734 replies to this topic

[mjr]

Ok here's another video.  Hopefully this is better.

 

https://youtu.be/eDiEHYrZfPQ

 

I moved the connection wires away from it and that seemed to fix it a little, but I've noticed that the ball will just shoot off the plunger, even before I pull it back.  Is there some number I can edit to fix that?

 

The sensor image looks pretty good there.  Based on that I'm not sure why it would be jumping around.  But if it is jumping around, that's probably why the ball is shooting - VP is interpreting the rapid changes in position as actual movement and translating that into a collision with the ball.  If you can fix the jumpy behavior you'll fix the VP problem.

[kruuth]

How do I do that? 

[Les73gTx]

Just as a note I have been following and working on my install also .... I found the "happy spot" for the light placement but I also had too much clearance or distance between the plunger and the CCD. I adjusted them closer together and now I have a very steady movement ..... I believe I was getting light "leaks" on the sensor.
One thing I found you need to do is calibrate the sensor every time after you change the light source it really makes a difference in VP

Sent from my SAMSUNG-SM-G870A using Tapatalk

[mjr]

How do I do that? 

 

Sorry I can't be more helpful, but I've given you all the advice I can think of.  I think you'll just have experiment a bit and see if you can figure out where the image instability is coming from.  The problem is probably the optical quality of the image, so I'd focus on trying to get as clean an image as you can.

 

In your latest video, I notice there are a couple of dark lines in the bright area.  Those could be responsible for some jumpy action in the position reading because they could be scanned as edges.  You should try to eliminate those as much as possible.  See if you can figure out where those are coming from.  I'd guess they're a reflection or shadow from a dangling wire or other nearby object.  Watch the sensor image in the config tool as you make adjustments and try to figure out how the image is affected by different adjustments.

[kruuth]

I can try to pack away the wires as much as I can.  As far as I can tell there isn't anything blocking it.  

 

Is this something in the software that can be tweaked a little in software?

By software I mean would it be possible to alter the sensitivity of the sensor?

[mjr]

I can try to pack away the wires as much as I can.  As far as I can tell there isn't anything blocking it.  

 

Is this something in the software that can be tweaked a little in software?

By software I mean would it be possible to alter the sensitivity of the sensor?

 

The code is all open source, so feel free to take a look if you want to try that approach.  There are unlimited possibilities to how to process an image, so yeah, nothing is ever set in stone there.  The image processing that the code does is pretty simple-minded edge finding.  More sophisticated algorithms would be possible if you have something you want to try, although the M0+ isn't a very powerful CPU, so it's a bit limiting because of the need to keep a high frame rate.  I personally think it would be more fruitful to attack it on the optical side, since the existing software is known to get good results you can get the right lighting conditions and sensor placement.  For what it's worth, it's rock-solid on my setup - the jumpy behavior you're seeing is not normal and should be possible to eliminate with the existing software.

[kruuth]

Well I'm trying to find what is casting a shadow.  I've masked off all but the LED facing up to the sensor right now.  The only wires are behind it.  What should I do to hide the wires so that there is no shadow?

 

Is it possible that the sensor is defective?

 

Actually I was thinking of something here...during calibration, mark the dark spots on the way down, and throw out those lines.  How tough would that be?

Edited by kruuth, 31 March 2016 - 01:19 PM.

[mjr]

Is it possible that the sensor is defective?

 

Anything's possible, but I doubt that's the problem, since it seems to be working overall.  I'd expect a defective sensor to produce either no image data at all, or 100% random pixels. You're clearly getting data from the sensor that's mostly good, so I think we can safely rule out any sort of complete sensor failure.  I imagine it's possible for a sensor like this to have individual dead pixels, which I'd expect to show as stuck-on or stuck-off pixels.  The pattern you're seeing doesn't suggest that to my eye.  It looks like you're getting some fuzzy shadowy areas covering multiple pixels in a gaussian sort of way.  That makes me think they're real optical features that the sensor is faithfully reproducing in the pixel data.

 

 

Actually I was thinking of something here...during calibration, mark the dark spots on the way down, and throw out those lines.  How tough would that be?

 

Probably harder than it might seem at first glance, and I'm afraid it's not something I'd have the time to do right now.  But like I said earlier, the code is all open source, so you should give it a try if you think it's a promising approach.

 

Before diving into that, you might want to wait and see if the upcoming new version of the software improves things at all.  It has some tweaks to the edge-finder algorithm that make it a tiny bit more tolerant of noise in the image.  It won't be vastly different, so don't expect miracles, but it could be worth a shot.  (Sorry, can't give you an ETA on the next release yet.  Hopefully soon.)

[kruuth]

I am going to try getting the wires out of the way and moving the light.  What else can I do in order to get rid of those lines?

[kruuth]

Ok.  I moved the light slightly, and obscured all of the wires.  I also moved the light slightly up and it's WORSE now.  What am I doing wrong?  Here is the video.  

 

 

If I cut the light I still get the same problem.  Do I have a bad sensor?  Can someone post a video of what it should look like?

[mjr]

Ok.  I moved the light slightly, and obscured all of the wires.  I also moved the light slightly up and it's WORSE now.  What am I doing wrong?  Here is the video.  

 

The sensor view looks pretty good to me, but as I've said, it's hard to see enough detail in the video to know what the software is seeing.  All I can tell is that it looks like it's supposed to at the level of detail visible in the video.

 

By "moved the light slighty up", you mean closer to the sensor?  Further away is better, closer is worse.  It makes the shadow edge fuzzier to be closer.

 

 

Do I have a bad sensor? 

 

Same answer as before ("no").  From the videos, it looks like your sensor is working perfectly.

 

I don't have any other recommendations based on the data so far other than to continue experimenting with lighting.  Work scientifically: make small changes, carefully note the effects, repeat.

 

If you can't make any progress that way, wait until I get the next release out and try that.

 

One outside possibility is that you have a loose solder connection somewhere.  It's possible that the instability you're seeing is coming from occasional glitches in the signal that aren't visible in the videos because they're too brief.  Maybe go over your solder joints with a fine-tooth comb (or better yet a magnifying glass) and see if you can spot anything that looks suspicious.

 

If all else fails, after I get the new version out, I might be able to create some instrumentation that will let you collect raw sensor data so that you can send it to me to look at.  That might either help debug your setup or provide test data for improving the noise tolerance in the software.

 

I know it's frustrating trying to debug something like this and not making any visible progress.  Sorry I can't be more helpful - it's like working in the dark to try to debug someone else's setup remotely, so the best I can do is throw out ideas based on my own experience.

Edited by mjr, 01 April 2016 - 01:28 AM.

[kruuth]

I understand.  What I find strange is that when it's white, visual pinball doesn't recognize it hardly at all, and all it does is instantly shoot off the ball.  It seems like it's almost an issue in visual pinball.  The calibrator tool seems to work great but it's the windows one that is throwing bad data.  I'm probably going to try going to a dimmable light source. 

 

When I said I moved it, I moved it only left/right about an inch.

[kruuth]

Tried firing you a message on the code.  I have a theory on a simple way to fix the ribbing.

[Mace]

Just a thought, is the light source powered by a decent DC supply? i know that it's a basic question but a lot of PC psus can be quite noisy and ripple on the LEDs will produce some odd effects.

[kruuth]

I'm using a PC psu to drive it but I can rig something off a switched PSU instead just to test.

[Mace]

I'm using a PC psu to drive it but I can rig something off a switched PSU instead just to test.

You can try a torch as a light source just to prove the point, on my table I found that two blue LEDs running on 12v through a resistor worked best with a plastic tuble covering the plunger with a slit in one side for the sensor and the leds side on in the other. Black plumbing tube slides perfectly over the boss that the plunger passes through and needs no mounting as its self supporting.

Edited by Mace, 01 April 2016 - 06:46 PM.

[kruuth]

I've tried a flashlight.  Same exact thing.  The wiring looks pretty solid.  I'm trying to move everything so it's out of the way as much as possible of the sensor but right now it's no good.  Same banding thing.

 

In the code, would it be possible to just throw out any dark spot between the min/max?

[NobodyYouKnow]

@kruuth - This is me going out on a limb - you have probably done this a hundred times by now.

 

I would ensure there is no dust on any optically significant portions of the sensor. It seems that could cause a similar artifact to the lines you are seeing. This type of sensor is also used on multi-function printer / scanner / fax machines. I had a problem with a couple of thin lines at persistent spaces across scanned (or faxed) pages. They were there regardless of which paper feed options I selected (flatbed or sheet feeder), and did not go away with changing brightness / contrast. I finally gave up, disassembled the printer and cleaned the optical sensor. To my pleasant surprise, it fixed the problem. I suspect a scratch or similar defect, even if microscopic, could cause the same artifacts you are seeing.

 

I am planning on going the route you have with the optical plunger sensor, and hope to learn from your experiences. Thanks for helping blaze the trail.

 

Regards.

Edited by NobodyYouKnow, 05 April 2016 - 09:58 PM.

[kruuth]

I'm trying to clean this.  There shouldn't be any scratches or anything.  I'm going to try changing the light bulb that I'm using.

[kruuth]

Ok I've ordered a dimmer bulb to work with.  Maybe the bright bulb I'm using is the problem.