you say at 12:50 that the scan version was faster but have the time for CAD at 35 mins and the time for 3d scan at 43 mins. Is this an error or are you discounting some of the time you used trying to figure out the best way to add the wing?
What I like to do is to make a photo of the part. Use that as a background in CAD and trace it. If you scale the picture in CAD you will have all the dimensions in one go.
Heard a tip on YT maybe a day ago using a camera - move far back from the item being photograph and zoom in to the subject. Removes the perspective distortion giving an (almost) orthographic image. Personally prefer using a DSLR over a scanner due to the higher resolution, pretty much the same idea though - can't have too many methods for these things! Think the channel was "Makers Tales", might be wrong though.
The first lesson of this video is to avoid the proprietary licensing crap and use an open source 3D scanner. Great concept for a video, as always, flawlessly executed for maximum educational value. Thank you. We need resin printers without the toxic mess and the hassles of slopping chemicals, drying and post curing. This part would be very strong printed in ABS-like resin.
@@TheTimeCatcher - I'm a big fan of FreeCAD. It's not as slick and polished as Fusion360 or SolidWorks but the time I spent learning FreeCAD and the designs I created in FreeCAD will never be used to extort money from me later. I also won't spend hours on the end user licensing agreements and installation hassles of proprietary software. Open source software just feels right.
I prefer CAD over scanning every day. Unless you want to print an organic object like a plant, bone or ring. But beyond that, CAD! Especially because everything's parametric and you can adjust and sometimes even improve the original part.
Typically 3d scans are used to build an ordinary parametric CAD file, all serious current implementations of 3d scanning for functional components are using the scans as reference geometry rather than using the scans themselves to create parts.
Yes, it depends on the type of geometry you want to capture. Unless you have a professional scanner with geometry recognition that is suitable for such small parts, CAD is much faster. If I have very organic shapes, I like to use photogrammetry, because it works in any dimension, as long as the photos are good.
I look forward to the day that consumer 3D scanners are affordable, reliable and work well. The world of home repair will never be the same. The effort of measuring and recreating parts is doable for some things, but if it could be automated.... that'd be awesome!
The FaceID sensor of modern iPhones can do a decent job at scanning. Used it to make a camera adapter for my turtle's shell. Don’t even have to jailbreak the phone, there are apps on the App Store that work
The interesting thing is that this can be done with a regular phone, doesn't even need a fancy depth sensor. It's a bit involved, but here's a possible approach: 1.Capture phone orientation and relative position by using the IMU in combination with one of those special AR QR codes. 2. Take those image+orientation data and train a NeRF model or neural radiance field model. The gist is to have a network predict the color and depth of any given ray of light. 3. From this model, we can generate a point cloud which we can turn into a mesh. Now, Nvidia created something called instant-nerfs which makes the training time for these networks to be seconds. So, if you're dedicated enough, this isn't out of reach to create. There's probably a random GitHub repo with a similar idea out there :)
If you just use standard photogrammetry software it becomes much easier, without any of the proprietary cloud lock in BS that for me at least makes that company a non-starter.. but with off the shelf sideways, the hardware to take a ton of pictures on a turntable has to be diy unfortunately
Einscan-SP or Einstar. Never had his activation problems. Maybe his device was a review unit? I have the Einsxan-SP and a Revopoint Mini. For this part I would have used the Revopoint Mini because of how small it is. With the mini I probably could print it without editing the STL in CAD. But I usually import the 3D Scan in Fusion and reverse engineering it.
A cheaper (and much less toxic) option than scan spray is dry shampoo, it gives the same light powder finish (and smells a bit flowery..) and is massively cheaper.
I have no clue either. Maybe if we saw the internals we could have found a replacement part? Maybe there it is a stash area? Or maybe it was modified and no longer meets regulations? I don’t get what is so private about a flush box.
@@Firstleakereither it's a joke, because of our 'fancy german flusboxes' or, and that's what I guess: it's dirty 😁 yes there is just water in there, but that doesn't mean it's clean 😜
Sometimes on stuff like electrical boxes there is printed documentation with your address. I have no idea what a flush box is though (we don't have em in America)
One thing I can recommend, idk if you have tried this. You can put the part (considering its not too big) stick it in a 2D scanner and scan it. You can add a ruler for scale, Inventor and solidworks allows you to import an image, which you can scale to your sketch plane. It works beautifully for parts that are small enough to fit in a 2D scannner.
This. I quite often use this method to copy parts that we need. Only difference is that don't use ruler - I just measure 1 dimension (usually THE IMPORTANT one) and than scale whole scan accordingly in Inventor. Solid edge also is nice for this purpose with it synchronous modeling approach.
I prefer cad modeling, you can identify necessary features like you did in the video, and optimize for 3d printing. Scanning sometimes leaves weird areas that require supports or a lot of smoothing and loss of dimensional accuracy for smooth surfaces and small details.
Thank you sir for another excellent video. I still find it incredible that if I need a part for something, I can fire up cad (or Blender) and, using some calipers and notes, I can make my own parts faster, and cheaper, than ordering and waiting for them. Or I can make something entirely unique that I cannot find already made. Or make a part that is no longer manufactured (like a vintage or antique part). I had an old pickup in my 20s. Like many old pickups, the tailgate latch didn't work right because one of the retaining clips that held the latch rods in place on the handle had broken. I tried for months to find a replacement, even going to the dealer, and had no luck finding a part that would work (the parts guy at the dealer basically told me to pound sand because finding that part in inventory for a 20+ y/o vehicle was a waste of his time). Now, I could model that clip and print it for pennies (some places wanted $45 bucks for a little plastic clip a cm long), and then share it with other makers.
These results are pretty much what I expected. You did address what I was thinking in your wrap-up too, that this flusher part is suited to CAD, but that something larger and more organic that's difficult to measure directly would be much better suited to scanning. The example of the wheel well and suspension hadn't even entered my mind, scanning a difficult-to-measure environment and then taking those measurements in software. Great use-case. Thanks for the video!
It's always fascinating how different countries have different methods for achieving the same goal. I'm from the US and our toilets have the flush mechanism attached directly on the toilet bowl tank as a small lever although some have a push button on the tank cover. With that first scanner, if the company ever goes bye bye, you have a very expensive paperweight. And as you've experienced, when the activation server is having issues, you still have an expensive paperweight.
hrm toilet company not in business any more. better 3d scan a replacement part. oh noes, 3d scanner company not in business any more. sorry actually that way they have locked down the 3d scanner makes me angry to the level of going psychotic.
I've recently had to this exact same thing with something that broke. Did it in CAD and also did the whole tracing on paper for the angles. Seing your result gave me a lot of confidence that the part will probably be good, cheers.
Nice video. What can help with the Calipers and CAD sometimes is to just take a photo as reference and drop that into your CAD tool as a background. Once you scale the picture with a measurement that is easy enough to make, it helps to have some reference to trace odd angles and curves with sketch lines.
Great video. I’ve been doing this type of thing for a very long time and all methods in the video and discussion have a use and place. Unless you are using a metrology grade scanner (and trust the scanners discussed here do not touch them) they are best used as reference geometry and then just mode over them in CAD. They are best used for high complexity models (complex curvature, highly faceted, non-intuitive feature easy to capture in top/side view pics). It’s also good to remember that most consumer and prosumer scanners are terrible at capturing black and shiny surfaces… if the part has this, the prep time goes up related to this step (use talc or the such, plus cleanup after) but scan is way better. If the part has relatively simple features (lots of basic shapes and the like) calipers, micrometers, or just a ruler can work well and much faster. For mid-complexity (still pretty simple shapes but straightforward angles that may be difficult to measure, picture reference works well (others already mentioned the flat bed scanner trick if the part/features fit or you can stitch ). But again, all the methods have their use case. One good thing to do is to verify one with the other. Like take a scan and measure what you can with calipers or ensure you have a known reference item I your scan or pic to scale off of. Glad to see this info is getting out of just the more hardcore engineering circles:).
Great video. I agree with you on what 3d scanners are best used for. That's part of the reason that i still use an old Xbox 360 kinect as my scanner instead of upgrading to something more high definition.
Am I missing something? You showed the CAD as taking 35 minutes and the scan as taking 43 minutes, and then said that the scan was faster? Was there an editing mistake here? Also, it’s awfully generous to say that the scan won when you unnecessarily made a replica in CAD, when in practice, all you need is a functional replacement. In a real world scenario, I’d expect to be done with the CAD approach in _maybe_ 20 minutes, and that’s with revising the part to avoid having to print supports.
I found when measuring parts with digital calipers that the designers of the part in question will often use whole numbers in its mm dimensions, making it easier to recreate in your 3D editor, and it especially helps when trying to measure circular dimensions when you can't directly measure the radius or diameter, ie curved inner/outer edges of the part. I would hope there's software out there that can take a 3D scan and apply the same whole numbers usage thing to end up with an easily editable highly accurate CAD file.
Right? I thought that was a really odd choice. And if he'd simplified the model to take out all the artifacts from the original manufacturing process, the replacement part would still have been about as close in appearance to the original as the lumpy scan was.
I loved your optimized part at the very end. I've got a scanner and calipers and agree that in a lot of ways caliper measurements are better than scans. Best if to combine both to compliment each other. Angles are easier with a scanner while hard points are easier with calipers. Every use case it unique though. Good video.
in CAD, you can take a photo, scale it and simply outline it, especially such simple details of the mechanism. at the same time, there is no unnecessary work with cropping scan defects.
I already did something along what you said at the end to fix a couple of stuffs at home. I used photogrammetry instead of a 3D scanner because I don't have one. As I can see photogrammetry does not not seem really worse than the scanner you used in terms of results. I mount the part I want to replicate on a support (3D printed) with known dimensions (so I can set the correct scale that photogrammety typically doesn't get) then I use the 3D model as a reference for CAD modeling.
I would say measurements are the way to go, since you can edit everything element of the part to make it work best with the manufacturing process you have. 3D scans are awesome tho.
Cool video Tom. But the cold hard truth is this. CAD skills are required for both methods. In fact, you need far more CAD skills to "clean up" a point cloud scan than you do to solid model something. However, you simply can't get away from the fact that scanners are awesome. Oh, and CAD coming in at 35 mins vs SCAN at 43 mins, what am I missing here????????
Parametric CAD skills and being able to make changes with CSG is two different skillsets, but both require that you feel comfortable manipulating 3D objects on a 2D screen, which may just be a barrier that takes a lot of power to cross for some. CSG has shown itself to be super friendly towards beginners though, thus the unwavering popularity of Tinkercad.
You could quite possibly have simplified that part and it would have come out stronger. A lot of the weird body shape is purely down to the manufacturer trying to reduce the amount of plastic being used.
For altering the scanned part I found that 3d Builder works great for simple stuff, kind of the ms paint for 3d models, doesn't have anything super fancy but what it can do it does well.
I've wondered the same thing. Thank you for going through this. My mesh and scanning skills are much lower than my CAD skills, and both much slower than yours, so I wondered if I was somehow missing out on not being able to scan better. So now I know "perhaps", but not so much that it's important. I love the idea of using scan to feed CAD, and that's what I've gradually tried to do.
Have no idea about 3D scants but in regards to CAD manually tracing a 2D scann (bitmap) and the addition of the calipers would help avoiding mistakes and make the process more pleasant, maybe even faster. Just import 2D scanned picture to CAD :) Cheers.
100% agreed. I've got a creality scan lizard when it came out and had the same experience and conclusions you came to :) it's good for certain tasks, but most of the time it's just easier/faster to draw the part in CAD.
As long as you have a reference. If some part broke, 3D scanning both halfs and joining them again might be more trouble than measuring the rest. I know of someone that is using 3D scaning and digital calipery, lasers etc to measure specific parts that are already worn or organic etc to mount sensors for various kinds. So there still is cad afterwards. But they use the scan as a positive reference
That part, and how it functions is definitely a CAD and near enough is good enough kind of solution, it even luckily aligns with the strength needed being along the layers of 3d printing
Your on point in the end of the video. I use the scanner for getting references that are almost impossible to measure with the caliper, then reverse engineer or use for references only. Even on a very expensive scanner you still get a dumb solid and need to reverse engineer it to get a real cad model. Scanning and directly printing is more for the people doing figures and stuff.
I spray in a container that seals wire spraying outside. I then use a water color bush with the over spray powder for any stubborn regions. I'll also usually just use the brush because it's cleaner and quicker for small parts.
I always go for CAD. For some reason I always heavily struggle with editing meshes/objects, let alone cleaning up a messy 3D scan. With CAD you just design what you need while taking 3D printing limitations in mind at the same time.
Interestingly enough. The 3D Makerpro Lynx, seems to be a rebranded Creality CR-01 even down to the softwear. I've used the CR-01 for about a year now, it can get decent detail, especially with the turntable for smaller parts.
I was wondering the same thing. What is so private about showing the inside of a flush box. I don’t get it. Unless it was modified, and no longer meets regulation standards?
It'd be interesting to see 3D scanning & 3D printing used in a loop, like deepfrying a jpg. Starting with a low-poly model would probably serve as a good noise seed as the verts & edges will round over to various degrees, where a smooth model may bias noise around detail more than on surfaces
Out of curiosity, what is so special about the flush box that makes it unsuitable for sharing? I cannot imagine there being anything personally identifiable
Great video Thomas, love your contributions! Let me give you one, in most CAD you can put an image on the background and planes, what I do in reverse engineering a part is take a picture of each profile and put on each plane and use a sketch feature to size it so it is the right size. Once you do this you can cut your measurement time down greatly and get really accurate parts, last set of parts I did were correct on first printing, and they had some weird shapes.
My personal perference is these situations in to take a couple of photos and either scale as image in CAD and trace it (or create an svg to speed the trace a little) and then combine with measuring with calipers.
What I would often do is take a photo of the part with a ruler (from high enough above to minimize distortions) and import it into the CAD. I then measure the fine details and precise lengths with calipers where it matters and use the photo (properly scaled) for the overall shape. But I like the approach with extending the angles too! Still, the photo is a great way to quickly spot any potential mistakes (like the rubbing nub) right away, most of the time my parts work in their first revision 🙂 Also, CAD gives us a very easy way to strenghten the parts where the extra mass would not block proper operation. With 3D scanned mesh this is more tedious. I tend to do that a lot, especially to prevent the original failure mode.
I think that the results of scanning of very product/software dependent. I have tried a few different ones and found that Polycam works best for me. It is a different process than what is shown in the video but it works well for me. I also do use the take a photo and calibrate it to use as a background to draw over. That works well, but there can be an additional step of lens compensation. There are many ways to it it, but it is important to be aware of. If not parallax effects can happen and lines will not be straight.
Thanks for showing up the online problems. This will always be a problem for me because a few years down the road it may fail completely. I think these companies have some good reasons for doing it this way but they’re not good for me. I’m sure that they’re aware that recent phones can scan with their fancy cameras and without all the online hassle. What advantage are they offering? My calliper and I do a lot of bonding but one area where I fail is in the compound curves. I can’t measure those well enough to print a conforming fit. Haven’t tried a scanner yet.
Been getting some good scan results with a Creality CR-01 3d scanner of full size car parts and guitars and orthotic foot arch scans to make better fitting shoes or printable flipflops that wont damage the arch. Its not good for smaller objects but what a vast improvement over previous scanners. And theres a bunch of newer chipset version scanners making for some very useful reasons to have one or a few.
The CAD advantage is clear. It also gives the user the ability to create parts rather than just replication of broken ones. I am struggling to learn how to model in CAD but I consider that a set of skills that unlocks the capabilities of 3d printing.
I just recently also compared EinScan HX to simply making my own modle. I tried a house key, a circuit card for my car fob, and a visor for a GPS screen. A scanner is neat, but it's not for duplicating objects and more of a fancy reference maker or for artistic use only. I much preferred calipers to capture all primary features and then taking pictures with rulers on the primary features to capture secondary and third features. Also, dont sleep on document scanners. Not only are they ridiculous, accurate, but they do export dimensions metadata that can be picked up by some CADs. I did the circuit card and key like this..
For the Einscan SE, try to get an older version of the software from around 2019. The application is stand alone and works offline after activation. I have my scanning computer disconnected from the internet for security reasons and this arrangement works. Another benefit is you can make a copy of the Einscan Folder with the application and keep it backed up on a flash drive. You can move it to any other computer and the Einscan will work with that version of the activated application. I had to do this when my previous computer finally died and I had to move it to a newer computer.
I usually trace the part or redraw the part then put measurements like old school drafting then I take those measurements to the computer. To me it seems smoother that way you can figure out how you are going to actually start the CAD process.
Thanks, very interesting comparison. Phones have such excellent cameras these days; is there such a thing as a pure software 3D scanner app that works?
I tend to trace the part on a piece of paper and then scan that into a jpeg to import into fusion 360. And then trace, using calipers to help confirm scaling and any tricky parts
If we’re 3D printing the part because it broke, going the CAD way is an opportunity to improve it, strengthening where it broke-and optimizing it for printing as well.
3D scanning is also awesome in areas where you can not simply measure. I needed the inside of an container with wild angles and curved planes. I didnt even get my tool inside to measure.
Depends, some parts I like to reverse build just via measuring, but for example I build a MagSafe adapter for my car using a 3d scan of the dashboard to fit it perfectly. Know your tools and limitations 👍🏻
I am still a fan of doing it the CAD way. I find it really relaxing. Thanks for the demo of scanning software. Anyone have any bets on how many generations of scan/prints you could get before the object isn't recognizable anymore? 🍻
A alternative way to get overall perimeter details is to take a picture of the part using a long focal length to eliminate DOF and then to scale the image to a known dimension. This workflow could result in a little faster reverse-enginering of the part.
You don't need to be "fair to the scanning". In fact your goal here is to get a functioning part which can be 3D printed and that doesn't break like the original did. That original part was once designed with the process of injection molding in mind as the method of manufacturing. But that's not what you're doing, so you might as well ignore unnecessary details and optimize the part for 3d printing right away. You dont want to copy what didn't work anyway, right? ;-)
At my internship I've been trying to recreate small turbine blades so we could recreate them out of composite. I've had trouble scanning thin flattish parts like turbine blades. Recreating it by hand is near impossible. I've done a mix of both. CADing the tolerance specific parts like the blade mount. Then using the 3d scan points to help me make the surface of the blade. It's not the best.
Scanning with the lynx I noticed you were in texture mode not geometry. You might get better results with geometry. I have the mole and that's my experience with small parts like that.
I am not sure if that vertical rod feature in the 3D-Printed part will hold up especially to repeated use. There is also the fact that PLA creaps significantly under constant load.
I find CAD is also well-suited for making design improvements to the replacement part - suppose the rest of the system has worn out around the part such that an exact copy won't have quite the right dimensions anymore. (Or it wasn't right to start with.)
In fairness to the Lynx, it was advertised for larger objects. Although I've used mine fairly successfully for smaller parts after a bit of love and patience
Durring the resconstrctions, many times I had an idea, that normal paper office scanner could be also also helpful. It can create very precise 2D contour in orthographic projection that can be very quickly reconstructed as sketch. Unfortunatelly never had it at home to test it. Also in this case it could speed up the reconstruction.
I use my digital caliper for parts. However had a delta wing project that required the model of the EPP foam wing that did not exist and the frame was 500mm so I found Polycam for iOS and Android did the job remarkably well compared to the $300+ 3d scanners.
I used an IPhone to scan a part of an rc plane to then design a new canopy for it. The plane has quite irregular shape so getting the 3D scan as a model helped a lot.
I prefer CAD, too (maybe because it is my only option :) ). But I take a bird's eye view of the part to see how it may be simplified or improved. Parts that are made by injection molding often have a different set of constraints and material saving measures that result in part failures at critical points. Beef them up at the weak point.
Scans never seems to have "stable" surfaces for me. They always have weird artifacts, are slightly off, or just straight up wrong. I far prefer measuring for CAD. Because then at least the model will be off by a common amount.
The Lynx scanner is for larger objects. The Mole scanner is for medium sized objects. This model is perhaps too small for the Lynx - better suited for the Mole scanner. This is a personal observation - not a commercial - I have both scanners and would use the Mole in this case. Better resolution for this size. :-)
I am surprised you didn't also use a background image in cad while doing your 3d modeling. As it might of helped you both get the scale right and avoid the error you had.
I would have placed the part on the flatbed scanner, model the basic shape ontop of the image in FreeCAD and add the details by hand. This was much faster than dealing with the post scan model. Another side effect is that I end up with a proper parametric model, too.
I was curious what you used to make the one-to-one replica of the mechanism: CAD or 3D scan? Based on the ending I suspect I know the answer. I just think it's worth pointing out the perfect natural experiment: once you've tried a couple techniques, see which one you /want/ to use when you actually need it
I´m not sure if the Lynx was the best option for this small part, because even the manufacturer write it on the product page that its made for bigger objects. Maybe the 3dmakerpro Mole or a Revopoint Mini would fit better for this kind of parts. The einscan-se should handle it very easy. Mine works fine with small parts, but ya the activation is sh*t.
I know the Einscan coule most likely have gotten me a better result (which is why I tried it first), but I'll take a 3D scanner that actually lets me use it over one that locks me out any day.
i understand that. The activation process is from the hell :D But my point is, that the lynx is not the perfect choice for this type of parts ^^@@MadeWithLayers
Great video! I’m not sure what that flush box is though? We don’t have those in the US as far as I know? I feel that a 3D scan is good for reference OR for amorphous objects like a face. But 3D scans lack the precision required for technical parts.
@MadeWithLayers Thanks for the heads-up on Shining3D's misauthentication process Tom! I already have one large paperweight in the form of a NextEngine where the software stack toppled permanently. I enjoyed the sponsor spot knowing the 'window' is already imaginary!
So this convinced me to not get a 3D scanner, since often I have 10 minutes here, 5 minutes there to work on a CAD part. So spending money on a scanner doesn't look like a useful improvment for me. The scan the environment concept sounds cool though as a use case like you said! I'd be currious how effective generic photogrametry would be for that too.
Thanks to Private Internet Access for sponsoring this video! Check them out at piavpn.com/Toms3D
yep we still won't be out of the job for a long time, those scanners are pretty lame, even super expensive ones
you say at 12:50 that the scan version was faster but have the time for CAD at 35 mins and the time for 3d scan at 43 mins. Is this an error or are you discounting some of the time you used trying to figure out the best way to add the wing?
@@nakkajin exactly
Thank you for not trying to fearmonger people into buying a VPN that they don't need. That's really refreshing.
What I like to do is to make a photo of the part. Use that as a background in CAD and trace it. If you scale the picture in CAD you will have all the dimensions in one go.
If the part is flat, a paper scanner provides nice, hi-res reference picture.
Heard a tip on YT maybe a day ago using a camera - move far back from the item being photograph and zoom in to the subject. Removes the perspective distortion giving an (almost) orthographic image. Personally prefer using a DSLR over a scanner due to the higher resolution, pretty much the same idea though - can't have too many methods for these things!
Think the channel was "Makers Tales", might be wrong though.
Place it on the scanner of a paper printer. Works with a lot of parts and way more accurate and already scaled correctly
just made a similar comment, then saw yours. spot on.
The thing with this is you need to get an exactly head on view of the part
The first lesson of this video is to avoid the proprietary licensing crap and use an open source 3D scanner.
Great concept for a video, as always, flawlessly executed for maximum educational value. Thank you.
We need resin printers without the toxic mess and the hassles of slopping chemicals, drying and post curing. This part would be very strong printed in ABS-like resin.
Yes it is a god damn scanner and they protect that crap like they invented fusion or something.
The first lesson is: Always be prepared
Fusion 360 is the proprietary licensing crap as well btw.
@@TheTimeCatcher - I'm a big fan of FreeCAD. It's not as slick and polished as Fusion360 or SolidWorks but the time I spent learning FreeCAD and the designs I created in FreeCAD will never be used to extort money from me later. I also won't spend hours on the end user licensing agreements and installation hassles of proprietary software. Open source software just feels right.
ABS like resin isnt very strong though, nowhere near as strong as PLA.
I prefer CAD over scanning every day. Unless you want to print an organic object like a plant, bone or ring. But beyond that, CAD! Especially because everything's parametric and you can adjust and sometimes even improve the original part.
Typically 3d scans are used to build an ordinary parametric CAD file, all serious current implementations of 3d scanning for functional components are using the scans as reference geometry rather than using the scans themselves to create parts.
Yes, it depends on the type of geometry you want to capture.
Unless you have a professional scanner with geometry recognition that is suitable for such small parts, CAD is much faster.
If I have very organic shapes, I like to use photogrammetry, because it works in any dimension, as long as the photos are good.
Until you're dealing with surfaces. Then it becomes very difficult to reverse engineer parts without a CMM
Says a non advanced CAD user
@@limbeboy7 Who are you talking to? And be careful with making the wrong assumptions based on thin air ;-)
I look forward to the day that consumer 3D scanners are affordable, reliable and work well. The world of home repair will never be the same. The effort of measuring and recreating parts is doable for some things, but if it could be automated.... that'd be awesome!
The FaceID sensor of modern iPhones can do a decent job at scanning. Used it to make a camera adapter for my turtle's shell.
Don’t even have to jailbreak the phone, there are apps on the App Store that work
@@among-us-99999 A camera adapter for a turtle shell? Lol, sounds cool - are you trying to make Turtle POV videos?
The interesting thing is that this can be done with a regular phone, doesn't even need a fancy depth sensor. It's a bit involved, but here's a possible approach:
1.Capture phone orientation and relative position by using the IMU in combination with one of those special AR QR codes.
2. Take those image+orientation data and train a NeRF model or neural radiance field model. The gist is to have a network predict the color and depth of any given ray of light.
3. From this model, we can generate a point cloud which we can turn into a mesh.
Now, Nvidia created something called instant-nerfs which makes the training time for these networks to be seconds.
So, if you're dedicated enough, this isn't out of reach to create. There's probably a random GitHub repo with a similar idea out there :)
If you just use standard photogrammetry software it becomes much easier, without any of the proprietary cloud lock in BS that for me at least makes that company a non-starter.. but with off the shelf sideways, the hardware to take a ton of pictures on a turntable has to be diy unfortunately
Einscan-SP or Einstar.
Never had his activation problems. Maybe his device was a review unit?
I have the Einsxan-SP and a Revopoint Mini. For this part I would have used the Revopoint Mini because of how small it is.
With the mini I probably could print it without editing the STL in CAD.
But I usually import the 3D Scan in Fusion and reverse engineering it.
A cheaper (and much less toxic) option than scan spray is dry shampoo, it gives the same light powder finish (and smells a bit flowery..) and is massively cheaper.
How come you said that the scan was faster when the cad when the video shows the CAD being faster?
Your toilet flush button is WAY different than mine.
Would be also nice to see how long it would take with some open-source photogrammetry solution and a phone camera.
In CAD you can also modify some features in order to obtain a more printable piece
... and possibly a more durable one as shown in the very final shot. The original part isn't always as good as it could be.
Why is a flush box confidential?
I have no clue either. Maybe if we saw the internals we could have found a replacement part? Maybe there it is a stash area? Or maybe it was modified and no longer meets regulations? I don’t get what is so private about a flush box.
@@Firstleaker inside it usually looks bad with mold, lime, dust etc.
@@Firstleakereither it's a joke, because of our 'fancy german flusboxes' or, and that's what I guess: it's dirty 😁 yes there is just water in there, but that doesn't mean it's clean 😜
Sometimes on stuff like electrical boxes there is printed documentation with your address. I have no idea what a flush box is though (we don't have em in America)
Obviously it was dirty. Don't tell me you guys screw panels off the wall to clean behind them.
One thing I can recommend, idk if you have tried this. You can put the part (considering its not too big) stick it in a 2D scanner and scan it. You can add a ruler for scale, Inventor and solidworks allows you to import an image, which you can scale to your sketch plane. It works beautifully for parts that are small enough to fit in a 2D scannner.
This.
I quite often use this method to copy parts that we need. Only difference is that don't use ruler - I just measure 1 dimension (usually THE IMPORTANT one) and than scale whole scan accordingly in Inventor.
Solid edge also is nice for this purpose with it synchronous modeling approach.
Fusion 360 also allows importing images
I prefer cad modeling, you can identify necessary features like you did in the video, and optimize for 3d printing. Scanning sometimes leaves weird areas that require supports or a lot of smoothing and loss of dimensional accuracy for smooth surfaces and small details.
agree. 3D scanning can be a massive time suck.
That shot of you tossing that scanner into the trash, that's such a beautiful shot. Tells me all I need to know about that product.
the superfastmatt shoutout was cool to see. two very different channels but I expect more overlap in viewers than people might initially think
Thank you sir for another excellent video. I still find it incredible that if I need a part for something, I can fire up cad (or Blender) and, using some calipers and notes, I can make my own parts faster, and cheaper, than ordering and waiting for them. Or I can make something entirely unique that I cannot find already made. Or make a part that is no longer manufactured (like a vintage or antique part).
I had an old pickup in my 20s. Like many old pickups, the tailgate latch didn't work right because one of the retaining clips that held the latch rods in place on the handle had broken. I tried for months to find a replacement, even going to the dealer, and had no luck finding a part that would work (the parts guy at the dealer basically told me to pound sand because finding that part in inventory for a 20+ y/o vehicle was a waste of his time). Now, I could model that clip and print it for pennies (some places wanted $45 bucks for a little plastic clip a cm long), and then share it with other makers.
3:30 and this is why Cloud based hardware has to be praised with care. You literally have no control over whenever or not it will work next week.
These results are pretty much what I expected. You did address what I was thinking in your wrap-up too, that this flusher part is suited to CAD, but that something larger and more organic that's difficult to measure directly would be much better suited to scanning. The example of the wheel well and suspension hadn't even entered my mind, scanning a difficult-to-measure environment and then taking those measurements in software. Great use-case. Thanks for the video!
It's always fascinating how different countries have different methods for achieving the same goal. I'm from the US and our toilets have the flush mechanism attached directly on the toilet bowl tank as a small lever although some have a push button on the tank cover.
With that first scanner, if the company ever goes bye bye, you have a very expensive paperweight. And as you've experienced, when the activation server is having issues, you still have an expensive paperweight.
hrm toilet company not in business any more. better 3d scan a replacement part. oh noes, 3d scanner company not in business any more. sorry actually that way they have locked down the 3d scanner makes me angry to the level of going psychotic.
If you want to be frightened look up a German toilet bowl and why there is a shelf.
I would have scanned it *on a flat bed scanner* (with a ruler for reference) and traced it out in F360. That would probably take under 20 min
I've recently had to this exact same thing with something that broke. Did it in CAD and also did the whole tracing on paper for the angles. Seing your result gave me a lot of confidence that the part will probably be good, cheers.
Nice video. What can help with the Calipers and CAD sometimes is to just take a photo as reference and drop that into your CAD tool as a background. Once you scale the picture with a measurement that is easy enough to make, it helps to have some reference to trace odd angles and curves with sketch lines.
take three pictures in XYZ with a ruler. Much faster than scanning.
Great video. I’ve been doing this type of thing for a very long time and all methods in the video and discussion have a use and place. Unless you are using a metrology grade scanner (and trust the scanners discussed here do not touch them) they are best used as reference geometry and then just mode over them in CAD. They are best used for high complexity models (complex curvature, highly faceted, non-intuitive feature easy to capture in top/side view pics). It’s also good to remember that most consumer and prosumer scanners are terrible at capturing black and shiny surfaces… if the part has this, the prep time goes up related to this step (use talc or the such, plus cleanup after) but scan is way better. If the part has relatively simple features (lots of basic shapes and the like) calipers, micrometers, or just a ruler can work well and much faster. For mid-complexity (still pretty simple shapes but straightforward angles that may be difficult to measure, picture reference works well (others already mentioned the flat bed scanner trick if the part/features fit or you can stitch ). But again, all the methods have their use case. One good thing to do is to verify one with the other. Like take a scan and measure what you can with calipers or ensure you have a known reference item I your scan or pic to scale off of. Glad to see this info is getting out of just the more hardcore engineering circles:).
Great video. I agree with you on what 3d scanners are best used for. That's part of the reason that i still use an old Xbox 360 kinect as my scanner instead of upgrading to something more high definition.
Am I missing something? You showed the CAD as taking 35 minutes and the scan as taking 43 minutes, and then said that the scan was faster? Was there an editing mistake here?
Also, it’s awfully generous to say that the scan won when you unnecessarily made a replica in CAD, when in practice, all you need is a functional replacement. In a real world scenario, I’d expect to be done with the CAD approach in _maybe_ 20 minutes, and that’s with revising the part to avoid having to print supports.
I found when measuring parts with digital calipers that the designers of the part in question will often use whole numbers in its mm dimensions, making it easier to recreate in your 3D editor, and it especially helps when trying to measure circular dimensions when you can't directly measure the radius or diameter, ie curved inner/outer edges of the part.
I would hope there's software out there that can take a 3D scan and apply the same whole numbers usage thing to end up with an easily editable highly accurate CAD file.
I am bit confused. You said that the cad took 35minutes. And that the scan took 43minutes. Then you concluded that the scan is faster?
I think he misspoke. Clearly the scanned version took longer. It also looked terrible, with bad dimensions even after cleanup.
Bro, not having to deal with not important features is a benefit of cad. You canot take that out of the equation.
Right? I thought that was a really odd choice. And if he'd simplified the model to take out all the artifacts from the original manufacturing process, the replacement part would still have been about as close in appearance to the original as the lumpy scan was.
I loved your optimized part at the very end. I've got a scanner and calipers and agree that in a lot of ways caliper measurements are better than scans. Best if to combine both to compliment each other. Angles are easier with a scanner while hard points are easier with calipers. Every use case it unique though. Good video.
in CAD, you can take a photo, scale it and simply outline it, especially such simple details of the mechanism. at the same time, there is no unnecessary work with cropping scan defects.
I already did something along what you said at the end to fix a couple of stuffs at home. I used photogrammetry instead of a 3D scanner because I don't have one. As I can see photogrammetry does not not seem really worse than the scanner you used in terms of results. I mount the part I want to replicate on a support (3D printed) with known dimensions (so I can set the correct scale that photogrammety typically doesn't get) then I use the 3D model as a reference for CAD modeling.
I would say measurements are the way to go, since you can edit everything element of the part to make it work best with the manufacturing process you have. 3D scans are awesome tho.
Cool video Tom. But the cold hard truth is this. CAD skills are required for both methods. In fact, you need far more CAD skills to "clean up" a point cloud scan than you do to solid model something. However, you simply can't get away from the fact that scanners are awesome. Oh, and CAD coming in at 35 mins vs SCAN at 43 mins, what am I missing here????????
yeah I noticed that too, dunno if his script was reversed
Parametric CAD skills and being able to make changes with CSG is two different skillsets, but both require that you feel comfortable manipulating 3D objects on a 2D screen, which may just be a barrier that takes a lot of power to cross for some.
CSG has shown itself to be super friendly towards beginners though, thus the unwavering popularity of Tinkercad.
You could quite possibly have simplified that part and it would have come out stronger. A lot of the weird body shape is purely down to the manufacturer trying to reduce the amount of plastic being used.
For altering the scanned part I found that 3d Builder works great for simple stuff, kind of the ms paint for 3d models, doesn't have anything super fancy but what it can do it does well.
Video drop from Thomas makes my weekend. Thank you for the information.
Yes, I think the point is to take the scan into CAD to use as reference. That's what I do. I use photogrammetry, having the textures can help too.
I've wondered the same thing. Thank you for going through this. My mesh and scanning skills are much lower than my CAD skills, and both much slower than yours, so I wondered if I was somehow missing out on not being able to scan better. So now I know "perhaps", but not so much that it's important. I love the idea of using scan to feed CAD, and that's what I've gradually tried to do.
Have no idea about 3D scants but in regards to CAD manually tracing a 2D scann (bitmap) and the addition of the calipers would help avoiding mistakes and make the process more pleasant, maybe even faster. Just import 2D scanned picture to CAD :) Cheers.
100% agreed.
I've got a creality scan lizard when it came out and had the same experience and conclusions you came to :)
it's good for certain tasks, but most of the time it's just easier/faster to draw the part in CAD.
As long as you have a reference. If some part broke, 3D scanning both halfs and joining them again might be more trouble than measuring the rest.
I know of someone that is using 3D scaning and digital calipery, lasers etc to measure specific parts that are already worn or organic etc to mount sensors for various kinds.
So there still is cad afterwards. But they use the scan as a positive reference
That part, and how it functions is definitely a CAD and near enough is good enough kind of solution, it even luckily aligns with the strength needed being along the layers of 3d printing
Your on point in the end of the video. I use the scanner for getting references that are almost impossible to measure with the caliper, then reverse engineer or use for references only. Even on a very expensive scanner you still get a dumb solid and need to reverse engineer it to get a real cad model. Scanning and directly printing is more for the people doing figures and stuff.
So Scanning with 43min was faster than CAD with 35min?
Was that measured with an employer clock?
I spray in a container that seals wire spraying outside. I then use a water color bush with the over spray powder for any stubborn regions. I'll also usually just use the brush because it's cleaner and quicker for small parts.
I always go for CAD. For some reason I always heavily struggle with editing meshes/objects, let alone cleaning up a messy 3D scan. With CAD you just design what you need while taking 3D printing limitations in mind at the same time.
Can you please do a vid on filament drying as far as what to get and or how to make your own?
Interestingly enough. The 3D Makerpro Lynx, seems to be a rebranded Creality CR-01 even down to the softwear. I've used the CR-01 for about a year now, it can get decent detail, especially with the turntable for smaller parts.
Why didnt you want to show the inside of the flush box? Just wondering…
I was wondering the same thing. What is so private about showing the inside of a flush box. I don’t get it. Unless it was modified, and no longer meets regulation standards?
It would be interesting to see a follow-up on how this part holds up. My only concern area would be the pin that extends out.
A nicely timed video, thank you. I was half considering a scanner. For now, I think I will stick with Vernier callipers, a micrometer and Sketchup.
It'd be interesting to see 3D scanning & 3D printing used in a loop, like deepfrying a jpg. Starting with a low-poly model would probably serve as a good noise seed as the verts & edges will round over to various degrees, where a smooth model may bias noise around detail more than on surfaces
Out of curiosity, what is so special about the flush box that makes it unsuitable for sharing? I cannot imagine there being anything personally identifiable
I'm dying to know this, too! WTF??
Great video Thomas, love your contributions!
Let me give you one, in most CAD you can put an image on the background and planes, what I do in reverse engineering a part is take a picture of each profile and put on each plane and use a sketch feature to size it so it is the right size.
Once you do this you can cut your measurement time down greatly and get really accurate parts, last set of parts I did were correct on first printing, and they had some weird shapes.
My personal perference is these situations in to take a couple of photos and either scale as image in CAD and trace it (or create an svg to speed the trace a little) and then combine with measuring with calipers.
How would this change if you are someone that is not so good at cad, would the clean-up of the scanned part still need the cad skills ?
Think a fair comparison would be a new person tackling both workflows. Then compare the times.
Not gonna lie, I thought CAD would be a dominant runaway winner, but it's impressive that it was so close. I really enjoyed this video!
3:36 is as far as I made it before I realized I will never have to worry about buying one of those scanners.
What I would often do is take a photo of the part with a ruler (from high enough above to minimize distortions) and import it into the CAD.
I then measure the fine details and precise lengths with calipers where it matters and use the photo (properly scaled) for the overall shape. But I like the approach with extending the angles too! Still, the photo is a great way to quickly spot any potential mistakes (like the rubbing nub) right away, most of the time my parts work in their first revision 🙂
Also, CAD gives us a very easy way to strenghten the parts where the extra mass would not block proper operation. With 3D scanned mesh this is more tedious. I tend to do that a lot, especially to prevent the original failure mode.
I think that the results of scanning of very product/software dependent. I have tried a few different ones and found that Polycam works best for me. It is a different process than what is shown in the video but it works well for me.
I also do use the take a photo and calibrate it to use as a background to draw over. That works well, but there can be an additional step of lens compensation. There are many ways to it it, but it is important to be aware of. If not parallax effects can happen and lines will not be straight.
Thanks for showing up the online problems. This will always be a problem for me because a few years down the road it may fail completely. I think these companies have some good reasons for doing it this way but they’re not good for me. I’m sure that they’re aware that recent phones can scan with their fancy cameras and without all the online hassle. What advantage are they offering?
My calliper and I do a lot of bonding but one area where I fail is in the compound curves. I can’t measure those well enough to print a conforming fit. Haven’t tried a scanner yet.
Been getting some good scan results with a Creality CR-01 3d scanner of full size car parts and guitars and orthotic foot arch scans to make better fitting shoes or printable flipflops that wont damage the arch.
Its not good for smaller objects but what a vast improvement over previous scanners. And theres a bunch of newer chipset version scanners making for some very useful reasons to have one or a few.
The CAD advantage is clear. It also gives the user the ability to create parts rather than just replication of broken ones. I am struggling to learn how to model in CAD but I consider that a set of skills that unlocks the capabilities of 3d printing.
I just recently also compared EinScan HX to simply making my own modle. I tried a house key, a circuit card for my car fob, and a visor for a GPS screen.
A scanner is neat, but it's not for duplicating objects and more of a fancy reference maker or for artistic use only. I much preferred calipers to capture all primary features and then taking pictures with rulers on the primary features to capture secondary and third features.
Also, dont sleep on document scanners. Not only are they ridiculous, accurate, but they do export dimensions metadata that can be picked up by some CADs. I did the circuit card and key like this..
....most important though, you got it fixed! And spot on about why 3d printers and the stuff we have are so great to have these days, diy all the way!
For the Einscan SE, try to get an older version of the software from around 2019. The application is stand alone and works offline after activation. I have my scanning computer disconnected from the internet for security reasons and this arrangement works.
Another benefit is you can make a copy of the Einscan Folder with the application and keep it backed up on a flash drive. You can move it to any other computer and the Einscan will work with that version of the activated application. I had to do this when my previous computer finally died and I had to move it to a newer computer.
Where can I find an older version? Einscan don't provide legacy versions of their software.
@@MadeWithLayersif only it was an open source project with a GitHub page 😂
I usually trace the part or redraw the part then put measurements like old school drafting then I take those measurements to the computer. To me it seems smoother that way you can figure out how you are going to actually start the CAD process.
what was that about open source being bad, where did your scanner go.
Thanks, very interesting comparison. Phones have such excellent cameras these days; is there such a thing as a pure software 3D scanner app that works?
KIRI Engine works surprisingly well. You can try it for free.
@@duanestrong9538 thanks, I'll give it a try.
I've used a polycam scan to capture the area where my cad designed part will go, so I can make sure my design will fit where it's supposed to go
I tend to trace the part on a piece of paper and then scan that into a jpeg to import into fusion 360. And then trace, using calipers to help confirm scaling and any tricky parts
If we’re 3D printing the part because it broke, going the CAD way is an opportunity to improve it, strengthening where it broke-and optimizing it for printing as well.
A great example of a use case with a good comparison! Nicely done Thomas!
For complex parts I use 3d scans to use as reference for drawing the CAD part.
3D scanning is also awesome in areas where you can not simply measure. I needed the inside of an container with wild angles and curved planes. I didnt even get my tool inside to measure.
Depends, some parts I like to reverse build just via measuring, but for example I build a MagSafe adapter for my car using a 3d scan of the dashboard to fit it perfectly. Know your tools and limitations 👍🏻
I use regular flatbed scanner.add a coin for scale. then import that image into cad. Then i trace it. Works wonders for angles.
I am still a fan of doing it the CAD way. I find it really relaxing.
Thanks for the demo of scanning software.
Anyone have any bets on how many generations of scan/prints you could get before the object isn't recognizable anymore?
🍻
A alternative way to get overall perimeter details is to take a picture of the part using a long focal length to eliminate DOF and then to scale the image to a known dimension. This workflow could result in a little faster reverse-enginering of the part.
You don't need to be "fair to the scanning". In fact your goal here is to get a functioning part which can be 3D printed and that doesn't break like the original did.
That original part was once designed with the process of injection molding in mind as the method of manufacturing. But that's not what you're doing, so you might as well ignore unnecessary details and optimize the part for 3d printing right away. You dont want to copy what didn't work anyway, right? ;-)
sod using online/cloud and huge lockdowns, as you just showed I learned this the hardway never again.
At my internship I've been trying to recreate small turbine blades so we could recreate them out of composite. I've had trouble scanning thin flattish parts like turbine blades. Recreating it by hand is near impossible. I've done a mix of both. CADing the tolerance specific parts like the blade mount. Then using the 3d scan points to help me make the surface of the blade. It's not the best.
Scanning with the lynx I noticed you were in texture mode not geometry. You might get better results with geometry. I have the mole and that's my experience with small parts like that.
15:00
The crossover we didn't know we wanted
I am not sure if that vertical rod feature in the 3D-Printed part will hold up especially to repeated use. There is also the fact that PLA creaps significantly under constant load.
The rod is a simple bumpstop that just keeps the lever from bouncing up too far. There's no permanent load on any area of the part.
I find CAD is also well-suited for making design improvements to the replacement part - suppose the rest of the system has worn out around the part such that an exact copy won't have quite the right dimensions anymore. (Or it wasn't right to start with.)
In fairness to the Lynx, it was advertised for larger objects. Although I've used mine fairly successfully for smaller parts after a bit of love and patience
Durring the resconstrctions, many times I had an idea, that normal paper office scanner could be also also helpful. It can create very precise 2D contour in orthographic projection that can be very quickly reconstructed as sketch. Unfortunatelly never had it at home to test it. Also in this case it could speed up the reconstruction.
I use my digital caliper for parts. However had a delta wing project that required the model of the EPP foam wing that did not exist and the frame was 500mm so I found Polycam for iOS and Android did the job remarkably well compared to the $300+ 3d scanners.
I used an IPhone to scan a part of an rc plane to then design a new canopy for it. The plane has quite irregular shape so getting the 3D scan as a model helped a lot.
I prefer CAD, too (maybe because it is my only option :) ). But I take a bird's eye view of the part to see how it may be simplified or improved. Parts that are made by injection molding often have a different set of constraints and material saving measures that result in part failures at critical points. Beef them up at the weak point.
Scans never seems to have "stable" surfaces for me. They always have weird artifacts, are slightly off, or just straight up wrong. I far prefer measuring for CAD. Because then at least the model will be off by a common amount.
The Lynx scanner is for larger objects. The Mole scanner is for medium sized objects. This model is perhaps too small for the Lynx - better suited for the Mole scanner. This is a personal observation - not a commercial - I have both scanners and would use the Mole in this case. Better resolution for this size. :-)
I am surprised you didn't also use a background image in cad while doing your 3d modeling. As it might of helped you both get the scale right and avoid the error you had.
I would have placed the part on the flatbed scanner, model the basic shape ontop of the image in FreeCAD and add the details by hand. This was much faster than dealing with the post scan model. Another side effect is that I end up with a proper parametric model, too.
I was curious what you used to make the one-to-one replica of the mechanism: CAD or 3D scan? Based on the ending I suspect I know the answer. I just think it's worth pointing out the perfect natural experiment: once you've tried a couple techniques, see which one you /want/ to use when you actually need it
I´m not sure if the Lynx was the best option for this small part, because even the manufacturer write it on the product page that its made for bigger objects.
Maybe the 3dmakerpro Mole or a Revopoint Mini would fit better for this kind of parts. The einscan-se should handle it very easy. Mine works fine with small parts, but ya the activation is sh*t.
I know the Einscan coule most likely have gotten me a better result (which is why I tried it first), but I'll take a 3D scanner that actually lets me use it over one that locks me out any day.
i understand that. The activation process is from the hell :D But my point is, that the lynx is not the perfect choice for this type of parts ^^@@MadeWithLayers
Great video! I’m not sure what that flush box is though? We don’t have those in the US as far as I know?
I feel that a 3D scan is good for reference OR for amorphous objects like a face. But 3D scans lack the precision required for technical parts.
@MadeWithLayers Thanks for the heads-up on Shining3D's misauthentication process Tom!
I already have one large paperweight in the form of a NextEngine where the software stack toppled permanently.
I enjoyed the sponsor spot knowing the 'window' is already imaginary!
There's some harsh audio issues during the PIA advertisement section. Particularly around 1:25.
So this convinced me to not get a 3D scanner, since often I have 10 minutes here, 5 minutes there to work on a CAD part. So spending money on a scanner doesn't look like a useful improvment for me.
The scan the environment concept sounds cool though as a use case like you said! I'd be currious how effective generic photogrametry would be for that too.