วีดีโอ

@glennmcgurrin8397 well then, have I got good news for you! :) The receiver side of the opto is just an NPN transistor, and the output voltage (in my setup) is just the voltage drop across the load resistor. So you can change the voltage range pretty easily by just changing the Collector-Emitter voltage.

@AK-xb9ke Thanks, and glad to hear the model is helpful for ya! Although sorry you have to deal with my terrible modeling practices :)

@ulamss5 Not sure I'm correctly envisioning what you have in mind, so please feel free to correct me! But my read is that you are asking about having nesting cylinders with a running fit to allow them to slide relative to one another....? If so, it's certainly possible to build a sensor this way (for example, these spring-loaded LVDTs: www.omega.com/en-us/control-monitoring/motion-and-position/displacement-transducers/p/LDI-G19 ), but they have some limitations compared to a flexure-based design. The first being that there will be a limit to how tight of a running fit can be reasonably achieved. Even with precision ground parts, you'll still have a clearance of 10s of microns (and this would be a fairly pricey set of parts in comparison to something printed). So there will be some unconstrained lateral movement in that gap, and it will also allow for some pitch and roll rotations (the amount of which depending on the length of the bore.) And because it's not constrained at all in that gap, the movement around that gap will be somewhat random and so can't be calibrated out. You could add a preload to force the probe to always sit on one side of the gap, but that is only going to make issue 2 worse... The second issue is the friction. Unless you go with something like an air bearing, there will be a non-negligible amount of bearing friction, both static and dynamic. The static friction is likely to cause some 'stiction' any time there is a change between static positions, and the dynamic friction is going to result in additional hysteresis. Depending on the application, these can be dealt with, but they can definitely present problems. For example, in applications where I've used sensors like the LVDT referenced above, we would only use them for measurements from a single direction. So they work just fine for something like bed leveling, where you can lift the sensor, move to the point you want to measure, and then lower the sensor to collect a measurement while moving only in negative Z direction. Then lift, move, repeat. However, they wouldn't work well if you wanted to leave the Z in position, and move the sensor across the surface to collect values continuously. If I totally missed the mark on what you meant, and just ranted for no reason, please feel free to say so! :)

@yousefosman8094 the current set of mounts are pretty stable. When intentionally cranking up the loading, like when I put my full-sized camera on it, there seems to be more deflection coming from the Prusa Y-axis carriage than from the mounts...which I suppose makes sense, given that the carriage is a narrower aluminum cross-section than the mounts in places. So I'm not sure that would buy much in performance, and it may make the fan clearance issues a bit worse. But if you have a specific offset and angle combo you'd like to see, I'd be happy to run it and upload the model for ya! I'd just need 1) the offset from the bed edge in Y, 2) the height offset from the bed in Z, and 3) the pitch angle, thetaX, for the camera mount. Cheers!

Thank you, that sort of feedback is VERY helpful! It's hard for me to gauge, since by the time I'm adding music I've heard the words a good many times already in editing :) Unfortunately there's no way for me to fix it in this one (YT doesn't allow for 'revised' uploads, as far as I know). But it definitely with help me in setting levels going forward. Really appreciate it!

Haha, thank you YT for finding me my kind of people! :) And seriously! Why the hell is it so hard to pick a naming convention! Thanks!

Oh, don’t even get me started on naming! I use a ton of data for my analysis work and every bit of it has been renamed by IT to meet THEIR standards. Note 1, the IT standards don’t match industry standard names used by the people relying on the data. If IT was building systems for the military they would call the Army the Land Force, and the Navy the Sea Force, ignoring Navy planes because that doesn’t make sense as there is already an Air Force. Renaming the Marines would involve months of discussion because they go about everywhere but space. Note 2, “Standards? Sure, we have lots of them.” Unfortunately, different groups within IT have different standards, and no interest in standardizing on one standard. IT has fairly vague understanding of what any of the business terms mean, so after they’ve renamed something to their standard they write a vague description of it - obscuring what it really IS. I need to expand my rant out to an article one of these days. Thanks for reading this far!

Why thank ya, good sir! Yep! I bought myself some steel stock to make some more rigid fixturing. My plan is to upload some follow up tests as shorts...although it'll depend on when I get the chunk of time to turn that steel stock into something useful. Do I get to attach it to the bed on the Bambu? If so, I think I like my odds :) .... but guessing that's not quite the challenge you had in mind. Although hell, with their machines, I'm not sure my GoPro has the frame rate to keep up!

Hahaha, well, now I can't unsee that! I guess I'll say it was an intentional public service thing. Given how overwhelmingly male my viewership seems to be, probably not a bad idea to make sure they know what one looks like :)

@@bubsbuilds hahaha love it

😁

If it makes you feel any better, I've spent much of my life wishing I could write some half-decent code :) But thank you, that is really quite the compliment, and I love to hear that you're interested in expanding to hardware!! I am a mechanical engineer, by both education and current job title. As to the 'how', hmm, I can do my best guess, but please add a 'grain of salt' since it's purely a guess from self-reflection here. I think part of it comes from a combination of being naturally curious and having that same sort of desire that you do about building hardware. But the biggest part, in my opinion, has been the incredible good fortune of crossing paths with a lot of people MUCH smarter than myself that were gracious with their time and attention. There are few things I enjoy more than being the dumbest person in a room, and I've found that if I embrace my ignorance in them, many of those folks are more than happy to indulge my curiosities. To be honest, that's a huge part of why I decided to try sharing the projects I dive into in my spare time. I by no means have the depth of knowledge or skill of those I've pestered with questions over the years, but my hope with this channel is very much to be able to pass along at least SOME of the things they've been kind enough to share with me along the way...and hopefully also convey my passion and gratitude for it as I do. Sorry, very long-winded and non-specific response :) But in short, get the printer, try the thing, ask the questions...just like you already are! I'm happy to try my best to answer questions along the way (if I can), but I may ask you to review code snippets in exchange!

Thanks! And yep, I used the Fusion360 "Simulation" Extension (they bundled the GD stuff inside of the sim package at some point.) Unfortunately it's VERY pricey to buy a license, but they have a two week trial for it. If you do decide to give it a go, I'd be curious to hear how it goes for ya. And if you run into any questions or problems, I can't promise I'd be of any help, but feel free to send em my way!

You get me :)

Thanks! Very much appreciate it, and I'll see what else I can dig up :)

Definitely agree that it's unlikely we'll ever see additive being the tool of choice for any high production parts, and that generative design tools to date can't seem to produce parts well-suited to other options. But disagree a bit on that opening 'disadvantage' being something unique to generative design...although it can do a hell of a job of highlighting those oversights! I think that's the case for anything being 'engineered' for purpose, especially if that thing is going to be produced in any volume. That 'margin' is coming out of somebody's bank account in material costs, no matter what process is being used to make the parts. It's that trade-off of upfront engineering costs vs. long term material costs...that damn Coke can has had some bucks spent on it! And that margin can be had with generative as well, just crank up that 'Safety Factor' number and let it go to town. But couldn't agree more that generative design seems like a bad choice for prototyping something out of sheet metal! And likewise for anything where it's not worth the upfront effort, since generative is indeed much less forgiving on designing without good part requirements. If nothing else, I think it could be a great training aid for young engineers, then maybe I won't end up throwing out so many poorly designed injection molded parts! Oh, and I would make a part like that, and I did :) Very much appreciate the thoughtful, detail-rich addition. May have a couple of "agree to disagrees", but lots of really good points, thank you!

Take a look at "Synera" if you want to see where the future might be heading. This tool addresses many of the problems you describe.

You’re spot on. DFM becomes problematic. Cool to look at and design but not very practical.

@@bubsbuilds I think we have more of a misunderstanding situation than disagree situation. Let me explain. With the error margin I din´t mean that you can´t simply crank up the thickness a little more I mean if you only design for a force in X and Y direction because you think that´s all that need be but due to vibration, temp. humidity, unforeseen events etc. the part need more load bearing in Z as well you have no margin. But you are right this can happen with a traditional design as well, so there shouldn´t be much difference except ease of manufacturing for high volumes and price without generative design. Injection molded parts are usually also 3D printed first, well at least that´s what I do and all my co-workers before we commit to that $100k mold. But yeah, as long as you are aware of it´s limitations it should be fine. On the other hand it´s pretty cool looking (could be great for marketing), it saves a lot of material and you can manufacture it easily via additive methods. Mercedes has done something similar lately on one of our latest projects, they molded a whole section on their car with one piece and skipped the 5-8 sheet metal parts which needed to be joined. They then made sure they can cast the part out of metal instead of producing it by additive manufacturing and voila pretty cool looking material saving mass production ready part. So it has it´s purpose for sure. In the end one can argue it´s even a better process. Usually I create a design first from past experience. Then comes the FEM-analysis and then countless interations of design changes. Here it´s the other way around. Generative design starts with the FEM-analysis and let´s me choose the parts, it already creates a couple of iterations beforehand. I am not sure if it predicts the non isotropic behavior of FMD printed parts though, but that depends on the software you use and the margin you choose so it´s solvable. So all in all I have to agree you made a good point, the issues I pointed out can happen with traditional designs just as well.

@@thomas8719 Thank you I will check it out.

Soo?? Did you find yourself a solution?? :)

@@bubsbuilds I think my solution will be to design and build my own device leaning heavily on what you've done here. I'd probably have two silicon carbide bearings spaced close together so the cylindrical magnets would contact those and then a third carbide bearing in the flexing part of the device. I think I could just take a big stack of magnets stuck end to end and slide them slowly through, pausing just long enough to see the diameter. Centering the cylinder tangent to the 3 carbide balls would allow for some variability in the angle of the cylinder that wouldn't affect the reading much for the first 4-5 degrees. I think it can work.

Very true!....well, except for the 'nice' gift part, as much as I would love to believe otherwise :) The ones calibrated off of the first will be calibrated to the uncertainty of the original, so they'll all be a bit worse than the first. Choose recipients accordingly!

Would "need to" for what? :)

@@bubsbuilds contact hardness dude

Haha, yeah, I found that kinda funny too :) but hey, they lived to gauge another day If you want to calibrate down to microns, I think it's gonna be tough to not find an expensive tool somewhere in the chain. If you have access to a half-decent coordinate measuring machine, you could use it to get some good 'actual' values for a set of feeler gauges to be your "Cal Feelers"....admittedly, this does sound a bit excessive. But with tools available in the shop, you should be able to calibrate it to better than 50 micron pretty easily. Half that with some effort. Not amazing, but not useless :)

Hope the head space is back to a place for enjoyin! whether you ended up watching or not :)

Why thank ya! :)

I did indeed only calibrate it with a relatively small number of points, and very true! I don't have any tools on hand at home that I could use for finer cal, but if you do, that will absolutely work! One of these days I'd like to splurge and get myself a Mitutoyo Calibration Tester, that'd be great for exactly that sort of thing. I should clarify a bit, even though I only used the handful of points, the 5 microns I mention was assuming/using interpolating values between the cal points. So the linearity error impacting them is quite a bit less than the full scale linearity error I showed in the plot in the video. But it is still contributing a little under a micron of the uncertainty, and that's an appreciable percentage improvement if you can reduce it. One additional benefit you may be able to get from finer resolution on cal points would potentially be a bit of extended useful range. There's some additional noise that comes out near the edges, but even with it, you should be able to get some usable range in the highly nonlinear bits at both ends.

@@bubsbuilds Should you ever get those values but have no way to get a good fit (like if the few functions excel comes with do not happen to work): Just post the values here, I can do that.

Not sure we agree fully on the 'why', but hey, preference is preference, so you do you! Purely mechanical devices are something of beauty though, I'm pretty fond of them myself! Those simulations were just in Fusion360's Simulation Extension.

I'm glad I'm not the only one that had that reaction! I was first introduced to using them in their linear range a few years ago, and it's still a personal favorite!

It absolutely can! A lot of indoor lighting sources can also have some content in the sensor's range. If putting a shade over it is easy, then that takes care of it without much trouble. A more robust option can also be implemented on the electronics side by using an AC signal on the sensor along with a high pass filter. That will knock out any DC signal from the ambient light.

@@bubsbuilds that's a fine solution

Doesn't look like I can just toss a voiceover on it, but I was thinking about making an updated video on this and some other 2020 extrusion stuff. I'll add a chapter marker for just this one :) If you are looking to build something sooner than that and have any specific questions on it, feel free to reach out!

Depending on the ambient lighting, it definitely can! A sometimes easier option though is to do it on the electronics side. Using a pulsed signal on the detector side along with a band pass filter can be really effective.

@chargehanger I'm not sure if you could get much better precision out of it with a lever, but I've been wrong more than once :). The only configurations I can readily picture would either cause a fair bit of stiffness increase or introduce a pretty significant Abbe offset. If you've got some specific designs in mind, I'd welcome takin a look at em! Same with the coarse/fine idea! I think I can roughly picture the direction you're thinking, but each way I picture it, either they both read the same, or the fine sensor gets chopped in half :) Love the suggestions, thanks! I may have to just test out a version with a lever to see how my assumptions above pan out.

Sounds like a fun project! I've actually been tinkering on a mechanical tester project lately too, must be the thing to do :) For the accuracy* and range you're looking for, you need 1 part in 2000, which I think may be pushing it with printed parts and an opto-interrupter. I think you could probably get 10 microns on 10mm, or 20 microns on 20mm....but even that isn't likely to be an easy task. One challenge is going to be that you won't be able to get an opto that has a 20+mm measurement range (or at least I'm not aware of any, and it seems like to make one with a beam this wide would mean arrayed emitters and detectors, which I would think would make them less suitable for measurement.) You can definitely still use them, but you'll need to add some 'gearing' in the flexure mechanism. Since you only need 10 microns, and the opto can very much be made stable to 1 micron with a good knife edge (I'd glue in a razor blade, instead of printing the edge), you should be able to gear it 10:1 with a lever arm, and 1mm range optos are widely available. Adding in this sort of lever mechanism into a printed flexure should definitely be doable, but it will add a fair bit of complexity. Some alternative options: A few folks in the comments have recommended using Hall Effect sensors in place of the opto interrupter that I used in the video. I haven't worked with them myself, so I can't speak to whether they'd be well-suited, but may be worth a look. They generally do have a much longer working range, but given the fun that are field lines, I would guess they get pretty nonlinear. If you end up going that route, I'd very much like to hear how it goes! Folks have me intrigued. If you are somewhat comfortable with electronics (nothing too crazy, but does require a little bit of working with AC signals) then I think a good candidate sensor for an extensometer would be an LVDT (Linear Variable Differential Transformer... the name sounds way fancier than it is...it's a few coils of wire and a steel slug :) ). They can have extremely good dynamic range, well beyond the 2000:1 you're aiming for...but yeah, little bit of electronics involved. And an alternative option that you could use that also uses an opto interrupter. You set it up as a linear encoder. You could print a sawtooth pattern (or maybe even just embed a piece of actual saw blade), and have it move through the U of the opto. You can then also add more than one opto directly alongside the first (just make sure not to space them at an even interval of the sawtooth wavelength). If you're familiar with a quadrature encoder, it's the same basic idea. In addition to giving direction (which you don't care so much about) it will also help quite a bit in averaging out some of the noise. Circling back to that 'accuracy asterisk' above, my answer really somewhat depends on what you mean by that goal. My guess is that for an extensometer you're mainly concerned with the relative movement as it's loaded, and don't care so much about the absolute accuracy....? Meaning if the actual length was 10.03mm long when it starts, and is stretched to 15.05. My assumption would be you want that delta of 5.02 good to that 2, and if the sensor was off on the absolutes, that'd be ok (say it read 9.08 to 15.00, but still gives the same 5.02). If you are indeed looking for the absolute accuracy, then I'd probably push my "challenging" to "pretty unlikely" :) You can't reference an ASTM standard and NOT expect me to write a novella :) Hope at least somethin in that long-winded reply is useful!

Thanks, and that's exactly the sort of stuff I'm aiming for! And great point on the humidity, I hadn't really considered the swelling of the plastic! I've currently got some thermal stability/drift tests going....I may have to cook up some humidity sensitivity measurements too...also, ugh, that's gonna be a tad more of a PIA than controlling temp :) They really are surprisingly impressive little things. Someone first introduced me to them for this sort of use a while back, and I still think they're just great little tools to have in the box.

Wow, well if that doesn't make your day to hear, I don't know what will, thanks!

Absolutely! I have a couple paired with Arduinos for something similar (admittedly, a higher bit depth ADC would probably be wise) . Although I didn't go with a nonlinear curve fit or the like, instead I just put the calibration data points in to two lists in the program, and then interpolate for the Displacement (aka a "lookup table", I suppose). Probably not the best if trying to do something high speed...but then again, this sensor probably isn't either :)