Author Topic: New Pick and Place design ideas  (Read 56335 times)

0 Members and 1 Guest are viewing this topic.

Offline bootstrapTopic starter

  • Regular Contributor
  • *
  • Posts: 69
  • Country: us
New Pick and Place design ideas
« on: May 08, 2016, 08:55:50 am »
With all that experience, you should build one.

I am so, so tempted to do so, partly because I need one (lame excuse), but mostly because I am finding the various aspects of pick-and-place machine operation fascinating.  And as a few of my recent posts indicate, there appears to be enormous opportunities to create different configurations of pick-and-place machines to make cheaper precise [but probably not fast] units.

Frankly, if I had one or two talented partners on such a project, I'd probably buckle and go for it.  I wouldn't be surprised if we could build a new pick-and-place machine from scratch for roughly the same I will probably spend on the complete production line (stencil-printer, pick-and-place, and reflow-oven).  As with most projects, the problem is TIME.


REPLY #1 of #2 (hopefully).

Do you need a P&P machine to assemble PCB's in-house or an all new business? I would have a ton of fun with the challenge of designing and building a P&P machine. I have a lot of experience designing and building very high precision mechanics for imaging and motion control. Now I design electronics and write embedded software most of the time. It would still take a mountain of time and money to do it with no guarantees. It would literally be an all new career to go down that path with the remarkably lofty goal of making MyData performance for low cost.

Seriously. Skip the N4 based on what you seem to require. You will be looking for a tall bridge to jump off of if you really need very fine pitch assembly as a reliable process. I don't have an N4, but I do have a Quad 4000C. The design of my machine goes back to the 80's and went out of production in the late 90's. PPM purchased the remnants of the company and started an update program for the software. Without getting too distracted with details - they definitely improved the machine and it is capable of placing 1005's. I am not suggesting you buy the same machine, but something like it. It is ridiculous (in my humble opinion) to consider one of the lowest cost/quality P&P machines on the market and expect it to nail PCB's with 0201 and .3mm without constant fiddling and re-work. You really can't be surprised that the moco subsystem is open loop steppers either - there is no money to put them in there. The PPM machine is steppers with linear encoders and a side scanner for alignment on the fly. Up vision for big parts and BGA. It is not fast, but is precise. With all that said, the small passive parts require the super precision feeders  - there is no way around that. The placement accuracy of the machine is meaningless if it struggles to pickup the parts in the first place. If the nozzle picks up on the edge of the component, it will flip or be at an angle causing a pick failure. You can only solve that by have very precise feeders that put the part in the exact same location every time and it has to be gentle about it. If the feeder is bumpy, the parts will jump out of the pocket before the head even arrives. From what I have read so far in this thread, the biggest weakness of the N4 is the feeders. For the price it seems pretty good. The Quad precision 8mm feeders are about $800ea which is still rather cheap when compared to the high-end machines - but they can index 2mm very precisely and smoothly.

I would not try to satisfy your requirements for under 80k with feeders, printer, oven, training, vacuum/air, accessories, etc. . If you are industrious like I was, you can get a broken machine and fix it up. That allowed me to get a 1005 capable machine with about 60 feeders for under $10k in cash - but it took nearly 10 months of nights and weekends to get it up and running. To buy it from PPM with with all the feeders and other parts would have been about $50-60k and would be ready to go on the first day. I also looked at DDM Novastar and liked them in general. They are targeted towards low volume and only the bigger machines can deal with 0201 (so it seems). What I did not like is that they cannot hold very many parts at a time. The Quad I have now can handle a TON of parts so I can have 6 designs ready to roll (in the machine) at any time and still be able to do prototypes without tearing the machine apart. That is super critical in my case where I do small batches of a bunch of different designs and don't want to have to setup each one every time I need a batch.

How much is your time worth and what business are you in?



I discussed some of my background in message #1002 above.  The short version is, I've been a self-employed scientist, engineer, inventor, product developer since before I finished school.  Once in a while over the years I've also taken project-specific contracts from places like NASA, AirForceResearchLabs and astronomical observatories.  My background seems somewhat similar to yours.

Like mrpackethead, pick-and-place intrigues me.  I know I shouldn't spend the time that will be required for a pick-and-place project, but then again, I almost always do projects that interest me somehow and avoid money generator projects.  That's one of my weak points (from some perspectives).

A year or two ago I designed a manual pick-and-place machine capable of handling 0201s and 0.50mm pitch components, which would have cost only $500 or so to build (as memory serves).  But I decided not to carry forward and finish that project... though I can't remember why.  Almost certainly it was to do other work that was more important (to me and financially).  It was a clever design, which is obviously necessary to do what needed to be done at such a low cost.

Part of the approach of that design also involved cameras, which I think may be the key to designing a modest cost pick-and-place machine capable of 0201 or smaller and 0.30mm to 0.50mm pitch BGA/QFN/etc.

Years ago I used to do lots of "hardware projects" (meaning "electronics + mechanics").  But I've been busy with software projects for a few years, and in that time pretty much all non-trivial components that interest me have become SMT only... or the SMT packages are much better (smaller, etc).

So the answer to your question is this.  I want to make several hardware projects again, but can't assemble them by hand like I used to with through-hole components.  I was going to have the prototypes for my first device assembled by an assembly house, but when I finished the PCB layouts, purchased the components, and went to give the assembly house the job they had quoted a few months before... they demanded somewhere between 5 and 10 times more per PCB for assembly.  That royally pissed me off, and made me decide that if I was going to make several hardware projects, I should be able to do assembly myself.  That's what got me to investigate SMT technologies, equipment and devices, including pick-and-place.  So my answer is, the equipment is for assembling my prototypes over the next few years.

As per my usual practice, I wouldn't do a project like this unless we came up with a "design breakthrough" of some kind.  And I'm not interested in making something for the hobby market (in the sense it must handle 0201 and 0.5mm pitch to interest me).  But a "design breakthrough" doesn't mean something totally new (like the solder-paste jet printer from MyData seems to have been)... it could just mean a new approach that radically cuts the cost of a precision machine (but can be slower than existing commercial machines).  My target would be dozens or hundreds per year of real and wannabe independent electronics engineers (and startup/micro-companies) who no way can [afford to buy equipment to] assemble their own PCBs today (assuming they aren't just making trivial PCBs with large components).

In other words, no way do I want to even attempt to compete with MyData!!!  No way, no how, not even close.  And at the cost/price point I suspect we're both thinking of, I'm quite confident MyData won't want to compete with us either.  For years I could make very sophisticated electronics devices (including complete computer systems including CPUs that I designed myself (including instruction set) and built with MSI and SSI).  Maybe the new generation of young electronics engineers think "no way can I do those kinds of projects"... OR... "no way can I adopt the latest and greatest components".  My purpose would be to go "back to the future" where pretty much anyone could develop their own state-of-the-art devices... like I did years and decades ago.

Though I don't claim we need to stick to this notion, my general feeling is the target price is $5K to $10K but able to support just about any modern component.  I'd like to even go one step smaller than 0201, but for now that's optional.

Like all my projects, whether I happened to do it alone or in collaboration with one or more smart, experienced folks like you, I'd put off the final decision to "go for it" or not until we have a final design.  If it isn't a breakthrough in some way or other, then I definitely won't invest the time and money to build prototypes.  You seem like a realist, so you probably think the same way.

Incidentally, what I would like to do is create something that competes with MyData in terms of "what size and range of components it can reliably place", but DEFINITELY NOT in terms of speed (probably by a factor of 5x to 20x).  Well, unless we make a much bigger breakthrough than I envision, which I seriously doubt (but isn't 100% impossible).

Like you said, the current designs for pick-and-place machines were developed when there was no such thing as "very cheap cameras".  In fact, it appears to me like pick-and-place systems with cameras are essentially earlier configurations updated with cameras.  Which is a very valuable improvement, but didn't change the mechanical configuration or the essentials of how the whole device works.  Which is why I see a HUGE opening for new approaches (for anyone designing with a totally clean slate, not just DIY replicating what already exists).  And like I said, either we come up with something novel and legitimately capable of calling "breakthrough" or we give up.

... continued in next post ...
« Last Edit: May 09, 2016, 07:14:19 am by bootstrap »
 

Offline bootstrapTopic starter

  • Regular Contributor
  • *
  • Posts: 69
  • Country: us
Re: Pick and Place Discussion
« Reply #1 on: May 08, 2016, 08:58:21 am »


... continued from above ...

I definitely prefer DC servos for many reasons, but as long as the system is closed-loop somehow, steppers are an alternative.  Note that in the past I developed devices that had high-count encoders that I made by copying encoder disks onto super-high-resolution microfilm and building the encoders into the mechanics itself.  So I got around the expense of high-count encoders that way.  Not sure the tradeoffs are the same today, but perhaps.

However, my hope is, we can leverage relatively inexpensive cameras into replacing expensive linear encoders (or rotary encoders for that matter).  I did that on my mechanical version a couple years ago, and I'm fairly confident we can figure out ways to do the same (or analogous) with a device we design from scratch.  This WILL require we add new code path to OpenPnP if we adopt that (which I assume we would, though not necessary), but that's not such a big deal.  Incidentally, another fellow here and I already mentioned introductory ideas for how to replace expensive encoders and/or mechanics by means of [somewhat] novel application of cameras in the design.  Maybe that "other fellow" was you (not sure).

Let me just point out one opportunity that addresses the exact problem you mention in your message... the pick-up nozzle/nozzles doesn't come down precisely in the middle of the component.  Well, that can easily be solved with a down-looking camera plus a slight change in technique.  How so?

If the "downer camera" (forgive the invented term) attached to the moving nozzle/nozzles assembly is at a fixed and exactly known x,y distance from the pick-up nozzle/nozzles, (and mounted close to the pick-up nozzle/nozzles), then the machine can move the camera over the component first, find or note the exact position [and rotation] of the center of the component still in the tape (and for conceptual and visualization purposes, then move the exact center of the field directly over the center of the component), then the exact center of the nozzle can be moved over the exact center of the component simply by moving the nozzle assembly the exact x,y distance/distances we know it is from the nozzle/nozzles.  Problem solved... at the expense of speed, because we need to perform this extra step for every [small] component the machine needs to place.

But... what about precision of x,y motion that we will need?  Well, one answer is to do what I did before... create super-cheap 1~2 micron linear encoder scales on microfilm.  But, there is another way that also takes advantage of the characteristics of cameras.  For visual and conceptual purposes, assume the camera has no "field distortion" (meaning points on the PCB imaged 1000 and 2000 pixels from the exact center of the field have a linear relationship).  In other words, that point on the PCB that is imaged 2000 pixels from the exact center of the field is precisely twice as far away from the point on the PCB at field center as the point on the PCB that is imaged 1000 pixels from the exact center.  Hopefully you intuitively know what "field distortion" means so you don't need to untangle my extremely clumsy description!

BTW, this does not need to be true (we only need to know what the field distortion is for the camera and lens), but for visualization and conceptualization let's make this assumption.

My claim is, this can replace linear encoders with no loss of x,y position precision.  How so?  Well, to visualize and conceptualize the situation before I cut to the chase, assume we DO have linear scales on the machine, but instead of being inside mechanical housings along with LEDs, light sensors and quadrature masks, we just "glue" the scales to the x and y "rails" or shafts the nozzle assembly slides against as it moves in x and y.  Now imagine we have an "x-axis upper camera" and "y-axis upper camera" attached to the moving "nozzle/nozzles assembly" pointing at these linear scales that are fixed to the x and y "rails" or shafts.  If we slide the "nozzle/nozzles assembly" back and forth on the x and y "rails" or shafts and look at the images these cameras send to a display monitor, we will see the lines on the linear encoder scales move back and forth as the "nozzle/nozzles assembly" moves.  If we were truly demented (and the motion wasn't too fast for cameras), we could write software to count the scale lines that pass, and even perform quadrature encoding on appropriately separated pixels.

Now that's a rather stupid idea so far, because the LEDs, light sensors and quadrature masks will probably cost no more and be no more hassle than these two cameras.

BUT... notice this.  We don't need the linear scales at all !!!!!

How so?  Well, let's now assume we simply placed very narrow marks along those two x and y "rails" or shafts approximately (but not precisely) 4 inches apart.  There are much better ways to accomplish this than scratch or paint marks, but we'll ignore that practical issue.  Essentially what we've done is remove 3999 out of every 4000 marks on the scale, and leave one out of every 4000 marks alone.  Except unlike the linear encoder scales, the marks won't be exactly 4 inches apart, they'll be 4 inches plus or minus something like 0.0100" apart.  In other words, not at precise intervals (which means, easy and cheap to accomplish but not precisely spaced like the encoder scales).

So... what is this supposed to do for us?  Well, think about it.  Before one mark exits one end of our 4096 wide/high pixel image, the next mark appears on the other end of the image.  Now, let's say that even though the marks are not separated by precise distances, we nonetheless KNOW how far apart they are (measured after assembly and before we test the machine).  Since we now know the precise separation of those marks, and the 4096 pixels on the image have a fixed relationship to distance on the x and y "rails" or shafts, when we watch any mark move from camera pixel to camera pixel on the camera images on the monitor, we know precisely how far the "nozzle/nozzles assembly" has moved in x and y.  And since the next mark always appears in the camera images before the previous mark vanishes, we can ALWAYS keep track of the exact position of the "nozzle/nozzles assembly"!

In other words, once we know how many microns on the x and y "rails" or shafts corresponds to 4000 pixels on these two camera images, we can keep track of the exact position of the "nozzle/nozzles assembly" as it moves by reference to these camera images!

Essentially what we need to do is establish that relationship... how many microns on the x and y axes corresponds to 4000 pixels on the camera images.  We can do this in several ways, so I won't wear you out by mentioning the ways I've thought up so far.  Maybe you have even better ideas for this.

Do you see how that works?  We now have the precision of super-duper precise linear encoders for the price of two cheap cameras.  In other words, we have a fully closed-loop scheme.  While we might (in some versions) need a precise gig at the factory to establish the distances between the marks, nothing expensive needs to ship with each machine!

BTW, there are several tweaks on this principle, including the fact it might be cheaper, more lightweight and bulky, and easier mechanically if the sensors in those cameras are 1D sensors instead of 2D sensors.  In other words, they can be 4096 by 1 sensors instead of 4096 by 4096 sensors (or whatever).  The example I gave generates 0.001" AKA 25u (25 micron) resolution without quadrature tricks, and about 0.00025" AKA 6.25u resolution with quadrature and/or mark versus pixel estimation/interpolation.  Personally I'd prefer to shoot for 1u, 2u or 4u resolution (which requires the marks be closer together and/or more pixels on the sensor (perhaps 8192 or 16384 pixels if linear sensors).

The feeders are another area where we probably need to "think far outside the box" and invent some better scheme.  That may be made easier by the "downer camera" I described above, since the position of components no longer needs to be precisely located.  But hopefully we can do much better than that!  For sure we need some way to prevent feeders from increasing the cost of real systems very much.

I have at least one idea about that, but I'm sure there are many others.  Rather than wear you down now, we can discuss this idea in our next message if you want to continue with this brainstorming.  I'd like to support both desires in this topic.  You want to support boatloads of feeders without excessive cost, which we should support.  While other folks might be willing to change feeders after each part.  PS:  What I mean to suggest here is that the software places ALL instances of each component on the PCB, then beeps to tell the operator to install the next component reel.  Though actually there should probably be two reel positions instead of one, so it can move on to the next component reel while the operator changes the other position to load the subsequent component reel (a classic "pipeline" approach).

PS:  I don't like used equipment... too potentially problematic for me!  Plus, to place 0201 and smaller components, most non-current machines are not capable.

How much is my time worth?  Hahaha.  No idea (other than smart aleck responses).  I guess my answer is this.  I'm in a position to do this project if I want to (or just retire and live a frugal but comfortable life).  So the real answer may be annoying, but truly is "if it seems worthwhile" (in "coolness" and "potential" at the very least).  I am annoyed that so many aspects of the hardware development process have been taken over by huge corporations by means of "scurvy tricks".  This project would defeat one of those... assuming we develop a sufficiently novel, cost effective and capable device.

Your turn!

FYI, the first two PCBs have already are:  8-layers, 4-mil traces, 0201s, 0.50mm BGAs/QFNs.  One is 200mm square and the other 100mm square.  I figure they are fairly representative in character to others that will follow in the next 3 or 4 years (at least).
« Last Edit: May 09, 2016, 07:13:17 am by bootstrap »
 

Offline glenenglish

  • Frequent Contributor
  • **
  • Posts: 450
  • Country: au
  • RF engineer. AI6UM / VK1XX . Aviation pilot. MTBr
Re: Pick and Place Discussion
« Reply #2 on: May 08, 2016, 08:58:59 am »
This is not the right discussion for the NEODEN4 thread

this thread is for NEODEN4 owners, and aspiring owners to discuss specific associated items. This is not a PnP general discussion thread....
 

Offline mrpackethead

  • Super Contributor
  • ***
  • Posts: 2845
  • Country: nz
  • D Size Cell
Re: Pick and Place Discussion
« Reply #3 on: May 08, 2016, 08:59:25 am »
Can i suggest you start another thread so you are not hijacking this one abut the N4.
On a quest to find increasingly complicated ways to blink things
 

Offline bootstrapTopic starter

  • Regular Contributor
  • *
  • Posts: 69
  • Country: us
Re: Pick and Place Discussion
« Reply #4 on: May 08, 2016, 09:14:44 am »
Can i suggest you start another thread so you are not hijacking this one abut the N4.

Yes, I think that's appropriate.  Since I'm new here, and don't frequent forums very often, does anyone know what that involves (to bring along those existing posts that are relevant to this, perhaps without losing the author identities)?

What's an appropriate title?  "New PnP design ideas for an inexpensive, high precision, modest speed machine for developer prototypes and small batches"?

As an aside... who cares whether some people are hard-core skeptics?  Doesn't matter.  Of course, statistically speaking, they're usually right, especially in the context of forums.

PS:  Yeah, wouldn't want neoden getting any radically new ideas!
« Last Edit: May 09, 2016, 07:12:59 am by bootstrap »
 

Offline Koen

  • Frequent Contributor
  • **
  • Posts: 502
Re: Pick and Place Discussion
« Reply #5 on: May 08, 2016, 09:39:50 am »
This really is novel, brilliant and ground breaking. Any chance we'd get to see some of your previous work ?
 

Offline bootstrapTopic starter

  • Regular Contributor
  • *
  • Posts: 69
  • Country: us
Re: Pick and Place Discussion
« Reply #6 on: May 08, 2016, 09:47:02 am »
This really is novel, brilliant and ground breaking. Any chance we'd get to see some of your previous work ?

Hahaha.  You forgot "/sarcasm", right?  Note that spikee had what is essentially the same main idea independently, so really, being unconventional isn't so impossible.  If I do say so myself, I don't do projects unless I have at least a couple radically unique approaches (or the device/technology is totally unique).  But, I have to admit I really don't know whether you're being sarcastic or not, so forgive this comment if I fell into your trap.   :scared:

I also suspect a few creative people can come up with at least 2 or 3 more ideas (hopefully to solve feeder issues).

PS:  Maybe two or three of us with good ideas should offer to collaborate with neoden on the neoden6 model (based on these and subsequent ideas) in exchange for free units and a small royalty.
« Last Edit: May 09, 2016, 07:12:30 am by bootstrap »
 

Offline Koen

  • Frequent Contributor
  • **
  • Posts: 502
Re: Pick and Place Discussion
« Reply #7 on: May 08, 2016, 09:51:40 am »
No, I'm genuinely interested in the radically unique approaches of your previous projects.
 

Offline mrpackethead

  • Super Contributor
  • ***
  • Posts: 2845
  • Country: nz
  • D Size Cell
Re: Pick and Place Discussion
« Reply #8 on: May 08, 2016, 10:24:06 am »
Feeder problems well solved, but you don't need our help.
On a quest to find increasingly complicated ways to blink things
 

Offline mrpackethead

  • Super Contributor
  • ***
  • Posts: 2845
  • Country: nz
  • D Size Cell
Re: Pick and Place Discussion
« Reply #9 on: May 08, 2016, 10:26:30 am »
PS:  Maybe two or three of us with good ideas should offer to collaborate with neoden on the neoden6 model (based on these and subsequent ideas) in exchange for free units and a small royalty.
[/font]

Actually i'd like to Introduce you to Micheal Bruch, who is the co-owner of SmallSMT.  I think you go will hit it off fantastically.  Micheal goes by the name SmallSMT on this forum..   Hes into the same kind of things you are.
On a quest to find increasingly complicated ways to blink things
 

Offline bootstrapTopic starter

  • Regular Contributor
  • *
  • Posts: 69
  • Country: us
Re: Pick and Place Discussion
« Reply #10 on: May 08, 2016, 10:32:24 am »
Feeder problems well solved, but you don't need our help.

Given I've never worked with pick-and-place (or stencil-printers, feeders, reflow-ovens), I wouldn't be so sure of that.  It often helps for people to explain the "problems" they have, or present their "wish lists" (even if as simple as "feeders are too damn expensive").  Though I have substantial experience with robotics and precision mechanical and optical systems, I haven't faced the practical frustrations of working with SMT equipment.
« Last Edit: May 08, 2016, 10:44:11 am by bootstrap »
 

Offline bootstrapTopic starter

  • Regular Contributor
  • *
  • Posts: 69
  • Country: us
Re: Pick and Place Discussion
« Reply #11 on: May 08, 2016, 10:41:42 am »
No, I'm genuinely interested in the radically unique approaches of your previous projects.

Previous projects... but not this one?  See, I'm still not sure whether you're being sarcastic or not.  Most of my projects have been in the fields of computer architecture, software techniques, optical systems (usually related to telescopes or instrumentation somehow) and advanced (smarter than human) inorganic consciousness.  I don't mind showing previous projects except... I've always been a very private hermit type, and don't like attention.

Here is one technology that may or may not mean anything to you.  NASA wanted to determine the 3D position of GEO satellites to much higher precision (with optical instruments, not radar and such).  They also had what they considered an "impossible wet dream" that they could determine the size, shape and orientation of those satellites from observations with ground telescopes.  Due mostly to atmospheric turbulence (and other factors), this was totally impossible, even with the biggest telescopes on the planet.  I invented a technique that let them determine the 3D position as well as the size, shape and orientation of satellites in GEO orbit to a precision of 50mm ~ 100mm with tiny telescopes (on the order of 200mm aperture).  Their PhDs swore this was impossible, but that's because they tend to think in conventional terms.  And indeed (being an optics expert myself), they were correct that conventional imaging could never achieve this.  Where I'm different is, I'm pretty good at thinking in very fundamental terms AND very comfortable searching for totally different (but usually simple) approaches that somehow "avoid the impossibilities".

Here is another totally trivial "technique" that's much too simple to call a "technology".  How can a cheapo microcontroller keep track of position from optical encoders when the microcontroller is dozens or hundreds of times too slow to read the encoder signals (A, B, I channels) before dozens or hundreds of encoder increments pass by (when the device they are attached to is moving fast)?  To be clear, the A, B, I signals are simply fed into inputs bits of general purpose I/O pins and not into counters or any other kind of internal logic, so the technique is 100% pure software.  The technique works entirely by the microcontroller reading the 8-bit I/O port that contains the A, B, I signals from two encoders per 8-bit I/O port (or up to 4 encoders if no I (index) channel is needed).

This is one of those "oh, man... this is simpler than drinking water" techniques... once you "see it".  The trick, of course, is thinking up these crazy techniques when nobody has before (I assume).  Can you figure this out?  I'll bet someone here does before I post the answer.  It really is simple, but does require a bit of "outside the box" thinking or perspective.

Another is a faster and superior memory management approach for CPUs that AMD almost adopted long, long ago in a galaxy not so far away.  It speeds up and simplifies simultaneous multi-task and/or multi-thread processing while greatly reducing the extent and complexity of circuitry on the CPU.  Can you guess the key idea in this one?  Very simple.

Frankly, most of my "great ideas" are fundamentally simple and/or straightforward once you comprehend them.  This is one reason I prefer not to study existing approaches and technologies [in detail if at all] before I try to solve known problems (or "impossibilities") on my own (without the biases and assumptions of conventional approaches).  Maybe I'm "simple minded" in some unusual way that works.  OTOH, I don't want to mislead... these advances are pretty much always within fields that I had much previous experience, so I did have context.  For example, I was designing my second or third "from scratch" CPU and was annoyed by various aspects of conventional memory management approaches when out of frustration I decided to look for "a better way"... and found one.

PS:  Now I'm getting myself into real trouble... for being even further off topic if not more.  Maybe someone can copy the appropriate messages to a new thread so we don't further pollute this neoden thread.
« Last Edit: May 09, 2016, 07:11:50 am by bootstrap »
 

Offline bootstrapTopic starter

  • Regular Contributor
  • *
  • Posts: 69
  • Country: us
Re: Pick and Place Discussion
« Reply #12 on: May 08, 2016, 11:00:51 pm »
PS:  Maybe two or three of us with good ideas should offer to collaborate with neoden on the neoden6 model (based on these and subsequent ideas) in exchange for free units and a small royalty.
[/font]

Actually i'd like to Introduce you to Micheal Bruch, who is the co-owner of SmallSMT.  I think you go will hit it off fantastically.  Micheal goes by the name SmallSMT on this forum..   Hes into the same kind of things you are.

Sure.  Perhaps you should invite him to comment on the ideas people have been presented so far.  That might be a good way to get started.

Or if this forum supports private messages, perhaps have him send me one.  But first, tell me/us what you're thinking?  Do you think a better approach than starting a new open-hardware project (of some sort) is to work with an existing [open or proprietary] manufacturer?  Or just involve him as the manufacturing part of the process?

In case it isn't obvious yet, I'm not proprietary with most of my ideas, at least not in this field.  This is not my focus in life, I just need this technology to work on my overarching "ultimate" project, which has a few subsystems, two or three of which involve subprojects (including that high-performance robotics vision system camera I mentioned before).  Plus, I strongly support open-source, open-hardware, open-core, open-technology, so I'm happy to contribute ideas, projects and technologies when I can.

Finally, if one or more of you out there is interested in exploring and developing these ideas further, we could schedule a skype conference call and do some real time brainstorming.
« Last Edit: May 09, 2016, 07:11:27 am by bootstrap »
 

Offline bootstrapTopic starter

  • Regular Contributor
  • *
  • Posts: 69
  • Country: us
Re: New Pick and Place design ideas
« Reply #13 on: May 09, 2016, 03:20:51 am »
Some recent messages in this topic, plus some straggler messages in the neoden4 topic these messages were extracted from, discussed ways that cheap cameras can perform some of the functionality of pick-and-place machines, and make the machines easier and/or cheaper to make.

I want to carry that idea further, and perhaps describe a technique close to the best approach.

To remind everyone what my goals are in this discussion, they are:

#1:  precision placement of small, medium and large parts with fine, medium, coarse pitch contacts.

#2:  inexpensive as possible, and hopefully something that can be built without exotic equipment.

#3:  can be slower than expensive commercial alternatives, but hopefully only 2 to 5 times slower.


What I describe here addresses the main pick-and-place machine, but only some aspects of the feeders.  The feeders are very important components of the pick-and-place machine in terms of cost and complexity, and will need further consideration later on.

-----

The basic idea is this:  cameras can look at various parts of the "workspace" and thereby replace some of the elements of conventional pick-and-place machines.

The "workspace" includes:

#1:  the PCB.
#2:  the nozzle tips.
#3:  components on the feeders.
#4:  components on the nozzles.
#5:  the x axis scale or reference marks (to measure x position).
#6:  the y axis scale or reference marks (to measure y position).
#7:  anything else that can help us increase functionality or speed.

Previous messages explained the following steps or facts in our provisional proposed pick-and-place process:

process A : pick up component from feeder (components in trays and other holders will be similar).

#A01:  The center pixel in the down-looking "nozzle camera" on the nozzle head is focused at a point on the PCB that is an exactly known x,y distance from where the exact center of the nozzle (or each nozzle if we choose to support 4+ nozzles) will touch the PCB when lowered.

#A02:  We have ways to establish the exact relationship of the "nozzle camera" image to the PCB surface.  In other words:  Assume the point on the PCB surface at the exact center of the image is 0,0 (offset in microns from point on PCB in center of image).  The software knows the precise x,y distance from 0,0 on the PCB that is imaged on each pixel of the nozzle camera image.

#A03:  We design the mechanics of the machine so the top of average small components in feeders is at the same height as the surface of the PCB.  The height of large components is not important because there is no possibility the pickup nozzle will come down off the edge of a large component (the machine can't be that imprecise).

#A04:  To pick up a small component from a feeder, first the "nozzle camera" moves over the expected location of the desired component and takes an exposure of that component.  If the component is offset in x or y, the camera moves slightly to center the component and takes another exposure to confirm (more likely software notes the x,y offset and factors that into the next step).

#A05:  The "nozzle head" then moves the known x,y distance between "nozzle camera" image center and nozzle tip center, which positions the center of the nozzle precisely over the middle of the component.

#A06:  The "nozzle head" lowers and picks up the component, which is exactly centered on the nozzle tip (no failed pickups).

The main purpose of the above process is to assure the nozzle does not come down slightly off center of tiny components, which can cause the component to tip over or the vacuum seal of the nozzle tip to the component to be insufficient to hold the part to the nozzle tip when the nozzle rises.  The precision we achieve by this process prevents failures to pick up tiny components, or failures to hold onto tiny components as the nozzle head accelerates and decelerates during subsequent steps.

Larger components can bypass this step because the precision of the machine is sufficient to assure the tip will come down near enough the center of the component that these kinds of failures cannot occur.

The component should now be picked up by the nozzle.

-----

Now let's continue with subsequent steps of the proposed pick-and-place process (some not addressed in previous messages).

process B : center component on nozzle then rotate to desired placement orientation.

We may or may not decide to implement this step depending on further consideration and tests.  We may also decide the perform this step on some but not all types of components (certain package types and/or sizes).

#B01:  The "align/center jaws" close on the component to center in x and align the component edges with the x,y axes.

#B02:  The "align/center jaws" open.

#B03:  The nozzle rotates 90 degrees.

#B04:  The "align/center jaws" close on the component to center in x and align the component edges with the x,y axes.

#B05:  The nozzle rotates to the angle the component must be placed on the PCB.

The component should now be exactly centered on the nozzle, and rotated to the appropriate angle.

-----

process C : determine exact position and angle of component on nozzle tip.

#C01:  move center of nozzle tip over center of up-looking "component camera".

#C02:  take exposure of component on nozzle tip with up-looking "component camera".

#C03:  if component not exactly centered or rotated properly, move and/or rotate nozzle to fix.

#C04:  take exposure of component on nozzle tip with up-looking "component camera".

#C05:  go to step #C03.

PS:  Note that the previous 3 steps are done based upon the contact pads or balls on the underside of the component for BGAs, QFNs and many other package types.  This may also be true for some or all discrete components, but some if not most discrete components could alternatively be centered based upon the outside of the package.

The center of the component is now directly over the center of the up-looking "component camera" and rotated to the orientation the component must be placed on the PCB.


-----

process D : center down-looking "nozzle camera" over exact point on PCB to place center of component on nozzle.

#D01:  move "nozzle head" in x,y to center down-looking "nozzle camera" over point on PCB where center of component on nozzle should be placed.

#D02:  take exposure with down-looking "nozzle camera".

#D03:  software finds and measures the positions of several "sync points" on the PCB near and surrounding the component to be placed.  The software knows the precise location of these pads and marks on the PCB relative to the point the component center must be placed from the gerber file and/or images of the PCB taken just before component placement process was started.  Therefore, the software can establish the exact point on the PCB image where the center of the component must be placed by offsets of where these "sync points" are in the image relative to where they should be when the component is precisely placed.  The software can also determine how much adjustment in component rotation is required to compensate for any [unexpected] rotation of the PCB axes relative to the machine axes (which should already have been known and compensated for, but can be double-checked here).

#D04:  move "nozzle head" slightly in x,y and rotate nozzle slightly in a to compensate for the offset and rotation errors found in the previous step.

#D05:  go to #D02.

Now the center of the image of the "nozzle camera" is precisely over the point on the PCB where the component center must be placed, and the component is rotated at precisely the desired orientation.

Note: we did something unconventional but very desirable above.  Other pick-and-place machines base their positions entirely on the basis of moving the "nozzle head" in x and y (which is done on the basis of stepper-motor steps or glass-scale increments).  True, they are based upon the original x,y positions of the two fiducial marks established ONCE before the PCB placement process began, but after those fiducial marks are observed, all subsequent positions are assumptions based upon presumed precise motions of the "nozzle head" along the x,y axes.

In our case, we determined the position the component center belongs by inspecting nearby features on the PCB surface immediately before the component is placed.  Any motion, rotation or bending of the PCB in its holder since the fiducial marks were scanned will not cause errors in our placement.  Also, any "missed steps" in stepper motors or "missed increments" in encoders will not cause errors in our placement.

In effect, by finding several key features near the component, we treat those key features as "local fiducial marks", and thereby determine placement position directly from the surface of the PCB.

Note that we cannot do this same process reliably by inspecting the pads of the component itself.  Why?  Because the solder paste may extend beyond those pads, and make their locations indeterminate by examination.  However, the pads around tiny through-holes and certain other features will not be covered by solder paste, so we can determine positions on the PCB from them (3 minimum).  We can even work off things like "pin 1 markers" as long as they are in the etched copper, not silkscreen.

-----

... continued in next message ...
« Last Edit: May 09, 2016, 07:11:03 am by bootstrap »
 

Offline bootstrapTopic starter

  • Regular Contributor
  • *
  • Posts: 69
  • Country: us
Re: New Pick and Place design ideas
« Reply #14 on: May 09, 2016, 03:22:09 am »

... continued from previous message ...

-----

process E : center component over PCB.

#E01:  Move the "nozzle head" the known x,y distance between the tip of the nozzle and down-looking "nozzle camera".  Note from previous steps above that we precisely know this distance.

Now the center of the component is precisely over the point on the PCB where the component center must be placed, and the component is rotated at precisely the desired orientation.

-----

process F : place component on PCB.

#F01:  Lower the nozzle until component touches PCB.

#F02:  Switch nozzle from vacuum to positive pressure.

#F03:  Raise the nozzle and keep positive pressure to "blow" component off the nozzle tip.  While the natural adhesion of the solder paste may be sufficient to hold the component down when the nozzle tip rises, we cannot count on that.  The exact optimum timing for when the positive pressure should be applied, and with how much pressure, is presumably something to be learned from experience and tests (and probably varies with component package, component mass, etc).

Now the component is placed.


-----

#####  COMMENTS  #####

The above description still assumes we can move the "nozzle head" precise distances along the x,y axis.  While the approach we describe above only needs to be precise over relatively short distances (perhaps 100mm or so), it still needs to be as precise as any highly precise pick-and-place machine over those distances.

Previous messages described how we can essentially create the equivalent of very cheap but ultra-precise linear encoders with two up-looking cameras (one "x-axis camera" and one "y-axis camera").  We do not repeat those descriptions here.

Instead, here is an alternative that completely eliminates the need for highly-precise linear encoders, or highly-precise lead-screws, or highly-precise measurement of any kind!

The following does assume we can measure x,y position with moderate precision, but this precision is easily and cheaply achieved with rotary encoders on the shafts of the motors that drive the "nozzle head" along the x and y axes.

To prepare our brains to understand the following approach, we should remind ourselves of an important aspect of the processes described above, and how they are different from conventional pick-and-place machines in one crucial way.

#1:  At every key step we actually LOOK at the components and PCB to determine position.  We LOOK down at the component in the feeder (or tray or holder) with the down-looking "nozzle camera".  We LOOK up at the component on the nozzle tip with the up-looking "component camera".  We LOOK down at the key features on the PCB with the down-looking "nozzle camera".  In other words, our process is LOOKING AT REALITY and making its decisions based upon the actual components and actual PCB (and that's all we care about... to position the components on the PCB).

#2:  We only need ultra-precise x,y motion twice:

2A:  The first time is when we move the "nozzle head" to switch from "nozzle camera directly over component center" to "nozzle tip directly over component center".  That distance is a constant and known x,y distance, but it must be precise (actually, only fairly precise during this step, precise enough to reliably pick up the smallest components).

2B:  The second time is when we move the "nozzle head" to switch from "nozzle camera directly over desired component center point on PCB" to "nozzle tip with component attached directly over component center".  That distance is the same constant and known x,y distance as step #1.  But this time, the motion must be ultra-precise, so when the nozzle is lowered, the center of the component comes down precisely where the center of the component belongs on the PCB.

-----

Now that we have that fresh in mind, we might notice something.  The only times we need ultra-precise motion is when we move the "nozzle head" back and forth by the constant and known distance between the center of the nozzle tip and the center of the down-looking "nozzle camera" image.

But wait!!!  That is a constant distance!!!  Which means we don't even need to detect the thousands of intermediate increments between those two positions, because we always move the same distance.  All we need to detect are two specific positions for the "nozzle head" (or 4 positions if we have 3 nozzles and one "nozzle camera").

Let's assume for the moment we only have one nozzle, and we design the machine so the required ultra-precise motion is only along the x-axis.  That means we don't even need precise position measurement on the y-axis!

One axis down... and one to go!  :-)

But what do we really need on the x-axis?  Since the distance we move is always the same, why can't we just slide the "nozzle head" back and forth against hard, physical "limit stops"?  Then we don't need to measure anything, because the physical device itself determines the position.

Holy smokes!  Now we don't need ANY precision position measurement!


We still need ultra-precise motion, but that is handled by the physical nature of the "nozzle head slide".  So we have now totally eliminated the need for ultra-precise position measurement devices!  Yet we can still place components with ultra-precise accuracy.  Can't beat that with a stick!

But our "reliability engineer" raises his hand and asks a question.  Is this mechanical scheme 100% repeatable?  Maybe a flake of dust will float between those hard limits and stop the slide several microns short of the desired position.  Then what?

Three possible solutions.

#1:  We test and learn the limits have been designed in a way that the motion remains precise and repeatable [unless someone operates the machine in an environment vastly too dirty for pick-and-place work].  We note in our instructions that "our machine is not appropriate for operating in sand storms or next to milling machines throwing off metal chips"... or whatever.

#2:  We place LEDs and photo-transistors behind 1~4 micron slits (openings) at the desired separation on this "nozzle head slide".  Then when the "nozzle head" is moved from one limit stop to the other, the device aligns the two 1~4 micron slits so the position is detected before it considers the "nozzle head" moved.  Essentially we just created a super-precise linear encoder with only 2 increments.

#3:  We place a piece of clear glass in front of the "nozzle camera" lens and tilted at a ~45 degree angle.  This clear glass will reflect somewhere between 0.1% and 4% depending on what kind of optical coatings the glass has, if any.  This will produce a "ghost image" of whatever is at right angles from the direction this down-looking "nozzle camera" is pointing (and at the same distance from the lens as the PCB surface).  We then have the software watch marks or back-lit LED slits on the fixed x axis rail to measure x-axis position.

These basic approaches can also be applied to 3+ nozzles (with thought).

#####  IMPORTANT ALTERNATIVE TO CONSIDER  #####

This "constant and known distance" and motion can also be done in a different but analogous manner... an alternative to moving/sliding the "nozzle head" back and forth a constant and known distance (the distance between nozzle tip and "nozzle camera" image center).  This alternative might be better for many nozzle configurations (if we choose to support multiple nozzles, which I tend to resist for the first "model" or implementation).

In essence, we replace that constant, known, fixed linear motion with rotation.  Imagine the nozzle (or nozzles) are inserted into holes in a 10mm thick by 125mm diameter round metal disk.  By some means (physical/optical limits, encoder, etc) our software can rotate that disk precisely 180 degrees in either direction (or 90-degree steps for 3 nozzles and 1 "nozzle camera").  Assume the camera and nozzle are inserted into 20mm diameter holes in that disk at 50mm from the disk center (and precisely 180 across from each other).

We can now replace those "slide the nozzle head back and forth" steps with "rotate the nozzle disk 180 degrees" steps.

This has some obvious advantages (especially for multiple nozzle configurations), but also some disadvantages (probably mostly physical issues, like how wires and air/vacuum tubes to the nozzles move around while placing components).


#####  NOTES  #####

PS:  To see what "align/center jaws" look like (mentioned in process B above), see:


(see 06:00 ~ 06:30 in the above video).

  ... and ...


(see 00:40~00:50 and 03:20~03:30 in the above video).


#####  FINAL REMARKS  #####

I probably forgot to mention one or more assumptions and insights that the above depends upon, but you guys can grill me about any apparent flaws in the above and make me explain why everything above works properly.

I am fairly sure I have described a device that can place components very precisely, but doesn't contain any expensive components, and doesn't have to be massive (unless we try to crank up the speed too far).

However, I also assume we can make improvements on this approach and design.


#####  KEY ADVANCE  #####

If this approach has a "key advance", it is that we LOOK at the components and PCB with cameras to determine position.  This gave us ways to make our pick-and-place machine very precise without expensive parts like linear encoders.

However, this approach also has other inherent advantages.  For example, it inherently compensates for changes in the working environment during assembly, including:

#1:  PCB moves, rotates or expands/contracts due to temperature changes.
#2:  Parts of our machine flex, bend, shift or expand/contract due to temperature changes.
#3:  Our down-looking "nozzle camera" can determine whether components are placed or not.
#4:  Our up-looking "component camera" can determine exact position and rotation of components.
#5:  We take advantage of marks on the PCB surface near each component to determine component location.
#6:  The nozzles can pick up super-tiny components (always lowers the nozzle tip down exactly on their centers).


####################################

Your turn everyone.  Dive in... the waters are warm and friendly.
« Last Edit: May 09, 2016, 07:10:43 am by bootstrap »
 

Offline mrpackethead

  • Super Contributor
  • ***
  • Posts: 2845
  • Country: nz
  • D Size Cell
Re: New Pick and Place design ideas
« Reply #15 on: May 09, 2016, 04:26:39 am »
Quote
"If this approach has a "key advance", it is that we LOOK at the components and PCB with cameras do determine position.  This gave us ways to make our pick-and-place machine very precise without expensive parts like linear encoders."

Lots of folks doing this with sub $1000 home build machines, made from sticks and bubblegum. ( almost litterally ).. So, not quite sure what key advance this is.

go to openpnp.org read all that, then go to the google grups and read all that..
On a quest to find increasingly complicated ways to blink things
 

Offline bootstrapTopic starter

  • Regular Contributor
  • *
  • Posts: 69
  • Country: us
Re: New Pick and Place design ideas
« Reply #16 on: May 09, 2016, 05:00:31 am »
Quote
"If this approach has a "key advance", it is that we LOOK at the components and PCB with cameras to determine position.  This gave us ways to make our pick-and-place machine very precise without expensive parts like linear encoders."

Lots of folks doing this with sub $1000 home build machines, made from sticks and bubblegum (almost literally).  So, not quite sure what key advance this is.

Go to openpnp.org read all that, then go to the google groups and read all that..

Rather than just make unsupported assertions, wave your words around and ask me to spend thousands of hours inspecting (and understanding) a huge project, why don't you explain to me how sticks and bubble-gum can precisely place components as small as 0201 discretes and 0.30mm~0.50mm BGAs/QFNs on a PCB.

Just a few short paragraphs should be sufficient.

THEN, please explain to me why so many expensive commercial pick-and-place machines (and apparently every DIY pick-and-place machine) cannot reliably support 0201 discretes or 0.30mm to 0.50mm pitch BGAs/QFNs since all that's required is sticks and bubble-gum.

I mean, if sticks and bubble-gum is sufficient, why do companies spend so damn much time, effort and money to create their machines?

For that matter, why don't engineers place 0201s and 0.30mm pitch BGAs on their prototype PCBs by hand like before?  Who even needs sticks and bubble-gum?

BTW, you already know from past reading of these messages that we are specifically talking about a pick-and-place machines that DOES handle 0201 and 0.30mm to 0.50mm pitch components.  So please don't pretend AGAIN that this is not a requirement and the context, not to mention the entire reason for this topic.

PS:  I already looked at www.openpnp.org last night and followed the hardware links.
« Last Edit: May 09, 2016, 07:10:22 am by bootstrap »
 

Offline mrpackethead

  • Super Contributor
  • ***
  • Posts: 2845
  • Country: nz
  • D Size Cell
Re: New Pick and Place design ideas
« Reply #17 on: May 09, 2016, 05:59:31 am »
Have you considered that your approach on this forum, might not be getting you the helpful responses that you might want.

If you took the time to read the posts, look at the pictures, you'll see there are people who have used low cost 3d printers, paperclips, rubberbands, Cheap web cams, bits of double sided tape, and all manner of other bits and peices to get remarkably accurate and useful results.  Theres people who have used the $350 Makerbot xy robot as a starting point..   These are everyone from students to hobbists, professional engineers, and people who also use Commerical machines, collectively i'm guessing with a lot of experience, and they have turned up at maker faires and the like so we know they are not just spouting this stuff off on the internet.

Vision processing is well down the track, and using things like opencv people are getting great results,  with both top and bottom cameras.
Theres some $40 nozzle adaptors you can buy, to connect to a $15 stepper..  and with a bit of microstepping magic, you can rotate those with the required accuracy to get 0201's down..     the small passives are not so hard..  Even a N4 can do it, with its open steppers and it less than nice feeders..   You jsut have to mess with it to do it well.     

Why dont' all the high end machiens support it.. Because in many cases they dont' have to..  There is'nt the high desnity pitch requirment..   And they are trying to hit targets in the 10's of thousands of parts p/hour..   Slow your machine down to 100's / hour and it gets much easier.

As a friendly suggestion, your coming across really aggresssive,  and that just never will work in these forums..    Many of the folks already have written you off as a phoney..  I'm not sure... but this is the internet so we just dont' ever know..


On a quest to find increasingly complicated ways to blink things
 

Offline Bud

  • Super Contributor
  • ***
  • Posts: 7084
  • Country: ca
Re: New Pick and Place design ideas
« Reply #18 on: May 09, 2016, 06:15:52 am »
font=Georgia  ...snip... /font

@bootstrap - can you not change font, your posts are hard to read
Facebook-free life and Rigol-free shack.
 

Offline bootstrapTopic starter

  • Regular Contributor
  • *
  • Posts: 69
  • Country: us
Re: New Pick and Place design ideas
« Reply #19 on: May 09, 2016, 07:08:32 am »
Reply to mrpackethead post above.

Have you considered that not everyone on this forum has the exact same interests or requirements as you?

Haven't you noticed all the people explaining how much difficulty they've had trying to place 0201 discretes and fine pitch components?  They TOTALLY CONTRADICT your claims that a pile of sticks, bubble-gum, paper-clips and rubber-bands can precisely place components.

So I guess not everyone is as brilliant and manually dexterous as you are, I suppose.  Certainly their machines aren't.

I don't know why you object to people having discussions about this topic.  OF COURSE not everyone will be interested in this topic, just like every other topic.  So what?  What's your gripe?  What's your problem?  If you want to place 0201 and 0.30mm pitch components with sticks and bubble-gum, that's okay with me (and I assume everyone else here too).  So why do you care how we do so?

I did microstepping decades ago.  Am I supposed to care about that?  The issue is NOT "how to move a small distance".  The issue is how to position precisely.  Just because a motor can turn a small distance does not mean the freaking component gets placed exactly where it belongs.  There are many pieces between the motor shaft and the component and PCB, and many assumptions and requirements too.  So you're just blowing smoke and making noise.

You say the neoden4 can place 0201s, and indeed they claim that.  However, some actual owners have said in this forum that it doesn't place them all reliably, which is a problem... though only if they want their PCBs to work.  And in case you have not noticed, the neoden4 isn't just a pile of sticks, bubblegum, rubber-bands and paperclips.

The expensive machines that cannot place tiny components do not fail because they are moving too fast.  How do I know?  Because they could slow down for problematic/tiny components if that was all they needed to place them precisely.

And companies aren't going to omit whole slews of components just so they can say they place 5%, 10% or even 20% faster.  It is more important that the machine can do the job than how fast it cannot do the job.

YOU are the one being aggressive.  And your "suggestion" is anything but friendly.  Calling it friendly doesn't make it friendly.  Furthermore, I don't give a damn whether 99.9% of the forum "writes me off".  I'm only interested in the 1, 2, 3 or 5 people who are interested in what I'm talking about (if they exist).  The others have other interests, which is fine with me, but apparently not with you.  The fact you don't know me or my abilities is irrelevant.  Anyone who cares about what I am discussing can judge from my posts whether I'm a total moron wannabe or not.  Sure, they can't know precisely, but they can make an estimate, just like I can make estimates about you from the content of your post.

BTW, I am not offended when people are more advanced than me in certain fields, or who have more sophisticated requirements than me, or etc.  That's totally natural.  I also don't look down on people less advanced than me or with less sophisticated requirements than me.  That's also totally natural.  But you appear to have a serious problem with me and my posts just because I need more precision than some or most people here, and want to discuss that with the 1, 2, 3, or 5 people who also need or care about that degree of precision (personally or at work).  My friendly suggestion is you focus on topics that match what your interests are.

Incidentally, I'm not sure what motive you might have for stopping a potentially worthwhile open-hardware project, one that I'd be spending tens of thousands of dollars of my own savings on (and who knows how many hours), and one that might benefit some folks in SMT land.  Maybe you can explain that.
« Last Edit: May 09, 2016, 11:23:13 am by bootstrap »
 

Offline bootstrapTopic starter

  • Regular Contributor
  • *
  • Posts: 69
  • Country: us
Re: New Pick and Place design ideas
« Reply #20 on: May 09, 2016, 07:15:19 am »
font=Georgia  ...snip... /font

@bootstrap - can you not change font, your posts are hard to read

All my messages in this topic are now changed to the "verdana" font.

Hope that helps.
 

Offline forrestc

  • Supporter
  • ****
  • Posts: 701
  • Country: us
Re: New Pick and Place design ideas
« Reply #21 on: May 09, 2016, 08:03:12 am »
I did microstepping decades ago.  Am I supposed to care about that?  The issue is NOT "how to move a small distance".  The issue is how to position precisely.  Just because a motor can turn a small distance does not mean the freaking component gets placed exactly where it belongs.  There are many pieces between the motor shaft and the component and PCB, and many assumptions and requirements too.  So you're just blowing smoke and making noise.

I think what you're missing is that he was talking about using opencv (aka machine vision) to place the parts accurately.   Once you have decent machine vision running, then you don't need all the expensive positioning feedback crap to make this work - but instead a decent vision system coupled with a less than stellar (aka cheap) positioning system.  And yes, that positioning system can be made out of whatever low-grade construction materials will work - and in this day and age it's not uncommon to see things like "a pile of sticks, bubble-gum, paper-clips and rubber-bands" as part of that.   I don't need an expensive stepper with a expensive belt and high quality position sensors.   I also don't need precision mechanics.  All I need is something which moves a camera and nozzle around to the general area of interest, provided the machine vision system can figure out where you are once you get there, and adjust accordingly.     So what if I've got nasty backlash, weird z axis warpage, etc?  Correct it with the feedback camera.

Of course, using high-end mechanical components will improve the accuracy of the placements - or I guess more accurately improve the accuracy of the initial machine placement before correction with machine vision. 
 

Offline Kjelt

  • Super Contributor
  • ***
  • Posts: 6570
  • Country: nl
Re: New Pick and Place design ideas
« Reply #22 on: May 09, 2016, 10:05:56 am »
Point of attention, how do you visually recognize where you are exactly at the pcb without using fiducials or other markers with each and every component, which is a no go?
A clean pcb looks totally different from a pcb that was paste stencilled. The paste could be off a bit also.
And then each paste stencilled pcb will start to look different each and every 5 minutes you look at it, because the paste is oozing after its placed and also when it changes temperature.
Ofcourse when actually placing the components on the pcb, the visual aspect of that pcb will obviously also change. All those things can be pretty tough on the software to visually determine its exact position.

I've always been a very private hermit type, and don't like attention.
You could have fooled me. I think the advice from mrpackethead was not meant negatively but a relevant suggestion to succesfully socially interact on this forum.
The way you are communicating at this point seems to me more fit for a blog (one way direction).
 

Offline bootstrapTopic starter

  • Regular Contributor
  • *
  • Posts: 69
  • Country: us
Re: New Pick and Place design ideas
« Reply #23 on: May 09, 2016, 10:15:42 am »
I did microstepping decades ago.  Am I supposed to care about that?  The issue is NOT "how to move a small distance".  The issue is how to position precisely.  Just because a motor can turn a small distance does not mean the freaking component gets placed exactly where it belongs.  There are many pieces between the motor shaft and the component and PCB, and many assumptions and requirements too.

I think what you're missing is that he was talking about using opencv (aka machine vision) to place the parts accurately.   Once you have decent machine vision running, then you don't need all the expensive positioning feedback crap to make this work - but instead a decent vision system coupled with a less than stellar (aka cheap) positioning system.  And yes, that positioning system can be made out of whatever low-grade construction materials will work - and in this day and age it's not uncommon to see things like "a pile of sticks, bubble-gum, paper-clips and rubber-bands" as part of that.   I don't need an expensive stepper with a expensive belt and high quality position sensors.   I also don't need precision mechanics.  All I need is something which moves a camera and nozzle around to the general area of interest, provided the machine vision system can figure out where you are once you get there, and adjust accordingly.     So what if I've got nasty backlash, weird z axis warpage, etc?  Correct it with the feedback camera.

Of course, using high-end mechanical components will improve the accuracy of the placements - or I guess more accurately improve the accuracy of the initial machine placement before correction with machine vision.

If you read my comments, then you should also know that I'm trying to avoid expensive mechanics in my design... and explicitly doing so with cameras AKA vision.  So I obviously have no gripe with that idea.  However, I haven't seen ANY of those who build DIY machines or ANY inexpensive pick-and-place machines that reliably place 0201 and 0.30mm~0.50mm pitch components.  That is, except neoden4 --- which is how I ended up in this forum and in the neoden4 topic.  While in the neoden4 topic at least one owner said the neoden4 places 0201 components, but not reliably (if I remember, 2 failures out of 100, which means many or most of my PCBs would be defective due to 0201 placement, even before including bad placement of fine-pitch components).

I wasn't aware openpnp requires steppers.  What I read about openpnp implied to me that the software was abstracted from the hardware to the extent possible so the software could run as wide a variety of machines as possible.  If openpnp doesn't require steppers, then I don't see how his comment about steppers has anything to do with openpnp.  Obviously steppers can drive a precise pick-and-place... IF the machine is closed-loop with precise sensors and/or the machine is awesomely precise.

You (and he) seem to imply that openpnp is magic software, that because it supports some kinds of "machine vision" means every pick-and-place machine it controls can place 0201 and 0.30mm pitch parts.  That is NOT what the owners of pick-and-place machines driven by openpnp claim, so I have to assume whatever "machine vision" support openpnp has, it is not sufficient to achieve that kind of placement precision on any and all wimpy, wobbly, imprecise, misaligned mechanics.  Again, that's what all the comments I've read imply.

If I'm wrong, and openpnp is "magic software" or "already incorporates and assumes the kinds of machine vision that I describe in my proposed scheme", someone should tell me so.  Nobody has explained how openpnp can achieve that precision, yet none of the actually built DIY machines do.

By the way, I don't recall reading anything on the openpnp webpages that I scanned that claims the approach openpnp takes and requires is inherently capable of placing 0201 and 0.30mm pitch components precisely, even on imprecise machines and backlash in the motion transmission scheme.  I didn't read every word, but I didn't see that claim anywhere, and I would have expected something like that right at the top in boldface!

How can the openpnp machine vision correct for all the machine defects you mention?  As far as I know, no configuration of machine vision cameras I've ever seen in a pick-and-place machine are sufficient to achieve that.

Which is the reason why I wrote my most recent double post to describe a specific configuration of machine vision to achieve the required precision with less than perfect mechanics.

One reason my brain is skeptical about these openpnp claims is because some of what I read about openpnp implies it tries to be general software that can be applied to pretty much any pick-and-place machine, especially those with conventional design.  Even conventional [commercial] designs that DO have machine vision (even 2-camera machine vision) are not able to place that precisely with crappy mechanics.  It was not my impression that openpnp is only designed for one specific design of pick-and-place machines.  Am I wrong?  I suppose I could have missed that statement somewhere.  Frankly, if openpnp is mainly/only for DIY machines, that would be a smart move.

Explain to me how the kind of machine vision you refer-to can assure the machine precisely places components.  Because I don't believe that, even when we play all the known tricks of the trade to compensate for backlash (always move the same x,y direction to anywhere important) or misaligned x,y axes and so forth.  I know at least most of these tricks/techniques well from a long time ago.

I assume you must have read the many, many, many, many, many comments about placement precision problems in the neoden4 topic and other pick-and-place topics.  Because I sure did!  Endless warnings about how much trouble those tiny components are, sometimes even on commercial machines that claim to fully support them with sufficient precision to be reliable.  And lots of very expensive machines WITH machine vision don't claim to support 0201 and 0.30mm~0.50mm pitch components.  Why would they not if it is so easy with appropriate software?  Is openpnp some kind of magic software or radical breakthrough or what?  I'm not trying to be a wise-ass with that question, either.  I ask because it sounds like that's what is being claimed.

Having designed, fabricated and worked with advanced diffraction limited optics for decades, I've worked with precision ranging from 0.0001" (mechanics) to 0.000001" (optical surfaces).  So I understand what these kinds of distances mean, and what is required to achieve them.  Even in devices without [fast] moving parts they're not trivial.  Here we're trying to work in the approximate range of 0.001" == 0.025mm == 25 microns or better, and we're doing so with all sorts of moving parts.  Now, if you read my post you see I'm a massive believer in solving the problem with "machine vision"... but only specific forms of machine vision that I've never seen on any pick-and-place I've seen in real machines.  And the "machine vision" approaches I have seen on pick-and-place machines are not sufficient to overcome the problems introduced by crappy mechanics.  I may be missing something, but I doubt it.  If I am missing something important, someone must explain what, because I don't see it, and I see all sorts of evidence to the contrary.

Please understand that I know people can place 0804 or 0603 and sometimes even 0402 components with conventional pick-and-place machines with conventional 1 or 2 camera vision systems.  Most that can even reach 0402 are very heavy machines with high-precision mechanics and rotary if not linear encoders, though some owners of these machines moan and groan at length on forums about reliability problems with the smallest supported components.

BTW, it really is a bit hilarious to me to see the current situation, which is:

#1:  Some people warn and complain about how many problems and difficulties exist with small components.

#2:  You two guys confidently claim "get openpnp and you can place any components with any precision you want with crappy mechanics made with sticks, bubblegum, paperclips and such".

I hope you can understand why that seems rather strange to me.

PS:  If you have an openpnp machine, I'll send you a pair of my PCBs from a few years ago complete with gerber files, BOM and the works.  Just place the 0201 caps (none of which are rotated), then take some close up photos and send them to me as proof.  Okay?  Oh, also send me detailed photos of your pick and place machine if they're placed correctly.  Unfortunately, they don't have very many 0201s... about 45 on the 150mm square PCB and 16 on the 70mm square PCB.

FYI, here are some photos of both sides of those PCBs so you have a general idea:

http://www.iceapps.com/ice_vision_pcb_6704.jpg
http://www.iceapps.com/ice_vision_pcb_6706.jpg
http://www.iceapps.com/ice_vision_pcb_6721.jpg
http://www.iceapps.com/ice_vision_bom_v0000a.pdf
« Last Edit: May 09, 2016, 11:39:09 am by bootstrap »
 

Offline glenenglish

  • Frequent Contributor
  • **
  • Posts: 450
  • Country: au
  • RF engineer. AI6UM / VK1XX . Aviation pilot. MTBr
Re: New Pick and Place design ideas
« Reply #24 on: May 09, 2016, 10:21:38 am »
Bootstrap if you don't change your attitude I doubt you will get any ideas across.

We're frankly not interested in your way of communicating.  Maybe we'll play some jokes on you, pretend to ask questions, annoy you a bit...

 
The following users thanked this post: mrpackethead


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf