super-cheap pick-and-place device with ~1 mil accuracy Moderators: adafruit_support_bill, adafruit

[pstemari]

... Anyone know where I can find: ...
#2: linear track end blocks : I thought I'd find these, but so far no luck. These are simply thick metal plates (or cast equivalents) that have two 15mm~25mm round holes 50mm ~ 100mm apart to plug round linear rails into. Obviously I can get this machined up pretty easily, but better if I can buy off the shelf.

Get some 1-2-3 blocks, e.g. http://www.use-enco.com/CGI/INSRIT?PMAK ... TPG=INLMK3. The tapped holes are 3/8"-16, or about 9mm, a bit smaller than what you want, but should work or be made to work.

[bootstrap]

I don't understand why you need these sliders at all. Seems to me that the key parts of the system are:

1) The webcam, which can look down either at the template or at the PCB.

2) The pickup device, which must sometimes (but not always) be in a fixed repeatable relative to the webcam.

If you simply built a mount for the webcam, with a jig in the mount for the pickup device to fit into, wouldn't that work just fine? All you'd need to be able to do is remove and replace the pickup device without moving the webcam.

So the operation would go like this:

1) Bolt the webcam mount to the table.

2) Place the pickup device in its jig on the webcam mount. (Slide it into a square corner, or whatever. The pickup device can rest on the table and just align itself relative to the webcam, or can be clipped or set into the webcam mount somehow. It doesn't have to be fastened, just positioned; in fact it should be easy to remove.)

3) Position the template, with the device resting in it, on the table under the pickup device (so you can pick up the part without moving the pickup device sideways).

4) Pick up the part.

5) Remove the pickup device, with the part still attached. (This corresponds to the "Slide 50 mm to the right" in the original design.)

6) Take a picture of the template.

7) Put the board on the table under the webcam, and align the picture of the template with the video of the pads.

8 ) Put the pickup device back in its jig. (Corresponds to "Slide 50 mm to the left.")

9) Place the part.

No sliders, no precision machining.

I think I'll build one of these! It's a neat idea.

You know what? You are correct, though the best implementation might be a bit different due to nature of optics. Very cool that you recognized this simplification.

Assuming I understand your idea, the Z-axis assembly with pick-up device would be exactly below the camera. So when that assembly is removed, you'd see the template or PCB surface, and when that assembly was inserted below the camera (with their axes being exactly coincident), the pickup device could pick-up the component from the template or place it on the PCB. That is so cool.

The problem that makes that specific implementation problematic is the nature of optics (but it doesn't matter, because you can do something very similar and just as simple). When the camera is that far away from the component template, the image of the component template would be much too small (the component template would fill the center 5% of the image). I'm trying to think of optics that would solve that problem cheaply, but I don't see it. The webcam I have, which has a very conventional field of view, needs to be 50mm to 100mm from the template or PCB to make a 25mm to 50mm area on the template or PCB fill the width of the field. In your configuration, It would be more like 500mm away. Obviously there would be some focal-length telephoto lens that would make the template fit, but I'm not sure where we'd find those lenses. But hey, maybe it is possible. I do prefer your configuration over the alternative, which I'll mention now.

The alternative is to make the camera and Z-axis assembly have an identical mating piece. Then you:

#01: attach the camera
#02: capture the image of the tempate
#03: remove the camera
#04: attach the Z-axis assembly
#05: pick up the component
#06: remove the Z-axis assembly
#07: attach the camera
#08: align the image of the appropriate pads on the PCB with the holes in the component template image (still a ghost image on your display)
#09: remove the camera
#10: attach the Z-axis assembly
#11: place the component

Note: steps #1 to #3 can happend after steps #4 to #6 rather than before. Either way works, and each has advantages and disadvantages.

Of course, you gotta be careful you don't let the component touch anything while shuffling it back and forth. And the vacuum must still be active that whole time too. But that can be done.

I like your original configuration better though.

-----

I hope I'm not being stubborn, but I think there's a place for both techniques. For example, the PCB that made me go off on this tangent in the first place has 7 BGAs, 20 QFNs and 200 or more 0201, 0402, 0603 components. To assemble a PCB with that many components I think the time savings and simpler physical manipulation of devices in my original design will be worth the extra expense.

But hats off to you, because many if not most new developers and startups these days want to make very simple PCBs with maybe 1~5 BGAs, 1~5 QFNs, and maybe 20 small capacitors and resistors. For them, especially if they're only making 1 or 2 boards, your "cheaper to build" system sure looks attractive, especially if the price to buy is much lower, which it should be.

I've been beating myself up to find ways to keep the price of my design down without giving up any convenience or accuracy. Recently I found a rather high quality 2Mp webcam with variable intensity LEDs and ability to focus on objects very close (less than 50mm away) for only $10. And similarly after freaking out at the cost of vacuum pumps, I found a couple suppliers in china where I can get [what look like] very nice pumps for $10 ~ $25. So I figure I can still (hopefully, depending on CNC machining costs) sell the device for $1000 or less. Even at $2000 that would be about 20x cheaper than an automatic pick-and-place machine capable of 0201 placement, unless you're willing to take the major risk of trying a used (and maybe abused) unit.

I've been busy on this and the other project that I need this device for, and didn't see your message until now. Congrats on that brilliant idea. Have you made one yet? I expect both of these approaches to be very cool. I'm going to write another message and post the photo of my design (taken from my design in solidworks) so people can contribute more better ideas to improve my approach. Please contribute your great brain to my approach too!

Thanks for the great idea! I look forward to more. :-)

[bootstrap]

Well, after some distractions (a small contract for NASA, and struggling to learn solidworks (AARG)), I got back to designing this device, made major configuration changes to simplify, improve reliability and improve precision, and have almost finished now. Reasonable CNC quotes permitting, I'll have someone start making 2 or 3 prototypes in two or three weeks. So, last chance to add your two cents (or $64,000 worth) of ideas, and point out my worst ideas and design flaws.

I will attempt to post some photos here, and attempt to attach a PDF document that describes how it works. This is identical in principle to the original steps I outlined at the start of this thread months ago, but the configuration of the device has changed so the exact nature of many steps is somewhat different.

Sorry, but you're probably going to need to maximize your window to see the images. If you're willing to make a serious effort to critique the design, ask me for the solidworks 2010 files.

I usually think up really great, elegant names for things. This time I'm a total failure (!please help!), so the device is called "papec" for now (pick-and-place extra cheap). Sheesh. :-(

First, here is the description of the device as currently designed:
http://www.iceapps.com/papec_description_0003.pdf

The following is the overall view in solidworks:


The main components from bottom to top are:

#1: the PCB cradle
The PCB cradle holds the PCB and slides around on the benchtop or tabletop to position the component template or PCB surface under the camera or vacuum pick-and-place tip. The component template is the 60mm square plate near the front edge of the PCB cradle. The one shown is for all BGA components with 1mm solderball pitch with an array of up to 51 x 51 solderballs. These component templates will be ~0.50mm thick blackened (?passivated?) stainless steel sheets with laser cut holes, squares, rectangles, etc. Beneath the component template is a bright white square of teflon permanently attached to the PCB cradle. The component template is held stationary by four 3mm hardened stainless steel pins press fit into the PCB cradle, and sticking through 3mm holes in the 4 corners of the component template. A sample PCB is shown (a simulation of the actual 5.8" square PCB that I need to assemble for my other project, but only including the connectors that stick a significant distance above, below or beyond the edge of the PCB.

#2: the base or stand
The papec stand has four big fat sturdy legs to hold the main device at the appropriate distance above the PCB. As designed, components can stick upwards or downwards 36mm without hitting the bottom of the device or the feet of the PCB cradle (when the PCB is slid beyond the left or right end of the PCB cradle. I haven't added the leaf-spring clips that hold the PCB into the PCB cradle, but the three M3 threaded holes that will secure them are visible just this side of the front edge of the PCB. At the far left and right of the top surface of the papec stand are limit blocks. When the rotary table is rotated exactly 180-degrees, these limit blocks stop the rotation in exactly repeatable position. Because the limit arm is so long, any tiny error in either limit stop position causes an error only ~20% as large in the camera or vacuum pickup tip. I have not added the "tension device" that gently pushes the limit arm against the limit blocks in the two limit positions. These might be spring steel leaf springs, or perhaps just loops of surgical rubber tubing attached to each limit block for the operator to loop over the end of the limit arm. That's right, "fancy rubber bands".

What you can't see in this image is an 8mm wide by 3mm deep groove in the top surface of the main papec stand. In that groove, and in a similar groove in the 200mm diameter round rotary table plate above it, up to 48 (forty eight) 8mm ball bearings act as a thrust bearing to precisely and securely support the weight of the rotary table and the Z-axis assembly that holds the vacuum pickup device. The papec stand and the rotary table both have 56mm square holes the vacuum pickup device moves up and down through. When a component is lifted "all the way up", it stops in the square hole in the rotary table. Thus no component may be more than 12mm thick (oh, maybe 15mm).

#3: rotary table
The rotary table is the round flat plate on the center of the papec stand. As described above, the bottom surface of the rotary table has an 8mm wide by 3mm deep groove on near its outer periphery to act as a thrust bearing surface. A 1mm thick nylon or teflon sheet with 48 holes of 8.01mm diameter retain up to 48 hardened stainless steel balls and assure they remain equally spaced. The rotary table also has an ABEC-7 sealed radial ball bearing in its center that is bolted to the center of the papec stand to provide for rotation but prevent any lateral motion. The Z-axis assembly is bolted to the top surface of the rotary table.

#4: Z-axis assembly
The Z-axis assembly is a linear stage that moves the vacuum pickup device up and down about 80mm. Four linear bearings are bolted to the Z-axis assembly plate, and two 20mm diameter hardened stainless steel linear shafts slide up and down about 80mm through those linear bearings. A thin spring steel compliance plate is bolted to each end of the two linear shafts, and the rectangular vacuum pickup device is captured and precisely aligned by rectangular holes in the compliance plates.

Any one of 3 or 4 vacuum pickup tips can be screwed into the 16mm x 1.5mm thread in the bottom end of the vacuum pickup device to pick up components within a certain range of sizes and masses. The rectangular vacuum pickup device has a vertical 10mm hole from the vacuum pickup tip and upwards through about 1/3 of the length of the bar, where an M6 hole from the backside of the bar connects to a barb fitting, which connects to the vacuum tube to deliver suction to the vacuum pickup tips.

When a vacuum pickup tip presses against the top of a component, the thin spring steel compliance plates flex, but remain in perfect alignment (no x,y displacement or rotation) due to the geometry of the device. However, the top of the rectangular vacuum pickup device presses against a tiny pushbuttom "limit switch", which triggers a vacuum solenoid to toggle between "vacuum suction to the vacuum tip" and "light puff of positive air pressure to the vacuum tip" (depending on whether the component is being "picked-up" (from the component template) or "placed" on the PCB.

#5: camera
Not visible in this view is the downlooking camera, which is located 0mm ~ 50mm above rotary table plate, looking downward through a 55mm diameter hole in the rotary table plate. Someday maybe this will be a super high-resolution 5 megapixel camera (that's I'm developing), but for now I found a very nice $10 webcam with 2 megapixels, adjustable brightness LEDs, and fairly sharp glass lens that focuses even closer than I need (50mm ~ 100mm, which images 25mm ~ 50mm square components).

That's the description. I'll add a few more images below to help make sense of the above descriptions.


In the above image you can see the PCB cradle assembly, the component template at the front-center edge, and a sample 5.80" square PCB. The legs of the PCB cradle are cut very low so the PCB can slide off the left or right end of the PCB cradle without any component on the bottom of the PCB hitting the PCB cradle. This maximizes the width of PCB that the device can place components on.


In the above image you can see the groove for the thrust bearing, square hole in the papec stand for the camera to look through and vacuum pickup device to pass through, plus the limit blocks.


In the above image you can see the groove for the thrust bearing, the shoulder-shaft-bolt that bolts the radial ball bearing to the papec stand, and barely just a tiny bit of the radial ball bearing that is held in the 22mm diameter recess in the top surface of the rotary table plate.


The above image shows an exploded view of how the rotary table, teflon retaining ring, 6 of up to 48 balls, and papec stand all fit together.


The above image shows the vacuum pickup tip designed to pick up largest components (virtually all BGAs and QFNs).

Okay, everyone. Time to point out my stupidities before it is too late. Come up with some really great ideas and get one at cost once I get them into production.

Oh, and if anyone out there has a quality CNC setup and wants to make my prototypes, pipe up! I am especially interested in finding someone who wants to make batches of 20 to 100 units once I finish the software, create a video to show how to operate the device, and announce them as available for sale. I'm doing this much more as a service to the poor "left out in the rain" [potential] developers of really cool electronics devices who can't now prototype with BGAs, QFNs, 0201s, 0402s, etc. The way I see it, all the really creative new developers and startups are pretty much out of luck at the moment (or need to pay through the nose for PCB assembly services). Note: I got prices to assemble my prototype PCBs for my robotics vision camera device and the cost was $7200 for several sets of PCBs. I want to be able to sell this device for under $1000, so people can assemble their own. Compare $7200 to $1000, then realize that's just for one freaking project! PS: I was going to have 25 of the 5.80" PCBs and 50 of the 2.80" PCBs (which have only a few components), but that's still just insane if you ask me. And yes, I already have all the components, so they don't need to do anything but assemble. Oh, and that included them assembling ZERO of the through-hole components on the PCB. Their quote to run only ONE set of PCBs was $2000, which is still more than buying one of these suckers (CNC machining costs permitting). Obviously if the volume is significant, it should be cheaper to manufacture by designing some components as sand castings, die castings, or some other intermediate-cost castings. I've sorta lost track of what relative casting costs are these days. Comments welcome on every aspect of this project.

[alphatronique]

HI

Nice Design :mrgreen:

Qoute "Their quote to run only ONE set of PCBs was $2000,"

ummm ok assuming assembly line run a 200$/Hour whit 1000 part\Hour this make 10,000part job(10 hour job) .
that based on my own shop price , one of my local competitor run it manncorp 7720 @ ~50$/hour

so assuming you use your device ,place a may take ~12 second (look for refdef ,place part on template,manually align etc etc)
15 sec x 10,000part = 150,000/3600 so ~41 hour no stop + human induced placement error

you still cheaper but you really what to work a week for assemble a batch of pcb ?

and next problem i see was stencil and reflow ,for use 0.4mm pitch BGA you need excellent stencil and good forced air reflow
of you do it in batch oven you risk to have high failure rate and need rework , and rework was really hard on this

i prefer buy a old obsolete pick place machine for under 2000$ whit feeder .use good DFM technique (Design for maneuverability)
and only assemble by hand ic that machine cannot do during machine assemble rest of part ...

or maby i just not well understand your concept ?

i also curious i have try to find cnc shop that may make nozzle and none accept it since it tell me it to small for it lathe

Best regard

[bootstrap]

@Alphatronique:

The finest BGAs on my PCBs are 0.65mm pitch, not 0.40mm pitch, so probably good quality stainless stencils will be adequate. My mention of finer pitches is my desire for the manual pick-and-place device to be precise enough for 0.40mm pitch devices, or maybe even 0.30mm pitch. Personally I don't see any current need for below 0.50mm pitch, but just give the chip people another 3~5 years and 0.40mm and maybe 0.30mm will become common... I'm guessing. Plus, one of my "guiding principles" in product design is "design forward"... design my products to be novel today, and fully capable many years into the future without upgrade or replacement.

The time and effort involved in preparing everything for the PCB assembly companies is substantial. You've got to put all the parts of each part number in a ziplock bag, fill out one of their labels in exactly the manner they demand, and repeat for all components. I have a really great, super complete BOM, but they didn't like that, and wanted me to re-enter all the information in exactly their pre-determined spreadsheet format. Then they had a few other circles they make you run in. So we gotta count that in the overhead of getting PCBs assembled outside, which closes the gap somewhat. They make us do more than half their work for them.

All this nonsense take time. So does shipping the stuff to them take time (and costs money). Then we need to wait until they fit the job into their schedule, and add extra days for weekends because "they're a real company and don't count weekend days". Then they ship it back which is more delay, and more cost. Furthermore, they don't do the through-hole parts, so after the PCBs come back we still need to solder a couple dozen components (in each of the larger PCBs... only a couple each in the smaller PCBs). So one big deal to me is being able to assemble my prototype PCBs the same day I receive the PCB cards... not 2 or 3 weeks later, or more.

While I designed and prototyped the PCBs for my past 10 or 20 products without a single mistake (no cuts, no patch wires), that sounds like famous last words, even to overly self-confident me. :-) For example, this is my first PCB that has DC-to-DC voltage converter-regulator circuits on it. So I wanted to populate and test those circuits (and make sure their output voltages are correct) before populating the rest of the components. I don't think this issue is important when prototyping many PCBs, but it is relevant now and then (and maybe more important for [less careful] novice designers than me).

Many components on the PCBs don't need to be placed by this device. My experiments indicate 0805 and even 0603 capacitors are easy to place manually with my manual vacuum pick-up device (from zeph) - or even tweezers - and therefore these components can be placed much quicker by hand than with the papec device. Same goes for QFPs and all those packages with visible leads sticking out the sides. It is easy to visually see whether all the contacts are soldered well, and lack opens or shorts. So probably only half the components need to be placed on most PCBs, and this is also true of my PCBs. This reduces the total time to assemble PCBs compared to placing everything with the papec. My goal for the papec is to place components that cannot be reliably placed without it... not to "place everything".

You are correct about needing a reasonably good reflow oven. To assemble reliable PCBs with fine pitch hidden lead components requires a good oven. Hot air guns and blow torches won't do! Hahahaha. Not for BGAs and QFNs, that's for sure.

The problem with old automatic pick-and-place machines for me is... they don't handle 0201s and some other modern fine-pitch packages that I already need. I'm not sure what good is a pick-and-place device that can't place the most difficult on my PCBs! Of course, for someone who never selects tiny parts, who needs an automatic pick-and-place machine at all? I haven't seen any used pick-and-place machines for $2000, but maybe I wasn't even looking for machines that can't handle my parts.

About small nozzles. My plan is to buy "blunt end hypodermic needles" and shink-fit them into a small hole in my normal nozzle, and have the hypodermic needle part only sticking out 5mm or so - mostly for safety reasons. You're correct about machine shops not being able to machine super-tiny tips, but that's what hypodermic needles are for (in my opinion). They're very nice, actually. They're stainless steel and have IDs from 0.08mm to 0.40mm. That covers the range from the smallest caps all the way up to something capable of picking up large BGA packages. Try that.

[martij13]

Instead of combining a roller bearing with a custom thrust bearing have you considered a pair of angular contact bearings? I am concerned that the manufacturing cost of the thrust bearing will be high and in the long run wear could be an issue.

As for the vacuum tips depending on what material you want machining small holes is not impossible. If the tip is 304 or 316 it will be very expensive (these materials work harden and weld easily making them difficult to machine), but if you can live with 303 or even better brass for the smallest tips I don't think it would be too difficult. Drills down to .05mm are available as standard items so it can be done. You just need to find the right shop with the right equipment (namely swiss machines) and, if possible, experience making similar parts.

[bootstrap]

Instead of combining a roller bearing with a custom thrust bearing have you considered a pair of angular contact bearings? I am concerned that the manufacturing cost of the thrust bearing will be high and in the long run wear could be an issue.

As for the vacuum tips depending on what material you want machining small holes is not impossible. If the tip is 304 or 316 it will be very expensive (these materials work harden and weld easily making them difficult to machine), but if you can live with 303 or even better brass for the smallest tips I don't think it would be too difficult. Drills down to .05mm are available as standard items so it can be done. You just need to find the right shop with the right equipment (namely swiss machines) and, if possible, experience making similar parts.

If I find the thrust bearing is expensive, I will consider two bearings as you say. I prefer the thrust bearings because the rotary table is so large, and having the support out near the OD is very beneficial mechanically speaking. To make a pair of central bearings work well will require they be separated by a significant distance. Fortunately, I have plenty of room to do that in the current design.

The 8mm bearing-quality balls cost about $0.03 to $0.10 depending on which kind and where you buy them. That works out to about $1.50 to $5.00 for the thrust bearing, assuming I insert all 48 balls, which I plan to do. The nylon retainer will probably cost more than the balls because cutting a single 8" diameter circle out of 12" x 24" sheets is very wasteful (only 3 per sheet, and each sheet ain't cheap, as in $12~$20 or so, depending upon material).

However, if I make each retainer out of 3 or 6 pieces, I can fit a large number of them in each sheet and greatly cut costs. As far as I can tell, this works fine (as long as each section contains 2 or more balls, the balls keep the retainer in alignment).

As for wear, I'm considering whether I should have those grooves cut with a ball end mill so each ball is supported along a line rather than on one point. Or possibly a V groove. My only worry about the [almost-hemisphere] ball-end-mill groove is a possible increase in rolling friction. Another possibility that I plan to test is having tiny hardened stainless steel rings stamped out to put in the bottom of those grooves. Unfortunately, this has the same expense problems as the full-size nylong retainer. I'm not sure how practical it would be to get those made in sections - I worry each ball would "tick" as it went over the seam between sections. With 48 balls that might not matter though, since there are always other balls on solid surface very near by. Only tests will tell about that, I guess. Another possibility is to substitute rollers made to go in roller bearings, but they are much harder to find.

About the vacuum pickup tips. I think brass or bronze will work just fine. My only worry about drills as tiny as 0.05mm is their ability to start the hole in the center of the piece where it belongs without the tip wandering around and bending the drill shaft until it finally gets a depression started in the metal. I'll look into it though... even the hypodermic needle approach requires tiny holes to be drilled, though the holes aren't quite as tiny since the OD of the needles are larger than the IDs.

I keep changing my design for the vacuum tips. Every time I make that long rectangular vacuum device bar longer, and make the vacuum tips shorter. I'm getting to the point where I'm about to try to make the tips only 32mm long or so (from top to bottom, including the thread, a relief area, a locating OD to assure precise centering, a largish OD section to grip with the fingers to thread in and out, and about 4mm to 8mm of tip sticking out. I want to make the tip short to make the small tips safe (they are needles, after all), but I don't want to make them too shore because sometimes 0201s are placed right next to thick components (jacks, connectors, etc).

Thanks for the ideas. I will definitely keep them in mind.

[Rob_ski]

Greetings.
I am a Pick and machine Operator based out of the US. As a company we found a niche in small to medium sized prototype and production runs. The main driving point behind a pick and place (PnP) machine is not the ability to place 1 part with 100% accuracy. A typical tin/lead or lead free solder will draw the component if it is too far off (within 25% of the pad typically). What a PnP machine excels at is enabling an operator, with training, to mass produce assemblies. For smaller (qty) assemblies a great deal of my billable hours is programming and loading all of the different parts onto the PnP machine. It is more cost efficient on smaller (qty) assemblies to just do it by hand because of initial programming and loading time. Which is why most companies have minimum amounts they bill for ($1000 was mentioned in one of the other posts)

The machine we have which is ancient by modern standards can place one 0603 type component every second. The assembly i just completed has a little over 800 components and takes almost 18 minutes to finish top and bottom. There are others responsible for hand placing smaller assemblies. They are expected to place one 0603 sized component every 10 seconds.

I am not familiar with the design and manufacture of machined equipment, however I can give advice as to the practical applications of the design. I could use a thorough explanation of how you expect your machine to operate ideally. From the design i see thus far, I see equipment for picking up one part, and placing it on one PCB.

If you are designing this machine around the ability to place a component with a tolerance of 1 mil. I have to say that all BGA's and fine pitch PLCC's have legs that are .5 mil, if not smaller. With that amount of play in the placement it would be more practical to just hand place it, as you would almost certainly have to take the time to correct each placed component smearing the paste and giving an inferior joint. There is a way to hand place a BGA or PLCC by hand with 100% accuracy. I have X-rays of a number of them I have done.

[bootstrap]

Greetings.
I am a Pick and machine Operator based out of the US. As a company we found a niche in small to medium sized prototype and production runs. The main driving point behind a pick and place (PnP) machine is not the ability to place 1 part with 100% accuracy. A typical tin/lead or lead free solder will draw the component if it is too far off (within 25% of the pad typically). What a PnP machine excels at is enabling an operator, with training, to mass produce assemblies. For smaller (qty) assemblies a great deal of my billable hours is programming and loading all of the different parts onto the PnP machine. It is more cost efficient on smaller (qty) assemblies to just do it by hand because of initial programming and loading time. Which is why most companies have minimum amounts they bill for ($1000 was mentioned in one of the other posts)

The machine we have which is ancient by modern standards can place one 0603 type component every second. The assembly i just completed has a little over 800 components and takes almost 18 minutes to finish top and bottom. There are others responsible for hand placing smaller assemblies. They are expected to place one 0603 sized component every 10 seconds.

I am not familiar with the design and manufacture of machined equipment, however I can give advice as to the practical applications of the design. I could use a thorough explanation of how you expect your machine to operate ideally. From the design i see thus far, I see equipment for picking up one part, and placing it on one PCB.

If you are designing this machine around the ability to place a component with a tolerance of 1 mil. I have to say that all BGA's and fine pitch PLCC's have legs that are .5 mil, if not smaller. With that amount of play in the placement it would be more practical to just hand place it, as you would almost certainly have to take the time to correct each placed component smearing the paste and giving an inferior joint. There is a way to hand place a BGA or PLCC by hand with 100% accuracy. I have X-rays of a number of them I have done.

Hi Rob. Thanks for your observations and comments. Let me respond to your comments and set me straight if I am wrong about anything or stupid. Understand that I've assembled zillions of PCBs with through-hole components over many years, but other than a little playing around with vacuum pickup tip, hot air pencil and a couple of my prototype PCBs, I have no first-hand SMT experience... just extraplotion from what I've read.

It is also my impression that components will self-center when they are placed within 25% of the pad width. A 0.50mm pitch BGA or QFN has 20-mil pitch and 10-mil pads, so placement accuracy of 2.5-mils is sufficient (25% the size of the 10-mil diameter pad). I am attempting to make my device accurate to 0.025mm == 1-mil with the understanding that accumulation of random errors will likely lead to the actual placement accuracy being in the range of 0.050mm == 2-mil or maybe even 2.5-mils. I do hope it is closer to 0.025mm == 1-mil thatn 0.050mm == 2-mil because 0.40mm and 0.30mm pitch components are clearly in our future.

I totally agree with you about the purpose and economics of automatic pick-and-place machines. However, this leaves individual developers and startups in a very nasty situation. They often don't have $50,000 of spare cash lying around to buy an automatic pick-and-place setup (including feeders, oven and possibly soldermask printer and/or wash equipment). And the very substantial overhead of setting up conventional pick-and-place machines makes contracting for prototype PCB assembly rather expensive. I do sympathize with the companies who offer these services, but they royally banned me off when their final quote is 5 times what the "come-on" price quote matrix on their websites imply (my experience with AAPCB and others).

Of course the problem with "do it by hand" is... human beings cannot reliably assembly PCBs that contain BGA components, other components with hidden contacts like QFN, iLCC, etc, or super tiny discretes (0201 and 0402). Humans can reliably assemble QFPs and other packages that have visible leads and pads given pre-heaters below and good hot-air pencil from above, especially with good lighting and stereo microscope setup.

So it seemed to me that individual developers and startups were stuck. That's what led to me to develop an inexpensive ($1000) manual pick-and-place device that lets us reliably place hidden-contact components, fine-pitch discretes, and any other problematic fine-pitch components. Everything we can reliably place by hand will be easier and faster to place by hand. I figure that includes discretes larger than 0402 or 0603 or 0805 depending upon the tools, skills and vision of individual developers.

I figure my device will consume about 1 minute to "switch over" from one package type to another, plus 15 to 30 seconds to place each component. In other words, it is not fast... unless you compare to the time required to program the equivalent PCB into the automatic pick-and-place machine, load the components into the feeders, load the feeders into the machine, and do everything else required to assemble ONE printed circuit board. Not to mention waiting for your PCBs and components to be shipped to the assembly house, and then get shipped back!

My device sets the component down vertically from above, with component contacts exactly on the appropriate PCB pads (which already have solderpaste on them, except for BGAs possibly). Therefore, there is zero smearing of solderpaste. From my limited experience AND from reports by others who have lots of experience, it is NOT reliable to expect soldermask and the surface-tension aspect of solder to prevent shorts and opens caused by solderpaste smearing.

Because my drawing doesn't show the vacuum system or the limit switch that detects the vacuum tip pressing against the component, you may not be able to tell that my device works the same as conventional pick-and-place machines in that respect. But it does.

Of course a developer would be completely insane to assemble even small production runs of PCBs with my device. For that, automatic pick-and-place machines win hands down on almost every metric. No doubt about that.

I don't quite understand your final paragraph. I think you may have confused millimeters with mils in one case. A BGA with 0.50mm pitch (that's a very fine pitch BGA) typically has 0.25mm diameter balls. That's a pitch of 20-mils and solderballs of 10-mils. I have no idea where you think there are components with "legs that are 0.5-mil if not smaller".

I would love to see a video of RELIABLY placing a couple dozen BGA components on a PCB with 100% accuracy (especially 0.50mm pitch BGAs). I absolutely do believe it is possible to place one or two in a row, but no way do I believe you can "hand place" a large number of BGAs on a PCB with 100% accuracy. You need to prove that to me. I would be thrilled to know how you do that. But it must be reliable for dozens and dozens of components in a row. Otherwise you'll have one misplaced component and/or open-connections and/or shorted-connections on a PCB which renders the entire PCB worthless. What's worse, when you're assembling prototypes --- which IS the idea here --- you don't know whether the circuit design is correct or whether there are PCB layout errors, so you never know whether a non-working PCB is due to an assembly problem or a design problem! That is very problematic!

I looked into the stick-on stencils for BGAs, which probably do work. However, they are so expensive that the cost of the stencils for a PCB with several BGA components is quite a bit more than the cost to have a PCB-assembly house assembly the entire PCB. So much for that idea (though it does let you assembly prototypes immediately, which is sometimes a very significant benifit).

I'd love to not need to make this device. It took forever to design, forever to find all the components (to make it cost effective as a product), forever to learn solidworks, it will cost me thousands to have a couple prototypes made, and I still need to write the software that makes it work. Hell, I might even need to make a small PCB to make the air solenoid toggle the vacuum off and on. So if you have a way to avoid finishing this project, please tell me. No, cancel that... please tell us ALL. Because there are thousands of developers who "give up" or never start because they cannot afford assembly of prototypes with BGAs, QFNs, 0201s, 0402s and other modern SMT components.

I appreciate your experience and thoughts. Please continue...

[brucef]

There is a way to hand place a BGA or PLCC by hand with 100% accuracy. I have X-rays of a number of them I have done.

I too would love to know how this is done. I've never placed a surface mount part in my life, but I find myself reading about reflow ovens, P&P machines etc. and it's all very fascinating. What are the tools and tricks involved in manually placing such finicky parts?

I fondly remember drawing circuits with a sharpie, agitating the board in an etchant bath, drilling holes with a 1/4" hand drill and soldering in parts with a dual-heat soldering gun. How times have changed.

[designer2k2]

Hello,

just some tips for this:

• use "Luer Lock" Pickup tools, there are 1000´s tools avaiable at nearly no cost, dont try to reinvent there...
• Rotate the Chip around its center, otherwise your math will need endless calibration to compensate all your machine related problems.
• your Pickup should beside its roation also have a spring so that with overtravel you can create "Bondforce" (from 10-500gr is good)
• and your Pickup should have a Sensor to detect that it made a contact, to make precise Z measurements
• get a in X/Y/Z fixed "Upward looking camera" to see the chip from the bottom to compensate pickup errors
• have a "Downward looking camera" at the same gantry with the pickup in a fixed X/Y offset to it, to make a substrate adjust (focal point in Z deeper as the pickup tool, otherwise you crash during adjust)

and think about how you want to calibrate the machine, how you want to get camera scale´s, find the offset between the pickup and downward camera, find the position of the upward camera, find your gantry system origin and so on...

1mil is only a target for the software and your troughput, if you go slow and your math is correct you place it precise.
its only getting expensive when you want to be fast and precise...

greetings d2k2

[bootstrap]

@d2k2: Yeah, I definitely prefer to adopt existing components.

I did a search on the internet for luer lock pickup tips but didn't come up with anything that seems to work for me. I did find some parts, but not that cover the range of sizes I need (~4.00mm ID, ~0.50mm ID, ~0.10mm ID) and fit the same holder. Also, I'm not whether the holes are perfectly centered in the part that is grabbed by the device. For 0201 and smaller components the hole needs to be 0.10mm to 0.15mm diameter, and therefore must be centered within a small fraction of 0.10mm (say, 0.025mm). Otherwise this super-tiny tip will come down with part of the vacuum hole off the side of the component, and not pick it up because the vacuum leaks. If you or anyone knows where I can get a set of compatible pick-up tips in the sizes I need that fit my requirements, then I'll adopt them. So far, I don't see them.

I don't need to rotate the chip... around its center or around anything. If you understand how my device works, you'll see that matching up the captured image of the component template holes/cutouts with the corresponding pads on the PCB assure every contact on the component is placed exactly on the center of the pad on the PCB, no matter how much the component or PCB is rotated.

I think your next comment is saying the component should be pressed down into the solderpaste with 10 to 500 grams of force, and should be spring loaded to achieve this. That's what those two thin, flat spring steel "compliance plates" at the top and bottom of the vacuum pickup bar are supposed to do in my design. When the component is pressed down on the PCB those plates flex. They are "leaf springs" instead of "coil springs", but they achieve the same result, except they are inherently incapble of rotation (since component rotation ruins my system).

Yes, my system has a tiny button at the top of the vacuum pickup bar. When enough downward pressure is applied (tip against component during pickup, or component against PCB during placement), those two compliance plates bend and the top end of the vacuum pickup bar presses a tiny pushbutton (not yet shown in 3D images). The pushbutton toggles the vacuum solenoid between "vacuum" and "slight positive air pressure" to achieve "pickup" and "placement".

The configuration of my system avoids the need for an uplooking camera and image recognition software to detect the component contacts. That is one of the advantages of my design.

I'm not sure I understand your final point. In my design, when you rotate the rotary table 180-degrees (from one limit-stop to the opposite limit stop), you move the camera optical axis over the exact axis of the vacuum pickup device that goes up and down. The "component template" and "PCB surface" are at exactly the same distance from the camera, so one focus is perfect to image both component templates and contacts on the PCB surface. I'm not sure exactly what configuration you are describing, but my scheme has no "crash" scenario.

My device requires 2 calibrations.

#1: The camera needs to be focused on the PCB surface by turning the focus ring on the camera lens. That's easy enough.

#2: When the rotary table is rotated from one limit stop to the other, the optical axis of the camera must stop exactly where the vertical motion axis of the exact center of the vacuum pickup tips. This is necessary to assure the tiny little 0.10mm hole in the smallest pickup tip comes down exactly on the middle of the tiny 0201 (or smaller) component. The limit stops can be adjusted and tightened down. Once set they should stay put until somebody messes with the camera (which they shouldn't). This calibration sounds very difficult off hand, but in practice it is very easy. You just put some white paper on top of the PCB, touch the end of the smallest vacuum pickup tip against an ink pad, then lower the vacuum pickup tip until it touches the paper, leaving a tiny ring with a 0.10mm hole in the middle. You then rotate the rotary table to the other limit stop and view the image of that little ring on the paper on your computer display. Then you adjust and secure the limit stop when the ring is exactly centered between the four central pixels on the image (made obvious by vertical and horizontal lines that extend almost through the center of the screen). This assures the optical axis of the camera and the center of the smallest vacuum tip are exactly aligned.

I don't have an "upward camera" or "gantry system" (other than the rotary table).

Again, one feature of my system is... it doesn't require any math! It is inherently correct as guaranteed by the same camera in the same position focusing on the "component template" and "corresponding pads on the PCB" at the exact same distance from the camera. This avoids all need for "math". The process becomes simply "make the two images align on your computer monitor" (the image of the component template holes and the image of the corresponding contacts on the PCB). This is how my device can have NO encoders and NO measuring devices of any kind --- other than "pixels on the camera" and the limit stops which get calibrated once and locked down (until you move the camera, which you shouldn't).

Sorry if I don't correctly understand some of your comments.

[designer2k2]

oh, now i understand your system! this is no automatic machine... why not automatic? :wink:

yes luer lock is not centered, what is no problem on a gantry system like machine, but for you this will make no sense.

you maybe want to look here:
http://www.smallprecisiontools.com/
http://www.micro-mechanics.com/

then you just need to make some kind of clamping device for this tools and your set.


About the switch as touchdown:
in a perfect world your touchdown would trigger within 1µm from touching the surface. With a optical system you could get very close to this!

something like this: http://www.digikey.com/scripts/DkSearch ... =H21LOB-ND (alternatives here:http://forums.reprap.org/read.php?13,14858,14864)
you would get a very high repeatability and with some adjustment very little force to trigger.

looking forward to see this in action :)

[Rob_ski]

http://www.solder.net/stencilquik/default.asp
http://www.solder.net/stiknpeel.asp

There are tutorials on the website explaining how to use them. I've placed 10 or so BGA's using those having an X-Ray of each one taken after reflow. Each one has come back near perfect.
Although I have a 5 zone convection reflow oven. I imagine this could be reflowed with a heat plate and an air gun

I was confusing my Mils with mm

[bootstrap]

http://www.solder.net/stencilquik/default.asp
http://www.solder.net/stiknpeel.asp

There are tutorials on the website explaining how to use them. I've placed 10 or so BGA's using those having an X-Ray of each one taken after reflow. Each one has come back near perfect.
Although I have a 5 zone convection reflow oven. I imagine this could be reflowed with a heat plate and an air gun

I was confusing my Mils with mm

Yeah, I looked into those stickons, and I assume they do work (looks like they should). Unfortunately, for a PCB like mine that has a few (7) BGA components on it, the cost of those 7 stickers is more than the cost of the assembly house to assemble the PCB. So yeah, it does let you "assemble at home", but it isn't cost effective. It is potentially "time effective", assuming you order the stickers long before your bare PCBs arrive. Last time I checked they cost $7 each, assuming you don't accidentally mess any of them up. That's $50 per PCB for my PCB, and more for PCBs with more BGAs. And that still doesn't solve the same problem with QFN components (unless maybe they now have QFN equivalents... don't know).

My goal was to find a cheap and permanent solution that works for all mini-pitch, hidden-contact components. That's what my device is. I'm sure the stickers make more sense for some people, like if they only make a couple/few PCBs per decade. For others, not so much. In my case I need to make 25 of my larger PCBs (with 7 BGAs and lots of QFN-like components) to prototype larger systems and to provide [development] samples to others. So for me, even this one project justifies the manual place device and a reasonable reflow oven.