Nomad with deep DOC, shallow WOC

Yeah, I’ll definitely give that a try. I’ll wait for the compressor to show up and see how it does with some actual chip evacuation and if that doesn’t help I’ll try lubricants. If lubrication helps, I’ll set up a droplet system.

1 Like

Still stock anti-backlash lead nuts?

Yep. Would that be a problem even here though?

Ide put my money on it, especially with them being the low-cost axial type. The plastic wears over time and depends on the spring force. There have been some that replaced them with brass ones and found better “rigidity”.

With movement like this and/or low spindle power, imo its better to cut with a low axial/high radial. You’ll have less deflection and be able to take a bigger chip. I’ve seen this increase tool pressure to where it stabilizes the cutter. 67% step over is my favorite and when it can be pushed over 90% I increase step down.

This is all my opinion from running many different types of machines 80 hours a week. Sorry its not more science-y

2 Likes

Hmm, I should be able to test that with my luggage scale as well. I should be able to see how much the bed moves relative to the spindle when I apply varying amounts of force to it. I’ll give it a go.

I adjusted the endmill rake angles in Milling Analysis 0.9.1 to get it to match your HSMAdvisor results as shown here:

That would be really nice if that small chip thickness was achievable (it’s 1/6 what @Julien, @Vince.Fab, and others would likely use).

You might want to try reducing the axial DOC to smooth out the cutting forces as shown here:

Using Audacity, as shown below with you video, might also be helpful in analyzing forces and chatter.

I don’t know German, but isn’t nass (water spray) kuhlung (cooling) recommended for that endmill?

OOPs I forgot to mention that less cutter stick-out from the spindle’s connection to the gantry might help too (elevate workpiece too?)

5 Likes

That’s a smaller chip thickness than I used on a 0.020" Harvey 3F in 1018 steel, and it survived full width slotting.

1 Like

It’s what the manufacturer recommends - it seems they should know best - and it helps keep cutting forces down, which is good!

Second using Audacity. Once you figure out all the expected noise spectra it works great finding abnormal sounds.

1 Like

That looks like a really cool tool. Where can I get it?

That cut actually was at 7mm, I brought it right down for testing to see if it would help.

It works when you get down much lower, like 1-3mm. I think because the chips are much smaller and easy to clear, more like small flakes than the thin strings I get at 7mm.

Yes, I should actually know German but thought I “knew” that this meant “none” (which would have been “kein”) when really it means “wet”.

I’ll try with the air blast first and try adding some WD40 or ethanol or something if that doesn’t help.

I can do 25mm or so no problem, I’ll try it.

Are the noise spectra constant (or easily calculable) once you figure them out? If so, that could be really handy. I could trigger the E-stop after X seconds of chatter for example.

It is, you should ask @spargeltarzan about it.

Audacity is a lot faster displaying waveforms in real time (when recording) than spectrograms. It’s spectral analysis is non-real time. I think you would do better using your ears to detect chatter.

Did you ever check spindle and system runout?

I’m a software engineer, I don’t need no Audacity :slight_smile:

The idea would be that the machine could be left unattended with a little bit less to worry about, so using my ears isn’t an alternative there.

So are the noise spectra constant or easily calculable?

Spindle was ~7µm, haven’t checked on a tool though.

I’ve used power spectral density for machine analysis before.

The tricks will be figuring out (or testing for);

  • What fundamental frequencies you expect to be in the signal, flutes * rotational speed as a baseline, this changes quite a bit on a Nomad or Shapeoko
  • What harmonics of those fundamentals should be there and to what extent
  • Whether there are specific frequencies which are associated with machine resonance that are always present or always bad if detected

I had previously thought about putting a contact mic onto the spindle to isolate the airborne noise from things like a dust extractor / compressor.

Spectral analysis is done by performing Fourier analysis on windowed segments of sampled data. The number of samples and the type of window used determine the frequency resolution. IMO using accelerometers to sense the vibrations that generate the noise would be easier and better. You can use your cell phone to try that out (if it has adequate frequency response). It’s possible that a simple amplitude threshold would be adequate to detect the onset of chatter.

I’ve done that, the spectrum is still quite messy. I ended up using hall sensors and magnets to determine router speeds instead. I suspect that filtering or spectral analysis of accelerometer outputs would be cleaner and more precise for vibration sensing. Accelerometer IC’s are inexpensive and are available with either analog outputs (for old guys like me) or digital outputs.

Hmm, I wonder if I could have the first few minutes be a “learning” period of what’s acceptable (because an operator will be present) and then look for strong deviations later on. This and this show that chatter peaks in the spectrum are fairly prominent.

This also has an interesting setup where accelerometers are mounted to the spindle itself. That could be really useful data.

Sounds like AI to me! - you’d be joining Bob Warfield in that endeavor for CNC’s :wink: However, since chatter is so easily detected audibly, maybe it makes sense to use an old fashioned frequency analyzer (with microphones and/or accelerometers) that’s better matched to human hearing like TruRTA’s. The free version provides 1 octave frequency bands and there are versions up to 24 octaves (which is real nice). His WinSpeakerz program is really nice too - if you’re into speaker design/analysis. He’s the kind of guy that would probably be interested enough in what you might be doing to provide free advice too.

Do you think that it makes sense to compare conventional milling to climb milling to see if it makes a difference? How about changing spindle speed to see if you can avoid machine resonances? I know that it helps on my router table.

Since the aim was for good surface finish, I did some tests with some different reasonably new endmills that I had available. First lesson: Really nice finish is not so easy to achieve :frowning:, at least not dry.

The best that was possible on a belt-Z stock Shapeoko was this:

Not great by any means… This is ap 20 mm, ae 0.08 mm, vf 600 mm/min, 25k rpm with a cheap 3-flute. But at least it didn’t sound horrible, which it does when increasing feed, just like @Moded1952 said.

Watching it work close up shows that there really aren’t any chips to speak of. This is more burnishing than milling. And it looks like the 45 degree helix is transporting some fraction of the chips/dust through several turns, it’s ejecting material not only backwards (as it should) but even forward. Not good.

2 Likes

Hmm, maybe @Vince.Fab can give us some recommendations? Maybe Proven Cut has (or will have) recipes for surface finish on a Shapeoko? :smiley:

1" insert facemill run as flycutter going live soon! This is Z plus however, and it makes big difference.

The belt drive Z is super noodly so most end up facing at a 45 degree to load the v wheels more evenly to try to stabilize the cut.

Dry is possible but a little wd40 will help with the heat (1" is high minimum sfm) and give it a little more shine.

6 Likes