Monday, April 21, 2014

DIY Pick and Place Machine - Part 3

Be sure to check out the other posts in this series below.
Part 1
Part 2

Now we are starting to make some headway into making this stock ShapeOko kit into something resembling a CNC Pick and Place Machine.  I am quite pleased with the progress so far, and I am learning a lot along the way.  For instance, I have discovered that blunt tip dispensing needles with luer lock connections are non-concentric, but more on that later.  For now, let's talk about some of the parts that are making this PnP take shape; specifically the head unit and the PnP fixture.

Head Unit:

The purpose of the head unit on this PnP machine is to facilitate the following operations:
  1. Advance the component tape to present a new component for placement.
  2. Vacuum grip to pick up a component for placement.
  3. Rotate the picked up component to the correct orientation for placement on the PCB.
  4. Hold a syringe of solder paste for robotic solder paste dispensing.
When designing the head unit I wanted it to be made from primarily off the shelf parts.  This speeds up development and allows a distributed approach for duplication should any reader wish to replicate the same.  I was able to achieve this ideal save one part, which I turned on a lathe.  The rest of the parts for the head unit came from online retailers that are listed below.

Here are a few pictures of the head unit assembled.

Below is a photo of the head unit prior to assembly.  It is annotated with numbers which will correspond to a descriptive list.

Refer to the image above and the list below for part numbers &c.  I see now that I left off an annotation number for the bottle of loctite.  It is slightly optional, but probably a good idea for many of the fasteners as stepper motors can be a bit vibratey.  All of the parts ( except #9 ) came from the following vendors: McMaster-Carr, Servo City, SDP/SI, Clippard, and eBay.  

  1. A 10-32 threaded female luer lock adapter and blunt tip dispensing needle from McMaster Carr.  Respective P/N's: 51465k151 and 75165A682 and 75165A677
  2. 6-32 nuts.  I had these. They measure 0.245" across the flats.
  3. 6-32 x 3/8" Long,  Button Head Cap Screws.  I purchased these from a local hardware store.
  4. 6-32 x 1/4" Long, Button Head Cap Screws.  McMaster P/N: 98164a106
  5. #6 Washers.  McMaster P/N: 98032a436
  6. Wood Screws to hold the maple block ( #7 ) to the head.
  7. This is a v-cut block of wood for holding the solder paste syringe.
  8. A flat channel bracket from Servo City P/N: 585468 and a pair of Servo City 90 deg. dual mounts P/N: 585470.  All of this is held together with screws ( line item #4 ) and washers ( line item #5 ).
  9. I turned this shaft out of some 8 mm bar stock ordered from McMaster-Carr P/N: 1272t38  A drawing to make this part is below.
  10. A Clippard pneumatic cylinder, plus a hardware store washer.  This is for drag feeding the SMD tape.  Clippard P/N: FSR-05-1 1/2
  11. Two 1/4" I.D. bearing mounts from Servo City P/N: 535110
  12. 1" x 1" 80/20 Extrusion.  I cut mine slightly shorter ( < 3.75" ) than the Servo city bracket this is mounted to.
  13. 3.75" Aluminum channel from Servo City P/N: 585443
  14. GT2 Timing belt and 2 disparate pulleys from SDP-SI P/N: A6R51M093060, A6Z51M036DF0608, and A6Z51M036DF0605
  15. A NEMA 17 stepper from eBay, plus a stepper motor mount from Servo City P/N: 555152
  16. An arbor shim kit for setting the end play of the diameter 8 mm shaft ( line item #9 ).  McMaster P/N: 3088a928
In the assembled head unit photos you can also see a brass spool valve for diverting the CO2 flow betwixt the solder paste syringe or pneumatic cylinder.  I probably should have just ordered another solenoid, but the diverter valve is Clippard P/N: HTV-3F.

Lets begin assembling the head unit by attaching the aluminum channel from Servo City ( line item #13 ) to the 80/20 extrusion ( line item #13 ).  You will notice I used non-standard mounting methods here.  I drilled and tapped the 80/20 extrusion for 40-40 BHCS ( Button Head Cap Screws ).  This was necessary because the standard 1/4-20 BHCS plus t-slot nuts came too close to my pulleys and belt for my liking.  One way around this would be to select a smaller tooth number pulleys and belt, while maintaining the same center to center distance.  


Next up, attach the bottom bearing mount using some 6-32 x 3/8" BHCS's, washers and nuts as shown below.

Below, you can see the pneumatic cylinder installed.  Note the hole position, and don't forget to install the GT2 belt prior to continuing, otherwise you are likely to gain the authors experience in re-assembly.

In the next photo you can see the stepper motor installed with the stepper motor mount from Servo City.

Install the diameter 5 mm I.D. GT2 pulley on the stepper motor shaft.

Prepare the aluminum plate and brackets as shown.

And install the solder paste syringe holder using wood screws.  For this, I countersunk the aluminum plate so that the wood screw heads were below the surface.

Attach the solder paste syringe holder assembly to the front of the aluminum channel using 6-32 x 1/4" BHCS's as shown.  It is important to do this now, as it could affect the vertical clearance on the vacuum shaft.  The author was able to install qty 3, 6-32 screws during this operation before discovering the stepper motor was blocking one of the screw holes.  It was decided to move on with life instead of trying to increase the number of fasteners further.

Ok, the steps shown below are the most complicated, but I think you will manage to follow along just fine.  Temporarily install the vacuum shaft and top bearing mount.  You may elect to install this with one screw.  Slide the shaft axially to detect end float.  Measure, if you can, this displacement.  Mine was 0.015". Otherwise, you can attempt an iterative process of fitment.

Remove the top bearing plate, install the dia 8 mm I.D. GT2 pulley, and install one of the shims from the arbor shim kit.  Since I knew my end float was 0.015", I selected a 0.012" shim.  This would leave me with 0.003" clearance.

Bolt the top bearing mount in place, and check that the shaft rotates freely and there is minimal, but perceptible end float.  Mine was predictably measured to be 0.003" axial clearance after assembly.  If you cannot rotate the shaft, or cannot detect/measure end float, you should select a smaller shim and repeat.

Below you can see the luer lock adapter and blunt dispensing needle installed ( line item #1 ).

Here is the assembled head unit interfaced to the ShapeOko gantry via an 80/20 extrusion angle bracket.

Remember that spool valve I mentioned earlier?  here it is installed to the head unit with a Servo City P/N: 545424

This is where I discovered just how bent/curved/or otherwise not straight the vacuum needle was.  When I rotated the head unit with the stepper motor, the needle spun around and orbited too.  So, I bent the needle a little and was able to get it pretty concentric.  Then I removed the needle, and re-installed on the second start thread only to discover the needle was non-concentric again.  For now, I will continue with this setup, but it may be worth investigating some precision needles at some point - probably making them.

If you recall the shaft that needs to be fabricated, below you can find a drawing.  Basically the length of the diameter 8 mm shoulder is going to be determined by your particular setup.  As it is drawn below worked well for me, but you may do well to have a measure of your system first.

The PnP Fixture:

You have probably seen the aluminum plate with milled slots, clocking holes, a shoulder register and a bunch of tapped holes.  This is my vision for making a low volume, home pick and place machine.  I can make as many different plates as I need based on the circuit board I want to fabricate.  

The design is made pretty universal.  There are 4 tracks for resistors or transistors, two tracks for capacitors, one for a regulator and one for a mosfet.  I selected these parts and subsequently the track dimensions based on the boards I wanted to populate first.  If I have a new board design in the future, and this setup does not have tracks for the correct parts, I can make a new fixture.

Here is how I envision this fixture working.  I say envision, because I am still learning how to write G-Code, and I am not sure how to do what I want just yet.  At the time of this writing, I am currently working on the programming I am about to describe.

If you look at the fixture you will see two countersunk holes for flat head screws.  This fixture will be bolted to the ShapeOko MDF work surface here. 

Inboard and towards the front, you will see two round bores of different diameters.  They would have been the same size, but I messed one up while machining.  Since the fixture is unlikely to ever be bolted down, perfectly aligned with the machine coordinates, I am trying to do the following.  Probe each of these two holes, calculate the angle between the machine coordinate system and the line made by joining these two center points, and rotating a new set of coordinates for the machine to use that are aligned with this fixture.

There is an X-Y register for locating the circuit board.  This X-Y ( 0,0 ) is located 5.000" right and 0.500" toward the rear from the bottom left probe bore.  Using this new origin I know where my circuit board pads are ( from the gerber files ), and I also know where to tell the machine to look for each type of component to place.

To hold the circuit board in this register, orienting it's X-Y ( 0,0 ) I used a piece of an old coping saw blade.  Held in place on one end by a Jarrah wood clamp, the other end applies enough spring pressure to keep the board steady and registered during assembly. 

Things are starting to look like this might just work.  I still have a long way to go with G-Code and tuning the machine operations.  I haven't even fed any parts yet!  Never the less, I keep chipping away at the subcomponents on this job.  

Incidentally, I will probably be showing this machine off at the Ann Arbor Mini Maker Faire, so if you are in the area, be sure to stop by.

Tuesday, April 15, 2014

DIY Pick and Place Machine - Part 2

Be sure to check out the other posts in this series below.
Part 1
Part 3

Ok, so we are moving right along with this project.  Parts are coming in weekly, and other parts seem to be endlessly delayed, so we make do.  Since the last post I have been concentrating my CNC PnP education on the following areas:

  • Automagic Z-Axis probing for setting the machines vertical zero position relative to your PCB.
  • Wiring in an emergency stop.  This should have been done before starting the first time.
  • Configuring LinuxCNC's HAL for digital outputs.
  • Investigations into robotic solder dispensing.
Much of this work has been around the back of the machine, so here is a picture of that, and lets discuss.

Above is the back side of the machine.  Along the top is a piece of DIN Rail that extends from the left and right of the ShapeOko frame.  This is where the majority of parts are installed.  Below this, you see a grey vertical surface with parts mounted.  This is a long steel L-Bracket, surplus from some IKEA furniture, that I screwed to the ShapeOko MDF work surface.

Working ( approximately ) from left to right, here is a list of parts:
  1. Green parallel port breakout board.  This gives screw terminal access to your PC parallel port.  To this we attach stepper drivers, digital outputs, digital probe inputs etc. Phoenix Contact P/N: FLKM-D25.
  2. DIN Rail mount Terminal Block.  Multicomp P/N: SPC10564
  3. I colored some of the above terminal blocks red or black with a sharpie.  These are electrically connected with Mulitcomp P/N: SPC11891 jumpers
  4. Big Easy Stepper Drivers - sold by SparkFun P/N: ROB-11876
  5. The two yellow circles are 3-way pneumatic valves. P/N: EC-3-12 for the valve and P/N: C2-RB18 for a connectorized lead. One of these will be used for solder dispensing/component tape advancing, and the other is for the vacuum pickup system.
  6. Connected to the above valves, via blue tubing ( clippard P/N: 
    URH1-0402-BLT-050 ), is a pressure regulator, clippard P/N: MMR-1N.
  7. The valves and regulator do not come with fittings so I also ordered clippard P/N: 11752-5-PKG
  8. Next you can see a circuit board.  This custom job contains electronics for Opto-isolation, transistor amplification, and Back EMF suppression.  All of which allows us to switch inductive loads like relay coils and solenoids from our PC parallel port safely.
  9. Further to the right is a 12 V DC relay for switching the 110 V AC vacuum pump on and off.  The second set of contacts will eventually drive the vacuum solenoid to either create vacuum in the pickup needle line or vent to atmosphere. I'm not sure if these are available or not, mine is scavenged from an old project Omron P/N: MK2P-S-DC12
  10. On the very far right you can see a blue snubber that is wired across the aquarium ( vacuum ) pump.  I added this to reduce EMI on the 12 V Rail.  It was necessary.  ITW P/N: 104M06QC100.
So, that's kind of an overview of some components, that make the subsystems.  Let's talk a little more about some of the details of setting up the machine and how I intend to make this a useful robot instead of a list of cool parts.  This post will discuss setting up the Z-Axis probing and robotic solder paste dispensing.

Z-Axis Probing:

In case it wasn't already obvious, this entire project has been made so easy by those who have gone before me and were willing to write up their own findings.  Clearly, the best example of this is LinuxCNC, without which, nothing on my machine would be moving this early in the game.  LinuxCNC is powerful and customizable.  And thru adding some text to the configuration files, we can setup the probe functionality built into LinuxCNC, or perhaps more correctly G-Code. wrote up an article on how to setup Z-Axis probing.  This will allow me to know how high my pickup needle and solder dispensing needles are relative to the top of my circuit boards.  This is important because later on, when I write G-Code programs for the PnP I will need to reference the top face of the PCB.  Of course you could set the Z-Axis manually, but why would you want to?

Have a look at the picture above, and I will try to explain what happens when we probe the Z-Axis.  The aluminum block you see is 0.75" thick.  The wire connected to this aluminum block via a ring terminal is connected to the PC's parallel port ground.  The alligator clip, clipped to the same electrical point, is connected to the parallel port pin 13 on the other end.  What I just described is a closed switch as LinuxCNC sees it.  If we were to clip the alligator clip to the conductive needle instead, we would have a switch that closes when the end of the needle touches the aluminum block's top face.

If you were to call a probe command in G-Code it may look something like this:
G38.2 Z-2.500 F15

G38.2::G38.5 are probing codes.  In the case of the above code it reads: ( G38.2 ) Probe towards work piece, Stop on contact, Signal error if failure. adding ( Z-2.500 ) means that if you get to Z-2.500 before touching the probe surface, stop; something is awry.  The F15 is your feed rate.

So, F15 is fast for probing, right?  Especially when you consider that's tutorial implements debouncing for the probe signal.  IIRC, the debounce routine is 100 iterations of the base thread.  Say our base thread runs at 1 mS, well 100 iterations would be 100 mS, and 100 mS at F15 inches per minute equates to 0.0025".  In other words, if you just zoom down into the aluminum block and set Z = zero ( or 0.750" rather ), you will be 2.5 thou too low.  the answer is to slow down that feed rate and consequently decrease the overshoot.

Since we don't want to wait ages for a slow feed rate if we are a long way away from our aluminum probe block, I wrote the G-Code subroutine for probing as below.  If you read the comments you will see that essentially the code quickly probes down to the block, comes up a bit, then probe down slowly to get an accurate Z-Axis reference.

o100 sub

( Set current Z position to 0 so that we will always be moving down )
G10 L20 P0 Z0.0

( quickly probe down to touch off plate )
G38.2 Z-3.0 F15

( switch to relative coordinates )

( rapid up 0.1 )
G0 Z0.05

( probe slowly to get a more accurate zero )
G38.2 Z-0.2 f0.5

( set Z0.0 to be 0.75 above work surface - this is due to the touchoff plate thickness )
G10 L20 P0 Z0.75

( switch back to absolute coordinates )

( rapid to Z1.0 - probe tip is now 1" above work surface )
G0 Z1.0

o100 endsub

Below is a video of this Z-Axis probe action.

Now, after making my way thru the probe tutorial above, and modifying the G-Code subroutine for faster probing, I learned enough about how to make buttons appear on my LinuxCNC GUI, and more importantly, make those buttons do something useful.  So, I set out to make some more buttons to speed up the machine setup.  I made buttons for zeroing the X and Y axes as well as one to turn on my solder paste solenoid for 3 seconds to purge any dry solder paste from the needle.

To make this happen, I edited myCNCconfigName.ini such that the HAL UI MDI commands look like below:

HALUI = halui
HALFILE = 4axis_PnP.hal
HALFILE = custom.hal
POSTGUI_HALFILE = custom_postgui.hal

# add halui MDI commands here (max 64) 
MDI_COMMAND = o100 call 
MDI_COMMAND = o101 call 
MDI_COMMAND = o102 call 
MDI_COMMAND = o200 call 

Then I edited custom_postgui.hal to contain the following:

net remote-o101 halui.mdi-command-01 <= pyvcp.o101
net remote-o102 halui.mdi-command-02 <= pyvcp.o102
net remote-o200 halui.mdi-command-03 <= pyvcp.o200

And finally, I added the following to custompanel.xml

# add halui MDI commands here (max 64) 

So, when you press one of these buttons on the LinuxCNC GUI the respective subroutines below are called:
o101 sub

( Set current X position to 0 )
G10 L20 P0 X0.0

o101 endsub

o102 sub

( Set current Y position to 0 )
G10 L20 P0 Y0.0

o102 endsub

o200 sub

M64 P1 ( DO 1 ON )
G04 P1.0 ( dwell seconds ) 
M65 P1 ( DO 1 OFF )

o200 endsub

I will be revisiting this creation of buttons again soon.  I am thinking of how to probe the X and Y axes instead of manually setting the zeros.

Robotic Solder Paste Dispensing:

To start with this robotic solder paste dispensing problem, I decided to rig up a manual pneumatic dispenser.  With a switch connected to a solenoid, I was able to selectively apply CO2 to my syringe of solder paste.  The solder paste naturally escapes thru the dispensing needle, and the hope is, that the correct amount makes it onto the correct solder pad.

This sounds easy enough.  We have a fixed orifice, and a fixed pressure ( 40 PSI of CO2 ), so we should have a constant flow, and with a constant time, a deterministic volume dispensed.  What I don't have, however, is a fixed viscosity fluid.  My solder paste varies from dry to entirely too wet.  The reason for this, I think is that the paste is 2 years old, and it has a shelf life of 6 months.

All this means is that my time based experiments in this area are largely useless at this point.  I have ordered a new tube of solder paste.  This has proven to be a good learning experience.

Eventually the human operated switch was replaced by an opto-isolated, transistor amplified, and back EMF protected circuit, that means that I can now write G-Code to turn on and off this solenoid for a specific time period with the code:
M64 P1   ( turn D0 ON solder solenoid )
G04 P1.5  ( dwell seconds ) 
M65 P1   ( turn D0 OFF - solder solenoid )

Strap the digitally controlled solder paste syringe to the pick and place gantry, and you can do this:

Below is a picture of solder paste dispensing onto a piece of paper.  The 5 dots on the left were dispensed with a 1.5 second period, and for the 5 dots on the right the CO2 solenoid was open for 2 seconds.

Here is a video of the dispensing in action. You can notice that the first pad does not get enough solder dispensed onto it.  Given the viscosity disparity of this expired solder paste tube, and my elected method of application, i'd say these results are pretty good.

While the solder was being placed on the circuit boards, I was hand populating the components.  The resulting boards are just as good as they ever have been using a solder paste screen.  And certainly, one of the coolest features of robotic solder paste application is that I can make new boards while never having to purchase a ~$200 solder paste screen again.

Ok, well I can't think of anything else to talk about on this topic.  It's all pretty prototypical at this point, but I'd say its moving in the right direction.  And if you've made it this far into the reading, thank you; you're doing better than me.