P&P placement force

Chat about pick and place machines, reflow ovens, assembly techniques and other SMT tips & trix

Moderators: adafruit_support_bill, adafruit

Please be positive and constructive with your questions and comments.
Locked
ralphstirling
 
Posts: 18
Joined: Fri Nov 06, 2009 2:53 pm

P&P placement force

Post by ralphstirling »

How does the MDC machine (and P&P machines in general) control the force with which a component is set down on the board? Is the nozzle assembly spring loaded, air cushioned, or counterweighted? Or do they just try to be accurate about Z height and stick it into the paste and release the vacuum before hitting the board? Have you ever run into problems with a warped board?

-- Ralph

adafruit
 
Posts: 12151
Joined: Thu Apr 06, 2006 4:21 pm

Re: P&P placement force

Post by adafruit »

it does something kinda cute which is that first it knows the part height and the PCB location and then it also puffs out the part with air to 'launch' it downwards

johnyq
 
Posts: 10
Joined: Mon Nov 16, 2009 11:55 pm

Re: P&P placement force

Post by johnyq »

The MDC placement head is spring loaded in the down direction. Accurate Z is not enough
because PCB may not be level. You do spec the height of each part and on some models
use the video camera to measure the part height.

I have extensively redesigned the machine and re-written the software from scratch.

JY

adafruit
 
Posts: 12151
Joined: Thu Apr 06, 2006 4:21 pm

Re: P&P placement force

Post by adafruit »

johnyq wrote:The MDC placement head is spring loaded in the down direction. Accurate Z is not enough
because PCB may not be level. You do spec the height of each part and on some models
use the video camera to measure the part height.
in theory the machine could use the video to 'focus' on the PCB and determine the distance.
I have extensively redesigned the machine and re-written the software from scratch.
neat! you should post it up and we'll try it out :D what was the primary reason for rewriting/designing?

johnyq
 
Posts: 10
Joined: Mon Nov 16, 2009 11:55 pm

Re: P&P placement force

Post by johnyq »

I aquired an ECM 96 about 5 years ago. It was missing the computer and interface card and software. Manncorp
provided the necessary parts and I got it going. It has been extensively modified to the point of only slightly resembling it former self. I'll summerize some of the mods.

First we tested and calibrated it witha laser interferometer a distance measurement tool that will read a 10 millionth of an inch. One axis had some bad spots but reversing the belts improved it. When I was done each axis
would position within .001 inches everywhere. But the actual position was out .080 inches in some corners. How could that be? I discovered a major problem and sent a detailed description to the manufacturer in Japan. Briefly, the recomended procedure is to adjust the left/right motion (I call it X) belts to equal tension. This doesn't necessarrily work becase the belt pitch might not be the same. On this machine the Y axis arm would skew significanlty even though each axis measured separately showed very little error. I put angle measuring optics (.00001 deg resolution) on the arm and tuned the belts for no skew which was not equal tension. Then the machine would position properly.

We needed more tape feeders so we added 24 more on the front by replacing the rod and bearing with a THK linear rail
with taller supports to make room. We needed still more feeders so we installed a mount for the 3 vibe bases in the back. We also made a cut tape block to hold 20 more parts 12 mm size tape. We can and do routinely place boards with 600 parts in 3 sq inches with 120 unique parts.

MDC in Japan were kind enough to give the source code for this machine. It written in Borland Basic. If I remember correctly it was about 18K lines with 4 or 5 statements per line. So somewhere near 80K lines if one per line.
I have decades of system programming experience in both assembly and high level languages. I worked with this program for a little while before deciding to scrap it and write my own. In 2 wks and 3K lines of Pascal I surpassed the functionality of the original. The code is about 8K lines now. The format of the user interface is entirely different. There is only one screen with 8 windows. It is far too complex to describe here but it is very quick and easy to use. In fact jobs setup in much less than one tenth the time. It is also mathematically more complex, the skew and fiducial transforms are more accurate then before. The code is fully multi-taking so all motion and feeder activity overlaps. Even the slide motion is handled by a lookahead process so thay are always positioned in advance. The vibe feeders are controlled and run only when necessary.

Most of the calibration and setup procedures are fully automatic or nearly so. To setup nozzles for example you just target the left-most and then the right-most and the system fills in the rest. The Z calibration is done with vacuum sensing and a guage block.

The motors on the machine were trouble so I replaced them with US made SmartMotors. Ultimately the squaring jaws were replaced by a camera but that necessitated a entirely new head which I designed and built. If you haven't guessed by now I have an extensive machine shop.

We put optical sensors on all the tape feeders and on a new 8 nozzle rack. If you remove or just lift a feeder the system clears its part number and puts that feeder at the top of the tape feeder window. You just scan the bar code on the reel and the number is updated and all program updates happen automatically. You don't have to count or mis-count the feeders anymore. - A big mess when you have 72 feeders slots a mix of sizes. The feeder configuration can stay with the machine and you can load new jobs and they will find all the parts automatically and tell you if you need some changes. We leave the machine packed with parts all the time and run numerous different boards with little or no changes. There are 80 optical sensors in all. The nozzle rack is monitored and protected by the software and these sensors. I made changes to the nozzle rack to make pickup and storeage work every time.

I gradually eliminated various boards and interfaces until now none of the original hardware remains. We even got rid of the interface cabinet and replaced it with a new one we built. The wiring and cable complexity is a tiny fraction of what it used to be.

I'm in the final stages of camera automation. The original down facing camera will identify fiducials as well as teach. A side facing camera will do part alignment. An upfacing camera is for larger QFP's. I didn't use a video capture board like everyone else but designed a 2"x3" card with its own MPU on it to handle all the video image data. It will handle 4 cameras so there is room for one more but I don't have plans for that yet.

neat! you should post it up and we'll try it out what was the primary reason for rewriting/designing?
I have probably already answered this question in that my machine is so different that the code won't work anywhere else.

in theory the machine could use the video to 'focus' on the PCB and determine the distance.
You can measure distance with focus. I have an optical Coordinate Measurement machine that does that and it works very well to .0001 inch accuracy. The optics are very fancy and costs about $125K.

I measure height on the P&P with the vacuum sensor. It repeats within a thou. My software will automatically calibrate
anything that requires a correct Z value with the vacuum technique.


I have attached a 5 meg .doc file with photos that describes the machine as of 4 years ago.



By the way we make aircraft instruments for a living.

JY

johnyq
 
Posts: 10
Joined: Mon Nov 16, 2009 11:55 pm

Re: P&P placement force

Post by johnyq »

The file was too big for download. Sorry.

ralphstirling
 
Posts: 18
Joined: Fri Nov 06, 2009 2:53 pm

Re: P&P placement force

Post by ralphstirling »

johnyq -

That is a very interesting story! I have been toying with the idea of
starting an open-source P&P machine project, and your experience
confirms my thought that it is doable. As most people (me included)
lack interferometers, I would not use belts, but packaged linear
actuators (such as those from IAI), replacing their controllers with a
LinuxCNC system. I would be very interested in more information
about your feeders and vision system.

-- Ralph

blogger
 
Posts: 43
Joined: Tue Nov 24, 2009 5:59 am

Re: P&P placement force

Post by blogger »

what was the primary reason for rewriting/designing
i would imagine, un-usability (is that even a word?) of original software.

adafruit
 
Posts: 12151
Joined: Thu Apr 06, 2006 4:21 pm

Re: P&P placement force

Post by adafruit »

blogger wrote:
what was the primary reason for rewriting/designing
i would imagine, un-usability (is that even a word?) of original software.
not disagreeing. although im wondering if rewriting it exchanges one evil for another :) I do think that it would probably benefit MDC to release the source code for the pick & place...its likely that many bugs and annoyances could be fixed by just the 4 people on this forum :)

johnyq
 
Posts: 10
Joined: Mon Nov 16, 2009 11:55 pm

Re: P&P placement force

Post by johnyq »

I don't if anyone should take the time write their own P&P code. I probably shouldn't have. But then I did and now we have a machine that can become anything that I want it to be providing I have the energy to do it. I have done other stupid things like write my own assembler for the 8051 processor that I have used for decades. In fact I have written several assemblers. The current one unlike any other and does thing most programers have even dreamed of yet. My programmers and I use it every day. We write large applications the current one (not P&P) about 700 pages long.
I also wrote a drawing compiler that converts CAD drawings to NC programs directly. We used it for years but have switched to MasterCam. (what a mess).

One of the goals to writting the P&P was to reduce the setup time as much as possible. I have a way to convert our PCB CAD drawings direclty to P&P programs. No effort and no errors. This puts a serious demand on absolute accuracy as opposed to teaching each part position.

As far as machine design goes it shakes down to belts, leadscrews, or linear motors.

I was very sceptical of the belts. But after measuring their performance to a millionth of an inch I developed a newfound respect for them although they still strike me as Mickey Mouse. The biggest problem withthe belts is the 2mm pitch is very hard to find on this continent. .080 pitch is easy but .079 is not. If the machine had to sustain any load in position they wouldn't work. But its not a milling machine so it does work and works very well.

Leadscrews are of course the mainstay of many positioning systems like lathes and milling machine. I have measured our good quality machine tools and found they will repeat to few 10's of millionths. They will produce enormous forces, thousands of pounds in a machine tool. Itis not commonly understood that they impose a speed and acceleration limit which is not as good as you would like. Our best mill will do 2000 in/min but at great cost, but then it moves a table that must weight a good fraction of a ton. Accuracy of the leadscrew is limited only by what you are willing to pay. Ball screws are required for the upper end applications.

Linear Motors are fastest method but not cheap and cannot produce huge forces like a lead screw. But big force is not a P&P requirement.

Positioning speed is just one factor affecting placement rate. The high end machines use multiple nozzles and multiple heads to get the real big numbers. Chasing high positioning speed like most things has diminishing returns. One often overlooked but serious concern is the cost of feeders. My collection would cost about 100K at list price. I didn't pay anywhere near that. But at least in my world 100+ unique parts on a PCB is common. I'd like a feeder for every reel of parts I have (a thousand might do it).

An open source P&P is an interesting idea, but there are so many machine dependent requirements that it would be tough to port to different systems. I not talking about language compatabilty here. Real time multitasking code is required and not easy to do or easy to adopt from someone else.

A far as un-usability of the original code goes, I assume customers of the ECM machines used it every day. I have talked to a few. If you aren't a programmer and thus don't have a concept of what a program could be you just accept it and build boards. I made the decision to start from scratch by weighing the balance of effort to learn enough about the Basic source code to modify it and the effort to just start clean. Clean won. I have faced this sort of decision over the years and have usually lost when I didn't start fresh if the code wasn't really good to begin with. But then I started programing 40 years ago and what might be practical for me might not be for some one else.

I use assembly for MPU's and Pascal on the PC. I hate C and won't use it under any circumstances.

John Y (905) 295 2755 Niagara Falls Ontario

sirket
 
Posts: 128
Joined: Fri Oct 26, 2007 8:46 pm

Re: P&P placement force

Post by sirket »

I have to ask what kind of aircraft instruments you make? Just wondering if I've encountered them in one cockpit or another :)

johnyq
 
Posts: 10
Joined: Mon Nov 16, 2009 11:55 pm

Re: P&P placement force

Post by johnyq »

To see what we make see insightavionics.com

Locked
Please be positive and constructive with your questions and comments.

Return to “SMT (Surface Mount Tech)”