Critical adjustment for MDC ECM P&P machines

[johnyq]

The pick and place machines made by MDC in Japan are belt driven and work well when properly adjusted.
But they are prone to perplexing behavior which is not commonly understood. I encountered the situation a few
years ago after buying a used machine. The machine symptom was very poor positioning. It worked if you
taught each position because it would repeat to that position, but it wouldn't go where you told it.
Position errors exceeded 1/16 inch. This was after I calibrated each axis with a laser interferometer
within about .0015 inches. How could each axis be so precise and still mis-position by
such a large amount. I suspected software so I setup to measure both axis' but they both
checked out. I carefully adjusted belt tensions but still no luck. I put angular measurement optics
on the gantry arm. I got a big surprise. The arm skewed significantly as it moved left &
right. No wonder it wouldn't position correctly. This meant that each end of the arm was traveling a different
distance under control of the drive belt. These belts were driven by a common shaft so they must differ in
tooth pitch. You can't tighten the belt without stretching and you stretch the tooth spacing when you do.
When I tuned the belts for identical pitch using the angular interferometer the skewing of the arm stopped
and positioning accuracy was vastly improved. Note the belt tension now differed but the pitch was the same.

I wonder if others are struggling with the same problems.

I don't suggest you rush out and drop $70K on an interferometer to check your P&P machine. All you need
is a mirror and a laser pointer. Mount the mirror on the gantry arm and bounce the laser off it to a distant
wall. Jog the arm back and forth if the dot moves the angle is changing and your belts need work. The laser
and mirror will reveal a tiny angular change that can guide the adjustment process too. Invest a few dollars
and a few minutes to check your machine if you are having positioning problems. It worked miracles for me.

JY

[adafruit]

are the ECM machines different than the 7722? for the 7722 there are in fact two Y motors but when it 'homes' it calibrates any Y delta using the two dots, perhaps thats why?

[johnyq]

I have not been near any of the later 77xx machines so I cannot give you a specific answer. Are there cal dots at
each end of the arm (4 total)? If so the software might attempt a calibration for each belt. But if not and the two motors
drive individual belts and the belt pitch is different then the same error will occur. You could try the laser test just for fun.

When we use the angular interferometer to tune the belt pitch we can see a huge improvement but the angular resolution
of the interferometer is such that the error is never zero.

We have achieved better calibration first by reversing (end to end) the belts to move a possible wear point. We also replaced them with new belts from Manncorp and even replaced all the pulleys to accomodate wider stronger belts. The pulleys were either worn or did not match the new belt profile causing backlash in positioning. You need to careful because all belts of identical pitch don't necessarily have the same tooth profile.

There are belts designed for linear positioning with steel cores but the smallest pitch is 5mm where the ECM belts are 2mm. I considered revamping the machine with huge motors and these belts but haven't yet.

I leaving for the Apex show in Las Vegas to-day. If Manncorp have such machine there I will have a look.

JY

[blogger]

7722 also uses 2 motors/belts to drive the Y axis. X is single motor.
They do calibrate, but the software is such junk, I'm sure 99.999% of the problems are due to extremely poor software.

[johnyq]

I'm back from the Apex show in Las Vegas. It was very interesting. Manncorp was there with an empty
booth. They brought absolutely nothing. I guess they came for a good time in Vegas not to actually
sell something. In fact I had to look up their booth number to find them. There were only a few
other machines there under $100K so they could of had that market segment to themselves.
It was not even evident that they were in the SMT business.

The booth was empty of customers every time I walked by.

If they aren't going out of business, they sure looked like it.

I did have occasion to study numerous machines up close. I still marvel at the diversity
of ways chosen to do the same job. If you put enough nozzles on a head you can put down
30K parts per hour. With 2 gantries and 4 heads you can have 120K parts per hour. Think
about consuming 300+ reels of components per 8Hr shift or maybe even 1000 per day!!
You have feed this thing with a forklift !

The price is a big hole in half a million. Feeders extra.

Feeders (8mm) seem to range from $600-$1200 each some that come 8-12
in a fixed package approach half that but have limited flexibility.
One Swiss feeder actually fed the tape by pulling on the cover tape -
no sprocket drive - really. When fed under keyboard control parts routinely hopped
out of the pockets. I wonder how well it worked.

One manufacturer offered a way to bring a board image from a flatbed scanner to program
or verify a board. Interesting idea.

One thing all the machines had in common was complexity...lots and lots
of complexity and lots to go wrong. No wonder the prices are multiples of $100K.

Most of the booths were staffed with sales and field service people so even
my most technical questions were well answered.

I studied the user interface of many of the machines. The Windows interface
or something like it was everywhere. Most of the screens were filled with
nondescript cave drawings giving little clue what they did. Most often they
looked like little PCBs. I noticed that even commonly used functions were buried
many levels deep. A flurry of mouse activity was required to almost anything.

It would drive me crazy.

Primitive man communicated with cave drawing before evolving language,
alphabet, and vocabulary. I guess we are in the cave man stage of machine control.

[Alphatronique]

Hi Johnyq

that may explain wly i love and stay whit my old machine
and i prefer put money on update it whit my own software that buy new machine
new machine was really complex and rely on vision for compensated lack of mechanical repeatability
and all that complexity = costly down-time + airplane ticket for the tech ..

the good think is that still have company that keep old hardware running like mydata and quad
TWS automation was also look good and simple ..

Best regard