Challenge scanning minis with the Mini

[JamesR]

I have a personal project where I want to scan some very old pewter 28mm models so I can scale them up, modify them and print them for my own use. I backed the Revopoint Mini in the hopes that it would be able to accomplish this task.

I am completely new to scanning so I understand it takes a measure of skill to produce good scans. I have read though lots of threads here trying to pick up as many tips and tricks as I can but I still cannot seem to produce a satisfactory result.

I seem to have reached the limit of my current capabilities but the real question is; have a reached the limits of the Revopoint Mini’s capabilities? I could really use some help in either making more progress or coming to the conclusion that this is as good as it gets.

My main challenge is getting “resolution” and crispness in the details. In short it the model it too “blurry”.

So, What have I tried so far:
Software and Firmware
I am using the latest firmware version and the latest versions of Revo Scan / Studio on Windows. Through all the steps (fusing, meshing) I have tried to preserve the highest level of details opting rather for noise than smoothness so I do not loose any details.

I have not tried any other point cloud processing software. If someone has a recommendation for a free tool or even a reasonably priced tool, I would be happy to give it a try.

Calibration
I have calibrated the mini to 0.0530 which according to threads here should be well within an acceptable range. I have tried to get better results but the calibration program tends to crash making it a bit frustrating.

Surface material
I understand the principles of light absorption and reflection. To eliminate the scanned material as a factor I bought a scanning spray, AESUB blue. I have tried using a lot of spray and very little spray but although it help it is not enough. I read that PUTV claims ATTBLIME 2h provides much better precision but honestly I am not sure I at a point yet were that will bring me the rest of the way. Or maybe I am wrong. If the consensus is that it will make all the difference, I will absolutely give it a try.

Lighting
I have tried everything from lots of light to no light at all (complete darkness). I am using a black felt background to eliminate interference. In general I have found that more light leads to more noise in the scan. So far I am getting the cleanest result having the model in a felt covered box with only natural sunlight coming in.
Sunlight is nice, but it is hard to control, especially far up north in the winter where there is very little of it… Could someone suggest a good lamp with an appropriate profile to be used with the Mini?

Turn table
I am using the turn table that came with the Mini but following general recommendations here to minimize the number of overlapping frames, I am using an external App to control it. I have also tried just doing partial revolutions in case the turning of the model around the edge throws Revo Scan off (like seen in the image below) but that has not seemed to make a meaningful difference.

Distance to model
I have tried very close to far away. Here closer seems to be better as long as the scanner can focus on the model.

Merging multiple scans
I have not tried merging scans from multiple angles yet. My assumption is that if I cannot get a single pass clean enough, multiple passes will not necessarily make things better.

Any tips this community might have would be very appreciated.

[Martijn]

I think it can be scanned a bit better, with more detail.

The things I can think of now are:

Try to get the distance as close as possible, on the edge of “too near” and just into “Excellent” about a bit more than 10cm, scanner to object.

The object seem to have a bit like shiny metal edges, that is something the scanner doesn’t like. The
use of a mattening scanning spray or powder would help a lot to get more detail.

Then one thing when using the Dual axis turntable, it takes a lot of frames (60 frames) at the same place before the table starts turning, this is not good for detail. Overlapping or multiple scans on the same surface reduces the final detail. One solution to this is to start the scan at a place where there is less detail, for instance the backside of the object. The loss of detail is less visible on flat surfaces, and the part with detail is only scanned once. This wil give much better detail.

Finally external lighting, this can cause detail loss or when the light changes a lot, it can cause differences in scans of the same object.

Hope this helps you get a better scan.

[PUTV]

Hi @JamesR

The results are not optimal and indeed too blurry and the only reason here I see is the distance , for the best result you really need to keep it at 10 cm from the MINI front face , do not lay it flat on the plate , but straight, you can use some support behind, do not use any light if you can , results will be better .
Use pitch point fuse setting at 0.02
Use meshing at level 6 , no denoiser ( on the phone or tablet do not mesh it or you lose the quality, just fuse it and export for future meshing in Revo Studio .

Do not over scan it multiple times , limit the capture to minimum for best quality .

You want more sharpness skip AESUB spray since it has 15 microns surface , use Attblime AB2 2h it has 3 microns surface you will get more sharpness from the surface , like comparing snow is to frost

Anyway keep practicing, when you capture a fingerprint on your fingers what are 7 microns then you reached the maximum .

That’s the basic for perfect sharpness

[penneyduck]

Try viewing in Meshmixer.
This will show more detail than Rev Studio.

[PUTV]

Hi @penneyduck he did in Zbrush

But I see what’s going on , this is not his fault… I just need to know what software version he used and firmware version of MINI.

@JamesR I just opened your Zbrush screenshot, the fusion is not correct, it is not your fault… but the software , please let me know what platform you use and firmware .

This need to get fixed … I will talk on Monday to the team after performing some tests to see if it is not calibration issue you have .

If you get a chance please perform Calibration test with your MINI and let me know your values , make 3 tests before calibration , and let me know .

[JamesR]

Thanks, this was a lot of good advice. Unfortunately I tried most of these things…

@PUTV. I imported the scanned model into ZBrush to see if I could improve on it. I realized that ZBrush was really good at presenting the details of the model much better than a normal 3d viewer.

The version of Revo Scan I am using is 4.1.1.20220930b on Windows 11. I have been monitoring the download page for new versions but this is still the latest version I can find there.

Firmware version 2.8.33.20220903
Algorithm version 7.6.9.0816

Testing updating the firmware now and it found a new version… Updating now.

[PUTV]

Ok great , please make test using the Calibration software after you updated your firmware , you find it under Starting Point/ Download . This will give me the final clue.

[JamesR]

Updated to the latest firmware. 2,9,23,20221031

I ran the calibration test a couple times in sequence:

0.0331
0,0405
0,0713
0,0596
0,0756
0,0719

[PUTV]

Ok thank you … the calibration values looks ok
I need to run a couple of tests .
I will let you know in about 2 hours or early when I am home , so you know what’s going on exactly .

But I am more than sure that the results you getting are not because you did something wrong.
The burlap effect on the mesh is result of a bad fusion algorithms.

MINI should have very smooth surface without any of it .

Even after fusing on the phone

[PUTV]

Hi @JamesR

I made today couple of tests with different devices , the latest firmware seems to be ok ,
Please recalibrate your scanner if you have a chance .

the results was fine and sharp at 10cm scanning distance with versions from 4.0.2 to 4.1.1

Fusing at 0.02mm
Meshed level 6, No Denoiser

[JamesR]

Managed to recalibrate the machine (after 4 crashes…) Post firmware upgrade and calibration the calibration test averaged about 0.05. But honestly, I am trusting it less and less. The variance is in the test results ±0.2 in subsequence runs even if I do not move anything…

I scanned the reference model so I would have a 1:1 comparison and this is the result. Your image is on the left, mine of the right.

I was going to write: “I am lacking the definition or the sharpness.” But looking more closely at my bust, it actually is smooth where yours is sharp, in for instance the ears. It must be a degradation of the mold they used when casting them…

In any case,
The new firmware does give me a “cleaner” result, that is it does not have the same graininess. But I am still not able to get more more detail from the scanner.

(Left is before the upgrade. The two right are after the upgrade.)

I have ordered a can of Attblime AB2 so I can eliminate the spray as a factor. Unfortunately not the easiest thing to get ahold of where I live so it will take a couple days before I can test it.

[PUTV]

Good that you made the bust scan , so I can see the surface , the fusing is still not 100% where I would want it to be in general . But the burlap effect got less visible what is improvement.

My bust was one of the first busts from the mold that came with my Beta prototype so it is the sharpest but not so perfect either as lots of air pockets.

Did you measured the distance correctly ? Try to around 8-9 cm in different positions to see difference.

The fuse pitch point settings is responsible for the scanned model resolution , but I remember scanning coins for the campaign, it took me some time to find the good angle , scanning relief is tricky.
Try marker mode and flat on the plate 360 degrees and see the result compared to standing straight, do not over scan multiple times , and see if there is any difference. when scanning with AESUB Blue is like scanning at 0.04mm pitch point in place of 0.02 , and when you get Attblime do not over spray it , even if you still see color it will be already enough , most important you get off the shine as that is all about .

[JamesR]

Thanks for the help.
I will report back in a day or two when I get a chance to test.

[Erked]

@PUTV

when you get Attblime do not over spray it , even if you still see color it will be already enough , most important you get off the shine as that is all about .

When scanning for high accuracy this information is of utmost importance for capturing as much detail as possible.

[JamesR]

I have been experimenting some more and I have made some small incremental improvements but I am still no where near to level of details I would like to see.

The Attblime AB2 spray arrived and it did help a bit. I tested being extremely conservative using only using a single sprits and worked my way though adding more and more. I think I found a good middle ground where I am getting the best results from the spray.

I tried both scanning the model standing and lying flat on the turn table in marker mode. With Aesub it did not make a difference but with Attblime I actually got noticeably more details in marker mode. My thinking here is either that it is the angle being more acute that is making the difference or that the tracking dots give the scanner more accuracy when the software fuses the point cloud.

At this point I am at loss what to do next. I have tried tweaking;
Lighting: Little to no external light. Scanning in a box coved in black felt.
Distance: As close as I can get without the scan cutting out. (~10cm)
Surface: Attblime AB2 scanning spray (minimal amount)
Angle: I have tried scanning from several different angles.
Scanning mode: Standing with Feature mode and lying down with Marker mode
Scanning frames: Scan a single revolution of the turn table. Control the turn table from an external app for I do not get any extra frames when it is still.
Fuse and merge: Use to values that preserve to most details (Fusing at 0.02mm Meshed level 6, No Denoiser)
Calibration: Recalibrated several times. The best I can get it is about 0.055

I am thinking the culprit is one or more of the following:

• Firmware: Before upgrading to the latest firmware I got a “burlap” texture but it actually preserved more structure. Since upgrading the results have been “cleaner” but more washed out or blurred. This leads me to think the device is capturing more detail than it is capable of translating into the model.
• Software: When looking at the preview it seems the machine can “see” more detail than it is able to preserve in the fusing. This leads me to think that the software is not capable of fusing with sufficient accuracy. This can either be a software problem or a source data problem. (Which may be because I am not giving it enough to track?)
• Hardware: Maybe I am actually getting results at the limits of what the hardware can deliver. This can either be because I have reached the limits of the product as it was designed or that my physical unit is subtly flawed.

Given the quality of the Revo Scan and Calibration programs I am leaning toward underdeveloped software / firmware. But given that PUTV manages to get good quality scans it may be hardware? Or maybe there are more tricks I am not aware of yet?

Are there any alternative programs I can test that will process the point cloud so I can at least exclude “software” as a factor?

[PUTV]

You did everything possible indeed … one last thing would be to scan something printed where the layers are visible like 0.2mm or 0.3 mm print layers to see if the scanner can catch it , it should without issues .

BTW sent you PM, check out …

[Martijn]

Did you try scanning it without the turntable, so only flat in front and then fusing it.
Sometimes minimal movement/overlapping frames gave me a better result.

[SphaeroX]

Can someone Upload an stl of a scanned 3d Print? I would really like to see if you can see the individual layers.

[PUTV]

@SphaeroX seems like you doubt in my words here.

there is already whole thread about people scanning 3D prints with MINI to capture printer layers and with success , sorry no link under my hand at this moment.
It was a thread about a vertical lines under Depth cam preview.

[MrVette]

Here is an image of an STL file. Unfortunately I can’t upload the file. it is not allowed as STL