Precision on a Budget: Finding the temperature sensitivity of my DIY displacement sensor.

[@JarheadCrayonEater]

I'm a former Metrologist and used to enjoy watching a gauge block grow on the granite surface plate during its environment soak time after we received it in the lab.

[@bubsbuilds]

@JarheadCrayonEater I dunno, I think once a metrologist, always a metrologist :) And I can absolutely relate to that sort of enjoyment!

[@lachlanlau]

As a person studying mechanical engineering it was pretty cool to see your test process and data analysis!

[@bubsbuilds]

@lachlanlau thanks and very glad to hear it! Also, not that I'm at all biased, but I think you've selected an excellent course of study!

[@arthurmoore9488]

@@bubsbuilds Troll comment of the day: Computer Engineering is better. It combines both Electrical Engineering and Computer Science. For real though, all types of engineering are great. My #1 advice is to not skimp on the basics. I "passed" my Calculus courses, but then spent years struggling because it turns out I really didn't learn everything I needed later on.

[@bubsbuilds]

@arthurmoore9488 If that's the worst trolling I'm subjected to for posting on YT, I'll consider myself incredibly fortunate! :) Couldn't agree more though (on the latter point of course...obviously the former is pure insanity...kidding....mostly...) I think engineering is more about a way of thinking and approaching problems than it is a specific discipline. A beautiful solution to a problem is beautiful regardless of whether it's in computer chips or machining chips. And couldn't agree more on the focus on the basics! My experience with respect to calculus has been frighteningly similar (and I'll just add linear algebra....that one has bit me more than I care to admit publicly). Love that you brought this point up, thank you!

[@ExtantFrodo2]

@@bubsbuilds Unlike other things, Math is a tool anyone can borrow from you, yet it never leaves your possession.

[@paulroberto2286]

You have a very soothing voice. Great for late night KZbin binges!

[@bubsbuilds]

@paulroberto2286 Never thought the comment section of a video would make me blush :) A very unexpected, but appreciated comment, thank you!

[@leonordin3052]

There are a lot of cheap low cte stuff. Invar expands or something with time so it is not that great anyways. I love this kind of content. A lot of clays are about 5ppm sometimes lower, pizza stone even lower. Alumina is good. Carbon fiber is extremely low in one fiber direction. Borosilicate glass which can also be melted by torch or microwave kiln is great. Wood in fiber direction is very low cte actually. And can be very stable indoors.

[@bubsbuilds]

@leonordin3052 awesome additional input, thank you! I feel like you've given me some homework, in the best possible way :)

[@patsauber4843]

Very interesting. Now wondering what it takes to drill a mounting hole in a pizza stone. I guess its possible.

[@lab_rabbit]

@@patsauber4843I'd imagine a masonry drill bit would work.

[@clement_roblot]

Glorious video! As an electronics monkey the PCB with though hole resistors is a bit spicy to my eyes. The mechanical aspect is well glorious! Love the content!

[@bubsbuilds]

@clement_roblot Thanks! And appreciate the relatively nice way of putting it for my ugly electronics making :)

[@BirnieMac1]

Magic Smoke Wizard* You guys gotta give yourself more credit for the wizardry of your work The magic smoke is magic for a reason

[@arthurmoore9488]

@@BirnieMac1 It sure feels like magic when you go down the rabbit hole of how hard it is to measure things accurately with electronics. Especially an Arudino ADC.

[@YSPACElabs]

As someone studying the dark arts of electrical engineering, the RC analogy for the temperature smoother was really good. It's funny that circuits are often used in EE thermal design because so many of the same principles carry over. A thermal resistance can be expressed as a resistance, where temperature is a voltage and rate of heat flow is current.

[@bubsbuilds]

@YSPACElabs I'm just glad to hear that my analogy wasn't too far off base! And I know, I also have always loved the analogies between circuits and fluid flow systems. That's one I've made use of many times over the years and have found extremely helpful!

[@mikemorrell7921]

You do know that a whestone bridge is not literally a bridge, right? Your resistor soldering looked like you where trying to make a suspension bridge :) Great video as usual. Love all the data. Your are my favorite mechanical monkey ever

[@bubsbuilds]

@mikemorrell7921 Hahaha, PCBWay messaged me asking if I intentionally assembled it that way. Took me a minute to even see the problem with it :) And aww, why thank ya! Means a lot coming from my favorite optical ocelot....if you'd prefer an alternative alliterative animal, please just let me know

[@InductorMan]

I like thermistors too, but I always use a simple voltage divider rather than a bridge. A voltage divider energized from the same voltage reference source that sets the scale of the analog to digital converter provides a ratiometric measurement, which produces a fixed ADC output value for each thermistor resistance even if the reference voltage varies (since both divider and ADC operate on a ratio basis with the exciting voltage). Usually on Arduinos etc the default configuration is for the ADC reference to be the 3.3V supply rail. Not super stable, but then it doesn't matter if you put the thermistor divider on the same rail. Using a separate voltage reference to excite a bridge as you've done here is not going to help matters, because now the ADC and bridge are excited from different references and relative drift between them causes error. There's one caveat which is self heating. Using a nice stable excitation source for the bridge or divider is nice to minimize variation in thermistor self heating, I suppose, but despite the Arduino's 3.3V not being super precise I think most voltage regulators on Arduinos are probably stable enough that self heating variations won't matter. Ideally self heating is only a couple tenths of a degree total, anyway. I am sort of coming into this half way, having not seen the first video yet. But are you using an LED/phototransistor optointerrupter as an analog position sensor? That sensor seems like it is going to exhibit an enormous temperature sensitivity, because the amplification factor of the phototransistor will be strongly temperature sensitive. Have you tried just subjecting an optointerrupter alone to the temperature swings, with your measurement circuitry attached, to see what kind of variation it produces? I feel like the only way to use a phototransistor as a precision analog sensor would be to somehow have it act as a null detector. For instance one could put a second, small LED in the field of view of the phototransistor which was NOT moved by the flexure. Then, one could alternate between illuminating the flexure-obstructed LED and the unobstructed LED at a couple hundred Hz or a couple kHz. Then, if one measured the difference between the two LED brightnesses, and used this difference to control a digital to analog converter or filtered PWM source via a control loop that changed the brightness of the unobstructed LED until the difference signal at the phototransistor was driven to zero, this would produce a signal that to first order was not temperature sensitive (or more specifically not sensitive to phototransistor gain variation). There would be issues with the changing luminous efficiency of the servo'd LED as its temperature changed as a result of changing drive current, but this could be compensated by shortening the on time of the servo'd LED as the drive current increased, to keep both LEDs at the same average current. Obviously this is complicated and electronics/firmware heavy... maybe not what you want to spend time on!

[@airman2468]

Good stuff, mate.

[@TheOpticalFreak]

And when working with plastics at micron level moisture matters!! 🧐☝️

[@bubsbuilds]

@TheOpticalFreak I definitely suspect that's true, good point! But I haven't seen any numbers on the swell for polymers common in 3d printing. Any idea if there's any good data out there (and where I could find it) that would allow me to get a rough idea of how much humidity will impact geometric stability of these sorts of things?

[@ethan_becker]

Awesome video. It is cool to see the error sources i this sensor being identified. It would be interesting to see what simple improvements to the electronics could be made to recuse their sensitivity to temperature. A point of feedback on the graphs. I love me some graphs, but the change in Y axis scale and range between every plot was making it difficult to follow. Would be nice to see them hold the same range if possible so it is easier to see the relative scale of the different curves.

[@arthurmoore9488]

That's a completely different branch of metrology, and electronics are annoying in that even when they're at temp they still can drift. Plus, need to be continually powered or you'll get different readings. The Marco Reps channel shows how deep that rabbit hole goes. The other thing is reference voltages on those small microcontrollers are, frankly, crap. Which is why they almost always have a way of using an external reference. However, if you only need millivolt precision things are much simpler. Just use a $10 reference voltage source and ovenize everything. Aka, use a PTC heater to bring everything up to 50C or so in a Styrofoam cooler.

[@bubsbuilds]

@ethan_becker Thanks! That would indeed be interesting....hmmm...wonder who I could ask about ideas for that? 🤔 :) My assumption is that the most 'bang for the buck' would be in stabilizing the current flow through the emitter LED, but as the video clearly showed, I've been wrong before! Great point on the plots, thank you! Looking at it now I can absolutely see that. I think I was too deep in the forest staring at trees. After so many days of staring at the data, I took the self-familiarity a bit for granted. Definitely something I'll try to keep an eye out for on future plot-including future videos, much appreciated!

[@BirnieMac1]

As someone on the tail end of their engineering degree; your work is literal engineer goals I spat out my breakfast giggling at “there’s plenty of data to prove me wrong”

[@bubsbuilds]

@BirnieMac1 I can't even express how nice that is to read, thank you! And congratulations on nearing the finish line!!

[@TheChillieboo]

Great balance between detail and pace from my perspective, very cool video!

[@bubsbuilds]

@TheChillieboo Glad to hear it, and thank you!

[@jemalibezeid5930]

Good stuff! Keep up the good work!

[@ExtantFrodo2]

Awesomeness! Subbed and instant fan. It's always nice to find people who delve deep into the nuances of their passion. Regarding the subject of this video, have you given any thoughts to mechanical temperature compensation? Like pendulum clocks utilize the difference in expansion coefficients of wood and brass to auto-regulate through daily and seasonal temperature changes. The brass bob which rests freely on the bottom regulation adjustment nut expands upwards as the wood rod expands downward. Likewise with a balance wheel, the bimetalic rims are only attached to one arm each. As the arms heat up they expand, while the bimetalic rims draw in towards the center thereby maintaining a constant moment of inertia through a wide temperature range. In some cases one might simply employ V shaped arms since each arm would expand or contract by the same amount the tops of the arms would remain level to each other. It's something to consider since in some cases the mechanical versions are more reliable.

[@bubsbuilds]

@ExtantFrodo2 Thank you so much, the kind words are very much appreciated! I'm glad to hear that my love of engineering comes through, and even more glad that it finds its way to folks like yourself that seem to share that love! I was not aware that pendulum clocks utilize that trick, and likewise with maintaining the moment of inertia in bimetallic rims.... that's awesome!!! I've used tricks similar to that for work projects previously, utilizing the difference in expansion between steel and aluminum to null out the expansion in a frame is one I've used quite a few times, but I definitely hadn't considered something like that for this. The main thing I would worry about in trying to adapt it for a sensor is the difference in thermal properties that would change the time constants for the two materials causing less predictable drift. But plastics don't have great conductivity, so in reality this sensor probably tends to hold gradients anyway, so this sort of compensation approach may indeed prove a net benefit. I'll absolutely have to ponder on this one a bit more. Thanks for the awesome extra tidbits and suggestion!

[@ExtantFrodo2]

@@bubsbuilds You're very welcome. You deserve it. Keep up the good work.

[@ExtantFrodo2]

@@bubsbuilds Hey, just a heads up. I searched for you on ROKU and found no trace. The closest was a bdubbuilds who definitely wasn't you.

[@bubsbuilds]

@ExtantFrodo2 Oh really? Hmmm, not sure if that's something I can have any say over, but I'll definitely give it a look. Thanks for the heads up! Also...not gonna lie, kind of curious now what "bdubbuilds" is all about :)

[@Chazz155511]

So glad I subbed.

[@bubsbuilds]

@Chazz155511 me too! :)

[@marcorademan8433]

The opamp has input offset voltages and currents that are temp dependent. If you vary the temp (even in a controlled environment) you will see problems. An instrumentation amplifier (built from 2 opamps) should do the trick. If those offsets are not nulled out, I would not trust the results you got. Input offsets for a cheap opamp (like the TL08) are on the order of 10s of mV, which will definitely mess with wheatstone bridge measurements. Zereo-offset opamps are also an option.

[@bubsbuilds]

@marcorademan8433 VERY good to know, thank you! I'll definitely have to look into that a bit more, much appreciated! Any specific instrument amps or zero-offset amps you recommend for general purpose builds like this?

[@marcorademan8433]

​@bubsbuilds Thanks for the reply. I would recommend the OPA4188 for a quad package zero-offset opamp, and the INA818 instrumentation opamp. You can also build your own instrumentation amplifier with 3 opamps, but you'll need some high precision resistors for that and probably a very low input offset amplifier, so it's best to go with the zero drift one. Perhaps google "non-ideal opamp characteristics" and look for a TI app note or two to get an idea of what temp issues you may face with opamps.

[@bubsbuilds]

@marcorademan8433 Awesome, thanks! Building my own instrumentation amplifier is probably not something I should dive into (you saw those resistors...right? :) ), but definitely appreciate the suggested googlings and part reqs, thanks!

[@theknio1]

Great video! 0:32 wouldn't the temperature affect the whole center mass up to the sensor? looks to be about 60mm which would match up with the experimental result a lot better. 5:52 going to all that work to build a board and then using the arduino ADC 🙀 just use the differential channels on an ADS1115

[@arthurmoore9488]

Arduino ADC with the built in reference voltage at that! I'm glad he at least recognized the temp-co errors in the electronics. No shame in not knowing what part to use though. I've been there before and trying to find what to use is not fun. For example, I had never heard of that ADC before now.

[@bubsbuilds]

@theknio1 Thanks! I don't think that's a contributor. Since the frame is made of the same material as the moving element, there wouldn't be any differential CTE. So the sensor should be moving along with the growth of the center element, and I wouldn't expect that 60mm to result in any relative movement between the 'knife edge' and detector. As a bit of extra support for this, when I did do some fitting during the initial temp cycling (prior to adding the electronics into the cooler), I did get a drift in the ballpark of the expected range of 1.5 um/degC...but my confidence in that data and resulting numbers would be quite low, so grain of salt on that tidbit for sure. Also, I've certainly been wrong before :) Haha, no clue on the second point, but I'll definitely have to give that dedicated ADC a look, thanks! I tend toward Arduinos and similar since they are a 'known entity' for me, and also provide serial and/or wireless comms. Also allows me to be lazy and re-use old code. But if in looking into that ADS1115 proves to be an easy swap, I'll definitely consider making the switch down the road! Thanks for the suggestion!

[@theknio1]

@@bubsbuilds ah, makes sense about the frame on the sides too... assuming it was only held by the lower bolts and the rest was free-floating. I mentioned that ADC in particular not because it's the best but because it seems to be a really common arduino accessory. Adafruit has it as a breakout board with code (any many other clones)

[@KarlMiller]

Nice work.

[@stefanguiton]

Great work!

[@TenchoTenchev]

Maybe adding a Peltier element with temperature control will get you closer to industrial grades.

[@bubsbuilds]

@TenchoTenchev My past experiences with implementing thermal control on stuff like this have had me chasing my tail quite a lot to make them work, so I'd be hesitant, especially for the sensor body itself. It can be hard to get the heat in and out consistently across the device, and I think having thermal gradients can be even more problematic than just temp drift alone. But it has been a little easier on the electronics side, and given what a large contribution to the error the electronics proved to be on this, temp controlling the electronics enclosure may indeed be a worthwhile idea to pursue!

[@accidentalengineering]

LM358 isn't a rail-to-rail opamp, the output won't go to fully to GND or VCC, did that affect your measurements?

[@bubsbuilds]

@accidentalengineering Great point, thank you for pointing that out! It didn't impact my measurements for this testing, but it does mean that my stated output range of 0 degrees C to 100 degrees C isn't actually true. It will probably only output over the range of about 5 degC to 85 degC (losing more on the top end because of the non-linearity in the thermistors). Is there a rail-to-rail option that you recommend for projects like this that would have been a better candidate?

[@maximthemagnificent]

Fantastic.

[@MorsAdvancedTech]

Im not shure...using a wetstone bridge for a thermocouple is kind of overkill you could have gotten by with a pullup resistor and done the calculation that way

[@jaro6985]

The bridge lets you narrow the measurement range, so more resolution within the temperature he wants to measure. If you have a high 24 bit ADC, then it probably doesnt matter.

[@szymonjastrzebski2909]

​@@jaro6985 thermistors are used because of their high sensitivity. Using a thermistor with a properly matched pull-up resistor for the range I was interested in, gave me ~0,04°C resolution with just 12 bit ADC.

[@jaro6985]

@@szymonjastrzebski2909 You can see at 6:02 his measurements are well under 0.01C, whether or not that is needed, no idea.

[@GNARGNARHEAD]

aaaand subscribed, 👍

[@bubsbuilds]

@GNARGNARHEAD Glad to have ya on board!

[@Hubert8299]

What are these sensors used for? I mean where is needed to measure microns?

[@bubsbuilds]

@Hubert8299 apologies in advance for what I suspect is about to be a long-winded reply :) But great question! For this sensor build specifically, I built it to measure the runout on a rotary table I was building. In general, the majority of use-cases I've had for displacements sensors measuring microns (or below) has been for characterizing error motions in things like machine tools. If you want a boring (but in my opinion interesting and extremely useful) read that includes references to quite a few uses of capacitance probes for these kinds of measurements, the ASME B5.54 (www.asme.org/codes-standards/find-codes-standards/b5-54-methods-performance-evaluation-computer-numerically-controlled-machining-centers/2005/print-book) standard goes through characterization of CNC mills. The general rule of thumb I've usually gone by is that you want your characterizing tool to be 10x better than what you're aiming to characterize. So if, for example, I wanted to verify the stated accuracy of my Elegoo Saturn's Z-axis motion, which is claimed to be 0.02mm, I would need a sensor that's good to 0.002mm/2 microns. Some places I've worked for/with have had quality assurance rules that reduce this req. to 4x, but that would still require a sensor that's traceable to 5 microns if qualifying a machine that's supposed to be good for 20. Thanks for bringing up a great point though. Chasing decimal points without asking "why do I need them" is a great way to waste a lot of time and money!

[@jaro6985]

Measure surface finish of parts would be one application, certain molds, anything to do with semiconductors.

[@arthurmoore9488]

I've spent far too much time working with imperial threads so, here's how I think about it. Coming from someone who does not need that much precision it's scary how fast microns become relevant. 0.001 inches = 25.4 micometers. In my experience 0.010 in can be the difference between barely engaging and having to use a wrench. Which translates to 254 micrometers. Still a big number, but my tollerance has to be smaller than that for things to actually work. 3D printers typically have a resolution of 100 microns and that's what I shoot for when modeling. So just in 3D printing threads for protective caps I'm already in the double digit micron range.

[@gtsiam]

Definitely keep the rambling. I'm starting to despise people dumbing things down for "retention". Also those resistors have to have been bait. You can't convince me otherwise.

[@bubsbuilds]

@gtsiam I will gladly let that be the narrative on the resistors :) And love the feedback, thank you! If there's anything I'm bad at, it's retention, so I'll gladly keep on rambling!

[@clayflannery2608]

What is the thermal expansion of Antonio Brown?

[@bubsbuilds]

17