PRECISION SPINDLE METROLOGY Pt. 3: Measuring Error Motion

In which we start to get into how to actually measure spindle errors and process the data. Next time, we get to go into error separation.

[Dellpodder]

Your demonstration at 3:48 was so good! I'm blown away by the quality of this series that everyone can watch for free on YouTube! What a time to be alive.

[Trainwreck]

I really hope you cap off this series with a video or two showing the implementation of everything on your diamond turning lathe

[robertfontaine]

I am sure that I will come back to this series many times over the years. Thanks for going to the effort to do a good job oa this.

[realcygnus]

Quite nifty as always! I'm very much digging this series. & I strongly agree that Fourier anything(series, analysis, transforms & their inverses, discrete or not etc.) are among the most useful & powerful ideas ever. As they're used in engineering of ALL kinds. I'll never forget how totally blown away I was when some of these things started to really sink in. If only Jojo could see his rather simple ideas being used in our modern world as both conceptual AND actual machinery(especially software/hardware).

[robertfontaine]

Looking forward to the return of your air spindle and the applications of this series as well. thanks again... off to watch the fourier video.

[adamthemachinist]

I like the googly eyes on the millimess

[rafusy]

Awesome series! I'm making a list of all the topics promised "for later" - there is no escape.

[rodfrey]

Highlight of my week, thanks for putting the time and thought into these vids.

[lunchman]

Most certainly need to watch this again when all parts is out

[XxSteamStreamxX]

The question at 8:17, I'm not the best as describing things, but I am fascinated by these sorts of precision videos. My guess as to at least why its not a pure sine wave is because measuring it from a fixed point, it may be in line with the center of the sphere twice per revolution, but that means it will be measuring the artifact offset from the center twice as well. Your peaks and valleys of the sine wave would be correct for a pure sine wave but the middle portion would be off more the more runout your artifact has.

[bobbyshaftoe]

What a fantastic lesson in measuring actual error! Bravo!

[jemalibezeid]

Thank you very much for sharing all your knowledge! I hope you can get those busy guys (Adam and Josh) to make more microcast podcasts! Thank you thank you thank you!

[jasonhill]

I'm waiting for the collaboration project with Mr. Renzetti! This is mostly over my head, but well done!

[adamthemachinist]

Regarding putting an indicator on the shinny surface and it not telling you much about error, please pass that along to used machine tool buyers. That and the five puck test on grinders both seem to get used a lot but machine dealers to sell not great machines to the uninformed

[raindeergames]

Awesome.
... now I get to beat my frustrations out of me in the gym while listening to machining ASMR😊

We use LVDTS on our massive steam control valves upstream om our main steam turbines. And our condenser bypass steam valves. Because steam flow has such a direct an big impact on Reactor power be need to have high control feedback on these valve positions.

[pmcquay]

8:50 ish: is it because the axis of the measuring device does not always intersect the line through the center of the ball parallel to the axis of rotation?

[dittilio]

Furry signals = fourier transforms ❤

What's the title/source of Bob's thesis you refer to several times? Like the 22:26 "FFT result of the 128 rev. test"

[termlimit]

Which ADC did you use for your testing? Thank you, this series has been exceptional thus far.

[asingleoat]

you can also compute just the frequency bin corresponding to 1UPR and subtract off the sine wave of the resulting phase and amplitude; rather than doing the full DFT and IDFT. it's easy enough to use a library for the full DFT, but working with a single bin will build understanding of what the fourier transform is actually doing: a coordinate transform into a new basis via a bunch of dot products, one per component

[PeregrineBF]

WRT the low sample rate on the roundness tester: you might try software equivalent time sampling. Instead of trying to sample all the points you want in a single revolution, synchronize the sample clock to the encoder's output, with an adjustable phase offset. Say you sample once every revolution, that's not enough to measure anything! But if you take that same sample at very slightly different offsets each time the spindle goes around, you can get the effect of having one sample from each offset. You can get far more points than your encoder's number of steps, since you can have a precise & fast clock as your reference source.

This does have some issues: if the error motion changes (say, from heating) during the sampling some of the points will be before the change & some after. Also if your integration time for a sample is too high (slow ADC might take significant time to settle, not sample over a short enough instant) it'll average out the value over that time.