

I'm still missing some mechanical parts to get on with the robot arm, but expect to have everything within the next week. Not sure what I'm going to use the end result for, but nevertheless it's an interesting project in robotics (basically combining mechanics, electronics & software).

 

 

 

Edited October 1, 2016 by esaj

[Rehab1]

18 minutes ago, esaj said:

Didn't hear back from them this week... Maybe they already found what they're looking for, or are just otherwise busy. I don't stress over it though, as I don't really need a new job

 

Bummer! I'll keep my fingers crossed. When my daughter was looking for a teaching job after college we set up a website and added the web address to her resume that directed potential employers to her site. It made a huge difference in opening doors.

20 minutes ago, esaj said:

Piss poor solder job, especially next to the lower right corner of the text block (those are two ceramic disc capacitors that had very thin legs in comparison to the 0.9mm holes, and I tried to get the solder to fill the holes, but it just spread all over the place), but hey, it works   

I would never have noticed! I my view this diy project is a work of art, patience and persistence. I love the before and after photos of the board and how well it cleaned up.

 

25 minutes ago, esaj said:

You think Trump could actually win? Weirder stuff has happened...

Anything is possible in the USA. I could only stomach 20 minutes of the 90 minute debate. Embarrassing!

[esaj]

I spent a good seven hours last evening/night trying to figure out a toolchain for something relatively simple: cutting "2D" pieces out of plastic sheet. In the end, I found out that QCad for the mechanical drawing, then exporting as DXF and using bCNC's own profile -options + cut -toolpath got what I wanted:

 

The question then was, can a fishtail-end mill cut plastic well? Answer is no:

The plastic tends to melt and stick to the bit, which will then screw up the hole edges.

Other than the melting, the end result was exactly what I wanted, but I need to figure out a better way to cut this... maybe a straight bit? Which I don't have at the moment...

I tried if running the bit agains the wood sheet would work to clean it up (it kept throwing chunks of the plastic off the bit during the plastic milling), but nope... it only started to burn the sheet and melt the plastic even worse. Let's hope I can at least salvage the bit, it looks like the plastic dissolves in acetone. Hopefully it won't react with the volfram (tungsten) carbide or the titanium coating of the bit...

[Hunka Hunka Burning Love]

I wonder if freezing the piece of plexiglass might help along with spraying some component cooler at the cutting interface.  I've tried to cut various plastics with high rotational cutting bits, and the frictional heat tends to melt the plastic and goo things up.  Or even blasting some compressed air at the bit as it cuts might be enough to keep things cool enough.

[Chriull]

@esaj for cutting "plexiglass" there are some tipps: http://flugwiese.de/2014/01/plexiglas-fraesen-so-gehts/ - unfortionately just in german.

the main points: normaly water or other liquids are used for cooling, but they did not want the mess. Others used compressed air, which was not available for them.

so they found that with the following parameters they got quite nice results:

Fraise 2mm single edged

12.000 rpm

"Feed rate" 11 mm/sec

 Max 0.4 mm "depth" per pass

from the comments:

a sharp bit is important

plexiglas is cut without cooling and a "chip with" of 0.2mm, compressed air is used to remove the chips (clean cutting edge)

Different "plastics" - very different results. Makrolon (PC) is much easier to cut than plexiglass (PMMA), transparent plexiglas is easier to cut than the white one

one used "Acmos Trennspray 100-2450"

if you machine does not make 12.000rpm, one needs a bit with a bigger diameter (circumferential speed)

[esaj]

5 hours ago, HunkaHunkaBurningLove said:

I wonder if freezing the piece of plexiglass might help along with spraying some component cooler at the cutting interface.  I've tried to cut various plastics with high rotational cutting bits, and the frictional heat tends to melt the plastic and goo things up.  Or even blasting some compressed air at the bit as it cuts might be enough to keep things cool enough.

Thanks for the tips, freezing might help (I don't have compressed air available).

 

1 hour ago, Chriull said:

@esaj for cutting "plexiglass" there are some tipps: http://flugwiese.de/2014/01/plexiglas-fraesen-so-gehts/ - unfortionately just in german.

the main points: normaly water or other liquids are used for cooling, but they did not want the mess. Others used compressed air, which was not available for them.

so they found that with the following parameters they got quite nice results:

Fraise 2mm single edged

12.000 rpm

"Feed rate" 11 mm/sec

 Max 0.4 mm "depth" per pass

from the comments:

a sharp bit is important

plexiglas is cut without cooling and a "chip with" of 0.2mm, compressed air is used to remove the chips (clean cutting edge)

Different "plastics" - very different results. Makrolon (PC) is much easier to cut than plexiglass (PMMA), transparent plexiglas is easier to cut than the white one

one used "Acmos Trennspray 100-2450"

if you machine does not make 12.000rpm, one needs a bit with a bigger diameter (circumferential speed)

Thanks to you too. The spindle motor can do around 7000rpm max (although I have some coming that should go above 15000rpm max). The feed rate sounds surprisingly low (11mm/min, I've tried both 300mm/min and 1200mm/min ), as that would mean the bit stays spinning around the same location for a very long time (heating up the plastic).

It's not "actual" plexiglass, but something called SAN (styrene acrylonitrile). I've read about milling plastics on various sites and sometimes the tips seem almost the opposite of each other: some say to use higher rpm and slower feeds, where as others say that you should use low rpm and/or high feed to stay in one place as little time as possible and cutting away as big chips as possible without the end result becoming too rough, the idea being avoiding heating one position for too long, which sounds more reasonable to me. Also using the fishtail end mill at first probably was a bad idea

With lower rpm/higher feed it's also suggested to use a single flute tip with very large cutting surface (in relation to overall diameter):

 

The above single-flute bit is actually sold as "plastic cutting end mill". Also "straight bits" have been suggested:

 

My bits are the cheap chinese kind (surprise? ), plus what further limits my options is the 1/8" (3.175mm) shank size, I didn't find straight bits except with 1/4" shank (in Aliexpress that is, I'm pretty certain such are available elsewhere, but the price goes up a lot).

Edited October 3, 2016 by esaj

[Hunka Hunka Burning Love]

If you don't mind things getting a little messy, and if you have a spray bottle maybe shooting a stream of ice water at the bit while it cuts might be something to try.  I would do a small test section (eg.simple linear cut) to see what works best.  A small pancake compressor is about $75, and it's handy to inflate tires as well.  A couple of cans of component cooler probably run around $15 I'm guessing?  Freezing the drill bit might work for quick jobs, but it heats up quickly.  I find those crosscut bits tend to heat up the plastic, and it gets up gunked up in the grooves making it a pain to clean off.  I wonder whether freezing a thin sheet of ice over top (or bottom side) of the plastic might work.  The drill could blast through the ice and cut the plastic underneath or have the ice sheet underneath, but you would need a good sized chest freezer to do that.

You need to figure out how to lathe your own custom drill bits .

Edited October 3, 2016 by HunkaHunkaBurningLove

[esaj]

8 minutes ago, HunkaHunkaBurningLove said:

If you don't mind things getting a little messy, and if you have a spray bottle maybe shooting a stream of ice water at the bit while it cuts might be something to try.

I'd prefer to avoid using liquids for cooling or other stuff that makes a mess, as the machine is in my study, maybe in the garage it wouldn't matter.

8 minutes ago, HunkaHunkaBurningLove said:

 I would do a small test section (eg.simple linear cut) to see what works best.  A small pancake compressor is about $75, and it's handy to inflate tires as well.  A couple of cans of component cooler probably run around $15 I'm guessing?  Freezing the drill bit might work for quick jobs, but it heats up quickly. 

Probably need to just do test runs until I find something that works... The holes are probably the most difficult parts, as they're cut in a small area, the outline (straigth lines / couple of arcs) should be easier as the bit won't stay in one place for long (with higher feeds).

8 minutes ago, HunkaHunkaBurningLove said:

I find those crosscut bits tend to heat up the plastic, and it gets up gunked up in the grooves making it a pain to clean off.  I wonder whether freezing a thin sheet of ice over top (or bottom side) of the plastic might work.  The drill could blast through the ice and cut the plastic underneath or have the ice sheet underneath, but you would need a good sized chest freezer to do that.

I put the bits (1 fishtail + 2 single flutes) in a jar with acetone and left them there for a couple of hours, shaking the jar every now and then. The bits cleaned up completely, no need to pick off the gunk even from the "crosscut" fishtail, and it seems that acetone doesn't react with the carbide or the titanium coating (in the fishtail bit). I might try just putting the plastic sheet in the freezer, and taking it out just before starting to work on it. Also my cousin dropped off some other types of plastic sheets this morning that he'd found at the company dumpster

 

8 minutes ago, HunkaHunkaBurningLove said:

You need to figure out how to lathe your own custom drill bits .

That might be a bit out of my league...   Besides, the (cheap chinese) bits cost like 0.30-0.60€ a piece, so not much point doing your own bits, unless you need something more special (and are able to lathe it yourself)... I've got some ER-11 chucks & collets on the way, that should be able to use different shank widths with different collets.

[Hunka Hunka Burning Love]

On ‎16‎/‎09‎/‎2016 at 3:45 PM, esaj said:

 

 

Forgot to mention that the above looks vaguely like some sort of group sex symbol for a weird kind of foursome. 

Maybe a can of component cooler or canned air would do the trick and keep the mess down.  It's also handy to diagnose failing electronics components or cold solder joints as well as clean computer stuff up from dust.  Not to mention a great chewing gum remover.

https://www.amazon.ca/Electronic-Liquefied-Multi-Purpose-Anti-static-Electronics/dp/B01D3QKDB8

Edited October 18, 2016 by HunkaHunkaBurningLove

[esaj]

Another small "side-project" (I have so many things running in parallel that I can't even keep track of them myself anymore... ):

http://edtracker.org.uk/

EdTracker is a small headtracking device based on Arduino Micro (it has to be Micro or Leonardo with AtMega 32u4, AtMega 328P used in Nanos etc. won't work!) and MPU-6050/9150/9250 (the last two have magnetometers, ie. they're 9-axis gyros, and won't drift by themselves, like the 6050). Windows (and maybe Linux, haven't tried) will see it as a 3-axis USB joystick and it can be used in games (or whatever supports joystick control), usually for looking around in car/flying/space-simulations and such.

The project itself is open source (both hardware and software), although http://www.edtracker.co.uk/ sells a ready-made "Pro"-version and a ready-made PCB for DIYers who can't make/order their own design and don't want to use a protoboard.

Anyway, my cousin got one for himself for playing Elite: Dangerous, and after looking around for a while, I ordered a few sets of parts to build some units. They're mostly for friends, although I'll keep at least one for myself. I got the Arduino Micros, MPU-9250's and 2-meter USB cables from Aliexpress (if you're surprised, raise your hand... ), the Hammond-enclosures are from TME ( http://www.tme.eu ) and I designed & milled the board myself (if someone wants the KiCAD-, Gerber- and or G-code files, I can put them up somewhere), just finished the first one tonight:

The board, really simple (I just used the pins that are actually needed and left the rest floating, except for AD0 on the MPU-board). The corners are cut so it fits in the Hammond 1551G -enclosure (50x35x20mm). I also have some 1551F-enclosures (otherwise the same but with 15mm height), but didn't start with trying to cram it in the smaller space.

Didn't take a picture in-between, but I first soldered pin-headers on both sides of the board, then soldered the Micro & MPU-9250 on opposite sides. The button isn't strictly necessary (not soldered yet in these pictures), it acts as a quick way to "zero" the device (ie. push it, wait something like 15 seconds for it to do a quick recalibation and it sets the heading straight to current position).

A couple of pictures from the sides. I'd need to get about 1-2mm less height for the thing to fit into the smaller enclosure. Probably doable, but didn't attempt it here.

 

Checking fitting, need to check dimensions to see where I need to cut the hole for the USB-connector & the button. The button is now soldered in place.

USB-hole done, marked the button position.

Button hole done, testing fitting.

Hot-gluing the device in place and...

Done (at this point the firmware is already flashed in & the longer "steady-temperature" magnetometer calibration done). Probably try to drop this off to a friend tomorrow (unless he wants to wait for another version, if I can get it to the slimmer enclosure ).

 

Edited October 3, 2016 by esaj

[Hunka Hunka Burning Love]

Going back to the plastic cutting, I wonder if you can program the cuts in short segments like a dashed line to briefly cut in the plastic to help keep it cool and prevent the plastic from becoming molten and sticking to the bit.  A second pass could cut out the remaining dashes like in a square wave pattern up and down.  Even with a frozen piece of plastic, I wonder if the leading edge and sides would still heat up from the friction leading to a gooey mess.  With brief in and out cuts that allows the bur to stay relatively cool and the plastic from heating up too much.

[Rehab1]

On October 3, 2016 at 5:26 PM, esaj said:

 

Another small "side-project" (I have so many things running in parallel that I can't even keep track of them myself anymore... ):

 

I love coming over here to see more of your ongoing projects! I get a headache watching your headaches. Bet you 5  bucks I already know your next project: Edtracker wireless version. ?