Feasibility of a small run/prototype SMT assembly service

[Alphatronique]

Hi

not forgot that it not need only a pick place

you need of course a pick place whit feeder
good reflow
a wave for trough hole but that may skip it you staff for handle it
laser for make in house soder mask outsource it was to expensive and to long delay
system for clean board after assembly
etc etc so 50K$ was quickly exhausted

i have do myself many small program and tool for make it faster
but it hard for program to find desing issue that will make problem
as example a QFN whit thermal pad on bottom if hole in mask was not right
you may have to mutch solder or to littel that need to check thermal pad dimesion
stencil thickness etc etc and if you provide libraries you may have many chip
whit same dimension and pin cont that may fit but whit different thermal slug dimension
so hard to knot if used take the right one , if not you may end up have ton of rework to do
and lost lot of money , ya if smell the true story here..

but same may apply to many part and software cannot read datasheet

you also have issue of use cut-tape on pick place
and hadeling almost all tape size and tube
and as exemple a 12mm mydata feeder was 2900$ used
so got one of each size will ruin you ... so cut tape was mandatory

think only way was hire army of low wage ppl and hand assembly whit past stencil
for low volume human was more flexible and that require only laser for stencil and good reflow

p.s. i learn that try to hack old machine end-up to cost more that buy right away a god used machine that was old but no need hack except many tool for program it faster ..

Best regard

[digistump]

Certainly Marc, pick and place and a good reflow would both be needed.
We would leave thru hole up to the customer to do themselves, I realize you couldn't do that in your business model, but as we only want to offer a hoobyist service and not a production one I think most hobbyists/prototypers would find it more than acceptable to have to due thru hole components themselves. We're lucky enough to have the laser cutter.

Your experiences certainly make me think there is no viable business in short run PnP, and I may be naive but it seems like a standard part library could over come many issues, and the wrong part/footprint/etc issues would fall more on the hobbyist submitting the design (not that we wouldn't want to help when possible). I think it would be very important to make it clear that the hobbyist using our service was responsible for their design, we certainly wouldn't be offering rework on design mistakes, after all I'm not going to ask OSH Park to remake my boards because I routed them wrong.

By having a standard parts selection cut-tape, feeders, etc issues would be greatly reduced - we'd only be running full reels and we know the costs to buy all the feeders - they sure are expensive aren't they! Luckly most popular parts are 8mm which are the cheapest.

I have no doubt humans making low wages would be more flexible and practical, but humans making fair wages to operate machines is much more appealing and part of what I'd like to sell about it - that and the fact that nothing would be made in China or other low wage countries.

I'll keep that in mind about old machines, I think we're mainly looking at new ones, including the 722FV that's so popular on this form.

Thanks again for the feedback! Please don't take my determination to figure out how this could work as a lack of appreciation for your stories of caution!

[Alphatronique]

Hi Erick

as for the pick place you may got a used Mydata that rock solid for lower price that a 722FV
mydata was so solid and reliable i place fine pitch IC whitout use vision ! until now all
my board was do by use only mechanical centering and i use even 0.65 pitch BGA
and machine stay calibrated after moving and tilted @ 45 deg for pass stair and door way
not try it whit a MDC ...

you may find interesting to knot that i not do any big production ,actually i do
short run of prototype just like you try to make but i do it for major company
instead of home builder ,(have drop home builder it since to way match trouble)
have some "running" design but batch was near away under 100 PCB
so i work to try to make low cost assembly for long time and it not easy

the better idea i may take to you for go whit this and for home designer board was
kind of semi automatic hand a Assembly for low wage staff ...

let me explain

take cheap X,Y laser Galvo on ebay ex: 170876787324 ,small 5mw red laser pointer
and king of cpu like ardunino , banned lase and galvo over a table and use pick place X,Y part location text file supplied by costumer for put laser dot on the right part location on the pcb
ad write on small lcd scree part to place

so staff or you not have to find were place the part ,and that also avoid error
(you will find that most of time REFDEF was not very usable or confusing)
and this will speed up allot the assembly process

you may also use a second laser color for pin 1 marking for polarised part

i see that kind of system for toughole for sale on action for less that metal price
http://www.thebranfordgroup.com/dnn3/Em ... fault.aspx (ok you now knot were i find my stuff)

but build it may cost lees that 1000$ most expensive was galvo + driver for more in X,Y lased dot

and remember pick place was fun but pain to program and debug even if i put lot of
time for make tool for auto-program it still take 3-4 hour to unload \ load reel ,new part programming etc etc each time new job come , so semi auto was the way to go for start new shop on that market ..

Best regard

[digistump]

Marc,

I've followed the posts on your mydata machine, that was quite a task getting it in there - my space for equipment is also my basement, unfortunately it is even less accessible, which would be the appeal of a smaller lighter machine, though I certainly am not ruling out a dedicated building or running 220 to my shop.

Have you worked at all with using X-Y data for the PnP or do you always just program the machine directly/use teaching mode? If you have, what has your experience been when the X-Y data is correct and accurate?

I really like the semi automated assembly idea - unfortunately it is not for me, as I rather not have to hire a staff and deal with all that comes with it. Really cool idea with the galvo though!

I have a few boards getting ready for medium run production, I'm thinking perhaps I'll use the assembly of them to justify the PnP purchase (afterall even if my idea failed it'd be great to have a PnP!) and then see if I can write custom software to make it practical/explore the limitations of it.

Thanks again Marc!

[Alphatronique]

Hi

Mydata have quite easy ASCII input format
i put example on end of post ..

i make small program that take p-cad ascii input and output all file need
BOM in Xcell ,Digi-key uplodable BOM , Pick place , Optical Inspection ,etc etc ..
think that same may done whit eagel or any cad that have documented file format
i also use IPC footprint on can and on pick palce for make it simpler
and my program add 120 deg rotation for fit CAD to machine part rotation

then on mydata all cordinate was relative to fiducial
so load file by FTP then use joystick for move camera over first fiducial
and voila ! all done on my old zevatech will need to offset all cordinate
for match machine 0 and pcb cad 0 togeter that was pain ..

the old tp-9 may slit in half so save some weight ,but newer machine it more harder
i learn hardway hat forced my to send a newer machine to scrap since not able to
fit a doorway .. (ya new one have 1,5" more deep so not fit 33" doorway anymore)
problem is that it hard to find a small machine that work well and reliable
and fit a doorway all that i found was structural frame made so lack of rigidity
or to big so end up to find a way to enter a mydata ..
i past ~2 year and ~10K $ for try to put zevatech reliable so trut me you beter to have
relaible machine at beginning that will save you lot of frustration specially on small
run profit margin was low so machine must work and work good
if you end up touch-up many part or do extensive pcb inspection for find defect
you eat your profit quickly change of machine permit to make 4 X more whit same time frame

Best regard


BXYZ_top_side <<- boar mane
U 0.1N < unit of mesure
C1 95000 144400 199326 95227 << fidicial position
C2 2118 53165 106444 3992
Q1 1
M1 55900 54900 90 CK106JA << X position Y position , rotation ,internal part number
M2 C5 << refDEf of last part
M1 73450 47550 90 CK104JB
M2 C6
M1 73450 44000 270 CJ150BB
M2 C7
M1 83850 34000 180 CK104JB
M2 C8
M1 73450 40000 270 CJ150BB
M2 C9

[Alphatronique]

Hi

i forgot ...

in cad part attribut was good place for store lot on usefull information
that need for auto-program other machine / step

as exemple on my case i use digi-key part number at part creation

then program use it for retrive from digi-key website other related info
price , mfg ,pakaging ,picture for doc , datasheet ,rohs status etc etc
keep local copy of image and datasheet for bom and assembly drawing

so it a wonderfull way to automatize the assembly and production
and it just litel over head to cad guy to just enter digi-key part number
and system do the remain by itself

you may keep some feild to for tape width , inventory location , preferred feeder position
etc etc ..

Best regard

[digistump]

I had a message asking for an update on all of this so I thought I'd provide one for anyone watching this thread:

I had to put this idea on the back burner due to the success of another project - the Digispark: http://www.kickstarter.com/projects/dig ... -dev-board

But this project has provided the capitol for me to setup a SMT assembly line (we are having the Digisparks contract made, but hope to make them in house after the first run as well as many other products when we open our web store). As part of our web store we will be offering custom laser cutting, a PCB pool service, and lots of neat boards, kits, etc - so naturally we hope to add a small run/prototype SMT service to this eventually - so the idea is still alive and well, and once we have our SMT assembly line setup I'll start playing with this idea again and be able to test feasibility in a real environment.

Thanks everyone for all the great info shared in this thread!
Erik

[Alphatronique]

Hi

that good

i will post i coming week a small hack to mydata machine
for remplace computer whit mutch more powerfull pc
that not make machien faster since pc was only for GUI and verry hight level task
but it will fast the webserver ,permit to use solid state drive ,usb mouse / keyboard
(original pc was pentium 100 to 233mhz so it real shame)

so any way i hope you will have great sucess to setup your shop
and if need will happy to help

Best regard

[Alphatronique]

Hi Erick

did you progress whit your idea ?

on my side i begin to open a bit my shop to outside job
i worked on formula that permit online qoute ,that not to complex but wrok well
i do a average of time need for program a part ,and load/unload feeder
then got a price per part based on shop hour rate / part/h that machine actually place
and keep some extra margin for smaller part for permit more inspection rework

you may test it here http://www.alphatronique.com/?page_id=150
i force to put email so i may track and find bog whit formula i use

think that key point was to make programing tool that permit to convert native cad file to machine file format
for now i have done P-cad ,Zuken ,pulsonix (Pads and altium may imported whit zuken so handle by same time)
still have to work for a eagel ULP script for have a good coverage

the good thing whit use of original cad data was keep file format inteligence that gerber lost
so i may extract part position ,position X ,Y and rotation and part name whit footprint name
that make programming much more simpler ...

Best regard