Neoden 4 pick and place

[rx8pilot]

The shifted amount is not the same, the least shifted is the left top board while the worst is the right bottom board.

Tried using some different reference (first chip and mark points) still got some shifted placement, verified using step - vision method (works now, it need all the mark points recognized first)

Now i wonder if the pcb is not accurate in dimension, or the step / mm setting of the machine is not true.

I do not want to try with another pcb for now, this one is what i have to do asap.

-ichan

Do you suppose it's possible that at some point it is skipping a few steps. Maybe the decel/accel is too aggressive. Does it have rotary encoders on the X/Y steppers?

[thommo]

Ichan,

Elmood had this to say in his 'reply item #639' in this forum.
I had a look for it elsewhere, because I thought it had been mentioned before this, but couldn't locate it.

Button “to curr p”
The button "to curr p" (cannot be completely read on the button text) on the file edit screen is used to recalculate all of the part locations and rotation based on the PCB mark settings.  When doing this, the rotation of each part is calculated incorrectly, it seems to keep adding additional rotation each time the "to curr p" button is pressed.

Elmood, how is the Button “to curr p” works?

I always got error pressing this button.

-ichan.

[Ichan]

Thommo yes i read that, the term "based on the pcb mark setting" is not clear to me and i always got error message clicking it.

Do you suppose it's possible that at some point it is skipping a few steps. Maybe the decel/accel is too aggressive. Does it have rotary encoders on the X/Y steppers?

I think it is not loosing steps, because loosing steps usually have some strange sounds when it happen.

I quickly check the step/mm of the machine by printing a square of 200x200 mm on a paper then verify the dimension of that printed paper on the machine. The error i got is not more than 0.2mm for 200mm length - seems the setting is true.

The pcb should be in the right dimension too because it is ok on my other machine.

Headache i had now... 

-ichan

[rwb]

@Ichan  Did Neoden send you the practice PCB + PNP file so you can populate that demo board?

I'm wondering if they had the PNP file available if you could load it and see if the machine places it perfectly or if its off on that also?

If it does run it perfectly then look at the PNP file and see if there is anything you can learn from it.

Just talking out loud here.

[mikeselectricstuff]

Which raises the question:
Do NeoDen provide lifetime free software upgrades, as the features roll out?
I assume they supply lifetime updates for fixes and other mods - pls confirm if you're aware, both of these circumstances.

Another rather important question is: How long is "lifetime" for a product like this?  How long before they stop manufacturing a particular product and at the same time stop issuing updates and fixes for it? 

Probably until the next machine comes out, or their software guy leaves.

[Kjelt]

Probably until the next machine comes out, or their software guy leaves.

What software guy 

[Ichan]

@Ichan  Did Neoden send you the practice PCB + PNP file so you can populate that demo board?

Yes they provide the sample pcb and working files, but then i have to prepare the components for that board which i may not have it all handy. Besides, i have to solve the problem with this board.

I will step back a little trying a single board tomorrow, seems this software had problem with angled position. Any non zero angle will need matrix transformation calculation, it may suffer an accumulated rounding error - just a thought.

-ichan

[TankSparks]

Ichan  can you send me your csv file. I will be might able to spot what's wrong with it.   I've gone through the same type of problems you are having... but I have now mastered the software with single or panelized boards.
Are you using a conveyor?

[TheSteve]

I think it was mentioned before but TankSparks is the owner of the machine I have used, he has a very good understanding of the software. I think we've seen pretty much every bug there is other then perhaps conveyor related stuff as we don't have that option. If you recall we also had the parts drifting off location as it placed them. There are subtle operation selections/order of operation selections that can cause the problem.

[rwb]

What would it take to get @TankSparks to make a video or write up for how to properly setup the NeoDen 4 software so you get reliable placement?

Seems like that would help others out greatly when it comes to getting started.

[thommo]

The community support for the N4 is very impressive. It's a shame that NeoDen are not following this more closely and stepping in.

I have been speaking with them a bit lately and they keep reiterating that their Chinese users aren't reporting these sorts of issues. So, this points to language, and probably better direct phone access.

There seems to also be a few from outside China that also have managed to 'crack the code' (not literally), and got their machines working reliably now.

Whatever happens, on behalf of all users - current and potential, please publish a step-by-step procedure for this fix when it occurs guys (hopefully today).

Good luck Ichan, TankSparks and TheSteve
and THANKS from everyone!

I think it was mentioned before but TankSparks is the owner of the machine I have used, he has a very good understanding of the software. I think we've seen pretty much every bug there is other then perhaps conveyor related stuff as we don't have that option. If you recall we also had the parts drifting off location as it placed them. There are subtle operation selections/order of operation selections that can cause the problem.

[thommo]

IC Tray

Guys, I spoke with Hiami from NeoDen yesterday to clarify a point about configuration of their IC Tray.

In particular, I wanted to know how many different sized components the machine could handle when using the trays - or more specifically, how many different tray cubicle sizes it could handle.

Not surprisingly, she confirmed that all cubicles in a single row were required to be the same size - that's a given.

What I did learn is that the machine can handle up to 10 different rows. Each can be a different length, and located in any position that the head can reasonably get to.

My intention is to use this feature and customise trays for specific board builds. As usual, there are many items that only have 1 instance in many BOMs. So this will instantly increase the machine capacity to a further 10 components. Eg 48 X 8mm feeders + 10 tray options = 58 different components.

[rwb]

I was hoping it could handle more than 10 rows so you could get way more parts in these custom trays if you needed.

I wonder if what the custom tray limit is on the  TVM802A machine that @Axel has? I think he say you can do as many custom trays as you have space for.

Seems like it's just a Neoden 4 software limitation cause I see no reason it couldn't do more than 10 different rows.

[thommo]

I think the 10 row limit is a consideration built around their 'rail/conveyor' version of the machine - which is the feature that makes it attractive to us.

I imagine you'd be hard pushed to get many more than 10 rows of larger components/ICs in between the rail and feeders, but I agree that you could potentially get say 2x 10 [half length] rows in this space - if the SW allowed it, even it the max component count was say 15 pcs, we'd be happy to reload the trays every 16th board.

BTW - Ichan's reply #682 contains a video of a set up featuring a 10x row Waffle Tray

I was hoping it could handle more than 10 rows so you could get way more parts in these custom trays if you needed.
I wonder if what the custom tray limit is on the  TVM802A machine that @Axel has? I think he say you can do as many custom trays as you have space for.
Seems like it's just a Neoden 4 software limitation cause I see no reason it couldn't do more than 10 different rows.

[ServoKit]

@rwb: I can define up to 10 trays, single or multi row, 1x10, 2x5, 3x4, no problem as long as I stay within the limits of the work plane and the rods of the PCB holder.

If it's practical is another question. Single row trays might be less prone to user error when you load them up manually. If you have multiple trays you will definitely want to label them clearly or even better have them in different colors (no problem with 3D printing...)

Regards, Axel

[Ichan]

TankSpark, i attached the working file in here so the other can check it too, and here the reference points i use for that working file. Thanks in advanced.



Close up photo of the right bottom board where the worst shift happen.



I re-read the thread back and found your post in here: https://www.eevblog.com/forum/reviews/neoden-4-pick-and-place/msg857078/#msg857078

You said that the X value for left top and left bottom need to be the same, whew... that is what i thought, it can not handle angled position!

-ichan

[Ichan]

About the tray, 10 is the software limitation.

The software has 58 slot for feeders, 48 of it had been assigned to reel feeder, 10 more for extra feeder including trays.

Anyway, i found the machine always get the wrong edge of my board as picture below:



My board have a copper pour edge about 0.5mm before the board edge, the machine detect the copper pour edge not the board edge.

-ichan

[TankSparks]

@Ichan
Neoden doesn't ask you for the size of the boards in the panel  but it does want to know the location of first component in your file.  With that info it will make the necessary calculations.  I create a dummy component with x,y value of 0,0 and skip=yes so it won't try to place it.  I put this component first in the list.  It will make the following steps easier.

For the 1st chip, bottom left use the align button to enter value, so if I'm using my dummy component it will be the bottom left corner of the bottom left board.
Here's the twist....
For the 1st chip, top left board and 1st chip top right board  mathematically enter values as if your board was perfectly square. Type those x,y values in do not use the align button.    Its a bit of pain and requires a calculator,  but I found no other way.  The neoden video examples do not use  fiducials, or have an example with manual data(absolute values) . I see you also use the bottom left corner as your 0,0 reference as I do.
Don't worry if board is not perfectly squared in machine,  its impossible.  I verified that this method will work even if your board is at a angle
Click Create panel list button,  don't forget this,  if you change any values you must click this button afterwards

type in  the values of your mark points  bottom left,  then top right.  Two marks points for the whole panel,  these values will be relative to the 0,0 reference which in your case is the bottom left corner of your board.

I also noticed you have vacuum test set to no for each part,  which would be a good idea to have it checked or you may end up with missing parts on your board.


Mount file
Click Step   It will step through the two mark points you entered,  then machine will stop
Now align button will no longer be grayed out.  You can now use align button for each part to verify where it will place parts.   You can also select different boards in your panel

Tips
Make sure your x,y data is the center of the part,  I got caught on that one.
Do not use 3 mark points unless I made an error 3 mark points caused parts to be way off.   Use 2 mark points  bottom left and top right.  I have not tested with 4 mark points.
You can do a dry run if you change test no    to  test yes, it will run the job but doesn't turn on suction.   Unable to put it in that mode with their gui software, so you will need to edit file manually,  haven't really had any need to use it but its good to know.   

[Jefferson]

2pcs Mark point (FED) they must be made on IPC standards and not as a hole like you. This error is possible in the accuracy of the installation.

[thommo]

Hi Ichan
I don't think the fact that the N4 sees the wrong 'edge' of your board will have anything to do with what problem you are describing.

TankSparks can you pls confirm this?

Board edge detection should only be an aid to tell the rail/conveyor system when to stop advancing the board into the machine. From there the N4 should begin its 'search' for the pre-determined marks (fiducial).
So 0.5, or even 1.0mm ought not be an issue if valid mark points are used.

Once the fiducials are identified, the machine to board/panel relationship, including rotation, should be known and automatically calculated.

It will then use the 0.0 (or first component) reference to identify the relative positions of all the footprints on the first board.

The reason for identifying the same 0.0 or first component on the top left and top right boards should be to determine the offsets between one board and the next/others.

What TankSparks said about only requiring 2 marks or fiducials does confuse me a little however. Without 3 marks to form a right angle, I'm not sure how it determines the panel rotation unless you are equiped to actually enter the coordinates of each of the 2 marks.

I recall something in a video, which requires identifying a vertical line, but think this is for single boards only, without fiducials.

TankSparks, thanks for the detailed explanation.
Can you please confirm my understanding from your practical experience?
Will be interesting to see how it work out now.

I still think you'll continue to have inconsistencies without proper markings however, but I hope for your sake that not the case Ichan.

Anyway, i found the machine always get the wrong edge of my board as picture below:



My board have a copper pour edge about 0.5mm before the board edge, the machine detect the copper pour edge not the board edge.

-ichan

[ServoKit]

I'm with Jefferson and thommo here. The problem might very well originate from using holes/vias as fiducials (or "marks"). If the software expects the fid to be 1mm in diameter (a common size) a larger or smaller hole might throw it off. Likewise, if the software looks for disks rather than circles what's under the hole/via might confuse it.  One way to test this would be to close the holes with some tape on the bottom of the pcb.

Regards, Axel

[rx8pilot]

Holes are not good fiducials. On top of that, there should be more than 2 for a panel that size. Chips land on pads. The pads are made by the same process and the same time as a FID. Drills, are a separate process where the alignment may not be great AND drills wander unpredictably for every hole.

I got inconsistent results using holes for fids. It has been perfect since I started putting real fids on every board. Granted, I don't have an N4 - but it's worth noting. I also had some trouble with HASL boards if the pads are domed. The irregular doming, would make the pad appear distorted to the top camera and the fiducial would be shifted. The flat pads on an ENIG board are more consistent.

[Ichan]

I do really understand how a standard fiducial will be much more preferable, but the fact is this board has none of them - i have to find the solution for this board, even the machine now already fully loaded with all the components for this board.

Has no time with the machine today, plan for tomorrow:

- check and if required realign the rails
- verify the board size on the machine coordinate
- try single board
- try TankSpark recommendation
- slow down the speed
- try manual mark point recognition

The last one is Neoden recommendation i got by email.

-ichan

[rx8pilot]

- try manual mark point recognition

I did that on about 50 PCB's that did not have fids. Got me through the batch with perfect placement, and I learned my lesson.

[ar__systems]

If you don't have fiducials it is not a big problem for a small batch. You can use any pad. For example take one of the any 0603 resistors pad. You have exact coordinates for them from your CAD. Then do manual alignment. If you are lucky the s/w can use square mark to do automatic alignment.