New Pick and Place design ideas

[forrestc]

Using pads with paste is an unnecessary risk. Vias are not all that great either since they are crowded by solder mask and are mostly small. Traditional fiducial marks are super reliable and provide more than enough information for nearly all placements. In very critical areas, you can always put a pair of local fiducials to better ensure success.

They are so easy to put in the PCB design and so easy for the machine to see - why would you avoid them? I have used vias and other holes as fiducials on PCB's I did with only hand assembly in mind, but it was more accurate and reliable when I add fids.

When designing boards, I always include fiducials...  sized appropriately for the machine.    The machine I have will place fine pitch components just fine... for chips it's down to 0201's.  (I use 0402's on my boards with no problems other than an occasional feeder alignment issue after swapping a reel).   But, this is also a commercial-grade machine which I paid large for.  I can also place fiducials around critical components but so far it hasn't been necessary. 

The suggestion I was making was basically that if your intent is to use imaging to help with position accuracy, and fiducials weren't good enough for whatever reason, imaging a pasted board is not out of the question.  I agree it's higher risk than fiducials, but if the goal is to build a very low cost machine which may have questionable mechanical positioning accuracy to be corrected with vision, then this might be a reasonable solution.  Of course the devil is in the details, and I'd have to see it work before I was 100% convinced.  But if the choice is between imaging potentially covered vias on a board vs a pasted pad, I think I'd bet on imaging the pasted pad.   

[rx8pilot]

I can get that concept, but after spending some time ( 10 years) designing and manufacturing precision motion mechanics - it's just not all that hard or expensive to make a suitably precise motion platform for even the tiniest of parts. That, in general, should eliminate the need for any fancy footwork in the vision system.

[bootstrap]

I've never seen vias covered by soldermask.  My PCBs aren't that way.  Take a look at the images of two of my PCBs several messages back for examples (near bottom of page 1).

Every board I make has, by design, all of the vias covered with mask.  If you look at many professional boards they have them covered as well.  There are many good reasons to cover the vias, and few reasons not to.

However, if you know your placement machine relies on the vias then you can choose that at manufacturing time for the boards.

I also don't see why you can't/shouldn't use the footprint that you're placing the component on in your vision system.  Even with paste on it, it should provide enough contrast that you should be able to place using the pad.

I understand that a pick-and-place machine that requires exposed vias (or extra fiducials) may be considered less desirable by some (especially if they already have boatloads of bare PCBs in the back room they want to assemble).  That is a "negative" I am willing to accept if necessary.  Why?  Because the "target market" that I care about (and also think is an effective target market for a commercial venture) is real and wannabe electronics engineers who want to [design and] assemble non-trivial PCBs with the latest tiny and fine pitch components but can't afford $40K for a pick-and-place machine (complete with adequate number of feeders).

Note that I do not claim other target "markets" are not justified, but they don't interest me (unless they overlap the one I defined).  I am definitely not interested in competing with mydata or ddmnovastar or any of those existing medium-high end manufacturers (especially on the basis of speed, though we'll make ours as fast as possible while staying within our niche).

I don't think the advantages of covering vias is significant.  To everyone:  If I am wrong about that, please explain!

I very seriously doubt close and uncovered vias become shorted by solder more than once in a blue moon.  There shouldn't be solderpaste on the via.  When a via is super close to a component pad that does need soldermask, that could be an opportunity for smeared solderpaste and a short, but I believe the chances of that actually happening is extremely low.  So... I just don't see a significant advantage in covering vias, and not covering the vias has never caused me any problems or failures.

I do agree with your observation.  After I scanned PCB images on the internet, I did find that many PCBs have covered vias (probably half or so).  Nonetheless, if we need [some of] those vias exposed to make our machine more precise, cheaper or a better value, I'm willing to require this.

Note the following consideration too.  Engineers can trivially create two resist gerber files for one or both sides of the PCB (based upon whether components are placed on both sides or one).  So they can leave the vias uncovered on small runs of PCBs they intend to assemble themselves with our target machine, but switch to the other gerber file if and when they have assembly done elsewhere for high volume production.

The only negative I can identify so far is... if someone has a boatload of empty PCBs in the back room they want to assemble, and they already have the vias covered.  In those cases, if they can perform quality application of solder-paste, our software should be able to determine exact component position from the component pads themselves (which are definitely visible, since the component hasn't been placed yet).  In fact, I'm willing to be our software NEVER needs to look at via pads except for developers who have poor solder paste application equipment, or who make the openings in their stencils larger than necessary, or otherwise don't create properly prepared PCBs for assembly.

Note that all our software needs to find is 1~2 pad edges on each axis (x,y) that are not obscured by solder paste in order to precisely determine where the down-looking "nozzle camera" is located.  The chances we can't find that on components with many contacts is extremely low.  On components with only 2 or 3 contacts and poor solder paste application... that could happen, hence the desire to also have access to nearby vias (within the 50mm~100mm square centered on the component center).

Though it is tiny bit of extra work, engineers who hate uncovered vias can leave them all covered except 2~3 within the [50mm or] 100mm square around each component.  That's a pretty damn large area to only have 2~3 visible vias, so I can't imagine very many engineers will decide to spend $30K more just to avoid that requirement!  And if they apply solder paste properly, they won't need to leave any vias exposed at all.

BTW, the problem isn't contrast, the problem is confusing the software that determines position based upon the component pads.  To give an extreme example to make the point, imagine a component in which the solder paste extends beyond all 4 edges of every component pad.  Now the software cannot precisely determine the position of any one of those pads.  Who knows where the pads actually are?  The sloppy application of solder paste might be offset or rotated in any direction by any amount (up to some maximum)... no way to know.

The main problem here is software complexity.  If the software is made smart enough, it can look all over the place in that 100mm square centered on the component center, and try to find pads on other components to precisely determine position.  But if those components are already placed... no workie.  One could make the software even more complex and have the software scan the entire PCB to find components with problematic solder paste application (or any other reason that component cannot be precisely positioned).  It could then look around for other component pads in the 100m square field, and if it found them, delay the placement of those components until after the problematic component is placed.  But certain cases can turn into a multiple loop or non-resolvable dependency, which makes the software even more complex to recognize (and fail or seek alternatives).

BTW, if the component in question doesn't have fine pitch, it can be placed anyway, because lower precision will be possible by other means (either the mechanics themselves, or various alternatives I prefer not to waste your time on here and now).

Make sense?

PS:  If someone simply abhors covering vias, they are more than welcome to add a few more tiny "local" fiducials to their PCBs.  Just enough to assure 2+ are in each 100mm square portion of the PCB.  That's actually the best generic approach.  The only need for the via pad technique we're talking about now is to provide an easy, painless, automatic way to avoid the need for a few tiny fiducials on each PCB that are not required by more expensive pick-and-place machines.  On any PCB that has at least a few regions of the PCB that are not massively crowded with components, traces and vias, the extra tiny fiducials are trivial, and I'd say a "no brainer".  In fact, I'd call them "best practices" too.

Note that several of the neoden4 messages talk about errors that seem to increase in certain directions across the PCB surface (not necessarily x or y).  Can't be sure without knowing why this problem arises, but this problem might be solved for neoden4 too by adding a few more tiny fiducials on the PCB.  And I read (at least 2 or 3 times in the past 3~4 years) pick-and-place manufacturers encourage PCB designers to put more fiducials on the PCB if it is easy and convenient.  So... sounds like a good practice in any case.

[glenenglish]

Bootstrap says
"I've never seen vias covered by soldermask. " 
well i dunno where you have been....

The NORM is to tent your vias. Sometimes, selectively some vias are not tented for test points.


Yes NEVER NEVER except with NO OTHER option use via holes as fiducials.
They are drilled only well enough not to cause a plane short and nothing more....

The PCB fab people also will go to extra trouble also to ensure the fiducials are faithfully etched and not over/under etched.

g

[alexanderbrevig]

Maybe we should start a new thread on tenting vias? Just to add, when using vias under ICs for thermal relief you actually want some solder to flow down (and your apertures adjusted for extra paste). Using tented vias under ICs I'm told is considered bad practice. The only other time I do not tent my vias is if I want them to serve as a test point (as already pointed out by glenenglish).

[bootstrap]

Hi folks,

I'm the author of OpenPnP and I feel like I need to clear up a couple misconceptions about my software.

1. OpenPnP is not magic software. It's current vision algorithms are pretty basic and bottom vision is brand new. It's working well so far but I've only tested it with a few dozen different components.

2. I do not know of anyone using OpenPnP to place 0201s and < 0.5mm pitch components. Maybe someone is, but I haven't heard of it. That is not my target market. I'm primarily interested in 0603s and 0.5mm+ pitch components. The software is geared towards hobbyists, prototyping and small businesses. That isn't to say it can't do it, but for components that small it's more a matter of how precise your machine is versus how fancy the software is.

3. OpenPnP does not "require" steppers. As noted, the hardware interface is abstract. All OpenPnP cares about is whether your machine can move, pick, and place. It doesn't really care about how it goes about any of those steps and people have used it with machines that range from sticks, bubble gum, and hope to dual head 14 nozzle 2.5m/s linear servo monsters.

Hope this helps. If you have questions about OpenPnP or need anything clarified, we're a friendly bunch. Join the mailing list and introduce yourself: http://groups.google.com/group/openpnp

Thanks,
Jason

Jason, thank you VERY MUCH for posting your message.  To make wise decisions we need facts, not rumors... especially me!  I have a few questions for you.

If some of us collaborate to create a new pick-and-place machine, I assume we'll save a LOT of time if we can adopt OpenPnP.  Do you agree, or do you think we will actually waste time by adopting OpenPnP because we intend to do more by vision than you probably ever intended to do?  If you need to have a basic idea of the series of steps this configuration we're considering requires, search through my previous posts in this topic to find the one that lists all those steps (in terms of processes A, B, C, D, E, F...).

-----

After reading your description of your "target market", I'm not entirely certain whether ours is very similar, or outside (speaking of "in your opinion" here).  Let me say a few things that make me think our target is compatible with yours.

Note that the below is mostly my opinion, but slightly blended with comments from others in this forum who seem serious about being part of this project if we converge on a very worthwhile configuration.

#0:  So far we describe our "target market" something like this:  For independent engineers and wannabe engineers (hobbyists) who want to reliably place tiny (<= 0201) and fine-pitch (<= 0.50mm) components but cannot afford the cheapest commercial machines with these capabilities ($40K up) and are willing to happily accept up to a 2x to 5x slower placement rate than the cheapest commercial machines, and possibly other [optional] "minor inconveniences" (perhaps regarding feeders).

#1:  We learn towards making the machine "open hardware design" at the same time as possibly offering finished units (or maybe even kits, though I personally dislike that option).

#2:  Any added capabilities or features we add to OpenPnP would become part of the OpenPnP project (unless you object).  The nominal plan would be to NOT change how OpenPnP does anything, but to add additional options or paths to what you have setting a few variables (or equivalent) in OpenPnP makes it run the code paths compatible with our approach.  Others could subsequently design pick-and-place machines that adopt some or all of our approaches and set their variables accordingly.

#3:  Our "target market" is "cheap but not too cheap to achieve our goals".  I think somewhere between $10K and $20K is likely as a purchase price for a new unit.

#4:  Our primary goals are: "be able to precisely and reliably place 0201 or smaller discrete components" and "be able to precisely and reliably place fine pitch components with hopefully 0.30mm+ pitch but at worst 0.50mm pitch" and "cost much less than commercial machines" and "accept 2x to 5x slower performance than commercial machines that can place similar components".

Therefore, as far as I can tell, we are consistent with your target market in terms of price.  As long as you don't object to the ability to place smaller components than your hardware implementations currently support, then I see our "target markets" as mostly or entirely overlapping.  But what matters is what you think.

#5:  I believe we would strongly resist creating a "new branch" of OpenPnP to add support for the elements of our approach.  Why?  Because we'd want all future capabilities and enhancement that you and your collaborators make to become available to our machines too.  If we "branch off", updating our software from time to time would require much more work, and there is always a chance you would do something in the code or architecture that prevented such upgrades (or made them messy or extremely difficult).

#6:  We would probably adopt OpenCV whether we adopt OpenCV.  I assume you adopted OpenCV for OpenPnP.  Is that correct?  If not, what?

#7:  At least some of our cameras will have resolutions of at least 4K x 4K (though further brainstorming could reduce that to 2K x 2K).  I assume OpenPnP can work with high resolution cameras.  If that is a false assumption, please explain.

#8:  We plan to implement a couple unconventional approaches with vision in our approach as a way to enable extreme precision (to support 0201 and 0.50mm and smaller) with "crappy mechanics".  Maybe we should say "modest mechanics", but probably the more accurate term would be "variable quality mechanics".

-----

If you have lots of time, you can get an even more detailed impression of our goals by reading this whole topic (or at least messages I wrote, which include and reply to most other important messages).  But the above works as a general guideline if you want to save the time and misery of reading all those messages.

BTW, if you're willing to give us an rough estimate of how much time would be required to re-create what you've done so far, please do!  I hope we don't need to do that, but... who knows at this point?  Certainly not me.

Again, thanks very much for responding to this topic!

[mrpackethead]

if(glenenglish == bootstrap)
{
  eathat();
}
else
{
dielaughing();
}

[mikeselectricstuff]

I'd agree that using untented vias is a reasonable compromise to allow top-vision to improve accuracy, but you may still have issues with things like soldermask alignment affecting the accuracy.

If there was a way to accurately measure head position then that would potentially allow the use of cheap mechanics, but I don't think visioning a pasted board is it. Visioning vias might be, but I'm not convinced.

A reason to use tented vias is if there is any flowsoldering or wavesoldering of through-hole parts, so it tends to be the default.

However feeders are still the thing that really need solving for cheap pick/place, and as you get down to smaller parts, feeder performance gets increasingly critical

[bootstrap]

Using pads with paste is an unnecessary risk.  Vias are not all that great either since they are crowded by solder mask and are mostly small.  Traditional fiducial marks are super reliable and provide more than enough information for nearly all placements.  In very critical areas, you can always put a pair of local fiducials to better ensure success.

They are so easy to put in the PCB design and so easy for the machine to see - why would you avoid them? I have used vias and other holes as fiducials on PCB's I did with only hand assembly in mind, but it was more accurate and reliable when I add fids.

You posted some great messages in this topic I still need to reply to tonight, but I'll start with this short and simpler message to get myself up to speed.

The key of the approach I described previously in this topic is to apply vision on the real world (components, PCB pads, possibly the x,y rails, and so forth) more than conventional pick-and-place machines do... in order to make reliable and precise placement of very tiny and fine pitch components without massive, precise, expensive mechanics.

If we end up with a design that doesn't go that way, then much of what I say here and elsewhere may fly right out the window and become irrelevant.  However, assuming for now the previous paragraph applies, I'll reply to your message.

#1:  I suspect any good computer vision package can precisely determine position of small via pads (to roughly the same accuracy as standard fiducials).

#2:  I agree that explicit placement of a few additional fiducials when necessary is the preferred approach and practice, and we should encourage all PCB designers who intend to adopt our machine to place one or more [tiny] "local fiducials" AKA "secondary fiducials" on their PCBs.  If they do this, we can tell the software to completely ignore via pads and component pads.  Only when PCB designers are too stubborn to add "local fiducials" would the machine [optionally] fall back on code to attempt the equivalent with via pads and/or component pads.

#3:  I suspect components pads will usually be sufficient even without vias, but on PCBs where solder paste is badly applied and smeared in irregular fashion beyond the edges of most component pads, inferring the component pad positions will be problematic and unreliable (to varying degrees depending on each case), in which case uncovered via pads become our last fallback (other than refuse to continue or risk unreliable placement).

#4:  We are nominally assuming the camera has a 4K x 4K pixel field that images a 4" x 4" (or perhaps 100mm x 100mm) square on the PCB, making each pixel correspond to 0.001" == 0.025mm == 25 microns on the PCB.

#5:  Fairly straightforward image processing techniques should let us infer positions to about half of that (0.0005" == 0.012mm == 12 microns).  This should be sufficient for placement of 0201 and smaller, and 0.30mm pitch and maybe a slightly smaller given our approach.

#6:  I am very much in favor of "local fiducials" as primary method, with component pads being a secondary method (only when "local fiducials" cannot be found in the image), with via pads being the last fallback.

#7:  I am also willing to say to potential adopters and buyers, "don't be jerks, make sure the 100mm square centered on every tiny (<= 0402) or fine-pitch (<= 0.80mm) component contains at least 2 "local fiducials".  The whole "via pad issue" doesn't even arise when our adopters "do the right thing" and comply.  I don't think we're asking very much to save them tens of thousands of dollars on a pick-and-place machine!  I'm willing to be overruled on this one if others involved in the project want to require adopters/buyers include those "local fiducials".

[mrpackethead]

#1:  I suspect any good computer vision package can precisely determine position of small via pads (to roughly the same accuracy as standard fiducially).

OpenCV will do a very good job now.  The challenges are having good enough cameras and how long you are prepared to wait to process it. ( and/or how much processing power you want to apply to the problem ).  Theres quite a few algorithms you could use.   

#2:  I agree that explicit placement of a few additional fiducials when necessary is the preferred approach and practice, and we should encourage all PCB designers who intend to adopt our machine to place one or more [tiny] "local fiducials" AKA "secondary fiducials" on their PCBs.  If they do this, we can tell the software to completely ignore via pads and component pads.  Only when PCB designers are too stubborn to add "local fiducials" would be [optionally] fall back on code to attempt the equivalent with via pads and/or component pads.

Just point them to the relvenent part of the IPC standards ( http://www.ipc.org ).. SMEMA 3.1.   No need to reinvent a standard thats been well thrashed out already.

#3:  I suspect components pads will usually be sufficient even without vias, but on PCBs where solder paste is badly applied and smeared in irregular fashion beyond the edges of most component pads, inferring the component pad positions will be problematic and unreliable (to varying degrees depending on each case), in which case uncovered via pads become our last fallback (other than refuse to continue or risk unreliable placement).

DFM is such an important part of designing a pcb, that you don't need to worry about this.  If you have gone to the effort of designing a system like you are describing, why would you design a pub with such fundamentally important features missing?   This is not a consumer product, you don't have to design to the lowest common demoniator.

#6:  I am very much in favor of "local fiducials" as primary method, with component pads being a secondary method (only when "local fiducials" cannot be found in the image), with via pads being the last fallback.

Just refer to the above standards..

On feeders, spend $50 on some Yamaha CL Clone feeders for 8mm tape.. ( they work with 0201+ ) The are very reliable, very available and just work.     You can also achieve reliable pick up from cut tape in simply made holders..   They just take more time to set up, but are fine if you are only putting 10-20 parts down.. 

[bootstrap]

I can get that concept, but after spending some time ( 10 years) designing and manufacturing precision motion mechanics - it's just not all that hard or expensive to make a suitably precise motion platform for even the tiniest of parts.  That, in general, should eliminate the need for any fancy footwork in the vision system.

I think you and I have a similar "psychological" challenge on this project (a challenge that I may be finding a bit easier so far).  What I mean is, apparently both of us have made very... even extremely... precise mechancial systems in the past.  So we both know "how to design and fabricate that way", and we are aware of various tricks and methods that make this easier or more practical.

BUT... in this project we're trying to adopt an approach that can make inexpensive and simple-as-possible mechanical systems reliably and precisely place extremely tiny and fine-pitch components that many professional machines cannot do or have reliability problems with.

To be realistic, I think we need to [somewhat] disregard our expertise in precision mechanical systems.  Otherwise we risk finding out inexpensive simple mechanics just isn't up to the job.

What "target markets" we decide to support is relevant too.  If we want to put an "open-hardware design" out there based on our project, we have no control over how precise are the mechanics others build.  Obviously we do have such control over machines we build and sell, so we need to identify what range of "target markets" we intend to support before we know how to optimize the design approach.

I agree that a machine that can "do the job" of reliably and precisely placing tiny and fine-pitch components can either be based upon "machine design and machine precision" or "vision feedback and application of vision techniques to more of the pick-and-place processes".  I've been leaning towards the later because dozens of pick-and-place companies charge huge bucks for their [mostly] precision-mechanics based machines... and many of them still don't support 0201 and 0.50mm at all, or don't support 0201 and 0.50mm pitch reliably, or don't support 0201 and 0.50mm pitch reliably for all their customers.

I am probably just as confident as you in my ability to design and fabricate super-precise mechanics.  Maybe you know how to do so at somewhat lower cost than I do (based on your messages).  But I am still skeptical that leaning on mechanics is the wise and cheap path to such reliable and precise placement of tiny and fine pitch components.

We'll just need to brainstorm this and the other issues out more in subsequent conversations (which maybe should include one or more long brainstorming sessions by text+voice skype or telephone).

[bootstrap]

#1:  I suspect any good computer vision package can precisely determine position of small via pads (to roughly the same accuracy as standard fiducial).

OpenCV will do a very good job now.  The challenges are having good enough cameras and how long you are prepared to wait to process it. ( and/or how much processing power you want to apply to the problem ).  There's quite a few algorithms you could use.   

#2:  I agree that explicit placement of a few additional fiducials when necessary is the preferred approach and practice, and we should encourage all PCB designers who intend to adopt our machine to place one or more [tiny] "local fiducials" AKA "secondary fiducials" on their PCBs.  If they do this, we can tell the software to completely ignore via pads and component pads.  Only when PCB designers are too stubborn to add "local fiducials" would be [optionally] fall back on code to attempt the equivalent with via pads and/or component pads.

Just point them to the relevant part of the IPC standards ( http://www.ipc.org ).. SMEMA 3.1.   No need to reinvent a standard that's been well thrashed out already.

#3:  I suspect components pads will usually be sufficient even without vias, but on PCBs where solder paste is badly applied and smeared in irregular fashion beyond the edges of most component pads, inferring the component pad positions will be problematic and unreliable (to varying degrees depending on each case), in which case uncovered via pads become our last fallback (other than refuse to continue or risk unreliable placement).

DFM is such an important part of designing a pcb, that you don't need to worry about this.  If you have gone to the effort of designing a system like you are describing, why would you design a pub with such fundamentally important features missing?   This is not a consumer product, you don't have to design to the lowest common denominator.

#6:  I am very much in favor of "local fiducials" as primary method, with component pads being a secondary method (only when "local fiducials" cannot be found in the image), with via pads being the last fallback.

Just refer to the above standards..

On feeders, spend $50 on some Yamaha CL Clone feeders for 8mm tape.. (they work with 0201+) The are very reliable, very available and just work.     You can also achieve reliable pick up from cut tape in simply made holders..   They just take more time to set up, but are fine if you are only putting 10-20 parts down..

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

OpenCV will do a very good job now.  The challenges are having good enough cameras and how long you are prepared to wait to process it (and/or how much processing power you want to apply to the problem).  There's quite a few algorithms you could use.

Yes, yet another issue to consider carefully (like we don't already have enough).  How much processing to require from the hardware and microcontrollers in the machine (which we need to pay for and ship), and how much processing to require of the PC that the operator interacts with.

The more we require of the PC, the more complex the OpenPnP software (and/or the special-purpose OpenPnP "plug-in" for our machines) need to be.  But also, the less we require from our machine components, the less the machine will cost, which lowers selling price).  On the other hand, speed may be slower on slower PCs.

The more we require from the machine, the less complex the OpenPnP software (and/or the special-purpose OpenPnP "plug-in" for our machines) need to be.  But also, the more we require from our machine components, the more the machine will cost, which raises selling price).  On the other hand, speed should never suffer due to PC speed.

Just point them to the relevant part of the IPC standards ( http://www.ipc.org ).. SMEMA 3.1.   No need to reinvent a standard that's been well thrashed out already.

I have no desire to re-invent wheels that don't gain us substantial benefits.

DFM is such an important part of designing a pcb, that you don't need to worry about this.  If you have gone to the effort of designing a system like you are describing, why would you design a pub with such fundamentally important features missing?   This is not a consumer product, you don't have to design to the lowest common denominator.

Excuse me for sounding like a moron, but what do "DFM" and "pub" mean?

The target market for this machine is definitely not "lowest common denominator".  However, while we're not gonna end up anywhere near "the cheapest machine", one of the most important goals is to precisely and reliably place components too tiny or fine-pitch for existing "lowest common denominator" machines (and even some lowish-end commercial machines).

On feeders, spend $50 on some Yamaha CL Clone feeders for 8mm tape ( they work with 0201+).  They are very reliable, very available and just work.  You can also achieve reliable pick up from cut tape in simply made holders.   They just take more time to set up, but are fine if you are only putting 10-20 parts down..

I personally don't have enough practical experience with pick-and-place machines to make decisions about feeders yet.  But feeders do seem to be one of the two biggest issues we must tackle to make this machine cost-effective and "fabulous".

I have no objections to adopting inexpensive but reliable and highly capable feeders that exist already... if that turns out to be the best approach.

While I tend to shy away from "lowest common denominator", I would like to at least consider ways to ALSO support extremely simple (even "feeder-free") alternatives for engineers who only make small boards with limited number of components, but do need to support tiny and fine-pitch components... an optional sub-target market (advanced hobbyists)?

Question:  Why do you say those feeders work down to 0201 rather than "8mm tape"?  Why would the feeders not work with smaller than 0201 components as long as the components are still on 8mm tape?[/font]

[Koen]

Spikee, open a new topic with your build if you wish. Some of your points will cause you more problems then they'll solve.

[mikeselectricstuff]

Question:  Why do you say those feeders work down to 0201 rather than "8mm tape"?  Why would the feeders not work with smaller than 0201 components as long as the components are still on 8mm tape?[/font]

Repeatability of pick position, and not bouncing parts out.

[bootstrap]

Yes, NEVER, NEVER except with NO OTHER OPTION use via holes as fiducials.  They are drilled only well enough not to cause a plane short and nothing more....

The PCB fab people also will go to extra trouble also to ensure the fiducials are faithfully etched and not over/under etched.

I know it is too much to read every mention, so I'll just firmly agree with your claim that via holes are NOT positioned well enough for determining positions.

That is ABSOLUTELY CERTAIN.

But that is also why I always say "via pads", not "via holes" (or just "vias", unless I accidentally omit "pads").

The via pads are etched on the PCB along with all component pads and fiducial marks.  So they are all on the same "playing field" so to speak, and thus precisely enough positioned relative to each other to precisely determine position (with adequate image processing routines).

[bootstrap]

Question:  Why do you say those feeders work down to 0201 rather than "8mm tape"?  Why would the feeders not work with smaller than 0201 components as long as the components are still on 8mm tape?[/font]

Repeatability of pick position, and not bouncing parts out.

Yes, if the feeders bounce < 0201 components out, that definitely rules them out for < 0201 components!

However, if we adopt the technique I suggested, where the down-looking "nozzle camera" precisely locates where each tiny component is before it moves the nozzle into position and lowers it to the component, then how precise the feeder locates the component does not matter (unless it also wanders around).

Note that the above "extra" step will not be required on components above whatever size the machine and feeder mechanics handles reliably.

[bootstrap]

Spikee, open a new topic with your build if you wish. Some of your points will cause you more problems then they'll solve.

Just to clarify...

You say this because you believe the precision he claims will not be reliable and repeatable on every instance of hardware... and thus not appropriate to base a new design on?

If so, that's my take too, though I am impressed he can get so much from such a cheap machine.

[mrpackethead]

DFM is such an important part of designing a pcb, that you don't need to worry about this.  If you have gone to the effort of designing a system like you are describing, why would you design a pub with such fundamentally important features missing?   This is not a consumer product, you don't have to design to the lowest common denominator.

Excuse me for sounding like a moron, but what do "DFM" and "pub" mean?
[/quote]

DFM - Design for Manufacture.   The practice of considering how you will make something, as well as what you are making.     For example i can "design" a 60,000km diameter sphere in 3D cad, with 1um surface flatness but in all likelihood i'd struggle to make one...  good DFM practice helps you avoid such issues.

Pub, sorry typo.  should have been PCB.

Question:  Why do you say those feeders work down to 0201 rather than "8mm tape"?  Why would the feeders not work with smaller than 0201 components as long as the components are still on 8mm tape?[/font]

Because i'm talking from real experience and know they do..  i've not had reason to work with smaller ones, and not even gone looking...   

[Koen]

No, I believe Spikee could make more efficient choices. We'll see if he opens a new topic.

[bootstrap]

In a manner of speaking, the main point of this exercise is DFM (design for manufacture), albeit with a slightly difference focus than usual.  What I mean is, the main point is to "get precise and reliable" from "easily/cheaply/simply manufactured and assembled parts".

I have another principle of my own, which I suppose would be called DFF if it was a common term.  DFF would mean "design for future".  This can have more than one emphasis, but here the point is to design to support [as many as possible of] the most advanced components of today [and near future], so the machine is capable of most applications of today, and remains capable of most applications for years to come.  This can also imply "don't bother just copying existing designs and products".

What "fundamentally important features missing" from the PCB do you refer-to?

[Kjelt]

However feeders are still the thing that really need solving for cheap pick/place, and as you get down to smaller parts, feeder performance gets increasingly critical

I wonder if alternative designs where made for feeders.
It will take out of the box thinking.
Packaging such small components in paper tape with plastick strips is not a good design, probably cost efficient.
Are there machines out there that have a picking and sorting stage before the head picks it up and places the component?
How do these P&P shooters work with small components?

[bootstrap]

While not claiming to have the ability to jump ship and become a pick and place manufacturer - I do enjoy considering the issues.  One of the biggest things to consider for anyone wanting to design and build one is to have some experience with the practical issues.  I had designed and build a manual machine (very basic) and considered making a machine on a similar level to Lite Placer to help with my high density prototypes and short runs.

One day, I stumbled across my opportunity to buy and refurbish a Quad machine. I threw myself into the pick-n-place fire instantly and had to deal with a mountain of issues from motion control to feeders and everything in-between. After getting the machine running in tip top shape, I was faced with the challenge of managing the process as a whole.  This very steep learning curve totally changed what I would consider for building my own machine.  Like @bootstrap said, I don't want to build the same old machine - I would only be motivated if I could see a path to delivering an innovation of some sort.  Something that is uniquely helpful.

Yes, and I'm keenly aware that I have almost no personal/practical experience with SMT assembly equipment (solder-paste stencils, pick-and-place machines, reflow-ovens, PCB cleaning systems for SMT).  While I've designed and assembled literally dozens of my own PCBs for my own projects over the years, they were primarily components with leads or pins that are inserted into "through-holes".  After doing mostly technology/architecture/system/software design for several years, I return to hardware (what I typically call "electronics") and find pretty much all components I need for my projects are SMT only (or the non-SMT are huge or somehow lame).

So my major hope for being here is to interact with people who have the SMT experiences I don't.  While I've been searching and reading about SMT on the internet for 4~5 years, I know how valuable practical experience can be.

I would definitely imagine you learned a lot from that Quad machine experience!

Any sane individual who knows how much time, effort and expense will be involved in a project like we're discussing here will be very hesitant to "dive in".  I can live and develop projects on my savings from previous projects, which it sounds like you can't do... or more precisely, which you're not insane enough to do unless you are seriously convinced to quit or pull back on paying contracts.  I totally understand and relate-to that.  I've been self-employed since before I finished school, and have never been an employee.  I've done contracts for the likes of AMD, NASA, AirForceResearchLabs and various observatories over the years, but they were all special-purpose contracts, not employment.

But since you are also a contractor, maybe you can reduce the time you spend on contracts while you collaborate on a project like this... ASSUMING YOU BECOME CONVINCED WE HAVE SOMETHING SO COOL YOU CANNOT RESIST.  That also tends to be my criteria for doing a project or not!

From a business perspective - I have never gone down the path of designing products to be a cheap as possible.  I tend to go after the 'cost-effective' business solutions that are not the lowest cost, but have an excellent return.  With PnP machines, a LOT of value is correlated with reliability AND speed.  That is where the MyData class machines fit.  They are reliable and extremely fast at the same time.  Of course this also means that the entry level MyData machine is very expensive and only targets full time production.  Sacrificing speed does indeed save a lot of money and I have always wondered what is the market size of a $20k PnP solution.  At $10k, I think there are too many sacrifices to yield a machine suitable for a business.  At $20k, I believe there is enough to work with to make a machine that has the right ingredients for small business in-house production work with modest volumes.  That means the (in rough terms) the BOM plus labor would need to max out at under $10K if sales are direct and any profit is expected.  The volume of sales has a lot to do with how that ratio is derived, but it is at least a rough starting point.

I tend to prefer projects that are nearer the low-end than the high-end, but never too near the low-end.  I think this fairly well matches your position too.  My goal is for a $10K selling price that I don't want to allow to go above $20K during the development process.  It is my opinion that MAYBE the new forms of applying machine vision will make it possible to achieve the $10K goal, or at least substantially below the $20K price point (at the same kind of direct selling and markup you mention).  Since our target market will not be "rich corporations", you'll probably also agree that our support may be mostly on a forum, where adopters of our machine help each other, and we only need to spend time when we disagree with that help (or someone wants to pay us to get help).  I don't think we should charge for support that results from our machine suffering defects though.  You probably agree (more or less).  In other words, I envision NO showroom, NO salespeople, almost no marketing that costs us money, etc.

This is one reason to make a machine that satisfies a niche that NOBODY ELSE SATISFIES.  The fact is, most of my projects (that weren't on contract for existing corporations) were devices or technologies that DID NOT EXIST.  In other words, no competition!  If someone wants what we make, they have nowhere else to go.

We won't be in quite that situation, but we will be within the niche we define... if we meet our goals.  If we don't, we shouldn't even start making or selling anything.

I'll bet you agree with that much.

PS:  I am also a fanatic about simplicity and low part count (especially simplicity).  In some cases I've knocked that ball out of the park in into orbit in the past.  With this project, I don't yet see the opportunity to excel that far, but we should at least be able to KISS (keep it simple, stupid).

I would focus on a low cost feeder design.  The machines core motion is relatively easy.  A feeder scheme that can handle a very wide variety of parts with high-reliability and still be low cost will be key.  I do have a number of practical ideas (based on real concepts that I have already designed and built for other projects) to build a machine with very high feeder count, accuracy, and low cost to deal with low volume / high mix / fine pitch / low-modest speed. I have read a lot of 'overthinking' in this thread so far, the only way to get a low cost machine is by examining the real everyday issues and creating the simplest solutions possible to solve them in order of importance.  Every nut, bolt, camera, connector, servo, etc adds to the challenge.  Part presentation from the feeders is super important since the feeders are duplicated so many times on any given setup.

Yes, from my reading, feeders are a crucial part of the project.  For example, they are a significant portion of the cost if I buy an LE40V from ddmnovastar.  And I've read lots of complaints about feeder costs and feeder reliability over the past couple years.

I am hoping our machine can ALSO (not the common case) support that portion of our "target market" who makes [usually] small PCBs with small to modest quantities of only a few components on each PCB... BUT... wants to start adopting the latest, greatest tiny and/or fine-pitch components.  I sorta see this as a sub-niche that might be called "super-hobbyist" or for some nano-company to make higher-end "super-hobbyist products" (better, more capable, more flexible, or more elaborate than sparkfun or pololu PCBs).

I agree about what you say regarding feeders.  Without making the lower-end of our "target market" burdened with anything huge or complex, I would never-the-less love to be able to support a large variety of feeders, and even for the operator to be setting up feeders for the next job while the current job runs.  Though, if we find a way to support an insane number of feeders, that becomes less important.  Only a total moron of an engineer would not adopt the same caps, resistors and common components for all his PCBs without a damn good reason, so I have to assume only 50% of feeders on the current job would typically need to be changed for the next job (plus or minus 25%).

I have a couple seriously unique ideas about feeders that will definitely need a mind with feeder experience to brainstorm with me.  One of them that has a LOT of benefits more-or-less assumes (to an extent, though not rigidly) that all of each component is placed before other components are placed.  Ask me further about that later, either in text here, or in a skype or phone conversation.

My current line of business is totally unrelated and it would require a career change to do anything with that.  I have done a LOT of contract work in the past where my employers pay me by the hour to give up all of my concepts and ideas and they have literally made many 10's of $millions.  That is my motivation to design my own products and get the upside benefit.  My contract employers from the past essentially had the development capital and sales chain to take advantage of my ideas/designs/engineering/and implementations.  These were great learning opportunities and I am grateful to have been able to see how a product goes from zero to hero.  I am now saving all of my IP until a I can engage in a business opportunity that has an upside benefit.  My eyes are always open to a new opportunity, but it has to be significantly better than the path I am currently on.

I totally understand.  One instance of my IP could make me billions of dollars... easy and without question.  But I'd have to sell it to a government or military (or their contractor), and shortly thereafter (a few months or years) 99% of humans would be gone.  But the technology has an equally if not more awesome potential for advancement and benevolent application (which would also assure that disaster can never happen once someone else figures this technology out).  The fallout of that is that I haven't gotten investment, because those with a few million or more to spend always want control of the technology more than the benefits.  I went around in circles with Steve Jobs about this when he was still breathing.  Turns out, it might have saved his life.  I'm sure 99% of people with an opportunity to become a billionaire would jump at the chance no matter what the consequences.  I'm one of the few that riches alone will not sway, especially in the face of such extreme consequences.  But... now I've gotten way too far off topic.

What I wanted to ask is this.  Do you have to eject what you're doing totally out the airlock to work on a project like this... when you have one or two other collaborators with talents, abilities and expertise in the same ballpark as you?

With that said - I am scared of going down the path of building machines with huge BOM counts and complex supply chains.  Right now, I see more success in simpler projects (for me personally).

Yes, in general I favor smaller projects too.  Unfortunately, my "ultimate project" (mentioned above) isn't tiny.  It does involve a few rather independent subsystems that can be products.  One of those may not be too huge or expensive, but requires javascript, a language that drives me nuts, and I am thus not efficient on.  So I'll probably subcontract that one out.  But the other subsystems are not quite in the category of "simple"... though most of them do tend to be simple to extremely simple mechanically.  One of these is an advanced "robotics vision system", which is mostly just a fancy term for a small, simple, compact, rugged camera that performs a small to modest amount of very specific and extremely crucial image processing (of sorts) in "hardware" (largely in the FPGA, actually) inside the camera... and sends super-crucial information about each image in the video stream to the PC (along with the uncompressed or lossless compressed image).[/font]

[bootstrap]

However feeders are still the thing that really need solving for cheap pick/place, and as you get down to smaller parts, feeder performance gets increasingly critical

I wonder if alternative designs where made for feeders.  It will take out of the box thinking.  Packaging such small components in paper tape with plastic strips is not a good design, probably cost efficient.  Are there machines out there that have a picking and sorting stage before the head picks it up and places the component?

How do these P&P shooters work with small components?

Well, since I (and others) consider myself a world class "outside the box thinker", the issue of feeders on this proposed machine is one way to test that claim!  Uh, oh.  Set myself up there!

Obviously we need to assume whatever "mechanics" component suppliers offer their components in, which seems to primarily be "tape", "trays" and "tubes".  Anything else?

Nonetheless, you are absolutely correct that IF we can get a huge benefit of some kind, the components can go through a pre-processing of some kind (remotely or on-machine) to put the components into a better form or state for the machine.  What to do, however, obviously isn't obvious yet.

I've seen one or two "shooter" videos on youtube, but the ones I saw picked up the components from what looked like conventional reels of 8mm tape.

PS:  Now that you mention it, I don't see why they didn't just stack up a whole bunch of those components right against each other in a tube with the same ID as the component rectangle (which would have to have a spring-loaded pusher at the opposite end from where the components exit the tube.  Except, of course, if any slack in the pusher ever developed, components in the tube would rotate, get flipped, get jammed or otherwise messed up.  So I guess I see why that didn't happen.  Sure would have been a lot more compact though!

[Spikee]

Regarding feeders. just put a camera above the pickup places (or a few) and just measure the rotation / location offsets while the head is moving.
Using this method something like the cheap neoden 4 feeders can be used. No need for (ultra) precise >150$ feeders.

Or the other method:
Expose more than one component at a time, take picture, calculate offsets for them all (lets say 5 components) -> save those values -> pick and place
When you know in software that 5 are placed you can just expose another 5 components and repeat the process. This would be a mix between a tape pulling feeder and sticking component strips to the table.

[Kjelt]

Now that you mention it, I don't see why they didn't just stack up a whole bunch of those components right against each other in a tube with the same ID as the component rectangle (which would have to have a spring-loaded pusher at the opposite end from where the components exit the tube.  Except, of course, if any slack in the pusher ever developed, components in the tube would rotate, get flipped, get jammed or otherwise messed up.  So I guess I see why that didn't happen.  Sure would have been a lot more compact though![/font]

If I get you correctly you mean something like a bulletclip for automatic rifles, just stack m up and shove them out one by one. Replacing is just taking out the old clip inserting the new one instead of removing reels and peeling tape and so on and on. 
It should be possible for non emc components like resistors and capacitors.
But probably there is a risk the components will stick together, probably the reason they are individually packaged.