Notes on different versions of PLACE.DLL

Started by Mike, November 27, 2018, 06:32:12 PM

Previous topic - Next topic


There is an issue with place.dll. Basically, there are 2 versions..
The following extract courtesy Alex Wilkinson:

There are two versions of Place.dll that I am aware of: an older one dated 22/12/1999 size 1774k, and a newer one dated 12/9/2001 size 268k. Both have different issues, however the older, larger version seems to be more usable in most instances.

Place.dll, 268k version.
If you set the machine to 'auto fid correct' then the first component picked after the fid correction will fail because the head lingers longer than usual in the lowered pick position, and before it retracts with the part it commands the feeder to index, so the head retracts after the tape has started to move, therefore it loses the part.

Place.dll, 1774k version.
The issue above in the 268k version is resolved, however instead there are a whole load of other problems related to larger parts and multi-indexing.
If a part is too large to vision in a single shot, therefore requiring 2 or 4 images to be taken of it, then the feeder where it came from is sent one index command less than its set to (the first one is missing, not the last) regardless of the number of indexes setting in the software. You can only set a maximum of 3 indexes per part I think (larger numbers are accepted but treated as if 3). This means large parts that require more than 2 indexes (3-1) can't be placed without multiple mispick errors each time. If the part came from a tray, then the machine always goes back to the same pocket to pick the part, even though its just taken the part from there. Picking from a vib feeder works ok as no indexes are involved.
Setting 'force fixed' in the CDF settings for a package will mean the machine only takes a single image of the part, even if its too big to vision properly. This does mean the indexes work correctly again for that package, but in reality is only really useful for parts that are only slightly too big for a single image vision normally.

Also if a part requiring 2 or 4 images fails vision analysis, the machine will go back and retry endlessly, ignoring the 'number of retries' setting, and without feeding at all as the first index command is missing and a vision failure only produces one index command.
My solution to this in most cases is to use a new design feeder controller board in the larger lane feeders, and then you can set (faster) multi indexes using the 4 spare dip switches on the feeder and the PC is then set to only send one index command for that part.


268k DLL version used for most jobs as this avoids all the problems with larger parts on tape or in trays. You just have to manually do the fiducials each run or accept the first part picked will be lost. For boards with just small parts where auto fid correct is useful then the 1774k DLL can be used.


I would add the following :

A while ago when I was playing with place.dll versions, there was a permutation where the machine behaved as if it hadn't been calibrated, so the DLL version may have some dependency on how it was calibrated.  This was a while ago & I don;t recall details.

I use the 1774K version, and the only issue I see with large parts and multiple indexes is where the part gets visioned quickly, and at least one of the subsequent indexes gets missed.
This seems to be related to time taken to get to the placement position, and simply changing the XY speed to slow or medium, or moving the part to a more distant feeder seems to fix it.

Other than the above I've never had to fudge index numbers, but I don't use the machine that often, large parts less so, though I'm fairly sure I've used quad-index at least once. I do regularly use triple-index parts with no issues.

Just to emphasise that for parts that need multiple shots, the vision setting in the CDF MUST be set to "Automatic"

Remember that for large, rectangular parts from tape feeders, "Pre-rotate" (0 and 900 are the only valid settings) can be useful to reduce the number of vision steps as they better fit the camera's field of view.