Experiment design: mill frame vibration

Hey, folks. I’m in the middle of designing a frame for a DIY CNC mill and I’d like advice on how to run an experiment to choose between materials.

The mill is designed to cut steel at relatively low speeds compared to most DIY CNC machines. Think 1000 RPM instead of 10,000 RPM or above you’ll get from most spindles.

I plan to weld steel tubing to make the frame. However, several people have suggested filling the tubing with epoxy granite. Sand is also a common suggestion. I have also considered buying old table saws and cutting up their cast iron parts to use as a ballast of sorts.

So, I’d like to run experiments to find the best combination for vibration absorption. My current plan is to cut 4’ lengths of steel tube, configure one for each combination and use a vibration app on my phone to record their movement when attached to a running spindle.

Here are the combinations I want to test:

  • epoxy granite
  • packed dry sand
  • oil soaked sand
  • cast iron ballast

What do you think?

You can buy an accelerator breakout board for just a few dollars, which means that you won’t be testing the interface between your order of magnitude larger mass phone and the part under test. They are so cheap they come in multi-packs. An example common use is klipper input shaping configuration.

Wouldn’t you want to drive them from a transducer instead?

Also, consider whether a local auto repair shop can give you cast iron swarf from brake rotor turning.

I suspect that four feet of steel tube alone will have more mass than my phone. Screens have gotten big but they’re not that big! :slight_smile: Do you think that I need to make each test with a longer tube?

The reason I was thinking of testing using a spindle in motion was that it would more accurately simulate the frequencies of cutting than a transducer. Perhaps that’s misguided, though, since it would be just spinning without any cutting forces.

With the cast iron swarf, is the idea to use it instead of granite with epoxy?

I was thinking about how sometimes when I have a sketchy setup, I can almost completely cure harmonics just by clamping an adorable little vice clamp (that weighs less than my phone) to the part.

But also, the circuit board in your phone isn’t glued to the case (probably). You’ll in part be measuring some or all resonant modes of the circuit board within the phone.

Here’s what I bought; $10 for 3 boards:

I’d expect that regardless of the source of accelerometer data, you would plan to take the scalar (dot) product of the three axes at every sample and do an FFT on the resulting waveform, and have samples across the range of spindle speeds you care about?

Yeah, it’s loading each individual flute that I think is the problem, unless you have a really bad spindle, which is its own problem… And a transducer will let you control the input consistently so you know what you are testing more reliably.

Honestly, you could start by just testing ringing from whacking it with a solenoid! The main reason to test it with a transducer is to look for resonant frequencies, but a 4’ length of tube won’t tell you anything about the resonant frequencies in your final frame. The clearest signal about damping vibrations will be a single shot. The main reason to whack it with a solenoid and not a hammer is repeatability. Come to think of it, you could set up a fixture to swing a hammer a fixed distance to get that repeatability and it’s probably just as good, or maybe even better, since getting a solenoid that can whack it hard enough might be a pain.

You would still take the scalar product, but no need for an FFT, just load it up in audacity (e.g.) and measure something consistent. Maybe the number of cycles for the amplitude to drop by 10x, or log(x), or something — I don’t know that you care whether you are directly measuring a time constant, probably more that you can compare them between samples.

(Incidentally, If I were trying to do this using analog sensors, I’d put a voice coil on X and Y, ignore Z (defining that as the length of the bar), use an op amp configured as a summing amplifier, and then look at the waveform on a DSO. But that seems like a lot more work!)

Well, that was the idea where I saw it mentioned in a comment on the epoxy granite garnet video I linked to on Social… But if you are also testing oil-soaked sand, you could test oil-soaked cast iron swarf I guess?

Ok, as a first go to see if I can get usable data I think I’m going to test these combinations by consistently (using a mechanism) dropping a hammer on 4 foot lengths of steel tube. I think I have an Adafruit Feather that has motion sensors of some sort on it but if not I’ll order one of the boards @mcdanlj recommended.

  • epoxy granite
  • packed dry sand
  • oil soaked sand
  • oil soaked cast iron swarf, hopefully from a free source like brake rotors
  • cast iron blocks securely fastened to the tube

Now that I think more about the cast iron swarf, I wonder whether some of my friends who run machine shops would donate swarf of various types for the experiment. I’ll ask around.

1 Like