Software release 08/14/2019

[JuKu]

Build date 8/14/2019

Download link: https://www.liteplacer.com/Downloads/Li ... 8_2019.exe

New Features, Significant Changes

In this particular version:

* Reset tape only resets the count. There are other methods for
resetting z values, but there wasn't a way to reset count only
(= reload the tape)
* Movement tests do only one move per click to prevent crash
if motor stalls
* Added a "Measure and set" homing button, that relatively quickly
corrects drift: Machine goes to zero, and measures the homing mark.
Result should be zero, but isn't, so coordinates are set to the difference.
* Indicating the newest version in forum post header

Earlier this month:

* Z0 to PCB value can be manually edited
* Slack correction distance is now editable, correction always applies
* Show pixels is moved to be always visible
* Goto Park now happens with A=0
* Added a label for CAD file load and placed status reset
(The operations are still slow, thanks to MicroSoft
and the UI component used)
* For developers: Now using Visual Studio 2019 and C# 8.0:
this should improve stability.

Bug Fixes

In this particular version:

* Fixed unwanted moves in place assisted mode
* Fixed issues in rotation and already placed status handling in placement
* Better diagnostics on failed homing
* More robust camera stream handling

Earlier this month:

* If camera refuses to stop, program doesn't freeze anymore
* Zguard works again (was broken in previous version (08/01/2019))
* Placement method selection works again (was broken in previous version (08/01/2019))
* Show pixels doesn't reset when changing tabs
* Pump and vacuum statuses are now "off" when recovering a board reset
* Homing is now more accurate in certain conditions
* On tapes grid, several rows can now be selected, Z reset
only affect the selected rows
* It was possible to jog with numpad keys even when nozzle was down
* Clicking the picture now produced the expected amount of movement
with showpixels on or off (triangular error is still there)
* Manually moving Z with F11/F12 no more triggers nozzle down error
* Fixed issue on identifiers on panelisation with more than 10 repeats
* Camera zoom factors are now forced to be >1.0
* Motor power checkbox now more accurately reflects the true status
of the system
* Tickmarks now show properly with show pixels on

Notes

1: If possible, take a copy of the log window content for a bug report (click inside the log window, ctrl+A to select all, copy and paste to a text document).

2: Previous releases (Build date 8/1/2019, Build date 8/6/2019) introduced new of bugs, which are hopefully now fixed for good.
I goofed, I shouldn't have released software with that light testing. And since customers are having issues, this version still isn't tested as good as I would like, but I'll release it to help on those issues now, and continue working.

[wormball]

Thank you! Most of the bugs seem to be gone.

However i encountered another bug. When i try to re-measure fiducials, it measures two of them quite well, but on the third fiducial it moves the camera several times and says

---------------------------
Data error
---------------------------
Transform estimation failed.
---------------------------
ОК
---------------------------

. So i can not measure true coordinates of the board. I tried another instance of the same board, but it did not help.

[Oaklands]

I have noticed that too, as I am setting up my system...getting closer to a full live run.

I suspect it is related to the filter settings you are applying. Go back to the camera tab and adjust the filters for fiducials & verify they can detect circles on all of these headers/ fiducials. I also noticed that when using headers as pseudo fiducials, the results vary by time of day as the ambient lighting varies.

From your photo it seems you may not have a diffuser installed ??? (based on the LED reflections on the PCB). I found that the performance is much better with the diffuser having originally left the diffuser off the down camera. (Ignore this last bit if you have the diffuser installed)

[wormball]

I reloaded this file and this bug magically disappeared.

[wormball]

Go back to the camera tab and adjust the filters for fiducials & verify they can detect circles on all of these headers/ fiducials.

If it can not see circles, it says something like "no circle found", not "Transform estimation failed". I think it is connected to the fact that i populated bottom side of the board, so after loading X coordinates became negated, and i added 42 mm (board width) when panelizing it, so X coordinates became positive again. But when i reloaded panelized file, all coordinates were already positive, so no error occured. So i suspect there is a problem with signs in the coordinate transformation code.

* It was possible to jog with numpad keys even when nozzle was down

Now there is an opposite problem. When i move Z with +- keys, when it reaches about half the way down, it says "nozzle down" in the log and refuses to move in any direction including up and down (however i did not try f11/f12). "Nozzle up" button does the thing, but also moves the carriage to the zero point.

* Added a "Measure and set" homing button

When it homes by the end switches, but fails to recognize the homing circle, this button is still disabled. I think it should be enabled in this case. And also the home button should be uncolored when "Measure and set" process succededed.

Also to my feelings optical homing works worse than before (however it never was perfect). When i try circle recognition in corresponding tabs, it finds the homing circle pretty well, but says "optical homing failed" almost every time when i do actual homing. I think "preferred circle size" setting will be useful (like in nozzle recognition).

[JuKu]

To diagnose the homing issue, could you please send me screenshots and log where you clicked measure button of the case when recognition works, and screenshot and log of a failed homing?

Disabling the measure and set button is a chicken and egg problem. For this to work properly, the software needs to know where zero is, and if the homing fails, it doesn’t. But you are right, I think I can add an intermediate homing status, where mechanical homing is ok but optical isn’t; accurate position is not known but optical homing can be re-tried without redoing the mechanical part. In other words, set zero at the end of mechanical homing and enable the button, but don’t trust the position yet.

[wormball]

Unfortunately i already made better homing circle, and with it (and with other circles i have) it homes perfectly.

[wormball]

I discovered two more bugs when i tested the feeder.

1. If i want to place a board without measuring fiducials, it says something like "Component C1: bad machine coordinate" and refuses to do so, and there are nans in the machine coordinates of all the parts. When i did so in previous version of the software, all was ok.

2. In the "tapes" tab it is impossible to manually change tape coordinates in the table. You can either add new part and enter coordinates in the dialog, or edit these coordinates manually in the tapes file, or move the camera to desired coordinates and push "set hole 1", but in my case the feeder is beyond the reach of the camera, and none of the future feeders will be in the field of view.

[wormball]

I encountered "Transform estimation failed" another time, and it also disappeared after i reloaded the file.

Also i discovered that the tapes file is binary, so i can only change auto feeder coordinates by deleting the entry and creating it again.

[wormball]

I got another severe error. When i try to place single part (C2_2_2) from tape "0.4", i got

---------------------------
TinyG Error
---------------------------
TinyG error. Review situation and restart if needed.
---------------------------
ОК
---------------------------

message before any movement! But before this error it homes and measures fiducials normally. Moreover, i got not one this message, but when i click ok, the same message appears instantly, and so on, and i am able to end this vicious circle only by turning off the tinyG and pushing enter for about 10 seconds while about 100 similar message windows are opened and closed.

This behavior occured after i manually changed "last Y" value in the tapes table.

The beginning of the log file is the same as the previous posted log.