LitePlacer.com forums

danmcb wrote:I am afraid that I don't really understand all the points you are making.

I am sorry about that Daniel. For 100% sure my English isn't the best. I always try though. But I guess that in this case it can be a matter of not seeing through reality. For example :

Why would you set up your production to load two panels, with a total size of 150x150mm, when you could have only one single one? This seems to make things more complicated than they need to be, no?

Inherently you are correct. But in the same realm you could tell me to cut out the individual PCBs, load one at the time and be done with all the problems.

Somewhere I told about 3 of these "screen shots" next to each other and 3 above each other. So that is 9 squares of 15x15cm and feasable for the machine's working surface. In other words not 2 (as you suggested) as this would be a bit non-sense and indeed only creates problems. So 18 of them and a total square of ~ 45x45cm.

If you would go to an assembly house and ask them to process two panels at a time on their P&P machine, what would they say?

They will say that this is BS because they deal with it on a per panel base anyway, or otherwise per individual PCB. Those machines have other functionalities than ours ... (in this context the moving in/out of the PCBs).
Btw I hope it is clear that we (us over here) try to use the LitePlacer to avoid hiring personnel. It is not a hobby and it certainly is not the intention that the persons normally doing the soldering are now busy with the machine instead. Of course the human interfering can't be minimized to around zero, which is related to how long the solder mask lasts hence relates to the speed of the machine (and thus relates to how many placements can be done unattendedly).

OK, there can be offsets - but however your panel is engineered, they would be repeatable. The point is simply to include all your offsets in the calculations given in the "boards.txt" file.

Surely this will be the part I could not make clear;
The point is NOT about defining the (repetetive) data by some means. Instead it is about you need to KNOW the data ... and we don't. Remember, all I do is order a 1000 of such PCBs and nothing about panels. It is just done like that. All I know are the dimensions and component positions of one individual board. Read : this is all the data which is available. The further necessary data can also be made available, but it has to go my measurement ...

So, fine. I measure. One PCB is 12x25mm - which I don't need to measure. There's a component A at 3.1/4.7. Don't need to measure that either.
The panel is -measured by me- 120.22 wide. The fraction of .22 is a bit tough but I tried hard.
Now I put two panels next to eachother and I measure 240.50. Or something like that as I have no tools to be more precise.

Again I say/claim that it really won't be difficult to make a program for this and spit out the positions for the components for all the individual boards. All I want to make clear though, is that this surely will NOT work by applying 10 x 12mm = 120mm because the "glueing" of the individual PCB's take space. This is how I measure 120.22 and not 120.00. OK, you got that. But now you will still think that with the measures of one panel you can extrapolate to the others (like 2 more sideways to the right).
And I say it again : no matter how hard you try to measure the "element" - if you are off by only 0.5mm over the total width or height, nothing works ...

So in my view, the only base of working can be the physical measurement of the lot so things get consistent; Measure the panel, measure the offsets in it (think borders) and measure the whole grid. Give that to a program which by now is far from simple to make.

Peter

Peter - all of that is perhaps *why* professional PCB assembly houses work with panels that have very clearly defined dimensions ... ?!

I say unto you ... "Keep It Simple ..."

juku .. they did that a long time ago I think!

danmcb wrote:I say unto you ... "Keep It Simple ..."

Daniel, I surely agree !

Maybe it's a good idea that I emphasize that the discussion is merely about "please be careful because this is not as simple as it seems" than that I want to have something done in this realm.

The real merit, of course, is that if I - or anyone - wants something like my proposed 45x45cm etc. stuff, he should make a panel of that complete size. Or otherwise stick to something which is managable in well fashion (and your script can do the job nicely).

Subject not closed, but back to basics sure Yes.
Thanks,
Peter

Just to throw in my two cents:

If I understand correctly, many of the challenges Peter mentioned only exist if you want to either populate panels with unknown dimensions (i.e. panelization done by manufacturer but not telling you how exactly) or an array of individual boards (i.e. multiple boards with same layouts but already separated).

I have similar issues with boards sourced directly from China and those from PCB pools where I cannot specify the panelization.

I am currently considering to change my PCB fabber for exactly this reason - I want to be able to specify the panelization myself and that way I know all dimensions in advance.

To name just one example, Eurocircuits allows you to specify your panel layout in "Delivery options". You may either have them panelize a single board layout according to your their specs (but these are published, so you still know / can calculate dimensions) or provide your own panelized layout. I also came across aetzwerk.de in Munich, who are a bit cheaper but also give you very fine grained control about panelization.

Summary: For small quantities (< 10) I will continue to source from overseas or cheap pools, receive individual boards and either place them manually (by hand) or one after another. For larger jobs, I will pay a bit extra to get a well defined panel from a "local" supplier and save me trouble with pick and place in turn.

there are degrees of "unknown".

If you are talking about boards that might have slightly rough edges (say +/- 0.5mm) because of the way they are made, and you want to pack them in alongside each other, OK, that might be fixable by having the software look for fiducials for each board. But that does then require some modifications to the basic software of the machine itself.

if you are talking about a panel that you have absolutely no control of the layout - well, if each panel is *totally random* - I think that is a non-starter. Get a decent supplier.

Of course, there are always cheap suppliers - but sometimes they cost you more in time than you save in euros.

The whole concept of having a manufacturing process which allows you to to turn out decent quality product efficiently relies on things being repeatable and known. Even for small/medium volumes, this is true. If that is *not* the case, there will always be problems, trying to fix them in software is not really a good idea, in my opinion.

> But that does then require some modifications to the basic software of the machine itself.

This is not a problem, though. Supporting panels is clearly something my customers need, and software will be my main focus for some time from now on.

If I am not missing something, decent support of panels is "only" telling the software the number of rows and columns in a panel and the offsets, and building a combined job from that. Some details:
- It should be possible to save the resulting pick and place file.
- The camera could be used to measure the offsets.

How the fiducials should be handled? It is unnecessary and dumb to measure tens of fiducials, one in each corner gives the transformation. Do you have panels that have fiducials in the panel, not on each board, but the p&p file still from one board only? If so, how is the fiducials location data represented? (If you have a p&p file showing the panel as one big board, the problem is already solved.)

Anything else that I missed? (I will study the long posts in this thread more carefully later.)

well, I looked the other day at Eurocircuits Panelise Guidelines:

http://www.eurocircuits.com/Introductio ... ard-panels

And in this case, they propose fiducials in the border of the panel, not per board. Of course, than means that positions of boards in panels is well defined.

However what *might* be useful could be to have an give an *approximate* location of each board (perhaps using my "boards file" mechanism, which seems most flexible to me), and then have LT software look for fiducials at the correct place relative to each board origin, and then calculate an exact offset for each board in the matrix.

That would allow, for instance, a set of individual boards with rough-cut edges to be loaded in a grid, and the errors caused by the edges to be dealt with.

However, before adding the complexity that this involves, I would ask - is it a common enough problem to justify the effort? In this case would it perhaps be more profitable for you to spend your time on other tasks? Just a question.

Hello Juha,

I completely agree with your thoughts, I started to think about this approach in my post may 11th.
define the matrix, X and Y offset and your ready to go.

Okay if fiducials on the panelized board need to be used it's a different approach. If the board fiducials are unknown in the P&P file, what if the user moves the camera to these fiducials' approximately position and LT software detects it automatically, accurate and calculate this position, and/or move to an already known position of one of the components / connectors or any through hole parts which can be handled as virtual fiducials. I think this would do the trick?!

A more advanced approach would be if the LT software creates a complete picture of the board, overlays the known P&P positions (of one board) and calculate the missing positions by some (really clever) software routine. It could give the user the choice to populate one ore more boards if wanted (UI pop-up) this way no panelized board outline positions are needed. Of course this approach will require advanced software development, but I think this will eliminate any future questions or needs.
Such a routine looks a lot like the optical inspection feature which again would also detect any lost components after the placement is done

I hope this gives enough food for thought..... meanwhile the perl script of Daniel can help us all a lot (I still had no time to test it, sorry...)