0.05mm accuracy seems unobtainable?

A new video on youtube has been posted where accuracy of pop2 is measured and best case seems to be 0.2-0.3mm and not the claimed 0.05mm.
Is there a way to actually get the claimed 0.05mm accuracy with pop2?

At 0:23, he mentioned “single frame accuracy” having the most precise results, but is not very “realistic”.

Revopoint’s specifications for the POP 2 are listed here: https://www.revopoint3d.com/pop-2-spec-2/ . The 0.05mm accuracy is clearly stated as being “Single capture accuracy” and it has a footnote that this is a lab measurement that will probably not be repeated in real world usage.

According to the Accuracy test video by Revopoint and the official papers the actual lab accuracy of a single capture went down to below 0.026 mm , there is so many factors that depends on it .
To even get close to 0.05mm you need to have perfect stabilization between the scanner and the object . Not on some cheap shaki turntable .
There is also a huge difference between older Handyscan and newest Revo Scan that affect it on so many levels .

I wrote a lengthy reply but gave up, I’ll just wait for mine to arrive and then see…

I saw in the video that he showed the meshing settings, but not the fusing settings. Maybe he forgot to switch from the preset 0.2 to 0.1?

@Eldkatten I checked the video , I saw the scan pop up in the feature mode , this guy is measuring the ** mesh volume resolution** not the Accuracy , mesh volume resolution and scan accuracy are 2 different things .
All he measured is the quality of Revo Scan meshing the object LOL

Marker mode in Revo Studio are for visual tracking only and not for measuring the object

The best way to measure accuracy is having 2 object and measure its distance .
Please remember that not all 3D scanners with an Accuracy of 0.05mm will provide a precise volume resolution of 0.05mm it depends also on the software and many other factors. There are many scanners that provide the accuracy of 0.05mm but the volume resolution is 0.1mm or 0.2mm or 0.060mm
You can observe the phenomena using Handy Scan vs Revo Scan vs Revo Studio , same scan accuracy but different Volume resolution aka precision . You will get completely different results measuring the volume precision using all the 3 software .

POP2 Accuracy is the single distance between point A to point B ( point cloud) , you need to measure it between the scan ( one frame ) and the distance on a real object .

2 Likes

I forgot to add that if a 3D scanner Point Accuracy is at 0.1mm the 3D resolution should be around 0.2mm to 0.5mm what is excellent .

What this means ? this means if POP2 Point Accuracy is at 0.05mm the 3D resolution should be at least 0.1mm to 0.3mm ( if the guy from the video cleaned the noises he would get exactly the right masuments)

You will never find a 3D scanner that have 0.05mm point accuracy and 0.05mm 3D resolution , not even if you pay $25 K for it .

So please stop splitting the hair in 2 and educate yourself on Youtube !
We all know by now that Revo Studio or Meshlab can provide better 3D resolution of the meshes than Revo Scan software. I posted the difference before.

2 Likes

Thanks for the detailed spec page. The reference on the sales web page ( POP 2 3D High-Precision Scanner |handheld 3D scanner, 3d printer scanner – Revopoint 3D ) didn’t show that.

The fusing setting was set to 0.1mm. Just remember that Accuracy, Resolution, and Precision are three different things. For most people, the result will end up being a combination of scanner, turntable, software, and the object being scanned.

I did a follow up test ( Revopoint POP2 Single View Accuracy and Precision Tests - YouTube ) that shows that you can achieve the specified single frame accuracy. The turntable test was meant to show what you can expect when you align hundreds of images together. 0.2-0.3mm is still pretty good for my applications. I expect most people will need more than a single frame for their projects :slight_smile: For the consumer, Accuracy ends up being the combination of all the pieces (scanner, turntable, software).

@nafis Chris , you made again mistake not setting the pitch point to 0.1m , sometimes it jump to 0.2 m , so before you fuse your point cloud make sure the pitch point is accurate . Making accuracy tests at home the way you doing will never deliver proper results , one of them is Revo Scan meshing is bad, too much overlapping of points and lots of noises . You should export the point cloud , clean it up and then mesh it proper and compare , because all results now are based on a buggy software and not actually the device . Videos like that can misleiding users as it did in your preview tests of the volume resolution . Accuracy tests is between point A to point B , and all you doing is still volume resolution what in this case is 0.1mm or maybe it was at 0.2mm since you never check pitch point settings .
Try to measure the distance between 2 objects and check it with your software and that should be always accurate , volume resolution will be never the same as many factor play role here .

here I give you an example of the same unprocessed point cloud meshed in 2 different software


for the same reason any Accuracy tests can’t be done while meshing in Revo Scan .

I expect most people will use the scanner, turntable, and software as a package. I’m interested in trying to quantify what you can expect. If you want to improve something, then you need to know where you are. I expect the firmware and software to be constantly improving (hopefully). Sure If I had a couple of thousand dollars I would get a traceable ceramic ballbar and follow VDI/VDE 2634. I have set and checked that the point spacing was set to 0.1mm. I also checked the mesh. Sure I can export the points and mesh in different software, but that’s not what the majority of people are going to do. Even with the ball-bar, you still have to best-fit the ball scans to a sphere. Then the distance between the spheres is one measure of accuracy.

I have seen the video and read the posts in this thread discussing different settings and terms like ‘precision’ and ‘volumetric accuracy’ and so on. While I have a great deal of respect for Chris’s ingenuity, I’m not convinced that his was a final, definitive accuracy test of this product. I say this because I have also seen a video by “jonhimself” where he scans an engine block and reports on the accuracy of the results. Again, it isn’t a lab test but a real world application in which he validates the results.
Something I am sure of from my own experience with other scanning devices, is that the accuracy of any scan depends on much more that the potential maximum capability of the device😉.

1 Like

@nafis Chris , the Windows Revo Scan just updated this morning with fantastic results , you should check it out and make another test , ot may be much better . The point cloud is so much more uniform than before , less noises .

And one more thing Chris , if you want really tests things out , you should use different 3D Spray, the best are the Dental 3D Sprays , as it make sure the accuracy is on top .
You know the surface can influence the noise levels and so do the 3D spray, for this precise work , you need to use precise elements .

And talking about packages, are you testing accuracy of Revo Scan software or the device ? that is the question people want to know … because I do care what the device deliver not the software. You also have Revo Studio , it is not ending at Revo Scan only for those that don’t know nothing about 3D scanning .

@UK_design , I have nothing against Chris , but claiming something based on a early buggy software that could not deliver proper mesh is absurd . The precision of the mesh is 75% lower quality than the point cloud itself . How you can claim anything based on that . That why I have problem with it.

And that is true , everything will influence the accuracy , from a shaky turnable to reflective surface or heavy load of 3D Spray . In jewelry work and dental we don’t use this kind of 3D spray , as the accuracy is most important for the client , so very fine matte spray is used .

For me right now POP2 deliver what I need it for at the exact precision , so whatever the accuracy tests claiming , it does not affect my work.

1 Like

That’s great to hear. Any chance you have access to step blocks? It would be great to have other people try to gage/document accuracy. Maybe you could export the point cloud before meshing and try several of your other packages? Spray is a tough issue. I’ve had good results using magnaflux, but spraying in general is a bad idea. In test that I did at work, ten different people will put ten different amounts of coating down. That’s why the ceramic balls with the special coatings are so nice (and expensive).

1 Like

Jon’s engine block is a very nice example of a complex/large part. I love how he shows his approach of controlling errors. It’s a shame an independent lab just couldn’t run the VDI/VDE 2634 on all the available 3D scanners. In my tests, I’m trying to quantify things with inexpensive objects that normal users could afford. Hopefully these tests will show Revopoints constant improvements. I’m trying to be as open as possible. I showed what I tested and how I did it. I am always open to improved methods and constructive feedback. That’s how the 3D scanner community can move forward. I love seeing all the examples people are doing.





I just downloaded the latest version of Revoscan. I did a feature scan of the step block with the turntable ring to help control alignments. The scan looks very nice and clean, but the alignment is still off? The combined scan is about 0.3mm smaller than it should be. I used 0.1mm spacing, no denoising, and mesh quality 6. I need to do a better job constraining the alignments.

2 Likes

I actually don’t see a problem with your latest attempt at all! Now that everything is so consistent, you say 0.3mm smaller, just scale up everything from now on and you’re all set. Much more important to get the group on a target than a single bullseye and the rest spread out.
Anyway, I now almost feel sorry for having started this thread. All I wanted was to be able to scan a gauge block and see how far off it is on-screen and be able to do exactly what you did on your last attempt - scale up or down by % and get the 3d model measurements down to within 0.1mm of the gauge block.
Then I would be able to scan a part for someone and tell them with confidence that:
“Here a usb stick and yes, you can now CAM and machine this part and it will be a copy of the original down to 0.1mm and the critical dimension you can micrometer and machine to your desired tolerances”.
With 0.5mm here and 0.12mm there I wouldn’t be able to do that. And to most machinists 0.5mm tolerance is laughing stock no matter the application.
Just to bring a little back story - when and IF my pop2 arrives, my first task will be to scan an elaborate broken tube bender clamp support (operator error) which is unobtainable nowdays so that my friend who owns the bender can have it CNC-machined. And he was quoted $$$$ to have it 3d modelled for CAM. Yes 4 figures. And with the pop2 I was hoping I can scan it and just model critical surfaces by hand, ready for CAM in minutes.

The thickness of AESUB Blue simple layer is approximately 0.015mm
Then you have TiO2 powder mixed with Isopropanol around 0.005mm but toxic as hell .
Then you have MAXIDON for very fine dental scan jobs , it works great on metal , waxes, ceramic and other surfaces.
There is more fine 3D sprays used in jewelry making and precision work , but I am going to save the accuracy tests for the upcoming MINI Blue Light Scanner.