Ordered my micRo today.

Have a neglected hand-held solder paste dispenser sitting at work so that will be my initial goal (after playing around for a while atleast).

Brain, since your goal seems to be pick'n'place could we expect to see a tool head for this sold on Lumenlab?

Have you considered a setup that can automatically switching its tool head to run through all the stages of PCB routing, Solder mask, solder paste dispenser, pick'n'place (and hot-air in-situ reflow).
Or does it sound a little too much given pick'n'place is a bloody big job in itself? A budget system capable of that would be insane.

Elder and BC, I am starting to realize that a capable pick and place machine would require quite a bit of programming above and beyond coordinate generation from a PCB file. A reasonable machine would have to coordinate optical information to even accurately grab and place parts, whether you have them fed from reels or fed from a tray since nearly every reel I have seen allows for significant position displacement from tape centerline and twisting in side the tape reel, not to mention what sort of misalignments would be possible using tray fed parts (the optical information would require an additional 4th axis to twist the spindle and rotate the part into the correct position). The air side of the pickup assembly would have to incorporate both vacuum and positive pressure to grab and hold parts but also to give them a small puff off the placement tip (this is especially crucial with parts smaller than 0805). This air side of the pickup assembly would be possible with some valving and a few pumps. The reels themselves would each require a solenoid, and a reasonable machine would require at least 8 reels (parts could alternately be populated one component reel at a time with reels loaded serially). Control of the reel feed would be possible but would quickly begin to fill up the control ports of EMC2 interface software with a large number of reel feeders. Anyone out there have any experience programming optical recognition that also wants a pick and place machine? This basic machine would still need user intervention to change out placement tips. The micRo could be used to machine accurate solder masks and also a reasonable solder stencil printer frame. Detection of IC registration marks and proper placement of those would require additional optical recognition programming effort. So far the port count that I have is:

3 axes - micRo gantry
1 axis - component rotation
3 ports - vacuum, positive pressure and air solenoid control (maybe only 2 ports here)
8 ports - tape reel feeders (maybe only 1 here with frequent attention for switches)

I count 15 ports total, can EMC2 handle that? I only saw 9 axes of simultaneous control available when I quickly looked over the spec. Please correct me if I am wrong though, I'd love to have an 8 reel pick & place. It might be possible to do localized in-situ hot air/IR convection heating component by component but I think the chance of having heat related failures due to over application of heat or solder failures due to under application of heat would be extremely high. The localized heating would definitely have to be experimentally verified. Worst case, ovens could be purchased relatively inexpensively off of eBay. Here at work we do lots of small production run PCB fabrication and the evenness of oven heating has proven time and time again to be the make or break step in our SMT production line. Any additional thoughts?

I'll start by saying the extent of my knowledge in circuit building is on a bread board. I've never seen or used a pick and place...

...so I don't know if it'd be practical with that application since I don't know how the reels work, but it seems like you could put together a subsystem with the reels and have them all controlled with a microcontroller only pulling from one port on EMC2.

But I'm just a codemonkey...when things get hectic I goes as modular as I can.

For the 4th axis rotation that would simply be a stepper in place of the spindle correct? For multiple tools I've seen video of a 4 axis machine with a triple tool setup run off an additional stepper. Seems one of those could be setup as a subsystem as well and simply take a flag to switch tools. Might be more suitable on the rogr though.

Not sure on the options in EMC for this...but hell, it's open source.

Thats why I called it a bloody big job.. I'm guessing BC has some rough plans of what he is thinking already though. Doubt he's thinking machine vision sounds like to much work for the reward.

If you can assume general orientation out of the reel, I'd imagine its not too hard to correct to near perfect alignment with the typical right angle corner tilt/shake setup. I'm not sure how practical that is or not without machine vision but it sounds possible to me... I have no experience with pick'n'place machines though. I've seen some videos of machines that have flexibility / quality control that is obviously way out of hobbyist league but for micRo's market a few restrictions won't kill. I guess we just have to hope and not expect too much.

Like you said oven costs aren't that high so in-situ heating isn't that big of a worry if it doesn't work.

I don't know EMC just yet so no opinion on the controls yet.

I'm keeping it simple to begin with and using the micRo for solder paste dispensing on a professional PCB.

I have lots of goals for micRo, but yes pn'p is a good one. Making the PCBs isn't actually interesting to me..but making the robot that makes PCBs is!

I've given this about 10 minutes' thought, that makes me an expert, right? I always try to approach these types of challenges from the 'bottom up' rather than the 'top down'. 'Top down' approaches give me vertigo!

That said, let's talk about what we want; a machine that can pick up a dinky SMT and plop it down on some machine applied solder paste. The next step is to figure out the minimum amount of function which brings the maximum benefit. What if we could supply the components in an indexed fashion? This would allow us to eliminate lots of automation.

Say we hand sorted the components by pushing them into precisely routed channels (in a block of plastic). The channel locations are absolute and only the same type components are in each channel. The part size is also absolute, so we have an index from part to part. The channels can be flared at the 'loading end' to make it easy to load the parts. With this method the parts are pre-indexed so we don't need all of the complex automation, just pn'p.

I'm going on intuition, but it seems like the part placement and positioning on the PCB is 95% of the time and pain, making the manual parts-sorting an acceptable compromise.

So if we accept that manual sorting into a loading tube is acceptable then our machine vision system and 4th axis of rotation become non issues assuming that you lay out your board with all the components facing in a single direction. If you would like the flexibility to lay components in both directions then a 4th axis capable of 90° turns is necessary. Using a loading tube or variant would let you get away with vibratory feeders and remove the PITA that is solenoid feeding of tape reels. That's a pretty good bit of simplification that would also eliminate the need for a vision system (phew, that would have sucked!). If you leave out the ICs and add them to the list of through hole stuff that will need hand soldering anyway then we get a machine that can solder a whole bunch of standardized RC components in a set of typical sizes. This proposed machine will eliminate a whole bunch of tedious soldering and reduce manufacturing labor and component cost (assuming SMT is cheaper, which it should be). This definitely seems doable and reduces the main issues to 1) XYZ calibration, 2) vibration slot loading of SMT components and 3) operation of the vacuum and positive pressure mechanism. Not bad considering it sounds like BC is already interested in investigating XYZ calibration for the main micRo project. Lets get this started and try to find a few inexpensive vacuum and positive pressure pumps and solenoids to do the pressure side of things. I'll grab a handful of RC tape reels from work and start doing some modelling in Solidworks for component feeders. Any more ideas for possible avenues of exploration?

Are you planning on 'push feeding' after every component pickup, or to have the component holder automatically feed the components once they're inserted into the channels? I guess minimalist approach would be push feeding on every component pick up and horizontal and vertical components done in two stages with the requirement to turn your board 90degrees in between.

Is this picture close to the general idea (minus elevation)? It looks like it could be eventually be scaled up to 10 inches of parallel component lanes and then even multi stacked with each level set back a bit.. kinda like cinema seats (I'm dreaming of a full range of 0603 / 0805 components at the ready )

Just thinking any elevation is going to require another axis to pick the component up perpendicular to the elevated plane so its adding complexity, but if it could be done reliably 'bulk loading' would be a big step above 'per placement' loading. Any easier ways than an extra axis + elevation?

That is pretty much what I was thinking of, but with vibratory motion assist instead of relying on gravity alone (vibration breaks friction, part slides down.) Hopefully this can be done with a low enough tilt so that the tilt doesn't interfere with the placement tip picking up the part. If you can get the channels compact enough then you could have slots with components in each orientation to avoid having a spindle rotation mechanism to place components at 90 deg.

I like that plan, the elevation may well be low enough to not matter with vibration. If it proves us wrong I don't think its going too far to add a micro controlled motor and have it elevate, vibrate and de-elevate to near-flat off the same input. Would want to prove the core system works before bothering to go there though.

I'm a bit iffy on the 90deg components. I'd prefer to stick with slots that only allow the components to fit down using their shorter dimension because its easier and I wouldn't really want to be doubling up on components anyway.. turning a board 90 degrees would be easier to use in practice. I think I'll stick to rotating the board or go all the way with the extra axis.

Why not use an aligning guide to orientate and set the components to a datum? The guide would simply be a right angle that the components are placed against. Micro would pick up the component, place it against the guide, then pick the component up again at an offset from the guide. To rotate a component, a small lug could be fixed to one side of the guide. Micro would simply push the component against the lug and it would rotate 90°, place it against the two sides of the guide again and it’s done. You could also use the guide set in a tray full of flux.

DJ

Interesting idea with the rotation platform. You would reduce the number of parts trays needed but increase the placement time. Honestly, laying out boards with components all in one direction isn't that big of a deal, we do it all the time here at work for no good reason at all.

I have watched a few "How it's Made" episodes and many older manufacturing processes use a simple pick and place system just like your discussing. Why make it more complicated. The more complicated, the more things that can go wrong leading to even more complicated solutions.

The major difference seems to be that these machines are optimized for speed with each device doing one thing and doing it incredibly well (with or without CNC). If modified to do pick and place the speed of the micRo as currently designed would be painfully slow and overdesigned for strength.

If you come up with a good pick and place tool it will only be a matter of time before people are crying for a new gantry solution to mount it on.

Unless you're trying to layout anything involving RF where line lengths and impedence matching are critical...

answerguru, you do have a good point, though I don't know how many of the people interested in this basic paired down pick and place will be into DIY RF.

JPD, a simple modification of changing the leadscrew pitch could speed up the gantry by quite a bit. Speed and resolution are related though. I do not know what the step angle is for the selected unipolar stepper motors that BC is using, nor do I know the pitch of the leadscrew. I do think you are correct in assuming that BC optimized the design for strength and resolution rather than for raw speed.

The motors in the motor kit are 200 steps per revolution, or 1.8 degrees per step. I am pretty sure the ACME screws are set to 10 turns per inch. Assuming my 10 turns guess is correct, BC is currently running the motors at 1 KHz, or 1000 pulses per second.

The limit on how fast they can be pulsed is based on a few things:

1. The internal inductance and resistance of the motor.
2. The voltage applied across the motor windings (higher voltage yields faster charge times, but more complex controls, i.e. PWM, which I believe comes with the micRo)
3. The power performance curves of the motors (from the spec sheet).

For a motor to turn one complete step at full torque, the inductance, the resistance, and the voltage come into play. I don't remember the formula off hand, but you can calculate the charge time of the motor coil based on this. That will give you the ideal time.

You can push the ideal time to be even shorter, but then the motor windings won't get fully charged. The less they are charged, the less torque they have. Since they are rated at 64oz/in, pulsing them faster than the ideal time will lower that number to say 40 oz/in (you'd have to get the manufacturers spec sheet to see the power curves). If you are moving a sharpie, you can probably drive it this hard and not miss any steps. Using a robotic arm may be more resistant to movement and there will be a point where the lower torque values cause you to miss a step, which is dire.

You'll just have to experiment. But that's the fun part of DIY!


Edit: Ok, I found the formula: The time constant for charging the motor coil is L/R (Motor coil inductance divided by motor coil resistance). How could I forget that!!! Anyway, this value is assuming that you are supplying the exact voltage that the motor is rated for. Applying a higher voltage will decrease the charge time, as will adding extra resistance in series with the motor coil (but with a PWM controller, you won't want to add any extra resistance).

Possibly the only contribution I'll offer to this thread: the micRo screws are 12 TPI

I would never rely on gravity, I'd just index the parts flat, and since the part size is known, so is the position of the next part in-line. Now, I haven't built a board since the late 80's, so I don't want to make any big proclamations...but I've studied boards a bit in wonderment of the changes that map our accelerated progress. I'm familiar with the size and pitch of SMT, and I believe intuitively that the pain is not in placing the piece at its general location, but rather ensuring its position and of course doing so while soldering. These two factors seem to be crux of the deal. I need more information though....like, what is the average parts count on a "decent" DIY smt cb? How many "types" of parts are required? An example board would be something like a MC'd multi-axis h-bridge driver.


Keep running with this Dazz, flesh it out.


To believe that screw pitch determines speed is a mistake. Speed increases by pitch by only one factor out of about six that I can probably name. An example; motor X can achieve a reversal with a 12 TPI screw in 200ms with load L, yet motor X will stall trying a reversal at 400ms on a 6TPI screw. 6TPI is twice as fast and 400ms is twice as long, so what gives? Even if motor X could reverse in twice the time at twice the speed, if your job consisted primarily of reversals, it would take longer due to the reversal latency at higher moment.

A more primitive example: You can go twice as fast with your bike in gear H over gear L, but when a hill arrives, gear H is toast. Darn, you built your bike with only gear H so that you could go fast, so you are walking that bike!

As someone who DOES have a lot of interest in building boards, since it's something that I do.

I'd rather have a tray with components laid out on them. Relying on gravity is something that I simply would not do. If I wanted to feed in components, there are several other systems that can be used. One that comes to mind as could be relied upon would be something like the conveyor belts at the grocery stores. A simple control circuit with a photo sensor to tell when the part is in place and stop the belt. Maybe send an alarm to the program doing the pick 'n place if a belt is empty if you want to get fancy.

A lot of the components that I deal with are polarit sensitive as well. Diodes, transistors and such. ICs have to be oriented so that the machine knows where pin 1 is. Loading a belt with these is more reliable than a gravity fed tray.

I am thinking though that placing the components is less of an issue for me than the ability to machine the PCB and perhaps print on a solder mask.

It all changes with how much soldering experience you have really, but assuming people are far from incompetent or master solderers I'd say major time consumers are:
- Positioning the very small pitch IC's and ensuring they have reliable solder connections, particularly legless package versions that more and more components are only available in these days.
- General sorting/handling/positioning of standard size components.. mostly but not only resistors and caps (eg. 0603 size components). I might just add that 0603 is a desired size in my opinion, 0805 is probably a bit better for the less experienced or aging eyes but I find 0603 to be a good balance between trying to shrink designs and trying to stay with components that can be handled without visual aid. Smaller than 0603 and I find my head gets stuck to a microscope most of the day.

The thing is a largish board (by hobby standards at least) may have say up to 50 different components, but 40 of them would fit into 1 or 2 different standard package sizes. Of the remaining 10 components maybe only 1 or 2 would be high pin count small pitch components that take up most of the time. If micRo could even only do the standard size resistors/caps in a more convenient way it would be a huge start. Alternatively if it could manage to support some semi-common difficult packages, say if it could handle a few different pin counts of each of QFP (quad flat pack), TQFP (thin quad flat pack), and QFN I'd be impressed. The issue with these small pitch components is the solder past dispensing is going to be as critical/difficult as the alignment itself.

Example of part of a pcb to show what I mean, its mostly 0603 components, a couple of time consuming legless QFN IC's, and 1 legless 4 pin crystal that I wouldn't bother supporting by default with micRo. The part marked 32L is actually 24 pin so ignore that

Anyway what I was getting to is if your priority is to demonstrate the ability to pick up and position rather than produce a flexible system for time saving production then the small pitch high pin count components are what you should be aiming toward if you still want it to have some practical use. For those components hand loading one component at a time into a pre-cut slot or something would still be very useful.

In contrast, for 0603 type components I think if I have to sort them one at a time I might as well position them onto solder paste by hand while I'm at it, assuming I go with a pre-applied solder paste setup, which is what I'm aiming for. I don't like the video of dipping components in paste myself.

I'm not for dipping either. I hear you about placing 0603 by hand; if it's a two lead resistor it just makes sense. Placing 30 pieces like that would probably take longer with the bot, and just a few minutes by hand.

Just been thinking about the solder paste dispenser. The dispensing head will be relatively simple, light and naturally doesn't need to do anything that adds resistance. The average PCB I deal with is a fraction of the micRo's working area.. I'd say the vast majority of PCB's I deal with have a smaller dimension of <3". So I'm thinking of sacrificing per-board working area for a triple headed dispenser to do 3 boards at once. Is there anything I've overlooked that would make such a thing difficult because it sounds pretty straight forward and would look bloody neat in action.. or at least I think so.

Obviously I'd still keep a single headed version in case of big boards.

It seems as though the goals in this project are changing and I would like to take a step back and think about how realistic this might be. We initially wanted to do simple two contact, standard size, predominantly passive component placement without vision assist systems. Now the conversation is moving in entirely the other direction towards fine pitch, high pin count parts but still without vision systems. Every automated pick and place machine that we have used at work to attempt to do precision placement of high pin count parts has had very high failure rates when it did not have vision assist. I think it would be incredibly unreasonable to think that this would be an easy goal for us to achieve when so many commercial offerings have failed in exactly the same task with arguably more research resources at their disposal. A more modest and definitely doable project would be the placement of the infinitely simpler two contact miniature passive parts. Just because the device would take longer to place automatically than by doing it yourself does not mean that it does not save you time. The simple fact that it is doing it for you and freeing up your time is what saves you time. Delegation to a machine beats straining your eyes more that you have to. You are going to have to do through hole and custom pitch parts yourself anyway, so save the effort.

I am sceptical about the cost benefit of a pick and place tool for something like the micro, but we don't always do things for money. As a dedicated DIYers we often do things because we like messing arround with hardware. It's fun, challenging and gives us a chance to learn something new.

Let's see someone build something. It doesn't have to be perfect. Over time others will probably improve on the design. Thats how the DIY projectors got started and thats how these home grown CNC machines are developing.

It's time to revive this thread. What would be the "ultimate" PCB manufacturing workstation? What functions do we want?

I would love to be able to mill 5 thou traces and spaces, but I'm happy with a probably more reasonable 8 or 9 thou. I want to be able to do some fairly dense QFP parts.

Double sided boards are a biggie. One of the first things I want to do is figure out a good way to clamp the board accurately enough repeatedly that I can engrave one side, flip it over and engrave the other and have my vias line up.

I want a good way to do vias. If that means we source some plated rings that I can clamp through a hole size we decide on that could work for me. I'd just feel silly filling all the holes on my board with solder and hopefully not missing one or filling one I need to put a part through later.

Dream world I would love for an automated way to do soldermask. Silkscreen would be another cool thing to do. Could we just mount a special white pen for that?

I suppose after all that taking it to the next step would be pick and place, but I would be happy enough if I could make some really nice looking PCBs and hand soldering the parts.

Using the dep printer I've been working on (pneumatic, computer controlled syringes), you ought to be able to place solder paste wherever you'd require. The capillaries are available down to a few thou, and we could always make our own for even tighter tolerance. This would negate the need for solder masks, yes?

For a lot of the reasons you would normally want a solder mask, yes. I'll definitely look into setting up something like what you suggest eventually (doing solder paste without a stencil by hand is a pain).

Another reason that I would want soldermask is that I know I'm going to be a terrible person and not be as safe about where I place my boards or what is near my boards and having a soldermask makes shorting out traces just a little more difficult for me to do. Also in the case that I give someone a bare board to put together themselves, the micRo solder paste method probably wouldn't be available to them.

To be honest though, I don't even know what a soldermask is made of... yet.

Have you seen this video Brain? I really like the dispenser. Will the micRo be faster than that Taig he's using?

http://www.youtube.com/watch?v=HdqVt0jCBHk

-John

I'd say the traverse between ~0:22 and ~0:27 takes about 4.65 seconds and I'm guessing it moves about 1.5" on the X and about 0.75" on the Y (1.67" diagonally). If my math is right that is roughly 22 IPM.

My decision has been to forget trying to mill 'professional quality' PCB's.. I'm going to go for the basics of no solder mask.. no silkscreen.. no plated through holes... and aim for as low track/clearance widths as proves possible, and work toward double layer. I'm happy to just buy my professional PCB's... and just use micRo to mill out some basic prototypes... break out boards for IC samples mostly. If a good micRo solution ends up being on offer.. I may bite at it though.

I'm going to concentrate on solder paste / Pick'n'Place. I've bought a cheapo IC component vacuum off ebay (~$50) and have a solder paste dispenser at work.. also an ebay item (~$200 i think).. which I will "borrow". I need to modify the electronics of the vacuum to give it on/off control.. I ripped it open a few days ago and it looks pretty basic. The solder paste dispenser has manual settable pressure time and a foot pad.. so it already has an on/off interface and I'll just set the manual pressure time fairly low and use multiple dispensing increments based on how much solder I want on a pad. Electronically it is relatively straight forward.. mechanically it is mostly already done for me except component reel holder, tool clamping.. and clamping the PCB's. Software to get it all happening with little manual work each time.. well that part is daunting

edit: Part of my reasoning is also that I've just started renting in share accommodation and well.. it complicates things a little.. despite the novelty I doubt they'll enjoy the spindle noise or the mess milling will create.

No, I hadn't seen that. Looks like it works well. If Hirudin is right, micRo will be about 8IPM faster.

Have you considered lego boards? Lego's are milled to exacting tolerances if you have a board size that is standard Im sure you could attach a clamp that fits on the Micro and lego register it. by attaching the board as a spoil seeing as you will not be making through cuts. Inexpensive accurate and repeatable for front and back board registration.http://www.kk.org/cooltools/archives/000675.php