Hi
Just wondering where to start really, is there a guide somewhere explaining the steps to using a pick and place machine from scratch.
I.E I have a PCB design in eagle, what would the steps be to populate it?
Presumably I have to export a Gerber file from Eagle and then somehow tell the machine which part is in which feeder ?
Does the machine align the part automatically or is there are process required to teach the machine the orientation of the component?
Is there perhaps a beginners guide somewhere and maybe any manuals for the RV machines that might help me get started ?
Thanks
Darren
http://electricstuff.co.uk/forum/index.php?topic=170 manual is linked in this thread. It does a fair job of covering all the steps without someone gong over it again in detail. Having said that you will probably find that on here too.
Orientation is defined in two places -
- The component package type, which tells the machine how the part is stored in the tape
- Against the component reference for a given product, which tells the machine how many 90deg step its needs to rotate from the its start position.
Because these two are linked, adjusting one messes with the other which can be frustrating and is worth bearing in mind.
If you are using Eagle some people on here may be able to provide you with ULP's they have written that skips some of the steps otherwise required, otherwise the machine works using Gerber data and a BOM. Essentially the process is automatic once you have supplied this data.
There is no "Manual Teach" mode as you might find on some other machines, you have to have some data to get started. Having said that the CD does include some software that purports to let you setup a PCB using a scan, I am unaware of anyone who knows how or even IF it works, it is not documented.
I'm not familiar with Eagle - I think there is at least one user here that is.
What the machine needs as a minimum is a list of part type and name, position and rotation values. RV Gerber lets you import the silkscreen gerber to help creating the setup file (TFR) - I don't use it much so can't comment.
The part type refers to a type in the CDF file, which tells the machine the vision parameters (size, vision type, threshold offsets for body colour), height, tool type etc.
The value is used to select which feeder to take the part from.
Feeder allocation in RV setup maps values to feeders
It will save a lot of time if you set up your Eagle library so that all the rotations are consistent and match the RV's "natural" zero orientation, which is is as parts sit in the right-hand feeder bank, i.e. index towards you.
Make sure you define different part types for resistors and capacitors of the same size, as vision parameters may need to be set differently due to the darker body of capacitors
Defining stuff in your CAD package the same as the machine is a good tip, something I have never had control over as a subby.
Not making your package definitions too generic is also a good tip, caps do indeed look quite different to the camera to resistors. Higher value caps can also be quite a bit taller, in 0805 they are as tall as they are wide, whereas pf values tend to be similar in size and colour to resistors. Getting the heights right helps ensure consistent placement pressure, getting the size and colour right helps ensure you get low vision failure rejects.
Component height is also affects vision as it's used to set distance to the camera for best focus.
So do people end up with multiple C0805 in the CDF, one
for each height? I'll have to see how I can capture the
height attribute in Eagle
Thanks for the tip on orientation. I'd already stumbled into
that one as there is no consistent orientation in the Eagle
part libraries which is a PITA.
I'm in no mood for modifying all the Eagle libraries so I'm
going to build a map so each automatic export from Eagle
uses the map to set the rotation rather than the fixed way
that I inherited from Mike's PCAD exporter.
Quote from: Jason on January 19, 2014, 09:22:53 AM
So do people end up with multiple C0805 in the CDF, one
for each height? I'll have to see how I can capture the
height attribute in Eagle
I've not found it necessary to have more than just different CDFs for R's and C's (mostly for vision) , but the vast majority of the cap values I use are the same for each footprint so YMMV. Worst case you can probably get away with just "fat" and "thin" variants.
I do have different CDFs for different height caps. The larger they get the more variety you have too.
For example I have I think two different heights for 0805, but 1206's I think I have 3, 1210's 3 too, and although I only use one 2220, you could end up with quite a few different heights anything from about 1.8mm upto 7 or 8mm for that footprint.
The pain is when you buy a new reel of, say, a 4u7 1210 cap from a different manufacturer and find their one has a different height! Ideally you need to change your BOM, then re-place them in the transfer file, save it and output to job again, though there are some quicker and dirtier ways of achieving the same result.
Thanks for the suggestions but it all sounds a bit complicated. I don't think I'm going to bother after all when the cost of having things done by a contract manufacturer is not that high.
I wouldn't say it was complicated, the basic day to day stuff is pretty straight-forwards and any special cases you only need react to when you encounter them, examples such as varying heights are things that I adjusted when I observed higher fail rates in terms of vision or poorer placement rates. The benefits of in house vs sub-con however will be different for each user, how they value their time, what availability they expect/require etc. As a sub contractor I can acknowledge there are things it is not easy to achieve in that capacity that one might conceivably be able to do in house especially where lead times are very tight.
I think if you were making lots of different little projects in small volumes, it's pretty hard to argue with the benefits of a small cheap pick and place of your own, what you don't know you will pick up fairly quickly and you may gain knowledge that helps with bigger projects you may then sub out. Beyond that it is probably more a question of do you want to be a manufacturer, do you want to expend the time and space a production line requires etc.
In our case I think our subcontract work was a very significant part of why we have surface mount capability, we would have found it much harder to justify solely for our own products. However, having it brings benefits to those products, what is possible and what they cost.
Looking back at your original question, were we too specific?
Breaking the process down to non RV steps would be:
Create Design in Eagle.
Get some PCB's made
Export Pick and Place data from eagle (or Gerbers if its an RV)
Generate a paste layer in Eagle and send that to a stencil manufacturer of some description (they can do this for you but they can and will miss things)
Import data to pick and place machine
Create any new packages required
(define how any new parts are packaged i.e reel/tube/tray)
Setup any panelling information
Put a blank PCB in the machine and setup/check your fiducial information information, on most machines you can also verify the placement co-odinates are correct. (not an RV)
Load your components into your feeders and tell the machine in what lanes these parts are loaded. For machines with dumb feeders, tell the machine where those feeders are on the machine. (on an RV this is where a job is created)
Setup the pick location and height for each feeder lane. (allows for variations in tape thickness and location of device vs index.
Paste your PCB's using your shiny stencil and ideally a stainless squeegee.
Load the board, load the job and run it.
Check the first PCB for any mistakes in loading/orientation/location (you cannot fully trust pick and place data)
Reflow the PCB, for prototypes this might be a toaster oven or small benchtop.
Inspect it, load your THT parts.
Things that can make your life easier :
Some form of printer with a frame or tensioning system for your stencil, and alignment adjustment. (even a DIY hack)
Fiducials on your PCB
Buying your parts in sensible volumes or use RS/Farnells "Production Packaging" option, little tiny strips can be quite annoying especially if you are going to use them more than once.
Put your PCBs in panels with routed or scored 10mm (or at least 5mm) borders, it gives you something to hold onto as well as machine it easier to secure the PCB in the machine. It also means the panel is ready for production in an inline process if needs be.
A proper convection oven or vapor phase batch oven. If you start using things like surface mount switches or large devices then simple reflow systems will start to let you down. The increased thermal mass and of such things means other parts of the board can overheat while you try to get others to solder. This is especially true with lead free where 217C is the melting point you need to reach which can be quite close to the maximum rated temp of certain devices. YMMV here however, some people will insist they are fine without.
Thanks for the info.
I have the PCB and stencil production covered that is pretty straightforward in exporting Gerber files from Eagle. We also have a proper re-flow oven and as a component retailer a plentiful supply of components. Really the gap in knowledge is in setting-up and using a Pick and Place machine, having never used one.
Basically the situation is that many new IC's are understandably being produced in small pitch SMT packages because that is what the demand is for. The problem is that our customers who are mainly electronics enthusiasts building projects at home find them difficult to solder and obviously prototyping on a breadboard is not possible. So we were hoping to produce small breakout boards that have the main IC and any essential support components such as resonators, capacitors, pull-up resistors to make home experimenting easier for our custoemrs.
Ideally running our own machine would allow us to run small batches of perhaps 100 boards at a time that would be uneconomical for contract manufacturing. Most of the passive components would be the same for all boards and could stay in the machine, in most cases it would be just the main IC that would change.
Having investigated a bit further I am considering a slightly newer MyData machine that has quite a nice feeder arrangement, anyone have any experience of the MyData systems ?
Sounds like a good idea, bit like these https://github.com/maugsburger featured on EEVBlog, but populated?
I am not familiar with them first hand but they were certainly on the cards when we were looking at machines, so I did look at them quite closely but only had a full in person demo of a MY100. I'm guessing you are looking at perhaps an MY9 and the feeder arrangement you are liking is 'Agilis' system? There are lots and lots of people out there with MyData's so there is probably one near you to have a look at if you ask someone nicely. ~2 years ago I could get an older TP11 series with a healthy complement of TM feeders for a little over £22k, so they are way more expensive than an RV.
IMHO
Pro's :
Agilis gives you great, low changeover times (which you don't need)
Built in Vacuum Pump so no need for noisy compressor and dryer
Lots of them out there ( I can see 22 listed for sale right now)
You can swap feeders on and off during a job if you need more than you can fit on your machine at a time
They seem to retain value better than many other brands
Easy to get more feeders (especially the pre Agilis 'TM' ones)
You can update a TP machine to take Agilis feeders (but it's £££)
Genuine MyData expertise available outside of MyData
Cons:
Agilis wastes parts because it uncovers more than one pocket (this is only bad if you do lots of changeovers and value those parts)
If you need a lot of feeders loaded at once you need a big model as all feeders are on one side.
Without a HYDRA head they are quite slow as the head has a long way up and down the machine to travel
Split axis - more to go grong?
Official support is expensive
Official support on 'TP' series is only available at a base level SLA
Agilis 'elements' are really quite expensive considering they are a piece of machined metal with a chip and some contacts.
Lots of different vision system options to choose between
There are mechanical centering ones out there which it is probably wise to avoid
Its not a shiny point and click windows world from start to finish.
From the type of work you describe you won't spend much time setting up and perhaps most of your setup time will be spend defining new IC packages, it also sounds like a job for the smallest machine you can find http://www.autotronik-smt.de/products/bs281.htm or a Dima ATOZ perhaps?
Plenty of us sub contractors are quite happy to do 100 PCB's, 4 panels of 25 small pcbs, or even one PCB. It doesn;t sound like complex or time consuming work I would certainly recommend getting a goodly number of quotes before spending a fortune. Those that want the work will probably quote similar numbers, the sums after all are all the same.
Thanks for taking time to reply, much appreciated.
Regarding boards they are more like small ready made modules like these examples from Adafruit MCP4725 Breakout Board (http://www.tandyonline.co.uk/mcp4725-breakout-board.html) or this MPL115A2 Sensor (http://www.tandyonline.co.uk/mpl115a2-i2c-barometric-pressure-temperature-sensor.html). Most boards will just have a decoupling capacitor and pull-up resistors maybe the occasional LED or resonator. As the values will be pretty common across all boards most components can stay on the machine all the time.
Having said that once we get things up and running we would also like to offer our kits part assembled, take this for example http://www.tandyonline.co.uk/electronics/kits/multiface-kit.html it has quite a few through-hole resistors and capacitors and LEDs that we could replace with SMT versions so that the board comes part assembled and the end user then just needs to solder in IC sockets and pin headers for more instant gratification.
Handy, a 10*10 matrix of either of the first two would fit in an RV quite nicely I would think and even that would finish placement in 30 mins or less.
You missed a bit...
Quote
Setup the pick location and height for each feeder lane. (allows for variations in tape thickness and location of device vs index.
Before pasting, do a dummy run on any newly defined parts to get the vision parameters right, and to check general alignment, using double-sided sticky tape on the bare PCB
Quote
Things that can make your life easier :
Some form of printer with a frame or tensioning system for your stencil, and alignment adjustment. (even a DIY hack)
Tensioning, yes. Adjustment - depends. I use the Eurocircuits printer that has fixed alignment pegs - works well enough that I've never found the need for adjustment
On a std clamshell printer the adjustment is the alignment/registration method, with the eurocircuit one its the pegs, and even it has adjustment to ensure the stencil is parallel to the board from what I see. If you have neither system then you end up using blind luck judgement and -- sticky tape?
Going through a list of project ideas one has a 28-pin microcontroller that comes in SPDIP/SOIC/SSOP could an RV handle that, presumably it wouldn't manage 44-pin TQFP or QFN ?
Fine pitch stuff can be hit & miss - on a good day & following wind I've done 64 pin 0.5mm pitch QFPs and QFNs with minimal manual tweaking, but it doesn't always work out & sometimes needs quite a bit of post-placement nudgery.
I'd guess it's down to the right coincidence of calibration, general mechanical repeatability and/or being close to the limit of the vision system. 0.65mm QFPs and SSOPs are generally not too bad, most of the time.
Other users here can probably advise better on what to expect.
I think the RV, especially the RV4, still represents the best price/performance you can get at the lowest end, and is at least as useable as anything you'll get under £10K. Small size and no need for air is also a plus.
If you're debating whether to get into assembly, An RV is a pretty low-risk way to test the water, and will give you you enough experience to know if you need something better, or if in-house assembly's not for you.
Incidentally, on the subject of low-end machines, the Mechantronika looks quite promising, and probably the closest thing to a modern RV-class machine. I think I've seen a price of 20KEuro but not sure if that was with or without feeders
http://www.mechatronika.com.pl/content/view/21/4/lang,en/
Anyone have any experience of these?
Tandy - are you in Headington? I'm only in Kidlington. If you want to see an RV working you could pop round when I'm running some boards. I've got a 11k part build to do probably next week if thats any good.
Heh I think I pretty much just echo Mike with some of this.
The Mechantronika machines do look interesting, I haven't seen much comment on them other than other people asking the same question, the disty would almost certainly hook you up with an installation or two tho'. The main RV contemporary is the Quad 4c, the faults and fixes for which seem to involve bizzare codes in multiple files and rom chips all over the machine, fun.
I used to place such things with our RV it is certainly more than doable. When you get down to things like 0.5mm pitch , you may find you need to give some things a little nudge with some tweezers before reflow and you may also need to adjust the placement location to compensate for drift in the RV. It could well depend on how lucky you are with your RV. The same would be true of many older budget placement machines, they mostly lack any encoder for precision work. Where I used to find consistent issues was with much higher pin count devices, with those a little excessive twist could result in a lot of tweaking and a lot of shorts. At 0.5mm you really want to get your pasting right as well as the stencil and footprint design, always check the default libraries they can be utter tripe. You will find QFN's more forgiving than QFP's
What an RV does not do :
Rotations in steps other than 90 degrees
Panels of mixed PCB's (you have to treat the panel itself as a single board)
Large PCB's - you should find a diagram in the manual)
Devices under 0402 (Personally I would not like to have tried 0402 either but I was spared that)
Take into account device height when choosing its placement order
Use bad mark or badcheck fiducials
Placement simulation or manual teach
Maintain any form of part number database
What it struggles with:
Fine pitch devices
Tall wobbly devices (see Phono's query about Electrolytics)
Inconsistent lighting
Really accurate placing where you have parts close together
Large devices in tubes. (this isn't an RV fault, vibratory feeders are crap no matter who makes them)
You can place parts at angles other than 90 degree multiples. I build some boards (feeder controllers for Vivo/Intelligent Drives machine actually) that has a LIF connector at some odd angle, 25 degrees I think:
Tricksy! how did you do it? I only ever used RvGerber and at the time it never proved to be a common requirement.
Quote from: Gopher on January 30, 2014, 10:20:17 PM
What an RV does not do :
Rotations in steps other than 90 degrees
I beg to differ
Rephrase it is! RVGerber doesn't. You can add extra rotation in RVPlace, although I suspect that is not how these were acheived.
You can do it in RVgerber. Place the part as normal. Hitting space will of course rotate it in multiples of 90 degrees, but if you select the placed part and right click you can them enter the angle in tenths of a degree:
Hah! Now who feels dumb, was that documented? It sound like something I should have found by mistake at least once.
Is hardly anything useful about the machine documented??!
Hey if you've never needed it, you've never needed to find out how to do it!
I needed it in in 2012, the year we got rid of it.
Oh and I think it was suggested for a product that never happened in like 2004. I don't do circuit design but I understand layout programs tend to hide custom rotations deep in convoluted menus so they don't get used much. If you intended to flow-wave surface mount however it is in important feature, as IC's meeting a wave head on turn into a bucket of shorts.
Ah yes documentation and software completion, IIRC one of the buttons in that menu you show right there does not do anything.
For the record I will point out our £100k Essemtec machine has buttons in the software that when pressed say " This feature has not yet been implemented" but they have been releasing a version or 2 a year and I am at least 2 behind the current release, its quite possible they now do things.
There are a few other machines that may be affordable, a Dima Optimat has been on offer from quite a few brokers for quite some time it is quite possibly always the same one. It may be that by now a lower offer would be accepted, we were offered one on a must go quickly price with a pretty hefty collection of feeders for similar money to an old MyData. Dima machines don't seem to be very popular over here but they seem capable enough and I seem to recall the current Dima distributor saying their guys had been getting some training on the older kit.
You will also notice a few Contact 3's, I think the manufacturer website lists these as having limited available spares, they do seem fairly common.
Shawline also currently lists an Essemtec FLX, this is a popular Essemtec model and very compact, it probably depends on what feeders they can offer with it, feeders from Essemtec are not cheap and you will struggle to find 2nd hand ones. Sites I visited liked their FLX's but it is worth noting that the site that placed a lot of 0402 did say there was quite a bit of tweaking going on using that platform (they were I think pretty large heavily populated boards). The FLX was their high end model until they introduced the Paraquda and Cobra, lower end models take different feeders (slower and less accurate) and start dropping down the drive system quality (encoders, servos, motors etc).
of course if you do direct creation of the transfer file from your cad package using the PCAD or Eagle exporters then it's a whole heap quicker to do odd rotates.
Quote from: Jason on February 05, 2014, 07:50:19 AM
of course if you do direct creation of the transfer file from your cad package using the PCAD or Eagle exporters then it's a whole heap quicker to do odd rotates.
Absolutely - if you're running your own P&P for your own designs, so you have full control of libraries, part names etc. it reduces setup time to little more than loading feeders. If you already have parts in the feeders, you can be placing a new design in a few minutes.