Changing the size to account for the zoom factor, improved things a lot.
Observations:
- It is now set up to use the component as the 'hole' then it moves to the part.
- As they are both now in the same location, there is no need to move and activate the slack compensation twice. If the offset is 0,0 then no need to move. Can you add that to the fix list?
- I am not sure but it seems the angle is not taken into consideration on part recognition, in tapes tab. Reason: The blue arrow is always horizontal or vertical. The amount of time the blue arrow is visible varies or is sometimes not shown (timer related?)
- As mention before black background is much better. I suspect in my setup an algorithm will be required for each tape to account for variance in the threshold levels in different areas.
- Again not sure, when the parts checkbox is selected it seems to ignore the XY&A offsets. (see attached images for 2 views of the same part)
- I tried to define the tape with the parts coordinates selected. The behaviour between components changed...I will need to read up on that more. Obviously, this should avoid the hole -> part moves of zero. see also previous point.
Release candidate for new software version, 20.05.2021
Re: Release candidate for new software version, 20.05.2021
- Attachments
-
- 2021-05-26 18_31_50-LitePlacer.jpg (118.5 KiB) Viewed 7683 times
-
- 2021-05-26 18_35_11-LitePlacer.jpg (129.07 KiB) Viewed 7683 times
All about IR remote Control. Webshop: https://www.AnalysIR.com/blog/
Check out our LitePlacer Trinamic/TMC2209 TinyG Stepper upgrade kit: https://www.tindie.com/products/analysi ... r-openpnp/
Check out our LitePlacer Trinamic/TMC2209 TinyG Stepper upgrade kit: https://www.tindie.com/products/analysi ... r-openpnp/
Re: Release candidate for new software version, 20.05.2021
I compiled the dev version from github using the latest VS community edition.
The 2 fixes you referenced seemed OK for me using LP.
With the zoom scaling fixed the vision was better again.
I think I have now got the 'Coordinates for parts' working OK thus avoiding the 2 moves for slack compensation.
However, I see that lighting / threshold will be the ongoing challenge.
So either spraying everything except the tapes black or implementing 'software' masking as above (you could also call it an 'exclusion zone' or 'target zone') looks like the best way to get better consistency between tapes. Other challenging tapes may have to use pockets or holes.
Some tapes work fine. A very similar tape with a different resistor needs a different threshold. Capacitors in clear tape can suffer a bit from light reflections at the corner/sides of pockets (it messes with the algorithm..could be avoided with a mask/zone).
Filter feature by Size.
I am pretty sure the vision function still draws rectangles for features larger than min/max setting. I assumed larger boxes would not be drawn.
Blue arrow / orientation.
While testing I noticed on some components that although the part was recognized, the orientation arrow could toggle quickly between 0 and 90 degrees - even though the component was continually recognized with the same dimensions/angle. Surely once a component rectangle is recognized the orientation should be pre-determined based on either the shorter or longer dimension etc????.
X, Y & Angle offset
Are all of these taken into account during pick-up & placement?
I still have a few more panels to do, but I will wait until the next RC is released.
The 2 fixes you referenced seemed OK for me using LP.
With the zoom scaling fixed the vision was better again.
I think I have now got the 'Coordinates for parts' working OK thus avoiding the 2 moves for slack compensation.
However, I see that lighting / threshold will be the ongoing challenge.
So either spraying everything except the tapes black or implementing 'software' masking as above (you could also call it an 'exclusion zone' or 'target zone') looks like the best way to get better consistency between tapes. Other challenging tapes may have to use pockets or holes.
Some tapes work fine. A very similar tape with a different resistor needs a different threshold. Capacitors in clear tape can suffer a bit from light reflections at the corner/sides of pockets (it messes with the algorithm..could be avoided with a mask/zone).
Filter feature by Size.
I am pretty sure the vision function still draws rectangles for features larger than min/max setting. I assumed larger boxes would not be drawn.
Blue arrow / orientation.
While testing I noticed on some components that although the part was recognized, the orientation arrow could toggle quickly between 0 and 90 degrees - even though the component was continually recognized with the same dimensions/angle. Surely once a component rectangle is recognized the orientation should be pre-determined based on either the shorter or longer dimension etc????.
X, Y & Angle offset
Are all of these taken into account during pick-up & placement?
I still have a few more panels to do, but I will wait until the next RC is released.
All about IR remote Control. Webshop: https://www.AnalysIR.com/blog/
Check out our LitePlacer Trinamic/TMC2209 TinyG Stepper upgrade kit: https://www.tindie.com/products/analysi ... r-openpnp/
Check out our LitePlacer Trinamic/TMC2209 TinyG Stepper upgrade kit: https://www.tindie.com/products/analysi ... r-openpnp/
Re: Release candidate for new software version, 20.05.2021
Just completed some more panels using the latest RC (Compiled from Github).
- 0805s & SOIC8
- no new issues to report
- This time I used pockets for most and it went well. No interruptions for 'iffy' holes.
- With the stepper upgrade, the motors remain close to ambient temperature after the jobs & of course way quieter.
I have thought some more about recognising small parts inside of pockets. Rather than spend more time on improving that (Its pretty good already). The best solution here is of course getting the bottom camera to make final adjustments & probably deserves prioritization for any future dev effort invested.
So other than the minor issues already reported/outstanding this RC is a big improvement(once the vision algorithms are updated/used).
@Juku: Sorry to hear about your leg. I hope you recover quickly.
- 0805s & SOIC8
- no new issues to report
- This time I used pockets for most and it went well. No interruptions for 'iffy' holes.
- With the stepper upgrade, the motors remain close to ambient temperature after the jobs & of course way quieter.
I have thought some more about recognising small parts inside of pockets. Rather than spend more time on improving that (Its pretty good already). The best solution here is of course getting the bottom camera to make final adjustments & probably deserves prioritization for any future dev effort invested.
So other than the minor issues already reported/outstanding this RC is a big improvement(once the vision algorithms are updated/used).
@Juku: Sorry to hear about your leg. I hope you recover quickly.
All about IR remote Control. Webshop: https://www.AnalysIR.com/blog/
Check out our LitePlacer Trinamic/TMC2209 TinyG Stepper upgrade kit: https://www.tindie.com/products/analysi ... r-openpnp/
Check out our LitePlacer Trinamic/TMC2209 TinyG Stepper upgrade kit: https://www.tindie.com/products/analysi ... r-openpnp/
Re: Release candidate for new software version, 20.05.2021
> Juku: Sorry to hear about your leg. I hope you recover quickly.
Thank you! Looks like I have ample time to sit in front of a computer in the coming weeks. I hope I can actually do that and respond to your great posts soon. Right now I’m kind of busy hopping around with my splint getting things ready.
And yes, using bottom camera has high priority, it is the most requested feature for a long time. Using the new vision system that shouldn’t be a major effort anymore, most of the needed software is done now.
Thank you! Looks like I have ample time to sit in front of a computer in the coming weeks. I hope I can actually do that and respond to your great posts soon. Right now I’m kind of busy hopping around with my splint getting things ready.
And yes, using bottom camera has high priority, it is the most requested feature for a long time. Using the new vision system that shouldn’t be a major effort anymore, most of the needed software is done now.
Re: Release candidate for new software version, 20.05.2021
The latest version fixed following issues:
-Clicking on the image gave wrong size move with show processing on and zoom was used
-Filter features by size did not take zoom into size calculation
-Placing 0402 resistors with only 3 or 4 parts picked up from wrong position (fixed, sort of. Now, the first hole needs to be good.)
-On old tape files, an error about bad data in capacity was sometimes reported
-No crash if all nozzle change move slots are full
It is much better software now than the old release, so no more release candidates, it is now the "official" software. (Whohoo!) The documentation links on the web site now also point to the documentation of this software.
Installation file: https://liteplacer.com/Downloads/LitePl ... 6_2021.exe
-Clicking on the image gave wrong size move with show processing on and zoom was used
-Filter features by size did not take zoom into size calculation
-Placing 0402 resistors with only 3 or 4 parts picked up from wrong position (fixed, sort of. Now, the first hole needs to be good.)
-On old tape files, an error about bad data in capacity was sometimes reported
-No crash if all nozzle change move slots are full
It is much better software now than the old release, so no more release candidates, it is now the "official" software. (Whohoo!) The documentation links on the web site now also point to the documentation of this software.
Installation file: https://liteplacer.com/Downloads/LitePl ... 6_2021.exe