Milling small parts out of 1/8" cast acrylic

Keep in mind that vacuum hold down for a PCB has the entire area of the PCB for the vacuum to pull on. When one is cutting small pieces from a sheet of stock, vacuum doesn’t do so well - it depends on what “small” is and how much vacuum there is.


I’ve been doing quite a bit of machining on acrylic lately. I don’t know how much useful advice I can give for flat parts (I’ve been focusing more on 3D carving), but for what it’s worth…

If I’m cutting out flat parts with a 1/8" end mill, I hold down the sheet with carpet tape (the kind that came with the Nomad) and take 0.5mm cuts at 10,000rpm and 1200-1500mm/min. Largely that’s because I use the machine in my not-very-soundproof apartment, but it also helps with keeping the work held down. I’ve noticed (from the smell) that deeper cuts heat up the acrylic enough to melt it, but I’m not sure that really matters; it might not be good for the cutter.

Holding down acrylic sheet is tricky. It’s flexible enough that clamping is no good, but rigid enough that as soon as the tape starts to let go, it lets go completely. I believe it helps a lot to make sure that every point is stuck down securely-- I use a bone folder (or a thumbnail) to get good adhesion on the part, with no bubbles, and I make sure the spoilboard is completely free of dust and maybe lightly sand it with 1200 grit before sticking the part.

The good thing about acrylic, though, is that even one small tab will do a lot to hold the part down. If I’m creating the toolpath manually (I use a Rhino script), I’ll arrange it so that I end up with one tab on each part, then just snip all those tabs with the end mill; probably one part in ten comes unstuck at that point, but the cutter has already moved on so it’s not a problem.

I usually finish the cut surfaces by flame polishing-- just run the hot part of a blowtorch flame over it (in a ventilated area), and you get a nice gloss finish in a couple of seconds. It’s a lot easier than a vapor bath, though I don’t know how it will affect your tolerances. I’d guess it puts a < 0.1mm radius on sharp corners. It’s fun, anyway.

ETA: to remove the tape, I just dump all the parts in the sink, clean off the residue with a drop of paint thinner, and wash the parts with dish detergent and water.

I am actually in the middle of making my own vacuum hold down system. I’ll post a new thread with more info shortly, but it all seems to be coming together rather nicely. First few plenum cuts have gone well. Currently I’m working on a vacuum pump design that will maintain 24" to 27"Hg without constantly running the pump (hopefully quieter milling).

1 Like

I am actually in the middle of making my own vacuum hold down system. I’ll post a new thread with more info shortly, but it all seems to be coming together rather nicely. First few plenum cuts have gone well.

I’d love to see this… especially if there is a nice way to integrate it into the enclosure so that the vacuum tubing isn’t problematic (not getting in the way).

We set this up by placing 1" MDF on the bed, machining it flat and then creating channels through it… pretty standard.

I saw one design that combined a vacuum bed with T-Rails. One could clamp or vacuum.

Currently I’m working on a vacuum pump design that will maintain 24" to 27" Hg

That’s a nice vacuum! The lower the better! Adjustable?

without constantly running the pump (hopefully quieter milling).

The vacuum beds I’ve used are NOISY, like need hearing protection NOISY. Anything you can do to solve the noise problem will be welcome. The small size of the Nomad would make “vacuum storage” easier than a 5x10 foot monster CNC.

A good vacuum bed is very useful for many types of work. The vacuum has to cause enough force to be applied to the object being cut to prevent it from moving. That means that at some point, a small object is not going to stay put with the vacuum.


I’m running into an accuracy issue that I’m banging my head against the desk trying to figure out. I have several holes in a part that must be 4.8 mm in diameter. If I run a sheet I get 4.8 mm one time and maybe 4.7 mm the next time. The variance is dead on .1 mm from job to job, but granted I haven’t done dozens yet so my sample size is low.

0.1 mm should certainly be within the tolerance of the Nomad. My only guess at this point is shifting of the double sided tape. I’m going to try some Nitto tape on the next run but just by feeling the stuff I don’t think it will be as good as what I was using. Definitely less ‘grippy.’

I did gain some accuracy by running the job at a shallower depth. I’m currently running my passes at 0.50 mm per pass to avoid the shifting issue. I’m hesitant to cut it more than that just because of the time it takes.

Should I trust G-Wizard’s deflection calculator enough to reduce my depth per pass even more but on the flip side increase the speed? I feel like even though the depth of pass is shallower, increasing the feed rate would cause more shifting of the stock than running a slower but deeper pass.

Do I have it wrong? BTW, the product I get back when it is dead on 4.8 mm for the hole sizes is absolutely beautiful. Perfect glass like vertical walls with extremely hard to see tool marks right out of the machine. I just can’t understand how one run can be dead on and the next be off by as much as 0.10 mm when using the same type of double sided tape, processes, etc.

Anyone have any helpful tips they would like to pass on? Thanks in advance!!

0.1mm is ~0.004". You’re beginning to approach the limit of the Nomad. That said, the Nomad is better than that.

At these levels, just about everything comes into play or needs to be though about.

The same G-Code each time? The same batch of stock? End mill? Was the spoiler board machined flat? Same batch of tape? Same process to apply/remove? Double check things?

The Nomad and G code has no idea there is stock there. The commands are just run. Not likely they are the issue. That things are sometimes “spot on” agree with this and also that G-Wizard is giving good advise.

Tape will work when the “sticking force” per unit area times the area being held is above forces applied. Small pieces have less area and so are held less well. Moving slower or using less depth reduces the forces on the piece.

From the picture I’ve seen I can’t get an accurate sense of the scale (no ruler next to it) but it looks fairly small to me (also because I expect LEGO pieces to be small). Small enough that I suspect the piece is moving, not the stock.

With small parts one needs better tape, to seek a different fixturing method, or a different machining method.

I’m not a big tape booster. I come up with other fixtures but when I’ve been forced to use tape 3M (Scotch brand) #410M is what has worked for me. Admittedly the parts may not have been that small.

Can you edit you posting wth the picture of the part with a picture with a ruler in the picture where the scale can be seen please?

Do you clean the stock before applying that tape? The spoiler board?

Was the spoiler board milled flat?



I looking for a neat fixture for tiny part I saw.

The hole size in the photo I provided is 4.8 mm in diameter, that should give you a rough scale. My production piece is slightly larger but essentially the same design. Machinability wouldn’t have changed. The problem with other methods of holding things down on the table is that I have (at most) ~6 mm of surface area that isn’t being milled and on average the space is closer to 4 mm connected. We’re talking LEGO scale here so if you have a LEGO anything laying around, you can get a quick sense of scale just by looking at the knob size.

  • Same G-Code
  • Same batch of stock sort of…same manufacturer but different colors. Cast acrylic varies in thickness a little from one sheet to the next, but I don’t think it matters at this point. I ran one job with a depth of 2.72 mm and the next batch had the same depth (2.72 mm) and got the .1 mm difference. LxW dimensions were exactly the same as was placement on the table.
  • Spoiler board was machined flat but I don’t machine it between each run since I’m only running these, the tape comes off of the table without residue and I’m never cutting past the backing paper on the acrylic
  • Same exact process. I’m actually using a little acrylic finisher before starting passes 2 and 3. I found this helps the cut a little and polishes as it goes along…no buffing afterwards is needed.
  • I’ve double and triple checked until my head is spinning!

I’ll see what the Nitto tape does and will look into the 3M #410M stuff that you mentioned. I’m so close on this I can taste it, but I just need to get the accuracy down a little closer. It keeps getting better as I try new things and gain more experience, but I feel as if I hit a brick wall with this last mile.

I’ve double and triple checked until my head is spinning!

I understand your frustration. You’re learning and pushing the edge of the envelope.

Spoiler board was machined flat

Just checking… Necessary for edge of the envelope machining.

If that’s 4.8 mm, that’s a small part to be depend on tape to hold AND one wants to machine in reasonable time.

There is a type of fixture that is used for this situation and for life of me I cannot find it… Essentially, it’s a set of adjustable springs that push down onto the stock and the gantry moves up and down. With some felt on the bottom - to maintain your surface finish - it would apply pressure on the tiny part, pushing into the tape.

I will find it…


I would be very interested in seeing that!

Tape is the best solution I can come up. My only potential causes that I can come up is the stock moving but I have no clue if it is moving above the acrylic backing, below it (double sided tape side), or maybe a little bit of both. I could probably diminish this a little by slowing the job down again. I did get close to 100% accuracy in earlier experiments, but I think I was cutting 0.25 mm per pass. That takes forever! I could go back to doing that, but I would want to compensate with a higher feed rate…which in my mind would be worse on a stock moving situation.

Better tape is another solution however this assumes one can find it and that it will work with such a small part.

FOUND IT! It’s called a “pressure foot”.

With a bit of thought a derivative of this can machined on the Nomad. Nothing like using a machine to make a machine. The bottom piece I would use Delrin - a super slippery one - or maybe Teflon. Even so I would consider felt or soft cloth to protect some finish.

Definitely worth thinking about 3D printing the bottom part. Material choice is a bit issue.

This works with sheet goods - not much 3D - and provides localized pressure on the part.

I’m busy right now but designing and machining one isn’t too big a project. I could help with the design and machining. The only real challenge is how to attach this to the Nomad. I not in front of mine so I need to stare at it for a bit… already have some simple ideas though.


@mbellon I’m having a hard time figuring out how this wouldn’t inflict some minor movement of the stock underneath. If it has enough pressure to hold the part down then how is that pressure not being forced in another direction (outward)?

Also, would you mind educating me on which is better in terms of accuracy: faster feed and smaller depth per pass or slower feed but greater depth per pass?

@mbellon I’m having a hard time figuring out how this wouldn’t inflict some minor movement of the stock underneath. If it has enough pressure to hold the part down then how is that pressure not being forced in another direction (outward)?

The idea - I’ve not used this device myself - is that the pressure is directed downwards. There still is tape. Between the two it prevents things from moving. You’re free to speak the vendor and ask them about it.

Which is better in terms of accuracy: faster feed and smaller depth per pass or slower feed but greater depth per pass?

With your part moving or not?


In this instance I guess specific to this thread where the part is moving ever so slightly, but I’m curious as to what the rule of thumb is in general.

In this instance I guess specific to this thread where the part is moving ever so slightly, but I’m curious as to what the rule of thumb is in general.

I posted an article, above and repeated here, that explains the general case:

Climb to rough, conventional to finish is the usual choice. Forces parallel to your motion as much as possible. Don’t push into the part.

Use the optimal feed and speed for the end mill and stock, ensure deflection is minimal.

In your case, there are many unknowns. If the part is moving, one would like to move as little as possible. If one goes too slow, one isn’t cutting - making a mess. Too fast, and finish can be ruined.

As you’ve observed, shallower - less bite - serves you better but the time goes way up. I’m not convinced that there isn’t a solution that is between the extremes that happens to minimize vibration and movement… I don’t know a simple solution except to experiment (under these circumstances).

In micro machining - where everything is vibrating like crazy - one just has to go very, very slow to get their part. You need overall production speed. It tape doesn’t work, we need to go to advanced fixturing (the pressure foot is one example but I can think of many others).

Fixturing will depend on how you’re cutting the parts - how many at a time? Can we move them around?


For anyone watching…I’ve been talking with @mbellon but don’t have a solution quite yet. I don’t believe that the stock moving is causing the (primary) issue now.

I got a new Onsrud 63-710 cutter which has a 1/4 shank and .125 cutting diameter. Very solid with almost no deflection. I also tried a different tape to hold the stock down. I ran my first pass over and over and over and while the stock did move ever so slightly, I don’t think it moved to matter in the end.

What I’m seeing from my most recent test is that .15 mm of stock is being left. I mentioned to @mbellon that I’m getting 4.7mm hole diameters but the dimensions of my CAD file have them at 4.85mm. In previous runs I noticed that if I overshot and altered my hole size to be 5mm then I would indeed get a bigger hole, but obviously that is too big to be usable. It’s almost like MeshCAM is creating g-code in inches despite my millimeter setting (I’m sure that isn’t the case, but that’s what it feels like).

I’m only doing a pencil finish at 1mm depth for pass #1. There is no place for a margin that I can tell with a pencil finish. While I’m waiting on Mark to get back with me I figured I would post an update and just see if anyone else had an idea to throw into the hat.

If you’re holding your part with tape you aren’t going to be able to get a better tolerance than +/-.010. It’s hard enough meeting +/- .003 on a Haas, with a vise programmed with mastercam using new endmills. This is why a good machinist working for Lego or Honeywell make $35-$45 an hour and are required to meet +/-.0005.

As simple as your part is, I would recommend you make a 2 part mold out of aluminum, make several parts rotated around a central sprue. You can machine the aluminum within +.002-.001 then polish the mold with aluminum oxide or super fine wet sand paper. Cast clear resin that can be pigmented with any color and use a pressure pot to remove bubbles.

(Edit: On second thought scratch this, you wont be able to meet lego tolerances without a high pressure injection mold.)

Another option since your part has a hole in it. Make 3 programs:
Program # 1 your part in a duplicate batch layout with many parts on one sheet. The program cuts everything but the center holes.
Program #2 The same batch layout only cutting the holes. (holes will be used to hold stock onto a fixture plate with cap screws and rubber washers.) Your stock is held down with clamps while milling the holes.
Program #3. The same batch layout as #2 but instead of holes, it makes studs with all material removed around them. Then ad a hole in the middle of every stud. Hand tap every stud hole to match the threads of a usable cap screw. (Make sure the cap screw will not interfere with your outside path.)

After you make your fixture you just use the #2 program to prep your stock for work holding, fasten it and all the holes on the fixture. Then use #1 program to finish the parts. You can also use tape on your fixture to keep your parts from the possibility of them shifting around the screws.


@dyelton and I have been emailing offline.

If you’re holding your part with tape you aren’t going to be able to get a better tolerance than +/-.010.

It’s well known that I hate tape and wax.

I’ve recommended a sea-of-holes bed to him previously. The part has holes in it - 3 in a good geometry - and drilling those first and using them to hold the part down would do it.

Unfortunately, these are very small parts and he wants 9-10 of them per 8x8 stock. A custom fixture plate would have to be created. Not hard, but it would allow this to go relatively fast.

Elsewhere @dyelton has said he doesn’t want to go to mold, at least not yet. It’s been mentioned that these types of parts are made with precision molds so volume is easy. Precision molds aren’t cheap…


1 Like

Yes mark, you have definitely gone above and beyond to help within the lines he has drawn for you. I think he is just trying to get a second opinion in hopes that it can be simple.

@dyelton You will have to dig deeper into the machining technical toolbox.

Remember the rules of physics, for every action there is an equal and opposite reaction. Yes, a mill with an up flute has downward force, but that force comes from it’s “press fit” rub on your part/stock. The press fit part/stock is experiencing an equal and opposite reaction just like the chip flying from the cutter. To expect that tape is going to nullify these forces just isn’t going to happen. You can dull the forces down like you have been doing but the effects of prolonging the forces at closer wavelengths creates another result of opposite reaction; one of a prolonged reaction at closer wavelengths.

I’ve been waiting to reply to this topic as I wanted to get as much data as possible but I’ve had a couple of PM’s from folks offering additional assistance and wanted to ‘clear the air’ so to speak.

@mbellon has been a huge help and I can’t thank him enough! Ultimately the cause was a few things being combined and I guess at this level of precision every little thing adds up to be a considerable amount.

The biggest issue was the inaccuracy of the end mills I’m using. Both the Onsrud 63-701 and 63-710 (one 701 is a .0625" and 710 is a .125") were slightly under the listed amounts. @mbellon contacted Onsrud on my behalf to get some details and the tolerance made up the amount that I was off. It is difficult to get a precise measurement from it since it has an odd number of flutes but after several measurements I figured out it was .115 which I later went up to .12 to compensate for a small wiggle I had in my part once it was milled. I additionally compensated for a slight inaccuracy for LEGO and adjusted from their spec’d 4.8mm to 4.85mm. Working at these small amounts is nuts but after a LOT of hours into this I feel good about having something that is repeatable.

I will now make a sample hole from time to time and get a measurement to see how much it is off and compensate for that. The same would go for wear on the end mill I’m using or for when I get a new one that may have a slightly different measurement but still within Onsrud’s tolerance.

I’m running an entire sheet at the moment so we’ll see how it goes. At some point I may very well make a custom fixture but for now I’m pretty happy.

It’s amazing with that little Nomad can do!

FWIW, I found that very shallow depths and higher feed rate offer a nice finish on the vertical wall. I’m doing depths of ~0.01 at ~14 IPM with nice results. I get the same accuracy by going a little deeper with a slower feed rate but the tool marks are much more noticeable. I’m sure this was stated elsewhere but you can never have too much info from lots of testing!