Neoden 4 pick and place

[elmood]

I wonder how everyone is doing with production boards on the Neoden 4? We are on our fourth run... 150 units this time. It's the biggest board so far with 248 parts, TQFP100 and a bunch of small resistor packs and things. So far the machine has been the least cooperative this time around. After it blew its fuse several times a few days ago it's been troublesome ever since... I'm assuming that's just a coincidence but thought I'd mention it anyway.

The TQFP100 part is actually placing really well, but we still get some shorts after reflow... is it common to have to touch up parts like this? Does anyone have suggestions to minimize rework?

The rest of the problems have to do mostly with pick error which seem quite frequent. I've tried different pick heights, pick delay, nozzle sizes, etc. and a few feeders are still being a pain. Specifically a Spansion flash chip that has a bit of a textured surface... I wonder if this is reducing suction. Also, CTS 1206 4x resistor packs which seem to have a bumpy epoxy surface often get a suction error. And most frustratingly the vibration feeders often fail to detect suction on even the most common SOP16 parts like 7400 series chips.

Would anyone mind sharing info about settings that they found work better for more troublesome components? I found that increasing the pick delay appears to give more time for suction to increase before sensing. Also, when using Neoden's recommended nozzle sizes for larger parts they just don't work. I often get pickup error and have moved to the second smallest nozzle on most parts like small SOP ICs and the like. The smallest nozzle seems to work great on 0603, 1206, SOT23, etc. parts and we rarely have problems with these.

Oh, and I should mention that two feeders have started acting up this week. One 8mm feeder just died and wouldn't respond to the system anymore... Neoden sent me a replacement for it. And just today a 16mm feeder stopped feeding. It makes a noise but the tape doesn't advance. Unfortunately the software doesn't let you skip a part when there is an error, so if you can make a feeder work you have to just cancel the run and start it where you left off later.

[TankSparks]

Fuse: According to Neoden For 110vac you should use 5 amp fuse.   Ours was shipped with a 2 amp

If it picks and gives error try  -20 vacuum level detect,  had to do that on some of our parts, no more retries.   I actually started using that value on all the parts instead of the default -40 or -50

We ran 8000 parts last friday.  Only intervention required was when a reel got stuck, that was a quick fix.   
Also had one part which wasn't placed accurately, it appears part moves on nozzle from the aggressive acceleration,  so we used a slower speed on that part, which solved the problem.

[mrpackethead]

If you are getting shorts on your TQFP's and they are placed well, then i'd almost certianly look at your solder pasting..

I reduce the apatures on a TQFP to 70%, ( the solder paste window vs the pad side )..  Have you done that?     Also check are you getting good gasketing on the board from teh stencil..

[48X24X48X]

but we still get some shorts after reflow

It's very likely your paste is too much on the pads. Even with hand stenciling and hand pick and place, I don't have shorts even with 0.5 mm pitch TQFP.
Usually for less crucial parts, I left the paste opening at -5% only with 0.125 mm stencil thickness.

[mrpackethead]

Stencil thickness coudl be part of the issue as well.

[bootstrap]

I have a few basic questions about the neoden4.

#1:  Are the x and y motions driven by DC brushed motors, DC brushless motors, steppers or some other type?

#2:  How is x and y position detected (linear optical encoders, rotary optical encoders, open-loop via steppers, other)?

#3:  Are there any SMT components the neoden4 cannot place?  I'm not much worried about components with coarse pitch that can be placed manually, but I'd like to know about them too.

#4:  Do any thorough reviews exist?

#5:  What other relatively inexpensive PnP machines are available that can accurately place 0201 components and QFNs and such with 0.30mm to 0.50mm pitch?

#6:  Before I saw this machine, I was seriously considering a dmmnovastar LE40V, but that costs 3~4 times as much.  Can anyone compare the neoden4 with the LE40V, or mention alternatives to the LE40V that are better and cheaper (and can handle 0201 or smaller)?

#7:  Are there any inexpensive (sub $50K) "solderpaste-jet" printers available or in the works (or rumor mill)?

Thanks in advance for answers.

[mikeselectricstuff]

I have a few basic questions about the neoden4.

#1:  Are the x and y motions driven by DC brushed motors, DC brushless motors, steppers or some other type?

#2:  How is x and y position detected (linear optical encoders, rotary optical encoders, open-loop via steppers, other)?

Open-loop stepper I think

#3:  Are there any SMT components the neoden4 cannot place?  I'm not much worried about components with coarse pitch that can be placed manually, but I'd like to know about them too.

Tall parts - don't recall the limit but it's mentioned somewhere here

#4:  Do any thorough reviews exist?

I think this thread is the most comprehensive source of info from users

#5:  What other relatively inexpensive PnP machines are available that can accurately place 0201 components and QFNs and such with 0.30mm to 0.50mm pitch?

I think these parts are at the edge of the capabilities of all the cheaper machines - I think SmallSMT was claiming to be able to do 0201

#7:  Are there any inexpensive (sub $50K) "solderpaste-jet" printers available or in the works (or rumor mill)?

Doubt it - I don't think there would be enough market for something like that - you can buy an awful lot of stencils for $50K
AIUI the Mydata is well into 6 figures, mostly used by assemblers with high mix, and/or special requirements like package on package and variable paste thickness, i.e. people who don't have any other realistic option.

[elmood]

Thanks for the info about paste.... both paste application and the reflow oven are areas that I initially underestimated. I think I was more enamoured by the idea of robots making boards. (it still is cool every time it's running) With paste I think careful adjustment of the stencil printing jig to properly support the board at the right height is critical. Also, the mixing and condition of the paste is something I've been trying to assess. When is it time to change the paste? It does end up sitting out for periods of time on the end of the squeegee while waiting for the next board to be ready.

As for bootstrap's questions:

- The machine has steppers with optical homing switches. So far we haven't found any problem with losing steps, but the motors are quite noisy.

- As for large parts, I am placing some 6mm inductors directly from 16mm tape... they are probably the most heavy / chunky parts and with proper tuning of the settings are working okay.

- For stencils: buy a stencil printer jig and find a local vendor near you that makes stencils. We found two in our area and chose one of them that had more responsive service. They charge $240 CAD shipped to our door and even let us put multiple designs on the same stencil for no extra charge.

- Parts that the machine cannot place would basically be stuff that you can't feed easily. - Tapes >16mm (although Neoden can make custom 24mm feeders for you) and probably some types of tube parts unless you modify the vibration feeder mouthpiece plate. Keep in mind that a bit of hand placement is not really the end of the world. We have an SMT crystal that we are placing by hand right now and it takes about as long as the machine to do it by hand.

Another thought about machine limitations... every process has limits even at the big PCBAs... you pay more for access to the best tech. But in engineering like in art, limitation is a form of inspiration... there are usually multiple ways to solve a problem and you get more creative when you can't use everything. If you're trying to make an iPhone then you probably have Apple-sized budgets for your processes. But if you're doing what most small shops are doing that need their own lost cost PNP machine, I'd bet there are options in how you design your board. And surely whatever you can place with the Neoden 4 will work even better on the next generation of machine. I've standardized on 0603 for most passives, and feel confident using pretty much any IC down to 0.5mm pitch or so. I have avoided QFN and BGA for now, but would probably try them if needed after getting to know our workflow a bit better.

BTW I wouldn't recommend Neoden 4 as a good machine for a mini PCBA job shop. It's great if you have full control over the entire design / build process though.

[mrpackethead]

Thanks for the info about paste.... both paste application and the reflow oven are areas that I initially underestimated.'

Dont' worry, you wont' be the first.. Myself included.  Everything else is dependant on the paste process. If you cant' get that right, you are stuffed.  If you are getting tombstones, its probalby because your paste is wrong.  if you are getting shorts, its probalby because your paste is wrong..  and so on and so on..

I think I was more enamoured by the idea of robots making boards. (it still is cool every time it's running) With paste I think careful adjustment of the stencil printing jig to properly support the board at the right height is critical

Absolutely..

. Also, the mixing and condition of the paste is something I've been trying to assess. When is it time to change the paste? It does end up sitting out for periods of time on the end of the squeegee while waiting for the next board to be ready

Check the paste manufacturers data sheets..    Having it at the right temp makes a big difference.. you can just pull it out of the fridge and start printing..    The stuff i use is suppoably good for 8 hours, but i've never left it sitting around that long..     

[bootstrap]

Reply to mikeselectricstuff and elmood:

I don't really know how to reply properly (with included quotes), so I won't try.

#1/#2:  Open loop steppers:  OUCH.  That makes backlash in the x,y drive mechanics a potential problem, as well as bends and warps in the general mechanics.  If they're smart they can attempt to compensate for backlash with clever software.  For example, if they know where the nozzle is over the camera, and they know where the final placement location will be, they can step the motors several steps (maximum backlash distance) towards the final placement position before they stop to take a photo of the part.  MAYBE (depends on how this might interact with PCB reference marks).

#3:  Tall parts:  Probably any part over a few millimeters high can be placed well enough by hand.

#4:  Reviews:  Can you (or anyone) who has experience working with pick and place machines try to estimate the real world practical differences between having one of these machines ($12K with some feeders) and the ddmnovastar LE40V machine ($35K with some feeders): http://www.ddmnovastar.com/pick-and-place/automated-systems/le-40v-pick-and-place-equipment-with-vision)?

#5:  Competition:  The smallSMT machine with servos appears to cost about $12K with some feeders, which is roughly the same as the neoden4.  They do have some smaller stepper-motor machines, but I doubt I'd trust them.  Also, they claim they support only as small as 0402 and I need to support 0201 or smaller (required for some of my BGA components that I need, which means I'm sure to run into this same requirement in the future).  I'm not sure why the servo-based machines are not accurate enough, unless it is what seems like the classic silly mistake that cheap pick and place machine designers make (200 step stepper to rotate parts on nozzle).

Do you know whether the smallSMT machines with servo-motors have linear encoders on the x,y slides, or just rotary encoders on the lead-screw shaft (which still leaves the backlash problem)?

Requirements:  I absolutely need to place 0201 discrete components and 0.50mm BGAs and QFNs (and maybe 0.30mm later on).  On the other hand, I do not need speed, as I'm only need to build small batches of prototypes and demos units (at the moment, anyway).

#7: Solder Paste:  A couple people I've communicated with claim most of their defective PCBs are due to imperfect solder-paste application.  That has made me strongly consider what I consider outrageously expensive automatic stencil printers like these:

A:  SPR-45:  http://www.ddmnovastar.com/stencil-printers/automated-systems/spr-45-stencil-printer

B:  SPR-45A:  http://www.ddmnovastar.com/stencil-printers/automated-systems/spr-45va-stencil-printer

What do other folks who have assembled SMT PCBs with 0201 and 0.30mm~0.50mm pitch BGA/QFN/etc think about this topic and these stencil printers?  Please point me to better value/quality stencil printers.

#8:  Reflow Oven:  They also said a good reflow oven is important, especially for lead-free which I probably need to adopt.  I've been looking at the following two products but also anxious to hear about experiences from other folks and learn about alternative products.  The price of the A unit below was something like $9000 last I checked (a couple years ago), and I'm sure the second one is at least $3K more (but never asked price).

A:  GF-12HT:  http://www.ddmnovastar.com/reflow-ovens/low-volume/gf-12ht-benchtop-reflow-oven

B:  GF-120HT:  http://www.ddmnovastar.com/reflow-ovens/low-volume/gf-120ht-reflow-oven

Costs:  I try to avoid "false savings".  In other words, I'd rather spend more up front for good equipment than find out my total costs are more for cheap equipment due to rejects, rework, debugging and other costs associated with lowering initial costs.  If I buy all this ddmnovastar equipment mentioned above, I'm probably looking at somewhere between $60K before shipping.  If anyone has suggestions for better value/quality equipment, let me know.

Collaboration:  I like the idea of collaboration, both the "sharing expenses" part as well as the benefits of brainstorming that come with collaboration.  So if anyone wants or needs to assemble fine-pitch PCBs and might want to collaborate on this equipment, let me know.  Since I tend to bounce between (Maui, Hawaii), (LasVegas, Nevada) and (Atacama, Chile), the equipment would probably need to be located in one of those places.  I have a fair amount of test equipment and "related junk" (like 4-channel 5-gigasample oscope), so we should be able to have a pretty nice development lab.  In fact, I'm also thinking about maybe opening the lab up as a "advanced/high-end private hackerspace/makerspace", but have only visited a couple of these places and they are way, way, way too "low-end" for what I do.  I'm not so much thinking of charging for membership as setting up a place for serious, advanced, experienced developers like me where more-or-less we all bring are great stuff for everyone to benefit from.  Or something like that.  Any comments along these lines will also be appreciated.  Note:  I've been developing hardware, software, optical and other advanced technologies for many years, but I've never collaborated, never been particularly social, and from my limited checks, kind of doubt there are very many advanced developers out there that haven't started formal companies.  In my case, I've always worked from home to keep expenses down (though I have had a partner or two on a few projects in the past).

Other equipment:  What else do I need for a complete assembly setup?  How about cleaning the PCBs after assembly?  I have a pretty good (~$500) stereo microscope, hot-air soldering iron and I forget what else, but it has been a while since I assembled PCBs (been doing mostly software for a few years), and now need to dive head first into PCBs with SMTs.  We are now FAR past the point where high-performance through-hole parts are available, and pretty much everything I make is (and always has been) fairly-to-very high-performance.

How I design:  Unfortunately, I don't design PCBs that look much like what I see on most electronics web-sites.  It seems I always design things that need to be "too complex" or "too small" or "too fast".  For example, one project is to create a high-speed, high-resolution robotics vision system camera that have a lot of fast components, FPGA, 10 gigabit per second ethernet interface.  In fact, an advanced version of that camera may have an 8/16-core AMD Zen CPU plus all components necessary to make a working embedded linux system.  And that board (and the camera board) need to be SMALL.  That's just the way things go with me!  OTOH, I'll be needing to make some simpler stuff in the near future too, like an advanced H-bridge DC servo motor controller for [brushed] pancake motors, DC-to-DC converters (for power supply voltages), and some other fairly straight-forward stuff.  But they all end up being at least a little bit advanced in some way or other, even if only because they are made as tiny as possible.  Incidentally, seems like most of my PCB designs are 8-layer PCBs, though I do have a couple 6-layer PCBs and might in the future too.  But knowing how things go with me, I may not be able to constrain myself to 8-layers someday in the future.

Thanks to everyone who helps me dive head first into the world of fine-pitch SMT.

[TheSteve]

For 0201's and 0.5mm BGA's I think you should look for something other then the Neoden4. Consider spending real money, not buying one of the cheap machines. The Neoden4 is actually pretty capable when dialed in and you figure out the software oddities etc but really doubt you'd get the results you want with it.

[mrpackethead]

- For stencils: buy a stencil printer jig and find a local vendor near you that makes stencils. We found two in our area and chose one of them that had more responsive service. They charge $240 CAD shipped to our door and even let us put multiple designs on the same stencil for no extra charge.

Wow, i pay about $40 for a very well laser cut, electroformed framed stencil and another $30 to get it here. ( $USD70 )..  and they just make it per the gerber i send them, so they dont' care whats on it.     Most PCB houses will now do this as a stndard service, so the stencil arrives at the same time as the pcbs.. 

[bootstrap]

For 0201's and 0.5mm BGA's I think you should look for something other then the Neoden4. Consider spending real money, not buying one of the cheap machines. The Neoden4 is actually pretty capable when dialed in and you figure out the software oddities etc but really doubt you'd get the results you want with it.

Thanks.  Do you have a neoden4 or work with pick and place machines?  I'm just curious what your reasons are for giving this advice.

[48X24X48X]

Thanks.  Do you have a neoden4 or work with pick and place machines?  I'm just curious what your reasons are for giving this advice.

It's best for you to read through the 40 pages of the thread then you will notice who owns a unit.

Wow, i pay about $40 for a very well laser cut, electroformed framed stencil and another $30 to get it here. ( $USD70 )

Got mine laser cut 47 cm x 37 cm framed, 0.12 mm thickness, laser cut USD15 shipped from China to here. Yes, shipping included!
Stencil usually take few hours to make but the shipping take like 4-6 days. You can add as many design into stencil as long it fits.
They do help to modify the opening if they think it is necessary.
Can't complain for the price.

[wraper]

Wow, i pay about $40 for a very well laser cut, electroformed framed stencil and another $30 to get it here. ( $USD70 )

Got mine laser cut 47 cm x 37 cm framed, 0.12 mm thickness, laser cut USD15 shipped from China to here. Yes, shipping included!
Stencil usually take few hours to make but the shipping take like 4-6 days. You can add as many design into stencil as long it fits.
They do help to modify the opening if they think it is necessary.
Can't complain for the price.

It's because you buy cheap mechanically polished stencil with a slow shipping. mrpackethead gets much better quality stencil in a few days with dhl.

It's best for you to read through the 40 pages of the thread then you will notice who owns a unit.

Yeah, spend half a day just to know who has neoden4, very productive time spending.

[mikeselectricstuff]

Reply to mikeselectricstuff and elmood:

#1/#2:  Open loop steppers:  OUCH.  That makes backlash in the x,y drive mechanics a potential problem, as well as bends and warps in the general mechanics.  If they're smart they can attempt to compensate for backlash with clever software.  For example, if they know where the nozzle is over the camera, and they know where the final placement location will be, they can step the motors several steps (maximum backlash distance) towards the final placement position before they stop to take a photo of the part.  MAYBE (depends on how this might interact with PCB reference marks).

You get what you pay for. You mustn't forget these machines are maybe a tenth or less the cost of anything you could buy a very few years ago.
They are creating a new market that simply didn't exist before - low cost, low(er) performance
You want sub-micron servos and granite bases, talk to Mydata, Juki, Essemtech etc. and bend over.

#3:  Tall parts:  Probably any part over a few millimeters high can be placed well enough by hand.

Unless you have a large number of them on the board. IMO this is one of the less necessary shortcomings of the Chinese machines.
 

#5:  Competition:  The smallSMT machine with servos appears to cost about $12K with some feeders, which is roughly the same as the neoden4.  They do have some smaller stepper-motor machines, but I doubt I'd trust them.  Also, they claim they support only as small as 0402 and I need to support 0201 or smaller (required for some of my BGA components that I need, which means I'm sure to run into this same requirement in the future).  I'm not sure why the servo-based machines are not accurate enough, unless it is what seems like the classic silly mistake that cheap pick and place machine designers make (200 step stepper to rotate parts on nozzle).

Small parts are not just about placement accuracy - things like feeder performance become increasingly important - parts not bouncing out. Also nozzle design and vision becomes harder.
Stuff like 0201s are still regarded as a bit exotoc by many assembly houses - Chinese are aiming at a higher volume market than the big guys,  I'm not sure we'll ever see parts that small being supported on cheaper machines.

Requirements:  I absolutely need to place 0201 discrete components and 0.50mm BGAs and QFNs (and maybe 0.30mm later on).  On the other hand, I do not need speed, as I'm only need to build small batches of prototypes and demos units (at the moment, anyway).

In which case I suggest you increase your budget or spend time looking for a good subcontractor with a high end machine.
Not needing speed isn't really going to save much money as most of the market for tiny parts is for high volume users, so there isn't enough market for a machine with high accuracy but low speed for anyone to put in the R&D to make something that can do it with a high degree of reliability.

#7: Solder Paste:  A couple people I've communicated with claim most of their defective PCBs are due to imperfect solder-paste application. 

Absolutely - paste quality is the biggest factor in final board quality.

[glenenglish]

"#1/#2:  Open loop steppers:  OUCH.  That makes backlash in the x,y drive mechanics a potential problem, as well as bends and warps in the general mechanics.  If they're smart they can attempt to compensate for backlash with clever software.  For example, if they know where the nozzle is over the camera, and they know where the final placement location will be, they can step the motors several steps (maximum backlash distance) towards the final placement position before they stop to take a photo of the part.  MAYBE (depends on how this might interact with PCB reference marks)."

Nothing wrong with open loop if it is done properly.....

backlash. where is backlash ? there are no  meshed gears and 2mm pitch belt is pretty good.  or ballscrews. backlash you can barely measure.....

If  you are designing for a board full of 0201s then IMO you are in a expensive league, and you will wear it.
it's the same reason I don't design with 0.3mm BGAs nor more than a couple (hand placed) 0201 per board. You put yourself into an expensive arena with that sort of construction. When a customer says to me "How small can I make it " . I reply "How much money do you have ?"

I'd suggest one of SmallSMTs ballscrew machines. try it. get one to evaluate. Only USD10k, if you have a board full of 0201s then the reality is your budget should be OK with that. it *may*  do the job. 

[mikeselectricstuff]

"#1/#2:  Open loop steppers:  OUCH.  That makes backlash in the x,y drive mechanics a potential problem, as well as bends and warps in the general mechanics.  If they're smart they can attempt to compensate for backlash with clever software.  For example, if they know where the nozzle is over the camera, and they know where the final placement location will be, they can step the motors several steps (maximum backlash distance) towards the final placement position before they stop to take a photo of the part.  MAYBE (depends on how this might interact with PCB reference marks)."

Nothing wrong with open loop if it is done properly.....

backlash. where is backlash ? there are no  meshed gears and 2mm pitch belt is pretty good.  or ballscrews. backlash you can barely measure.....

If  you are designing for a board full of 0201s then IMO you are in a expensive league, and you will wear it.
it's the same reason I don't design with 0.3mm BGAs nor more than a couple (hand placed) 0201 per board. You put yourself into an expensive arena with that sort of construction. When a customer says to me "How small can I make it " . I reply "How much money do you have ?"

I'd suggest one of SmallSMTs ballscrew machines. try it. get one to evaluate. Only USD10k, if you have a board full of 0201s then the reality is your budget should be OK with that. it *may*  do the job.

Once you're down to this level it becomes more like a game of statistics - to make a machine that will place small parts with a high reliability level gets expensive, and needs a lot of testing and refinement - cheaper machines may do it, most of the time, on a good day, but the manufacturers aren't going to guarantee it. Do you feel lucky?
 

[mrpackethead]

#1/#2:  Open loop steppers:  OUCH.  That makes backlash in the x,y drive mechanics a potential problem, as well as bends and warps in the general mechanics.  If they're smart they can attempt to compensate for backlash with clever software.  For example, if they know where the nozzle is over the camera, and they know where the final placement location will be, they can step the motors several steps (maximum backlash distance) towards the final placement position before they stop to take a photo of the part.  MAYBE (depends on how this might interact with PCB

Actually having watched one in depth and watch it place parts i'n doubt its missing any steps and it certainly seems to be mataining it over placement over several hundred parts.

#4:  Reviews:  Can you (or anyone) who has experience working with pick and place machines try to estimate the real world practical differences between having one of these machines ($12K with some feeders) and the ddmnovastar LE40V machine ($35K with some feeders): http://www.ddmnovastar.com/pick-and-place/automated-systems/le-40v-pick-and-place-equipment-with-vision)?

This is a low cost machine that is in a different class from a production machine.  You get what you pay for.. Mostly the issues are with the feeder system and the software.  The actual placement is not too bad.  I watched it place 0201's and it got most of them right.

#5:  Competition:  The smallSMT machine with servos appears to cost about $12K with some feeders, which is roughly the same as the neoden4.

you can buy the same machine in china on tabao for a whole lot less than that.   The servos are actually closed loop steppers so the use of the word 'servo' is true ( steppers are servos ) but its not whats commonly thought of. The feeder system on these machines is what really lets them down.

They do have some smaller stepper-motor machines, but I doubt I'd trust them.  Also, they claim they support only as small as 0402 and I need to support 0201 or smaller (required for some of my BGA components that I need, which means I'm sure to run into this same requirement in the future).  I'm not sure why the servo-based machines are not accurate enough, unless it is what seems like the classic silly mistake that cheap pick and place machine designers make (200 step stepper to rotate parts on nozzle).

#7: Solder Paste:  A couple people I've communicated with claim most of their defective PCBs are due to imperfect solder-paste application.  That has made me strongly consider what I consider outrageously expensive automatic stencil printers like these:

Solder paste is the root of so many issues.   But with a bit of practice and knowhow a $30 stencil and a $500 manual printer will do a very good job time after time.  I'm doing fine pitched QFN's ( 0.4mm ) and 0201 this way.  But like learning to solder with an iron, it requires some effort to learn.

#8:  Reflow Oven:  They also said a good reflow oven is important, especially for lead-free which I probably need to adopt.  I've been looking at the following two products but also anxious to hear about experiences from other folks and learn about alternative products.  The price of the A unit below was something like $9000 last I checked (a couple years ago), and I'm sure the second one is at least $3K more (but never asked price).

I have a great multizone oven in the factory.. I also have a converted toaster oven..  Honestly for doing 1offs, the toaster oven with the spark fun kit does just a good a job.. and its ready instantly to use.

osts:[/b]  I try to avoid "false savings".  In other words, I'd rather spend more up front for good equipment than find out my total costs are more for cheap equipment due to rejects, rework, debugging and other costs associated with lowering initial costs.  If I buy all this ddmnovastar equipment mentioned above, I'm probably looking at somewhere between $60K before shipping.  If anyone has suggestions for better value/quality equipment, let me know.

I've been doing this game for quite some years now, and i have a variety of gear.  You can build great boards with nothing more than a pair of $5 tweezers and a frying pan. It takes a lot of time, but it can be done.     I've also seen absolutely terrible results from million dollar lines made when people don't know what they are doing.    Good equipment does not guarantee good results.    My best advise to you is that i've you've had little or no experience in this stuff, find someone who has to help you out..   Otherwise you're just going to face frustration and not get the job your looking done.

Other equipment:  What else do I need for a complete assembly setup?  How about cleaning the PCBs after assembly?  I have a pretty good (~$500) stereo microscope, hot-air soldering iron and I forget what else, but it has been a while since I assembled PCBs (been doing mostly software for a few years), and now need to dive head first into PCBs with SMTs.  We are now FAR past the point where high-performance through-hole parts are available, and pretty much everything I make is (and always has been) fairly-to-very high-performance.

From what i read, here,  my advise is that you might be better of using a service like macrofab in Dallas.   They do outstanding work, will save you a lot of time and money..   and its relatively easy to use.

[mrpackethead]

Once you're down to this level it becomes more like a game of statistics - to make a machine that will place small parts with a high reliability level gets expensive, and needs a lot of testing and refinement - cheaper machines may do it, most of the time, on a good day, but the manufacturers aren't going to guarantee it. Do you feel lucky?

I watched the N4 place 100 odd 0201's.. it did about 98 of them properly... Is that good enough?

[mikeselectricstuff]

Once you're down to this level it becomes more like a game of statistics - to make a machine that will place small parts with a high reliability level gets expensive, and needs a lot of testing and refinement - cheaper machines may do it, most of the time, on a good day, but the manufacturers aren't going to guarantee it. Do you feel lucky?

I watched the N4 place 100 odd 0201's.. it did about 98 of them properly... Is that good enough?

Not really, if you are looking to buy a machine that needs to do them.
 With that error rate, how can you have any confidence that the next board might be 98% or 50% or 0% unless you completely understand the reasons for the problems.
My machine will do 0402s on a good day but I'd only do it for my own stuff, not a customer job as it's really hit & miss whether it will do it OK, or will need most of teh parts nudging.

[mrpackethead]

Once you're down to this level it becomes more like a game of statistics - to make a machine that will place small parts with a high reliability level gets expensive, and needs a lot of testing and refinement - cheaper machines may do it, most of the time, on a good day, but the manufacturers aren't going to guarantee it. Do you feel lucky?

I watched the N4 place 100 odd 0201's.. it did about 98 of them properly... Is that good enough?

Not really, if you are looking to buy a machine that needs to do them.
 With that error rate, how can you have any confidence that the next board might be 98% or 50% or 0% unless you completely understand the reasons for the problems.
My machine will do 0402s on a good day but I'd only do it for my own stuff, not a customer job as it's really hit & miss whether it will do it OK, or will need most of teh parts nudging.

in deed it means *every* board is a failure!

[mikeselectricstuff]

And of course pick position becomes increasingly important for smaller parts - you can pick a larger part pretty much anywhere on its surface and the position will be corrected by vision, but as the part size reduces, consistent positioning accuracy of your feeder becomes increasingly critical.
It is possible to fine-tune the pick position over a number of pics by looking at the offsets when visioning, but this only practical when the offset is consistent from part to part.

[bootstrap]

And of course pick position becomes increasingly important for smaller parts - you can pick a larger part pretty much anywhere on its surface and the position will be corrected by vision, but as the part size reduces, consistent positioning accuracy of your feeder becomes increasingly critical.
It is possible to fine-tune the pick position over a number of pics by looking at the offsets when visioning, but this only practical when the offset is consistent from part to part.

That's interesting.  So you're saying reels are designed such that component centers are exactly across from sprocket holes (or some other fixed relationship to the sprocket holes and tape edges)?  I guess that's one way to do it.  Another would be to have a camera on the pickup head (right next to the pickup nozzles).  Since the center of the camera field and centers of the nozzles should be fixed distances, that should assure every component is picked up very close to its center.  But I guess when they design these machines, they try to minimize parts.  These days fairly high-rez cameras are rather cheap, so maybe the tradeoff is different than when most mainstream pick and place machines were built.

This raises another point.  If a pick and place machine did have such a camera, the fixed relationship I just described could be a way to place parts super precisely even on not very precise mechanics.  How?  After the machine picks up the parts off the tape, it moves over an up-looking camera which flashes and takes a photo of the part/parts on the nozzle/nozzles, then slews to where the down-looking camera on the place-head is looking at the pads on the PCB where the part belongs, then centers itself exactly.  Then it only needs to move a known short distance in x and/or y to put the component over the pads.  Obviously this is one extra step for every part (look down at the pads on the PCB), but it should assure perfect placement of every part, every time.  If having zero errors is as important as many people say it is, the machine could also look at the placed parts on the PCB to make sure they are really there (and didn't fall off or fail to be picked up or something).  This step is probably more efficient to perform for the whole PCB after all parts are placed.

About your question "do I feel lucky"?

Well, that's an interesting question!  If I'm paying $10K for high-quality automatic stencil printer and another $10K for an adequate oven, the final question is whether to spend ~$12K or ~$40K for the pick and place machine (for a total of $32K or $60K.

If you look at it your way (as a "gamble"), there are three possibilities:

$32K - neoden4 plus novastar oven and stencil printer
$60K - novastar LE40V plus novastar oven and stencil printer
$72K - neoden4 plus novastar LE40V plus novastar oven and stencil printer

The last would happen if I tried to spend $32K and the neoden4 wasn't up to my requirements.  In other words, that's what happens if I start with the $32K neoden4 option, then had to upgrade.  Actually, maybe I could get $6K or $8K for the "barely used neoden4" if I had to upgrade, which would make the last option more like $65K.

Actually, I left out the "very cheap" option that I have no idea about yet because nobody here has given me any guidance about yet.  That would be a possibility if people can show me quality, reliable stencil printer and oven that aren't so expensive.  I definitely know I'm not going with a "toaster oven", but perhaps there are some quality ovens and stencil printers for $5K each or so.  In which case the cheapest package would be more like $20K~$25K instead of $32K.

Or I could design and build my own pick and place machine.  They are rather fascinating beasts.  But I'd really rather not, because I have other projects I want to complete.  An added irony is, some of those projects would be excellent components for a precision pick-and-place machine (servo controller/driver with encoder inputs and robotics vision system camera at the very least).

-----

One more question.  I assume there is a very significant difference in consequences for different types of assembly errors.  For example, if a part fails to place, a review of the PCB under a stereo microscope will catch that.  Actually, this could and should be done by the pick-and-place machine with the down-looking camera!  Is it?  If not, why not?  If a part is placed so poorly that the review under the microscope will catch it, that error can be fixed manually.

I assume the really huge disaster is when a part is placed in a way or position that smears solder-paste in a manner that causes either a short or open, but is not obvious before the PCB is run through the oven.  What other assembly problems cause PCBs to not function?

-----

Oh, at least some of my PCB designs need to have components on both sides.  Is there anything I need to know about the pick-and-place machine or stencil printer or assembly process?  For example, are the components soldered on one side first, then run back through the stencil-printer, pick-and-place and oven a second time for the other side?  Seems like that's the only way that works... or am I missing something?

[glenenglish]

Mr Bootstrap
you said

" An added irony is, some of those projects would be excellent components for a precision pick-and-place machine (servo controller/driver with encoder inputs and robotics vision system camera at the very least)."

You are clearly inexperienced with all aspects of this stuff- but making wild claims.... especially your belief that servos and closed loop is required. If this is from your own experience, your implementations have been flawed.
 I'd suggest you go and read the following forums end to end, yes all 20 to 40 pages  on eevblog
:
Neoden 4 forum
smallSMT forum
TVM 802B forum

and then go into say the DIY cnc forums and see what those guys are doing with regards to accuracy and precision and what they're using.

then go and consult all the tech manuals of all the mfrs of belts, belt drives, ballscrews, linear guideways.....

and then go and do the numbers on exactly what is required. You'll find if local datums are used and sensible motion practices are used, it is not a difficult job considering that the placement head requires no dynamic accuracy , only static.

You seem fascinated by needing closed loop and servos when the reality is, depending on the choices you make and the speed this may be entirely unnecessary  (or indeed entirely necessary ) if you understand the technology rather than... guessing.

I'm actually suspect of your motives in this forum. Can you let us know if you work for a high end manufacturer?