New Pick and Place design ideas

[Kjelt]

Unbelievable what a large and expensive setup that Martin is, just for dispensing paste and it still has an alignment error  .
It is pretty slow compared to a decent P&P machine, so you must also have multiple of these machines to keep up with real production.
I guess these are only used for prototyping but then a decent pcb manufacturer can also deliver a stencil for a few tenners.
I think the the target audience for these machines is only dispensing red thermal glue for two sided boards?

[ServoKit]

I think one application for a device like this is rework, where stencils are usually not an option. But you are right, it is fairly slow and looking at the whole setup (standard aluminum extrusions, simple steppers with belt drive, apparently no encoders), that's $1500-2000 for the hardware, tops. Obviously, there's considerable know how in stuff like temperature compensation for the paste etc. - but still.

Regards, Axel

[forrestc]

Unbelievable what a large and expensive setup that Martin is, just for dispensing paste and it still has an alignment error  .
It is pretty slow compared to a decent P&P machine, so you must also have multiple of these machines to keep up with real production.
I guess these are only used for prototyping but then a decent pcb manufacturer can also deliver a stencil for a few tenners.
I think the the target audience for these machines is only dispensing red thermal glue for two sided boards?

I have live, shipping products for which a typical 'restocking' production run is 2 copies of a board.  By the time you put the stencil on the printer, get the paste on the stencil, print the two boards, clean the stencil and so on, you could have the boards run with the dispenser.     

In my environment the dispenser is actually on the pick and place machine as an option, so you load the board and the P&P machine does the dispensing.  We have it pause between the dispense and placement step to verify a good dispense.  Then it proceeds to placement.   There is no special setup here - you load the program for the board and it does the work.  Yes it's slow, so we don't use it for higher-volume products, but for those we do a few (<10) a month of, it's perfect.

I really like this functionality because I also do first run prototypes on the exact machine which is used for production.  Because our standard component set is on the machine we don't have to load any feeders, etc., and any specialty parts can be quickly hand-dropped on the pasted board.

[Kjelt]

I have live, shipping products for which a typical 'restocking' production run is 2 copies of a board.  By the time you put the stencil on the printer, get the paste on the stencil, print the two boards, clean the stencil and so on, you could have the boards run with the dispenser.     

In my environment the dispenser is actually on the pick and place machine as an option, so you load the board and the P&P machine does the dispensing.  We have it pause between the dispense and placement step to verify a good dispense.  Then it proceeds to placement.   There is no special setup here - you load the program for the board and it does the work.  Yes it's slow, so we don't use it for higher-volume products, but for those we do a few (<10) a month of, it's perfect.

I really like this functionality because I also do first run prototypes on the exact machine which is used for production.  Because our standard component set is on the machine we don't have to load any feeders, etc., and any specialty parts can be quickly hand-dropped on the pasted board.

Ok you are the person with the experience , I would bet it takes time to put on a new nozzle, fill it through with paste change halfway for a smaller nozzle and afterwards clean all used nozzles also?
Or are you doing all the pads with the same nozzle, then you probably only have a very limited set of pad sizes in your design? I mean if you have to fill a 0603 pad with a nozzle capable of delivering a 0,1 drop it will take some time. But I can see it can be of value if you have a lot of different pcb's and very small series you do not have to switch stencils. Still what is the price of such machine? (oh Dave told us in the video if you have to ask you can't afford it  )

[bootstrap]

Working on a dispenser myself, I've been reading up on the topic and also stumbled upon the Martin devices. Dave made a video about them some years ago:



In the comment thread someone mentions that they are in the $30K range... I thought about using an auger for controlling the flow (see attached for a first attempt)

Regards, Axel

Thanks very much for that video!  I realize those are 0402 components and it appears like the BGA is 0.10mm pitch, but the implication is the device might indeed be able to apply solder paste for 0201 and 0.50mm pitch components (especially if that's not the smallest nozzle).

I don't doubt the machine costs $30K, since it is about 80% of a complete pick-and-place machine.  However, the idea of this project is to perhaps integrate solder paste application into the pick and place machine, so all we'd need is the nozzle.

I see Martin also sells the nozzle separately, and apparently I already posted the part number of the nozzle in a previous message.  They claim the dot size can be as small as 0.150mm, which I consider adequately small for 0201 (barely) and BGQ/QFN/etc pitches down to 0.50mm and possibly 0.40mm.  I am a bit surprised they claim they can support one size smaller than 0201, but... maybe.

Another question.  The video makes it appear like the tip of the nozzle actually pushes against the PCB, then emits the solder paste as the tip rises off the PCB.  Is this true, or is that an illusion.  If the tip really does push that firmly on the PCB, I am a bit surprised, and a bit worried about damage to the tip over hundreds of thousands of applications.

Another question.  Is the quantity of solder paste applied a variable?  If so, then larger pads can be accommodated by setting a higher volume of paste to be applied with the tiny nozzle.  Also, is there a way to tell the machine to apply multiple dots to rectangular pads.  That too is a way to accommodate larger pads without nozzle changes.

[bootstrap]

Unbelievable what a large and expensive setup that Martin is, just for dispensing paste and it still has an alignment error  .
It is pretty slow compared to a decent P&P machine, so you must also have multiple of these machines to keep up with real production.
I guess these are only used for prototyping but then a decent pcb manufacturer can also deliver a stencil for a few tenners.
I think the the target audience for these machines is only dispensing red thermal glue for two sided boards?

If you noticed, the alignment error was extremely constant on every pad.  AND, very importantly, he did not let the device choose the fiducials, he manually centered on two large pads.  That's probably where the error came from, and thus the machine is likely not the error source.  Let the vision system determine position, and do so on fiducials.

I'm definitely not sold on applying solder paste with the pick-and-place machine, but I want to be totally fair as we consider all options.  Plus, IF the pick-and-place machine applies solder paste, then the true cost must be judged on the basis of being both a precise automatic solder paste stencil printer and a pick-and-place machine.

[bootstrap]

I have a PCB conveyor and nearly all of my boards are double sided. I love it and would be bummed if the machine did not have it.

Explain what you mean by a conveyor belt?  I was thinking that means a wide flat belt surface that moves horizontally.

Do you consider the rails on the neoden4 constitute a "conveyor belt"?  If not, are those rails an acceptable alternative?  If not, why not?

[bootstrap]

Topic:  Mechanical structure and configuration.

So far this topic has mostly discussed approaches and specific details of the proposed pick and place machine.  Now I'd like to begin the discussion of various ways to implement the basic mechanics.

Rather than bias this, I'll open the floor for ideas.

PS:  The video of the Martin solder paste application system did make me wonder whether something nearly as lightweight and compact as that might suffice (not counting feeder system, of course).  Opinions?

PS:  Note that Spikee started discussing along these lines in reply #43 on page 2 of this topic, but... most of us weren't ready for that part of the discussion yet.

[forrestc]

Ok you are the person with the experience , I would bet it takes time to put on a new nozzle, fill it through with paste change halfway for a smaller nozzle and afterwards clean all used nozzles also?
Or are you doing all the pads with the same nozzle, then you probably only have a very limited set of pad sizes in your design? I mean if you have to fill a 0603 pad with a nozzle capable of delivering a 0,1 drop it will take some time. But I can see it can be of value if you have a lot of different pcb's and very small series you do not have to switch stencils. Still what is the price of such machine? (oh Dave told us in the video if you have to ask you can't afford it  )

We use 24Ga disposable deedles.  Same nozzle size for all dispensing, which does limit your your dynamic range somewhat.   We could probably get more repeatable on very small parts if we didn't have to also dispense larger ones.   Just to give you an idea of the dynamic range we DO do, is that we dispense all the way from 0.5mm TQFP's to TO263AB packages with the same needle.  The TO263 heatsinks have multiple, large, dots.

The entire setup time is less than 5 minutes.   The most consuming part is actually the z axis alignment.   In addition, the syringe just stays on the machine, so if I walk out there to do a prototype, it takes much less time - say 30 seconds to uncap the needle and manually trigger the machine to squirt paste through the needle to clear any old/verify lack of clog.

Remember, that we balance our line so that for those items which are dispensing, the operators are not waiting on the machine.  Instead they're completing their inspection and additional assembly/test steps from previous boards.

As to cost, I think it added somewhere between $5 and 10K to the machine.     I looked back but I don't have an itemized quote, and I can't find the emails discussing it.

[MicroBlocks]

This guy is getting close:

[ar__systems]

Wow so many posts.... there is no way I'll be read all of these, so I guess I'll just reply to the first one.

I think it is a strange target machine - slow, but with 0201 capability, and 1u accuracy? Why do you need 1u accuracy? Also, if we have to place 0201, I presume it would be complex high density circuit boards with let's say 500 components on them? Dropping speed to 500cph makes our throughput at 1 pcb per hour? Pointless, imho.

Using cameras as a replacement for positioning is not going to work very well. Did you ever try filming a moving target? Possible, but definitely not with a cheap camera.

[MicroBlocks]

Better start reading the rest. First post is just that. Initial ideas that needs to be discussed and improved upon.

[rx8pilot]

Wow so many posts.... there is no way I'll be read all of these, so I guess I'll just reply to the first one.

I think it is a strange target machine - slow, but with 0201 capability, and 1u accuracy? Why do you need 1u accuracy? Also, if we have to place 0201, I presume it would be complex high density circuit boards with let's say 500 components on them? Dropping speed to 500cph makes our throughput at 1 pcb per hour? Pointless, imho.

Using cameras as a replacement for positioning is not going to work very well. Did you ever try filming a moving target? Possible, but definitely not with a cheap camera.

To me:

500CPH = Slow death
1000CPH = Periodic prototype
1500CPH = Slow production
2500CPH = Passable for low qty of high value PCB production

My Quad 4000C is a 3600CPH rated machine, but in reality it achieves around half of that. If I optimize the feeders and programming - it would likely struggle to see 3000CPH. I only optimize for ease of use - not speed and deal with the hit in efficiency. Also, I run the machine fairly slow which reduces some issues. This gets me around 1500-1800CPH and that is generally faster than all the other processes like printing, inspecting, re-flow, etc. For my small operation where I make a low volume of high-value PCB's - it works out ok. Would I like it to be faster? Absolutely - but the P&P machine is not the slowest part of the process so I would focus on the ancillary steps first until I was waiting on the machine. After that, I would tweak the setup to run at 100% speed. After that I would optimize the feeders.

500CPH - too slow to be interesting. 1500CPH (real CPH, not 'RATED') at $10k-15k that can do 0201 would be very interesting.

[mrpackethead]

Has bootstrap lost interest.  No posts for two days

[ar__systems]

Absolutely - but the P&P machine is not the slowest part of the process so I would focus on the ancillary steps first until I was waiting on the machine. After that, I would tweak the setup to run at 100% speed. After that I would optimize the feeders.

Makes sense, but only if your process is fully pipelined. If you do all by yourself, then it is not fully pipelined and every stage becomes a bottle-neck.

[rx8pilot]

In general, it's a 'production line' no matter what. It cannot go into the P&P until it is printed. It can't go into the oven until it's placed. I can't do the 2nd side until the first side is out of the oven.

My process is FAR from optimal - no doubt. Even if I have more people, the machine would not be the first bottleneck even at 1800CPH (slow).

[mrpackethead]

In general, it's a 'production line' no matter what. It cannot go into the P&P until it is printed. It can't go into the oven until it's placed. I can't do the 2nd side until the first side is out of the oven.

My process is FAR from optimal - no doubt. Even if I have more people, the machine would not be the first bottleneck even at 1800CPH (slow).

but most importnatly, its working and doing a job.

[rx8pilot]

but most importnatly, its working and doing a job.

Yes! And I am thankful for that.
It's how my family eats.

[Corporate666]

Some thoughts, since I can't help myself

I have 3 Quad 4C's - same machine as RX8, but different software (same mechanicals for the most part).  Before these I had a Fuji IP-3.  Before that I had two Dynapert chip shooters.  Before that I had another gantry machine.  Before that I had some customized home-brew setups.

In my humble opinion...

1) I could not agree more with RX8 when he pointed out the folly of spending a lot of time trying to overcome shitty mechanicals with all sorts of software and/or vision tweaks.  A quick and precise underlying X/Y/Z mechanical system using the right parts is not expensive or difficult to build using established and accurate parts.  There's no reason to be dicking around with substantial backlash and such.  You can and should use the software to zero-in the accuracy, but starting with linear rails, closed loop positioning and a rigid structure is a no-brainer.

2) Mike said it earlier and he's right on the money.  Feeders.  Feeders.  Feeders.  Everyone ignores feeders and focuses on placement.  Forget placement - it's easy.  Feeding and picking is the issue.  There is no point whatsoever in ignoring the feeder issue and loading little cut strips of tape on the bed of a machine and/or pre-staging parts by hand.  You spend an hour setting up the machine so you can not spend an hour placing a board.  False economy.   

3) RX8 said it right when he said that the machine he would build today, after owning a "real" PnP is not the machine he would build before he actually owned one.  That hits the nail on the head.  Most of the issues people dote over when designing in their head are non issues and they gloss over the ones that DO matter (feeders!).  A useful machine should hold a lot of parts.  It should be easy to program.  I don't care about speed all that much compared to how much I care about reliability and unattended operation.  My machines now let me go so something else.  They are not fast by modern standards, but I can place hundreds of boards per day.  And even if I am making $10 per board in profit, that would be thousands per day, tens of thousands in profit per week.  If I needed 5 times the throughput, I could buy more of these same machines, or invest in a $300k Juki/Assembleon/MyData/Universal/whatever.

4) Forget paste dispensing, IMO.  I chased that pipe dream for a while - then I re-learned a lesson I already knew.  Things are done the way they are not because everyone doing it is too stupid to think of a better way, but because the way it's done works and works well.  Etched steel stencils without frames and pasted on a desk using masking tape to hold boards in place can support multi-million per-year operations.  We can paste 500 boards (50 panels, 10 PCB/panel) in maybe 20 minutes. And we can place them in maybe 2 hours.  And reflow in about 30 minutes.  There is zero economy to be gained with paste dispensing.  The only exception is prototype boards or glue dispensing for wave soldering, but those are specialized applications.  And even then, given the time to program the machine, I've never had a case where it wasn't faster to just do it with the desktop pedal-operated dispenser.

Vision is what it is - been discussed to death and I think it's well understood and well implemented in most machines.  The big discussion should be about feeders.

I've seen all kinds on the machines I have owned - varying levels of complexity and success.  I like the Quad feeders the most.  They are self contained - basically just power and index input.  The feeder handles holding the reel, indexing a given amount, peeling the plastic cover tape, presenting the part, and adjustable index amounts.  There will always be some gears or belts involved, but a simple servo indexing the tape with a mechanical linkage tying the cover tape capture to the tape indexing I think is the best solution.

But the big elephant in the room about these projects is that you can't sell a PnP that cost $5k in parts for $10k.  You need to sell it for more like $40k.  Because when you add in the cost of R&D that you have to spread over the sales, and you factor in the marketing and sales cost, and the cost of support and warranty, you will need every bit of that margin to run a business.

If you think you will open source stuff or have online forums for support - you are limiting your customer base to hobbyists who won't spend too much on the machine, but without a real business and support, you will never get business customers interested.

PPM (the company who took over the Quad line) is not very far from my shop.  I've been up there and toured the facility.  They charge prices for spare parts, support and whole machines that most people on this thread would consider outrageous.   They are doing OK by New Hampshire small business standards, but their offices aren't exactly luxurious buildings of glass and steel.  And they get most of their machines for FREE or next to free - and if it's beat to shit or missing parts, they probably don't even want it for free.  Then they refurb them and sell them for $25-50k. 

Those are the kind of realistic margins you need to be able to sell PnP machines to business customers.  And if you sell to hobbyists, it's a very limited market (in size and how much $$ they have to spend) and IMO, probably a support nightmare.

[mrpackethead]

But the big elephant in the room

I think the big elephant has gone quiet.

[mrpackethead]

On the feeder topic, im heading down a path of using Yamaha CL stuff.

The CL clone feeders cost around $50 for a 8mm, though to several hundred for bigger ones.

40 way plate ¥1999 ( USD 324.00 ) https://world.taobao.com/item/45596890497.htm?fromSite=main&spm=a312a.7700824.w4002-5269705.14.5HqNtz
20 way solenoids on Manifold ¥999 ( ~USD$164 ) https://world.taobao.com/item/522807850694.htm?fromSite=main&spm=a312a.7700824.w4002-5269705.22.5HqNtz

for $652 I will get 40 solenoids and feeder plates..    Add 40 feeders for another $2000.     Heck i might even double that and spend $5k on an 80 feeder setup.. And another $3-5k for drives, cameras, and put Open pnp on it, and i'll ahve spent about $10k + time, but learned a lot, and i susspect i'l have a machien that runs as well as our Yamahas... 

[mrpackethead]

Can i assume this project now is dead?

[rx8pilot]

Can i assume this project now is dead?

I don't think it was ever a project - more of a random grouping of thoughts about how to put parts on a circuit board. Interesting discussion to participate in though, especially since I have so recently learned the daunting task of PCB assembly by P&P machine. The nuances are plentiful and hard to anticipate until you have done it.

My machine is reliable, rigid, accurate, repeatable, flexible and a host of other good things - but PCB assembly is still a remarkably pain in the ass endeavor. The whole system and process has to be 100%. 99% accurate creates enough work that you may as well assemble by hand. Super unforgiving process. The key word is that it is a 'PROCESS', and not a machine. The P&P machine simply facilitates a complex and unforgiving process.

What I cannot wrap my head around is the interest in a 'hobby' P&P machine. The very nature of what the machine does makes it a commercial system. What hobbyist needs to make hundreds of PCBs' in a day? That is a business and a business (like mine) does not have time to fiddle around with a toy. You will go out of business trying to 'save' money with a consumer targeted solution. The Neoden4 is a barely passable machine and it costs $10k+ - well outside of the hobby range.

After reading about the N4 here on EEVBlog, I have very thankful that I don't own one. Too limited and too fiddly.

If I had money burning a hole in my pocket and a lot of extra time - designing and building one sure would be fun. As long as I don't have to make a business out of it.

[mikeselectricstuff]

I spent some of Maker Faire on the OpenPNP booth this weekend. There was a LOT of interest in the machine being demoed, which was a pretty basic setup feeding from tape strips on the bed, but with top & bottom vision (and using vision to detect tape holes).

Afterwards Jason told me that if he'd had kits available at about $2500 he could probably have shifted a couple of dozen  (BOM cost around $900, target spec 0402/0.5mm QFP)
Now I'm sure that a lot of those potential customers would have only later realised the time to set up would be marginally worthwhile, and I wasn't around enough to hear many of the conversations, but I was quite surprised at the level of interest.

[Kjelt]

What I cannot wrap my head around is the interest in a 'hobby' P&P machine. The very nature of what the machine does makes it a commercial system. What hobbyist needs to make hundreds of PCBs' in a day? That is a business and a business (like mine) does not have time to fiddle around with a toy. You will go out of business trying to 'save' money with a consumer targeted solution. The Neoden4 is a barely passable machine and it costs $10k+ - well outside of the hobby range.

Personally I don't think a hobbieists needs a high cph machine or a super duper 100% reliable machine 24/7. Small starting one person businesses as yourself might as long as they don't sell >100 pcb's a week that is an ideal market for these cheap machines. If you have a real business like >200 pcb's a week, I do not understand why you do not outsource this to professionals where the whole process is certified and the chance on productfailures is kept to a minimum. Why bother yourself any longer, economicially it does not compute.
So there is this grey area between high production and manual production where these machines fit.

But I guess what a hobbieist that makes 2 to 10 pcbs a week really needs is a faster manual camera assisted placer.
Especially when people get older there is a need for a mechanised system that can reliably pick up an extremely small smt part or very large high pincount device and be able to place it on the exact right spot on the pcb. Now there are little commercial manual p&p camera assited systems with a decent price and optimizing that further would be a nice market I guess.

I do agree, spending $100 or more on just one feeder, needing at least 80 or so of them alone is not viable for a hobby market.