Neoden 4 pick and place

[uncle_bob]

All good suggestions but seem to all focus on "make it heavier" instead of "make it more rigid". I'm convinced that some strategically placed bracing can make an improvement without creating a "widow maker" waiting for next time we move.

Will report back.

Hi

The best approach without a big block of something is to bolt a sturdy work surface to a wall or two. When you move, out come the bolts and the work surface gets moved without a lot of crazy effort.

Bob

[metalphreak]

Hello !
...
I look at that picture PC. And much disappointed for many reasons such as:
1) This is an old version of HW, which isn't supported by the market.
2) Issue of a very old or very old components and reskrs ended. (Hard to find in the garbage, such components are already recycled)
3) A huge doubt prolonged use of such PC.
4) Not supported industry standards for this PC, only as a hobby.
5) What are the problems with the BIOS? no information.
6) OS Windows XP no longer supported.
7) The cameras look at the components and their old driver for XP, also the problem of long-term work.
Regular failure, if you look at the installation inside. Can apply the standards, but don't, sorry.
9) S/B bugs in software for video, enough capacity to deal with this. For this reason, many inaccuracies when installing small-sized components 0402, BGA and the like.

I am sure that such a high cost is just business, and beautiful picture.
Why is this not a professional, and not even HAM choice?

Who buys buys big risks in the PC and ...
I wanted to warn seeing photos PC inside.
I Am disappointed :-(

Based on the bios boot screen from Ichan's video, this is the SBC in the machine:

http://iotsolutionsalliance.intel.com/solutions-directory/itx-d525-2chb-mini-itx-embedded-board-intel%C2%AE-atom%E2%84%A2-d525-processor-intel%C2%AE-ich8hb
http://www.zeroone.net.cn/?c=products&a=show&id=45

The atom D525 was launched in 2010, but is still in use in heaps of current generation products (my Lenovo EMC NAS has one in it). The intel lists that particular board as being added in Jun 2015. It's a perfectly fine & modern board to use in a product like this. A replacement ITX board is much easier to source than some custom SBC.

[thommo]

Yes, I agree Metalphreak - [David].
For what it's worth, that Mini ITX looks fine by me.

We use Mini ITX MoBo in a couple of our designs.
There are several models are available, from a number of manufacturers, which are specified as being long-life, with commercial support, for that reason.

[ServoKit]

I'm baffled by the workflow to establish the position of the fiducials. As I see it fiducials let the software map the positions in the part list (which are in PCB coordinates) to the machine coordinates. At my machine the fids are not expected to be in the part list as it seems to be here. Instead, after moving the camera over the fid you simply tell the machine in a dedicated dialog that the fid has (e.g.) coordinates x=5, y=5 of the PCB. Given that the camera position is known, that's all the software needs.

Regards, Axel

[mrpackethead]

I'm baffled by the workflow to establish the position of the fiducials. As I see it fiducials let the software map the positions in the part list (which are in PCB coordinates) to the machine coordinates. At my machine the fids are not expected to be in the part list as it seems to be here. Instead, after moving the camera over the fid you simply tell the machine in a dedicated dialog that the fid has (e.g.) coordinates x=5, y=5 of the PCB. Given that the camera position is known, that's all the software needs.

Regards, Axel

Ive never seen fiducials in a part list either.  thats odd

[mikeselectricstuff]

I'm baffled by the workflow to establish the position of the fiducials. As I see it fiducials let the software map the positions in the part list (which are in PCB coordinates) to the machine coordinates. At my machine the fids are not expected to be in the part list as it seems to be here. Instead, after moving the camera over the fid you simply tell the machine in a dedicated dialog that the fid has (e.g.) coordinates x=5, y=5 of the PCB. Given that the camera position is known, that's all the software needs.

Regards, Axel

Why would you want to have to enter this manually?
Having fids in the P&P location list makes a great deal of sense, and I'm surprised it isn't standard practice - you just create a part with the fid pad, resist clearance and pick/place point, and place it on the board. The offsets will then be baked into the p&p file, avoiding any need for manual entry.
It is stupid that (AIUI) the fids have to be the first items in the list - you should be able to either select fid parts manually, or tell it that parts called "fid" are always fids.
   

[ServoKit]

I don't have to do it manually; the file that the machine expects has actually two sections:

1. the actual part list in standard CSV format
2. configuration which can contain fids etc.

If I create a part list manually (which I did for testing during the first days), I simply leave out the 2nd section and set the fid coords manually. However, with a real PCB there's an ULP for Eagle that not only exports the part list but also the configuration part with the fids. As I see it that's the best of both worlds. Having the fids at some definite position of the part list or giving them a special name are both hacks.

Regards, Axel

[Smallsmt]

Why would you want to have to enter this manually?
Having fids in the P&P location list makes a great deal of sense, and I'm surprised it isn't standard practice - you just create a part with the fid pad, resist clearance and pick/place point, and place it on the board. The offsets will then be baked into the p&p file, avoiding any need for manual entry.
It is stupid that (AIUI) the fids have to be the first items in the list - you should be able to either select fid parts manually, or tell it that parts called "fid" are always fids.

Another good option is to place a fiducial at the origin.

[mrpackethead]

Why would you want to have to enter this manually?
Having fids in the P&P location list makes a great deal of sense, and I'm surprised it isn't standard practice - you just create a part with the fid pad, resist clearance and pick/place point, and place it on the board. The offsets will then be baked into the p&p file, avoiding any need for manual entry.
It is stupid that (AIUI) the fids have to be the first items in the list - you should be able to either select fid parts manually, or tell it that parts called "fid" are always fids.

Another good option is to place a fiducial at the origin.

I always ( just because i had no other better way to do it ), make the bottom left corner of the panel the origin.  I normally have fidudicals on the pcb frame, ( a bit that is thrown away ),   I always have three fididucals ( bttom left and right, and top right ). the reason for having no top left is so that you dont' orientate the board the wrong way accidently. 

Theres no definative right way, but there are probably some thigns that make it harder!

also, take some time to make decent fiducials that are easy for the camera to read reliably..  I use a 1mm dot of copper, with a gap and then a 5mm circle around it.    And make sure there is NO solder mask on it.

Hows those videos going Micheal?

[Ichan]

Below a small test placing 16x 1206 smd resistor from 4 feeders using 4 nozzles, with and without vision - speed is 100% both for run speed and feeder speed.

(Sorry the background music is beyond my control  )





14 seconds without vision and 18 second with vision, turned out as 4100 and 3200 CPH. Non vision speed is below my expectation but seems i still can try to speed it up above 100% speed setting, surprised by the speed with vision.

The placement result is almost the same, below is the result with vision - the pcb use transparent adhesive sheet provided by Neoden.



Funny i still do not fully understand how is the coordinates system works, seems it is different between manual programming and import from file.

-ichan

[elmood]

I found that the best way to do the fiducials is by creating a Neoden CSV file right away. I have a program that does this and puts these positions in the file right away. Their file format is pretty simple and a bit of testing can reveal pretty much what all the values mean.

Also, beware of using only 3 fiducials. I found that it made the positioning fairly inaccurate. It seems that the machine transforms all the coordinates to fit between the found points. Using four seemed to be the only way to get actually very good accuracy. And they should be spaced fairly far apart across the board.

Neoden responded directly to my question and I had to go back and read several replies to be clear. But it seems that whatever the size of the mark point there should be no other holes within a distance of 3x the mark diameter from that hole. There is a detection size range setting (min and max) which apparently doesn't affect the distance between holes... or that's how it sounded when I asked about it.

[mikeselectricstuff]

I found that the best way to do the fiducials is by creating a Neoden CSV file right away. I have a program that does this and puts these positions in the file right away. Their file format is pretty simple and a bit of testing can reveal pretty much what all the values mean.

Also, beware of using only 3 fiducials. I found that it made the positioning fairly inaccurate. It seems that the machine transforms all the coordinates to fit between the found points. Using four seemed to be the only way to get actually very good accuracy. And they should be spaced fairly far apart across the board.

Neoden responded directly to my question and I had to go back and read several replies to be clear. But it seems that whatever the size of the mark point there should be no other holes within a distance of 3x the mark diameter from that hole. There is a detection size range setting (min and max) which apparently doesn't affect the distance between holes... or that's how it sounded when I asked about it.

There really ought to be a way to manually select a mask region to avoid mis-recongnising

[elmood]

I asked about that, but the typical response to pretty much all my bug reports or feature requests is..........

[rwb]

@Ichan  Looking good so far   

This should be a interesting week for all of us as you get the machine up and running at full speed   

Keep up the great work with the video reviews.

[mrpackethead]

Below a small test placing 16x 1206 smd resistor from 4 feeders using 4 nozzles, with and without vision - speed is 100% both for run speed and

14 seconds without vision and 18 second with vision, turned out as 4100 and 3200 CPH. Non vision speed is below my expectation but seems i still can try to speed it up above 100% speed setting, surprised by the speed with vision.

Out of curiosity, the board came in from the left and then got spat out to the left when it was finished. Is this a configuration thing.. I woudl have thought for most applicaitons that having it feed out to the right, would be good..  ( i'd park my oven next to it on the right ) and have my stenciling on the left, in some kind of line..   

[thommo]

Another great update - thanks Ichan.

Although I'd like to achieve a super quick turn around, accuracy is my greatest interest with the N4.

If I have to redo a bunch of boards which I produced really quickly, I'll instantly be behind the eight ball.

I know you don't intend running much fine pitch stuff, but if there was any way you could run a fine pitch 0.4mm - 0.5mm test in your 'standard' review quality, you'd have me forever.

Happy to pay you for whatever out of pocket expenses might be involved and your time.

[uncle_bob]

Another great update - thanks Ichan.

Although I'd like to achieve a super quick turn around, accuracy is my greatest interest with the N4.

If I have to redo a bunch of boards which I produced really quickly, I'll instantly be behind the eight ball.

I know you don't intend running much fine pitch stuff, but if there was any way you could run a fine pitch 0.4mm - 0.5mm test in your 'standard' review quality, you'd have me forever.

Happy to pay you for whatever out of pocket expenses might be involved and your time.

Hi

+1

For some of us, the 0603 is a "giant" part and 0.5mm pitch is about as big as the main IC's ever get.

Bob

[rwb]

Yes same here @ichan , I am willing to send you some IC chips with .5mm pitch pins. One is a BGA with .5mm pitch balls that I would love to know if the machine can place properly.

I know your busy but when the time comes I'll send the parts over if your up for it.

[rx8pilot]

For some of us, the 0603 is a "giant" part and 0.5mm pitch is about as big as the main IC's ever get.

Bob

I have settled into an 0402 - .4mm average to have a chance at the necessary density. I would be bummed if I couldn't make 0201 and <.4mm work in the future. Do you think that is beyond the N4?

[TheSteve]

I highly doubt you'll successfully run the machine in production at 100% speed. The parts always look much better when using double sided tape then they will with paste. I'd aim for a top speed of 80%, unless Neoden adds independent control over the Z speed.

btw, we bolted our machine to the wall with a solid shelf bracket.

I wouldn't expect the machine to be super successful with anything below 0.65mm pitch but you never know.

[rx8pilot]

I wouldn't expect the machine to be super successful with anything below 0.65mm pitch but you never know.

Ouch.

[TheSteve]

I wouldn't expect the machine to be super successful with anything below 0.65mm pitch but you never know.

Ouch.

Maybe it will be ok, the smallest we have placed is 0.65mm. The placement accuracy seems pretty good overall but the parts tend to move when actually placed by the Z axis depending on amount of compression on the nozzle etc. We also saw the problem at times with our Juki, but it had an option to slow the Z axis independent of the X/Y.
Other then bug fixes I think the best feature Neoden could add would be control over the Z axis speed.

Also remember it is using stepper motors, not some nice linear encoders, so I don't ever see it matching a high end machine.

[uncle_bob]

For some of us, the 0603 is a "giant" part and 0.5mm pitch is about as big as the main IC's ever get.

Bob

I have settled into an 0402 - .4mm average to have a chance at the necessary density. I would be bummed if I couldn't make 0201 and <.4mm work in the future. Do you think that is beyond the N4?

Hi

I'm perfectly happy to accept a speed reduction to get one of these machines to do the job. If it runs at 1/2 the speed that others are after, but places 100%, that is an acceptable tradeoff. Letting the 0402's align with surface tension .. that's fine. Getting the IC's down right the first time, that pretty much has to happen. They don't do much self alignment.

Indeed this is a major hijack away from the main thread. My needs are "weird" compared to what most people are after. Since I've only gone into part of it, they are even more strange. My target is prototype / low volume / engineering run stuff. That's why speed is not quite the issue for me. The setup time and all that will be the biggest part of any job. Running ten boards in ten minutes plus an hour setup is not that much better than ten boards in twenty minutes plus an hour setup. If the extra ten minutes is the difference between 100% accuracy and an hour of rework, that is a fine thing. Yes, my world is different.

Bob

[thommo]

I had this exact discussion with Hiami from NeoDen this morning.

She pointed me to a board that they run for themselves - eg, part of one of their machines.

Looks good to me. I think it's worth the risk. If I thought I could repeatedly get this accuracy, I'd be in heaven.

[mrpackethead]

If the N4 had more feeders i'd be in like flin.  having less than 80 feeders is just problematic for me.  Just like the person up above, my requirement is not for runnign a small factory, its for supproting a team of design engineers who are constantly building stuff, ( 3 or 4 of something ) . And if we coudl build 10 or 20 of somethign it would transform how we could do a few other thigns.   I wonder how well two of these would behave end on end.