Everything a thermometer, everything a spring

[@Spirit532]

This is how some high end machine manufacturers measure spindle elongation directly in the machine! If you look closely at something like a Kern, you'll see a little Г-shaped lever holding a nub a couple mm away from the spindle nose. That's this thing. As others have said, the electrical noise is coming from the cheap switching supply you're using. For this low power consumption stuff, it's best to use a linear supply. You can use a 12V switching supply and a linear post-regulator with a bit of filtering(just capacitors should do). Maybe the jumping on the air bearing is due to slightly varying pressure/leak paths changing?

[@reps]

delicious stuff

[@cylosgarage]

The honorable Mr. Reps! Thank you for stopping by sir :)

[@serzl4660]

I have noticed you use switch mode power supply in the last test to power something. They are really noisy - will pollute with high frequency rf noise both your DC rail as well as the environment they are at. Consider using linear power supply as well as terroidal transformer for powering the whole set up. A quicker test could be to use batteries - say a 12V car battery to test the theory and see if it improves the noise.

[@RonStoppable102]

Agreed. Those power supplies are super noisy

[@donepearce]

@@RonStoppable102 I was going to say the same. I would stick with the power supply to get this set up, but any measurements should really be battery powered. Also what is the noise floor of that amplifier? If you unplug the capacitive gauge, does the noise vanish? If not, you can reduce the noise by dropping the measuring bandwidth; you will win 3dB for every halving. Finally, there is definitely a settling time associated with this stuff. Get it all set up, but don't bother even looking at it until the next day.

[@dougaltolan3017]

Then theres the emf at the capacitor from probe coax screen and green earthing wire. Better to earth the puck with a short wire from the probe case.

[@hasanzakeri8708]

I like your idea of “massive deflection”!!

[@BloodyMobile]

Only found the channel today, watched 2 videos and I'm already a big fan of this stuff. This is just incredible, not just how sensitive the probe in this video is, but also the "visualization" of /how/ sensitive it is.

[@selbermacher1267]

Jump ! This was the first video of yours i've seen and i'm impressed this worked, especially since you did it all by hand. Congrats!

[@bdykes7316]

Best of luck. I also find these thermal and elastic phenomena endlessly entertaining.

[@L0RD_F00G_the_2st]

It's quite mind blowing how accurate it is

[@sannyman13]

Dude, this demonstration is insane, never even thought that it is possible to literally bend steel by heating it so little... My only concern is the capacitive probe may be affected by the dielectric properties of difluoroethane in the cooling demo.

[@theromihs]

I don't have ANY experience with these types of sensitive measurement instruments, but I do understand electronics very well. That noise floor which you are seeing is very likely due to the green "ground" cable connected to the slug. The relatively short length of wire will experience some induced emf from all sorts of sources around you (that DC/DC supply for example). Keep all mobile phones and wifi routers as far away from the setup as possible.

[@piranha32]

Did you try listening to the signal coming from the probe? Some components of the noise on the oscilloscope plot can be resonances or high frequency periodic disturbances, and they will be easily identified after conversion to sound. Doing FFT on the signal will do the same, but connecting to a speaker can be more interesting. The probe can pick up noise from surrounding power supplies and other electronics. The floating of the reading on the anti-vibration table could be caused by differences in electrical charge between isolated components of its mechanical assembly relative to the ground. Make sure that all metal parts are solidly electrically bonded, and grounded. Just make sure to not create an acoustic feedback between the gauge and the speaker.

[@Ryan-yg5uq]

Excellent demo. I think if you want better resolution with that probe you may have to re-configure the 4810 to run it at a closer standoff distance. Usually this is done at the factory, but it's done with jumpers and BCD switches on the PCB, so provided you remember the original settings, you can just reset everything should things not work out. If you estimate the probe diameter to be approx. 0.5mm and the standoff to be 100μm, then depending on the set bandwidth (user selectable, but assuming 1kHz or 10kHz) the catalogue lists an rms noise of about 2-5 microinch for a similar probe and setting. If the board is reconfigured to run the same probe at a standoff of 10μm, the noise is reduced to about 0.1 microinch. The catalogue also suggests selecting a probe with the largest sensing area the target will allow to minimise noise and get the best resolution. You get lower spatial resolution with a larger probe as it effectively averages the target surface, but I don't think this would be an issue for your application. In your case using a larger sensor would still require adjusting the board, it would have a proportionally larger standoff with equivalent noise as it is. Also, in 'Know your machine tool' by Tim Sheridan, Section 6 describes a capped probe test used to determine the minimum noise level of a capacitive probe system.

[@ebrewste]

Really nice demos on the temp based distortion!

[@omars_projects]

Really cool stuff, keep the updates coming :)

[@はいこれはロボ子の婚約者]

splendid! i'm blown away!! this is witchcraft!

[@dan4653]

What's the noise floor of he data acq system? Try short the input and take a measurement. Typically high bandwidth measurements have higher noise floors, as noise is proportional bandwidth. It's very difficult measure small signals at high bandwidth. You may be bumping into the 1/f noise limit of the system. Might be worth running an fft to see what the noise floor looks like. You may still be able to see your oscillation of interest on the bearing in the fft, which may tell you what you need to know. All fun stuff!

[@extropian314]

10:20 Re the calculated distance: Isn't the noise floor of 20mV corresponding to about *400* microinches, not 4? (And then on the vibration isolation bench, doesn't the ~45mV "unknown deviation" correspond to about a milliinch, not microinch?)

[@dougaltolan3017]

Put it in a Faraday cage.. The electrical loop of probe coax screen and green earth wire wi pick up emf from everything. Maybe earth the puck with a short wire from the probe case. On the lines of everything is a spring... Whats the fundamental frequency(ies) of the anti vibration table? It's going to be low, about the period of the oscillations?

[@ashpalmer1365]

Try a ferrite bead on the coax to the capacitance probe and on the power supply and data cables.

[@carltauber2939]

Hi Cyrus, you have some nice toys there. You need to do the arithmetic to determine if what you're seeing is Johnson Noise. Recall that the Johnson equation is explicitly dependent on bandwidth as well as temperature. This is why your various gauges are heavily low pass filtered in order to give meaningful measurements. It may be impossible to detect the vibrations you're looking for with a broadband measurement. All is not lost however, if you connect the output of your cap gauge to a spectrum analyser you can effectively make narrow band measurements at various frequencies. Can you infer the target frequencies by examining the defects that you think are being caused by vibration? Also. it looked like you used a lot more refrigerant on the granite than on the steel bar. The difference in the time to return to equilibrium may be because you took more joules out of the granite than the steel.

[@omars_projects]

Thanks for this, it might solve an irritating noise problem I've been running into

[@DanielHeineck]

Yeah the Digilent USB scope he's using can do spectrum analysis (FFT of the incoming data stream). There's a few settings in the software that I found can really help with noise, and I genuinely worry most of the noise he's seeing isn't coming from the capacitance gauge as much as the digilent. Would depend on what sort of Vp-p are coming off the gauge amplifier. Another issue will be brownian motion in the gap. And tribolectric effects in the cabling/instrumentation. But my bet it's either the gauge amplifier or the Digilent scope. The shape of the noise will be telling.

[@carltauber2939]

@@DanielHeineck You make excellent points. I had assumed that the gauge amplifier has a good low noise front end, is designed for broadband operation and that it's output voltage is adequate. I also assumed that Cyrus's repair to the sensor left it fully functional and properly matched to the gauge amp. All of these assumptions need to be examined. You also raise the issue of Brownian motion which is a mechanical analogue of Johnson noise, thermal motion of atoms and molecules versus thermal motion of electrons in a circuit. But I don't know how to properly account for iti.e., is it already included in the Johnson noise, or does it need to be added. And, as you say the issues of cabling, shielding and grounding need to be considered.

[@carltauber2939]

@@omars_projects You're welcome, I'm glad you found it useful.

[@andrewphillip8432]

Good points here. Fortunately, based on product literature for similar cap gauges I’d guess that there is a long way down to the noise floor. The accumeasure 9000 advertises a resolution of .75ppm of full scale range per root hertz. So for a 1kHz measurement bandwidth and 25um range that would be sub nanometer

[@rigidrobot]

Have you tested the isolation table? Once a top of the range one at a national laboratory failed a simple glass of water test. Can be difficult to find the source of the problem.

[@cylosgarage]

yep, please see kzbin.info/www/bejne/pWSrmWCImJ6FZsUsi=jA5U3tLAhi6LP7E5.

[@realcygnus]

such nifty gear !

[@rolandjollivet38]

what about the hot air from the PSU wafting underside the granite plate...

[@robertfontaine3650]

Absolutely amazing. The drift is obviously solvable as you didn't have it in your first experiment. Is this the same technology that they use for global earthquake / explosion sensors? Maybe one of those guys can give you some clues. Could it be radio frequency noise with the cable acting as an antenna? Does shielding hurt or help the jiggle? I'm still at the point where I'd just like to lap a steel plate flat enough to see stiction. Your enginerding seems like magic.

[@Mister_G]

Fun stuff! Have you considered precession due to slightly unblanced workpieces as a source of your unwanted movement on the air bearing spindle?

[@Taygetea]

did you check against like.. cars going by outside for that unknown drift

[@spikeypineapple552]

Would love to know the prices on this stuff!

[@cylosgarage]

This model can be had for less than 100 dollars on Ebay right now

[@chrisyu98]

the noise could also be noise as in sound waves, your voice, Air conditioning, etc..

[@chrisyu98]

for more fun flood the gap with butane or propane. the change in dielectric constant should change the reading.

[@movax20h]

Try turning lights off. A lot of modern lights have bad emi.

[@prbmax]

It would be interesting if you would take your set up to some area of non-human activity just to see what you instrument reads. There appears to be some oscillation on your vibration isolation table. This in and of itself is a troubling problem. Hope you inform us of the issue should you discover it.

[@CATANOVA]

"Everything a thermometer, everything a spring" and everything 'resonates' with its own natural frequency when touched or excited by external stimuli. kzbin.infoVhSglueCYzs 😁

[@cylosgarage]

@@CATANOVA that is indeed a behavior of springs

[@CATANOVA]

@@cylosgarage Check out the link in my previous response and maybe give it some thought as to whether it may have any significance. In very high end CNC machining one can test the spindle by way of striking it and measuring its natural resonant modes. From this data one can formulate a map so as to machine in between any resonant nodes.

[@BigBoss-rh7zq]

May be not very stable setup