Liteplacer program crashes regularly

[tek]

Often when I'm switching between Basic Setup and Setup Cameras liteplacer will lock up completely and I will have to exit it to get it to work. Also, I'm regularly losing feedback from the down looking camera, e.g if the software crashes and I restart it, the downlooking camera will not respond when I next run the program, even though the up camera does. Or I'll simply switch tabs, and when I return the camera image just shows a black screen.

Is it only me having this issue?

[mrandt]

Which version of Windows are you using? And which version of the software?

I had some similar issues when TinyG and / or webcam were disconnected. But other than that, LitePlacer runs fine on my machine.

Maybe you have some intermittent connection problems on either USB connection?

Make sure you use properly shielded cables which have larger diameter signal and power wires (low resistance) internally. Many cheap USB extensions seem to suffer from tiny wires.

Also, you might want to measure supply voltage of your USB ports under load if you have the equipment. I have seen ports that were incapable of supplying 5V @ 500mA, especially with older machines. In this case, an externally powered USB hub might help.

If your problems persist, it might be a good idea to contact Juha directly.

[JuKu]

> Which version of Windows are you using? And which version of the software?

I would have asked the same, but mrandt got first. Also, I'd like to see what (if anything) is on the log window when you switch to camera setup page, try to select a camera but don't succeed.

[tek]

I'm running windows 7 with the latest version of software (Version: 1.0.5602.17122, 05-04-15). I believe the issue may be with one of the cameras, as the program hasn't crashed yet after reversing them. I want to test the uplooking camera, but I have another problem in that I can't proceed past the "finding circles" stage with the downlooking camera, regardless of which camera I'm using.

With the previous camera, after clicking measure, then home, it would go to the home XYZ location, and then fail to do anything else, with the message "Optical positioning: Can't find Circle".

With the second camera, it would go to the XYZ location, then would move up and down diagonally as if trying to locate the circle and then would fail with the message "Optical positioning: Process is unstable, result is unreliable!" or "Optical positioning: Can't find Circle" again.

[JuKu]

Can you send a screenshot of the camera image and the image processing algorithms you are using for finding the home mark? Either here or to my email; I'll sort this out.

[tek]

This is on the standard surface we're using

[tek]

This is using a white sheet of paper and the invert function

[mrandt]

Circle detection looks OK - it is marked in the camera image.

Have you measured and configured the mm to pixel ratio, i.e. Box X and Box Y size, for down cam? If the values are too large, machine might "overshoot" while trying to home in on the circle and give you an error - happened to me before.

Please also check your camera orientation (moving gantry left should move the image to the left, moving it away from you should move image to the bottom).

Last not least, make sure your homing mark is a bit up and right from the mechanical / limit switch zero. What I did is home all axis individually (Basic Setup Tab), then switch to down camera and place the mark a bit right and up of the center cross. In my case, the mark is on a small sticker so I can move it around as needed.

[tek]

1) I have the box configured for the downcam, but when I alter the zoom obviously the size of the box changes. Is this taken into account automatically when you zoom in or is there something I else I need to do? I've tried running the process with zoom at 1.0 and it made no difference to the outcome however so I doubt that's the issue.
How accurate does the box measurement need to be. If it's out by, say, 0.2mm, is that going to be a problem?

2) With regards to orientation, I'm a little confused. When I move the gantry left, the image moves to the right, and when I move it up, the image moves down. Can you confirm this please?:

(moving gantry left should move the image to the left,

3) The dot is about 7mm to the right and up from the XYZ location. Is that okay?


As an aside, we didnt realize this particular surface would be so rugged. From the above image, can you tell if this is likely to be an issue later on?

Thanks

[mrandt]

Hi Tek,

1) I have the box configured for the downcam, but when I alter the zoom obviously the size of the box changes. Is this taken into account automatically when you zoom in or is there something I else I need to do? I've tried running the process with zoom at 1.0 and it made no difference to the outcome however so I doubt that's the issue.
How accurate does the box measurement need to be. If it's out by, say, 0.2mm, is that going to be a problem?

I am not sure it does change, as the box also increases in size when you do that - doesn't it? I took the measurements with Zoom 1x and never had a problem afterwards.

For homing, it just needs to be in the correct range - so if it is a bit off that does not create a problem.

However, for part pickup you want this measurement to be as accurate as possible.

Another user, thereza is working on ehancements to the software to automate the calibration. Basically, his algorithms move the gantry by a known distance away from an optical mark and measures how many pixels the mark moved in the picture to figure out the ratio.

You can do the same manually:
1) Put a ruler on the table and align with Y-axis
2) Move gantry so that the left box boundary is exactly above the edge of the ruler
3) Note X-position
4) Move gantry to the right until the box boundary on the opposite side is exactly above the edge of the ruler
5) Subtract noted x-position from current x-position and you know the box size
6) repeat for Y axis

Depending on the height of your component tapes and / or additional trays underneath, there might be a perspective error as described here:
http://liteplacer.com/phpBB/viewtopic.p ... rt=20#p350

I have a different, higher-resolution camera system. I think with the kit cameras this effect will not be as significant. Just make sure to measure the box size at the right z-level

2) With regards to orientation, I'm a little confused. When I move the gantry left, the image moves to the right, and when I move it up, the image moves down. Can you confirm this please?:

Just noticed my terminology is confusing, sorry about that. See the following pictures - these should make it clear:

Note the small arrows for X and Y on the paper

cam_orientation_1.jpg (35.93 KiB) Viewed 4379 times

LitePlacer software should display it like this

cam_orientation_2.jpg (29.74 KiB) Viewed 4379 times

3) The dot is about 7mm to the right and up from the XYZ location. Is that okay?

In my case it is a little closer, like 2 mm up and 2 mm right - but as long as it is in the camera picture after mechanical homing (machine tripped limit switches), it should be allright.

As an aside, we didnt realize this particular surface would be so rugged. From the above image, can you tell if this is likely to be an issue later on?

I have a similar problem. My table surface also has a relatively rugged structure. Juha mentions in the build instructions that the surface should be smooth, but I must have missed that part

While the structure makes computer vision harder due to creating pattern noise, it has not been a problem for me so far. Around the optical homing mark, I put a sticker (self adhesive paper) with smooth surface large enough to cover camera field of view.

For component strips and single component pickup, I use custom trays anyways. If you run into problem with sprocket hole detection, putting a strip of paper or other smooth surface underneath them would probably help.

For fiducial detection it is no issue at all as the PCB covers the table surface anyways.

Out of curiosity: What material is your table surface?