Wanted to buy: Moderate pick and place line.

[IconicPCB]

Mrpackethead,

Visit Adafruit forums on assembly... You will find interesting info on the Japanese manufacturer there.

[Kjelt]

Cars have now been in existence for over a hundred years... everybody owns one.. they are a commodity.
Why are they as expensive as specialty industrial equipment?

They are as expensive as the sum of their parts + 30-40% which is a lot less than a P&P.

[mikeselectricstuff]

Another thing to consider is durabilty - even a low-end P&P will be expected to place millions of parts within its lifetime, whilst maintaining good accuracy.
As mentioned before, 99% accuracy is next to useless as even a moderate sized panel will often have 1K+ parts. As soon as you need to rework a few tens of parts the advantage of doing it automatically rapidly diminishes.

You also need to consider that the market for a small P&P is quite small, as it lies between doing it manually and shipping the work to someone with a big machine.

[Kjelt]

You also need to consider that the market for a small P&P is quite small, as it lies between doing it manually and shipping the work to someone with a big machine.

Good point! I myself would see a niche market for hobbieists. For myself I would not mind having to wait 10 minutes for a hundred components to be placed instead of twp thousand, so speed is for me less important.
Accuracy can be gained by rehoming after a few pick and placements , yeah it costs time but as said less important for amateurs. So no camera, no need for enormous large pcbs just reliable.
But then this topic is about the semipros with larger pockets, so I will shut up now 

[Precipice]

No camera means you place the part, at best, as it was positioned in the tape. That's not acceptable. If your board is that simple, do it by hand.
(Yeah, tweezers and other dreadful, slow, non-scalable solutions. Eww.)
(100 components, by hand, with a decent manual pnp machine, should be 100x3 seconds. That's 5 minutes. How much money are you willing to spend to get that time back, especially since machine setup time, feeder loading and general dicking about will be rather longer than your 5 minutes.  Your brain, your eyes and your arms / hands, in conjunction with a decent manual machine to stabilise your hands and allow easy placement, trump any automated machine, until you're doing decent volume)

[sweesiong78]

Cars have now been in existence for over a hundred years... everybody owns one.. they are a commodity.
Why are they as expensive as specialty industrial equipment?

They are as expensive as the sum of their parts + 30-40% which is a lot less than a P&P.

Are you sure a car costs + 30-40% over the sum of its parts?  I maybe biased since I'm in the US but it seems to me a car cost LESS than the sum of its (retail price) parts....a new prius cost about $20K, while a medium range Trek bicycle here cost $2K. Considering the prius has thousands more components and complexity than a bicycle it truly is remarkable how cheap the modern car is. I recalled reading somewhere a long time ago that some car manufacturer only made a profit of about  $20 per car....

[sweesiong78]

As a designer and manufacturer of specialty commercial products that mix complicated mechanics with electronics, I regularly get the "Why is this so expensive?" question.

My answer is....design and manufacture it for yourself, sell a very small qty and then you will know. The sales channels cost 30-50% of most industrial products because they are a tough and time consuming sale that generally needs specialized and experienced sales people. The company has to go to various trade shows, make videos, documentation, etc. Those massive expenses are on top of the R&D costs.

This guy is nuts making his own PnP...it looks like he made everything from scratch except for the rails, servos and camera
https://www.vbesmens.de/en/pick-and-place.html

[mikeselectricstuff]

There really is no excuse whatsoever to not have vision these days. As vision tasks go, P&P vision is very easy - well-defined field of view, fixed focus distance, easy lighting and limited range of shapes.
You don't even need an exotic lens as you can just move a normal one further from the sensor to get closer working distance.

Without vision, it's just a toy.
The thing that really needs solving is feeders.

[Kjelt]

I recalled reading somewhere a long time ago that some car manufacturer only made a profit of about  $20 per car....

You can not believe that? He can better stop and work in a supermarket then.
In my country you can with good negotiating get 9% off the MSRP, beyond tbat you see the guys desparity. They make money on the service and parts but they still have to make some decent margin on the car itself.

[Corporate666]

I get the sales pitch. In comparison, you can get decent german made cnc milling/routing tables that have much more expensive steppers and linear ballbearing guidance (due to the extreme forces milling metals) for less than €4k. Than look at a Polish made CNC machine and they ask starting prices (so not complete yet) from €15k.
You can tell me a lot of stories but no way that this is in balance.

Not many people need P&P machines, so the development cost has to be spread amongst a low number of units.

This is also a chicken and egg problem IMO, if that entry machine would cost <€5k many more hobbieists would order one and numbers would go up.

Oh well I just have to face it  that a P&P is not going to be in my reach.

Years ago, we built a CNC machine in the shop.  CNC machines are pretty simple... X/Y/Z and a spindle - the trick to getting it right is rigidity and speed.  There are lots of truly shit CNC machines out there using acme screws, stepper motors and aluminum extrusions for the frame.  That might be OK for routing out bits of acrylic, but CNC machines are all about side loads against the cutting tool, and that requires accuracy and rigidity, which means linear slides, ballscrews and servo motors at least.  IMO, anything less is a toy.

With PnP's, it's all about speed and accuracy as far as the machine goes, and that's not really any less complicated than CNC machines, so it's not really much cheaper.  You don't have the side loads of CNC cutting but you have to accelerate and decelerate pretty quickly and very accurately, which is not trivial.  But then you also have the added complexity of component centering, vision and component feeding - none of which a CNC machine has. 

We also built our own PnP machine.  First one was incredibly tricky... the better solution was to buy a run down old PnP and just replace all the electronics, so that's what we did.  Just the software was tricky because you have quite a few things going on between the pick and the place, and you're always trying to do it faster and faster to get good speed out of it.

The best solution is a used low-cost PnP (I mean one that was lost cost when new), still has factory support and parts, and is as simple as possible.

[mrpackethead]

(As it turns out, they were building something that promised to be rather less half-arsed. Firepick or something? OpenPnP based software, and an interesting placement mechanism.)

I'm actually building a fire pick Delta..  I don't expect that it will be anything more than a curiosity, but its certainly got potentially to be a step in the right way..   I know Dave doe'snt like it, but i'm holding my breath.     It might just be really good for Hobby DIY guys.   Though i can't see it being faster than using a good manual PNP set up. ( i have a Dima FP-600 )

[rx8pilot]

I think that there should be a distinction between wanting a basic P&P for hobby purposes and a basic machine for professional use. The big difference is reliability and good support - both of which cost quite a bit of money. I can find a lot of benefit in a low-end P&P that is relatively slow and has no vision. What I can't tolerate is a fussy machine that breaks regularly and hard to get serviced. Time is money in a business and it will cost me way more than the purchase price of a TM240A if was broken for 3 weeks. TM240A is about $6k USD with shipping, a similar DDM Novastar is triple that, but targeted to professionals with service and support.

I got a laser marking machine from Epilog a number of years ago. It was way more expensive than the cheap Ebay options, but it has been very reliable and easy to service when needed. Still a low-end laser, but good low-end. My friend got the random Ebay offering that has twice the part size and double the laser power for a third of the up front cost. The only problem is that it was a complete POS that ruined parts, caused production delays, and a huge distraction while trying to get it working even half as good as the Epilog. I am sure that given enough time, it could be a reasonable hobby machine but I would have been put out of business if I had that project to deal with.

I am looking for the low-end of P&P that is still suitable for professional use. The 3 PCB's I am working on right now have a ton of passives and a couple of fine pitch components. I could totally make use of a non-vision machine to put down 0805 passives and SOIC's which is the bulk of the placements. After P&P, I can hand place the fine pitch parts which I am already doing with success. I would not say a no-vision P&P is a toy, but most of the ones I have seem (TM240A comes to mind) seem to be built to toy standards. The non-vision DDM Novastar offerings seem to fit the bill for non-precision, low volume PCB's and certainly seem to be targeted to professional use.

For me, there is not a practical way to jump directly to a $150k solution so I spent some time working out a good manual system on my own. The goal is to be able to make enough products manually to get a low-end P&P line, then consider from there. Just about anything will be faster than me manually placing about 250 components per PCB set. Even if time was not an issue, I going crazy in the process.

Hoping this thread can continue, it seems to be revealing some ideas and experience of others.

[coppice]

Why does even a medium P&P machine cost the equivalent of a car? It is not rocketscience anymore?

It has never been rocket science, but, like rockets, not many people need P&P machines, so the development cost has to be spread amongst a low number of units.
There will always be an entry cost due to precision mechanical stuff - it's never going to get below a few K for a useable machine.

The really fast PnPs are a bit like rocket science, and just as impressive to see fired up. The real question is why a simple TM240A is so much more expensive than an simple Chinese CNC, when the volumes are probably not a lot different and the complexity is not a world apart either. The head clearly costs a bit, but the feeders are the kind of simple mechanics they almost give away in China, even for modest volumes.

[mrpackethead]

feeders seem to be a real killer line item thats for sure.   $500 a pop for 8mm feeders.. Mmm

[mrpackethead]

Options on my list now.

Mechatronika M10V.  ( or its bigger brothers )
7740 from MDM ( Japan )
BS281 from http://www.autotronik-smt.de


the $5k chinese machine isn't going to cut it.  Just aint good enough.

Any other suggestions. 

[rx8pilot]

For me the more feeders, the better. the BS281 would allow me to run a few different PCB's in a single setup which is worth the extra cost. Fully loaded is certainly the most expensive, but without surprise the most capable.
For what seems to be about $20k USD less, a fully loaded 7740 (Manncorp FVX in USA) which is also enough to get me fairly far down the road.
The M10V seems nice but 40 feeders would be to small for my work.
$50k seems about right for a P&P with feeders intended for short runs and the 7740 meets the criteria. I have not seen anything else in the $50k USD range that any more compelling.

[IconicPCB]

M10V does have a somewhat limited repertoire of feeder locations. Two feeder banks on either side of the machine frame good for a maximum of 36 8mm feeders
BUT...

It handles loose components, tray components, cut tape components ( mounted on double sided tape on a shelf within head locus in my experience) ski slope tube feeders so it is not a bad choice for low volume precision work. I have seen reports of M10Vs with cut tape feeders dispensing over a hundred different components.

[mrpackethead]

I've just got pricing for the MV10's bigger brother the MX70..  Its 66 reels and its not a super lot more.. ( 22,600 Euro vs 14,600 Euro )..

Need to of course add feeders onto hat, but its the same price for both the MV10 and the MX70.

Seems like a pretty useful machine.

[Kjelt]

I've just got pricing for the MV10's bigger brother the MX70..  Its 66 reels and its not a super lot more.. ( 22,600 Euro vs 14,600 Euro )..

And the MX80 will be around €35k probably a bit less if I see your prices.
http://www.printtec.nl/contents/en-uk/d411.html

[mikeselectricstuff]

As mentioned, precision and speed are the key.
One thing that could reduce cost substantially is if you could find a way to accurately measure the position of the tip of the tool in real time, so you could use cheap mechanics and discipline them with software.
Unfortunately visioning of the board isn't going to work as there's solder paste on the pads.
I think the feeder problem is solvable - it's just a case of designing something that can be made cheaply. 

[coppice]

One thing that could reduce cost substantially is if you could find a way to accurately measure the position of the tip of the tool in real time, so you could use cheap mechanics and discipline them with software.

Surely any serious disciplining of the hardware would make is desperately slow. Simple offset compensation for measured play in a belt drive or lead screw would seem to be all that's possible in a reasonably nippy machine. It isn't just a matter of the time taken to perform any analysis and correction. If the mechanics are not inherently tightly controlled you have to wait for the shaking to die down.

[rx8pilot]

I just spoke with Precision Placement Machines about the Quad 4C's they have. They just jumped to top of the list for me. They completely refurb Quads with new boards and software running on Win 7 and have full non-contract support available. $28k for machine plus feeders that are only a few hundred each which is amazing. A full ready to run machine for $50k with lots of feeders and full support. 120+ feeder capable.

These are not eBay machines, they can actually do 0201's, new linear scales, etc. Obviously the 4C's are proven, but I did not consider them thinking that a used machine is too much trouble. With this option, I could get a real industrial workhorse and support in the price range of a low-end new machine.

[mikeselectricstuff]

One thing that could reduce cost substantially is if you could find a way to accurately measure the position of the tip of the tool in real time, so you could use cheap mechanics and discipline them with software.

Surely any serious disciplining of the hardware would make is desperately slow. Simple offset compensation for measured play in a belt drive or lead screw would seem to be all that's possible in a reasonably nippy machine. It isn't just a matter of the time taken to perform any analysis and correction. If the mechanics are not inherently tightly controlled you have to wait for the shaking to die down.

Not necesarily as you only need the accurate position for the placement, so you could wizz to close to the right position, then home-in with a tighter closed-loop position. However I have no idea how feasible it would be to measure this - laser TOF/interferometry is probably the only viable option, which maybe rather tricky to do in practice.
 

[Kjelt]

Not necesarily as you only need the accurate position for the placement, so you could wizz to close to the right position, then home-in with a tighter closed-loop position. However I have no idea how feasible it would be to measure this - laser TOF/interferometry is probably the only viable option, which maybe rather tricky to do in practice. 

How about those capacitive guidance rails used in calipers, they can be had for cnc machines and are accurate within 0,01 mm, should be good enough for P&P.
They are not that expensive anymore. And you can also use closed loop stepper motors and drivers, also very affordable these days (<€300 for a motor and driver)
But as someone else said you do not know where exactly the head will pick up the component so the component placement should also be equally accurate and you're own remark about the cost of good feeders.

[rx8pilot]

Linear scales on the X/Y accomplishes this already - at least as well as any other position sensing option. I have designed for this kind of thing before [I primarily design mechanics with a little electronics and software on the side]. The machine can deal with much more error since it always knows where the position is regardless of mechanical slop. There are obviously limits to this and the support and guide structure still needs to be rigid and tight to deal with accel/decel forces. This is true even if you used some other exotic from of position sensing. If the table and the head are twisting and moving the information from the scales becomes inaccurate. Backlash is a real challenge with all the motion reversals and high precision. Much easier and productive to have a super tight mechanical than deal with it in software.

You can always create an offset LUT in as fine of increments as your please, but it requires the machine to be repeatable to be worth anything. Loose mechanics are not repeatable so you could not expect the LUT to help much. The machine would be hunting for position.

On our laser marking machine that uses linear scales is still not rigid enough to deal with the vibration which makes the marking fuzzy. When the machine is bolted to a very heavy table, the marking quality is better.