Scanything setup and configuration questions

[David Buchhauser]

Les,
Will the camera with fixed exposure help with the calibration (as opposed to the camera with variable exposure)?
Thanks,
David

[David Buchhauser]

I noticed the camera in the Arclight Dynamics auto tracer demo video seems to be a lot closer to the table compared to the distance I was using. Does this make a difference as long as the camera is in focus. Also the Arclight Dynamics camera does not seem to have any internal light source, or maybe I'm just not seeing it in the video.

Will more or brighter light help to calibrate more easily?
Thanks,
David

https://www.youtube.com/watch?v=TjKcyQaeus4

Arclight Dynamics Auto Tracer option.jpg (129.66 KiB) Viewed 5269 times

[Les Newell]

When calibrating and running watch the 'mask' view. This is what the edge follower actually sees. If you get flaky calibration and randomly losing the edge it is often due to the brightness setting. Again if you watch the mask while adjusting the brightness you will hopefully find the setting that gives the clearest view.

The camera distance mainly affects accuracy. The closer you run the more accurate the results. Conversely the further away you run the faster you can scan. Depending on the amount of ambient light you may also struggle to get enough light on the subject if you run too far away. Keep an eye on the video FPS. If it drops significantly below 30 you need more light. The general rule is that you can't have too much light, though reflections can be a problem with reflective surfaces.

By the way a useful trick is to deliberately run out of focus if your edge is a but rough, for instance if you are tracing a wooden template. Being out of focus will smooth the edges without having a major impact on accuracy. The biggest down side is that corners wll also be smoothed and rounded.

Looking at your video it appears to be tracking quite well, if a bit slowly. Try experimenting with the feed rate and servo gain. Servo gain controls how hard it tries to keep the circle in the center of the screen. Keep winding it up until you start seeing or hearing the machine hunting for position. You may want to deliberately crank it up too high just to get a feel for what that looks like.

As with camera distance there is a tradeoff between speed and accuracy. These cmos cameras take time to scan the image. This means that the top of the image is slightly older then the bottom of the image, causing distortion. You often see videos on YouTube where fast moving objects appear oddly distorted. This is the same effect. Again more light helps as this shortens the exposure time. The XBox Live Vision cameras apparently use a CCD sensor which does not suffer from this distortion effect. The ELP cameras are CMOS but they are pretty sensitive so as long as there is plenty of light they have a short exposure time.

I am wondering if going to the better camera (suggested/recommended by Les) and adding the led light ring around the camera (also suggested/recommended by Les) would possibly help it to calibrate better.

Yes it should hekp a bit. Scanything relies on detecting the transition between light and dark at the edge. Cameras with auto exposure are continuously adjusting the brightness of the image so that target keeps changing. When scanning a drawn line like your example the advantage isn't that great because the average ratio of dark to light in the image doesn't change a huge amount. However it is very noticeable if you are scanning solid shapes. Take for instance scanning a black square. When scanning along the edge half the image is black. The camera is trying to keep the average light level constant so it cranks up the exposure. Now you get to a corner. Now only about a quarther of the image is black so the camera drops the exposure. Cameras that allow a fixed exposure remove that variable. However you do also need a fixed light level, hence needing a bright ring light.

I really don't understand why using the ethernet/ess instead of a parallel port would cause problems with calibration. Is this due to some latency with the data transmission thru the ethernet/ess as opposed to a quicker response with the parallel port.

That is the main problem. The whole system is effectively a big feedback loop. Feedback loops really don't like any delays in the system. In previous tests I found the ESS had too much delay. It is possible they have improved things in that area.

[David Buchhauser]

Hi Les,
Thanks for your reply. The new camera I ordered should be here tomorrow. I will report back when I get it running and let you know how it works out. My question on the feedback loop is this: since the ESS ethernet connection may be causing some latency (compared to a parallel port connection) in the time it takes to jog the machine when commanded by Scanything, would it be possible to introduce a short dwell time into your software to compensate for this?
Thanks,
David

[David Buchhauser]

Hi Les,
I think I've got the new camera and led lamp ring working pretty well. It now seems to calibrate much easier and faster, and tracks pretty well once I get the settings right for a particular situation. I have experimented with the camera adjustable aperture. It seems to work best with it adjusted to wide open. I need the Exposure slider all the way to the left (minimum setting) for it to work. If I turn that setting up, then I lose the camera image on the screen. Here are some photos of the items I have scanned, along with the corresponding videos.

Les - I do have a question. If you take a look at the videos, you can see that the camera target circles seem to bounce back and forth along the path as it tracks the edge of the part. Any idea what causes this, and should I be concerned that it is affecting the trace file? It seems to decrease as the camera slows down around corners, and it isn't as noticeable at very slow speeds (but still there).

My fps stays around 30, and the CPU load around 45% regardless of my settings changes. Should I be concerned, or is this OK?

So far I am very impressed with the Scanything operation now that I have the better camera and lighting. There is one feature that I wish Scanything had, and that is the ability to save a partial trace without having to trace all the way around the part. I can see where it might become frustrating with larger parts where it makes it almost all the way around the part, but then loses the edge. It would be nice to have the ability to save this partial scan, and restart from where it lost the edge and finish the scan. Would this feature be something you might consider implementing in a future version of the program?

Thanks,
David


https://www.youtube.com/watch?v=PGVjuis ... e=youtu.be

20200306_005641_resized.jpg (446.67 KiB) Viewed 5219 times

https://youtu.be/fZyjGonJJgA

20200306_005006_resized.jpg (474.11 KiB) Viewed 5219 times

https://youtu.be/1A1QJEYGzi8

20200306_010306_resized.jpg (421.9 KiB) Viewed 5219 times

[David Buchhauser]

https://youtu.be/vl9FJ-JPF3s

20200306_011750_resized.jpg (399.98 KiB) Viewed 5219 times

https://youtu.be/3gkAuFm2104

20200306_011644_resized.jpg (378.07 KiB) Viewed 5219 times

https://youtu.be/2O7PQdEWUyY

20200306_004238_resized.jpg (1.39 MiB) Viewed 5219 times

[David Buchhauser]

I scanned an open end wrench (shown in the last video). Here is what the resulting dxf file opened in PlasmaCAM Design Edge.

20200306_004633_resized.jpg (818.22 KiB) Viewed 5218 times

Here is what the rabbit dxf file looked like after scanning. I could run the cut path, cut, and it would look pretty much like the original.

20200306_012108_resized.jpg (758 KiB) Viewed 5218 times

I traced a drafter's triangle onto a piece of white paper using a black Sharpie. I wanted to see how closely the resultant dxf file resembled the original drawing as far as straightness of the lines. I think it's pretty good. Obviously it would be better to just draw it in a cad program, but I wanted to see how it would turn out.

20200306_010247_resized.jpg (864.89 KiB) Viewed 5218 times

[David Buchhauser]

This is my camera/ring lamp setup.

20200306_004306_resized.jpg (546.48 KiB) Viewed 5217 times

20200306_004119_resized.jpg (667.07 KiB) Viewed 5217 times

After I scanned the open end wrench, I used Vectric Aspire to do some "node editing" to clean it up. Then I offset the outline of the wrench to the outside to create a slightly larger copy. This will be part of a tool holder board for the cnc router. It will have pockets for the two collet wrenches and an assortment of commonly used router bits. I'll cut a sample first in MDF to check the fit. I added finger slots to facilitate removing the wrench from the pocket.

20200306_025104_resized.jpg (1.61 MiB) Viewed 5217 times

[David Buchhauser]

Hi Les,
I did just purchase a license for Scanything from your website. I think your program will be a great addition to my CAD/CAM arsenal.
Thanks,
David
www.desert-hybrids.com

Business Card 1.jpg (23.33 KiB) Viewed 5211 times

[Les Newell]

Sorry for the delay in replying.

Les - I do have a question. If you take a look at the videos, you can see that the camera target circles seem to bounce back and forth along the path as it tracks the edge of the part. Any idea what causes this, and should I be concerned that it is affecting the trace file? It seems to decrease as the camera slows down around corners, and it isn't as noticeable at very slow speeds (but still there).

That's pretty normal. As long as the circle stays roughly near the middle it won't affect accuracy. The actual image to vector conversion is a lot faster than the machine movement. It tends to scan a bit then stop and wait for the machine to catch up.

My fps stays around 30, and the CPU load around 45% regardless of my settings changes. Should I be concerned, or is this OK?

The FPS is good. CPU load seems a little high but not too bad. Smoothing is a CPU hog so turning it down may help. Running slightly out of focus has a very similar effect to smoothing.

There is one feature that I wish Scanything had, and that is the ability to save a partial trace without having to trace all the way around the part.

Yup, that's in the next release, along with lots of other changes. It's taking me a lot longer than I'd like to get the new version working. At the moment I can't give an estimate for how long it will take.

You have been scanning some pretty tricky surfaces. Reflective surfaces are hard to scan due to stray reflections. If the item is waterproof, a few light coats of temporary black hair spray can help a lot https://www.ebay.co.uk/itm/Party-Succes ... Swt-BeEgsR. It can be washed off afterwards. Some hair sprays can be glossy but you can get around that by using several light mist coats. I haven't yet found a good matt black water soluble spray.

I'd recommend using mainly the mask or mixed views when setting up. You get a much clearer view of what Scanything is catually seeing.

[David Buchhauser]

Hi Les,
Thanks for your response. I'm happy to hear that the camera target circles "bouncing" is a normal phenomena, and not a result of my using the ESS instead of the parallel port. I do have one other question with regards to the camera/tracer "accuracy". I will illustrate this with the photos below. I have discovered that if I really slow down the Feedrate, I get very nice results with smooth contours around the trace. If I increase the Feedrate (but still less than 50% of the slider range) portions of the trace become jagged. Is this a normal phenomena, or perhaps caused by either the ESS or a problem with the rigidity of my camera mount? I don't have a problem running the Feedrate slow, and I would rather have a "clean trace" than spend lots of time doing node editing.

I was also wondering if you will at some point in the future be offering versions of Scanything for Mach4 or Centroid CNC12. When I purchased my CNC Router Parts (now Avid CNC) about a year ago, they were offering Mach3/ESS which is what I have. They have now done away with the Mach3 in favor of Mach4. Avid sells a "ton" of machines and I would expect that many of the buyers would be interested in using Scanything with their tables.

Thanks for all of your help!
David

20200307_024557_resized.jpg (797.08 KiB) Viewed 5153 times

20200307_024916_resized.jpg (900.21 KiB) Viewed 5153 times

[Les Newell]

Those jagged edges look like camera movement to me. How solid is your mount?
Another possibility is a timing issue with the ESS communication.
I am working on a version of Scanything that is easier to transfer to different controls. Mach4 is on the to-do list. I don't know if Centroid have a facility for directly controlling the machine movement.

[robertspark]

...... I don't know if Centroid have a facility for directly controlling the machine movement.

UCCNC now has an API installed with the software that allows you to directly control machine movement without a licence but only with UC motion controllers.

API is in c# or VB via .net (I believe..... Not used it though)

[David Buchhauser]

Those jagged edges look like camera movement to me. How solid is your mount?
Another possibility is a timing issue with the ESS communication.
I am working on a version of Scanything that is easier to transfer to different controls. Mach4 is on the to-do list. I don't know if Centroid have a facility for directly controlling the machine movement.

Hi Les,
The mount is not all that solid. It was temporary to test. I am making a more substantial mount from 3/8" thick Aluminum plate. I am thinking it is the camera moving with the mount flexing. I notice when I do a quick jog in the Y direction that I can see the image moving back and forth on the screen. Also I can tap the camera with my finger and see movement on the screen. I will keep you updated on my progress. I should have the new mount finished in the next several days.

I am using the Centroid Acorn controller on one of my other cnc routers. The control program is called CNC12. It is a free download for testing - but they limited the size of the g-code program with the free version. I don't think this would be a problem for testing. I am running their "Pro" version so I can run the Auto Squaring of Software Paired axis with auto homing for my machine. Alot of the guys seem to be switching from Mach to Centroid. I thought I'd try it out. It seems to be pretty well supported, and pretty much "open source" as far as the macros go.

David

https://www.centroidcnc.com/centroid_di ... oller.html

Centroid cnc.jpg (415.81 KiB) Viewed 5119 times

[David Buchhauser]

Those jagged edges look like camera movement to me. How solid is your mount?
Another possibility is a timing issue with the ESS communication.
I am working on a version of Scanything that is easier to transfer to different controls. Mach4 is on the to-do list. I don't know if Centroid have a facility for directly controlling the machine movement.

Hi Les,
I have upgraded my camera mount to a much more substantial 3/8" thick billet aluminum bracket.

20200314_234435_resized.jpg (587.91 KiB) Viewed 5013 times

I believe this has made a great improvement in the quality of my scan/trace results. On most solid objects, I can now run at the maximum feed rate with no problems. I am also very impressed with the dimensional accuracy of the resulting dxf produced by the trace. Here is one example. This is a large washer cut on my PlasmaCAM cnc plasma table. It was originally drawn with and outside diameter of 4.750". The cut part measures very close to this.

20200315_000342_resized.jpg (732.05 KiB) Viewed 5013 times

The actual part does have a very small amount of ripple on the plasma cut edges (outside circumference and inside the center hole edge) which may or may not contribute to any waviness in the resultant dxf after tracing. I assume that when I am have Smoothing set to 100% (or even 75%) this may help to eliminate this in the trace.

The dimensions on the trace dxf are very close to the actual measured dimension of the part. I have traced several other round objects with similar results.

scanything large washer trace.jpg (428.78 KiB) Viewed 5013 times