Author Topic: Questions on small batch PCB assembly - esp in UK  (Read 3779 times)

0 Members and 1 Guest are viewing this topic.

Offline 741Topic starter

  • Frequent Contributor
  • **
  • Posts: 382
  • Country: gb
    • Circuit & PCB Design (small PCB quantities OK)
Questions on small batch PCB assembly - esp in UK
« on: August 29, 2018, 01:20:41 pm »
I have some basic queries on what is involved, fixed cost wise and otherwise. I think I'd have the PCBs made in China, but I'd maybe like the assembly done in the UK.

Here goes...

  • Is is possible to PnP onto Vee-Cut panels? If not, would I typically need to supply a panel with routed/tabbed separation, or can I supply individual boards?
  • Presumably a multi-panel PCB needs a multi-panel stencil - at what point (# boards) would this typically be cheaper than individual boards? (eg size for 1 board: 5x5cm). Would it be say around 2000 PCBs or something similar?
  • Would a multi-panel PCB need repeated PnP data, i.e. per board? Or could the machine re-use the same PnP data for each board within the panel?
  • Would the PCB assembly house insist on their own stencil - eg to fit their equipment?

Suppose the parts cost is excluded, and the parts are 2xQFP64, 30x small passive (0603/0805).
Also, assume 100 pcs.
What would the approximate costs be for
(a) 1-off charges, eg a stencil that can be used on the next run
(b) Set-up costs for placing reels etc, down-time for machinery
(c) Cost per board (excluding setup etc and parts).

Finally, what other things should I be bearing in mind here?

Thanks
Stephjn

Online Ice-Tea

  • Super Contributor
  • ***
  • Posts: 3063
  • Country: be
    • Freelance Hardware Engineer
Re: Questions on small batch PCB assembly - esp in UK
« Reply #1 on: August 29, 2018, 01:48:04 pm »
1. Especially for small volumes, I've found it to be easier/better to supply my stuffing house with a single design gerber and let them handle how and how many to put in a panel.
2. Same remark. This is one of the ways they keep cost down: put multiple designs on a stencil. Ask them where the cutoff point is, will be different for different manus.
3. No. Any stuffer worth his salt should be able to program it in his machine. Better not forget fiducials, though.
4. Probably, especially for low volumes/protos. Same for PCBs, by the way.

a) b) and c) depends on the stuffer. For the same design I had my proto specalist quote 135€ startup cost and then a fixed board price and another 500€ one-offs, 135€ setup cost and a lower cost-per-board.

Again: depends ;)
 
The following users thanked this post: 741

Offline 741Topic starter

  • Frequent Contributor
  • **
  • Posts: 382
  • Country: gb
    • Circuit & PCB Design (small PCB quantities OK)
Re: Questions on small batch PCB assembly - esp in UK
« Reply #2 on: August 29, 2018, 03:16:08 pm »
Thank you.

Quote
...For the same design I had my proto specalist quote 135€ startup cost and then a fixed board price and another 500€ one-offs, 135€ setup cost and a lower cost-per-board...

-->So, first time ever
  • 500€ one-off
  • 135€ startup cost
  • n€ / PCB

Then, another batch:
  • 135€ setup cost
  • (n-x)€ / PCB

For your example - Approximately what quantities, PCB size and 'complexity' (#components etc)?

Thanks
Stephen

Offline SMTech

  • Frequent Contributor
  • **
  • Posts: 845
  • Country: gb
Re: Questions on small batch PCB assembly - esp in UK
« Reply #3 on: August 29, 2018, 03:28:20 pm »
OK as the Surface mount guy in a UK company that does this kind of work..

1)Yes V cuts are fine, although sometimes from China the cuts are too deep and you get a very floppy PCB. It is ALWAYS preferable to at add a border (~10mm) to any PCB you physically  ship/issue yourself. And yes Fiducial are something engineers of all levels forgets and they are quite important, best is to make them a component so their location is already in your centroid data output. Different people will have different size limitations, for smaller batches keeping things small and simple keeps your options open, remember small batches might mean some parts are being stuffed by hand and that is easier on a smaller panel. (No elbows or sleeves smudging something off the board)

2)In our case you could supply a pre-panelised multi-PCB gerber, or ask us to do that for you.  There are two small savings to be made here - a single stencil and a slight reduction in handling time. However it could cost you more the long run if at some point you no longer want some of the PCB's included in the multi-PCB panel as they will be wasted.

3)For setup purposes I would want the centroid data for the each design in the panel they are entered individually. I can then tell the machine the PCB is made up of several sub-assemblies and how they are located. Note this also means if you later want a panel that is only of one of these designs I can reuse the data I have easily.

4) Essentially yes, most would insist; however we would accept the Chinese framed stencils from places such as PCBWay for most designs although it is appreciated if the design is shared with us first. I wouldn't suggest such stencils were a good idea for prolonged use however, many of the ones I see a decidedly wonky to start with. Most EMS houses will use a frameless stencil tensioning system, there are several competing systems out there in several standard sizes and they would prefer to have some input on stencil design (or to take advice from their supplier).

a)A proper stencil foil can be had for <150GBP , some suppliers cost quite a bit more however. Data input (your centroid file, creating any new parts/shapes etc) would also be a cost here and will vary according to complexity. In our case we might offer to amortise some of these cost into the initial batch or absorb them (all depends on the numbers involved).

b) Job Set-up cost would be built into our per unit costing with any savings reflected in price breaks offered. Internally we would be using both the line count and component count to gain an estimate as to how long the job takes to load and run, others might use simulation etc to reach similar conclusions.

c)How long is a piece of string... See part 2 of b. There are other factors here too, recently I built 50 boards with ~600 parts per board over 120 BOM lines, all kitted by the client, several lines were issued with parts from RS (NOT selecting Production packaging is a way to make me really annoyed)  in strips of 5 or even less and added up to several 1000 components, no machine can handle parts supplied this way, this added over a day to the assembly process as well as potential human error.
For your very simple example (line count not known, but guessing it would not be many) the assembly cost would be quite low, workshop rate for like 2-3 hours should be enough time for almost anyone to build something like that. I'm sure you can draw some conclusions from there..
 
The following users thanked this post: Mr. Scram, 741

Online Ice-Tea

  • Super Contributor
  • ***
  • Posts: 3063
  • Country: be
    • Freelance Hardware Engineer
Re: Questions on small batch PCB assembly - esp in UK
« Reply #4 on: August 29, 2018, 06:51:20 pm »
Thank you.

Quote
...For the same design I had my proto specalist quote 135€ startup cost and then a fixed board price and another 500€ one-offs, 135€ setup cost and a lower cost-per-board...

-->So, first time ever
  • 500€ one-off
  • 135€ startup cost
  • n€ / PCB

Then, another batch:
  • 135€ setup cost
  • (n-x)€ / PCB

For your example - Approximately what quantities, PCB size and 'complexity' (#components etc)?

Thanks
Stephen

I explained badly. These are quotes from different companies for the same product and for the first production run.

Offline mikeselectricstuff

  • Super Contributor
  • ***
  • Posts: 13693
  • Country: gb
    • Mike's Electric Stuff
Re: Questions on small batch PCB assembly - esp in UK
« Reply #5 on: August 29, 2018, 07:00:01 pm »
Thank you.

Quote
...For the same design I had my proto specalist quote 135€ startup cost and then a fixed board price and another 500€ one-offs, 135€ setup cost and a lower cost-per-board...

-->So, first time ever
  • 500€ one-off
  • 135€ startup cost
  • n€ / PCB

Then, another batch:
  • 135€ setup cost
  • (n-x)€ / PCB

For your example - Approximately what quantities, PCB size and 'complexity' (#components etc)?

Thanks
Stephen

I explained badly. These are quotes from different companies for the same product and for the first production run.
You will see a lot of difference between companies for small jobs, due to varying attitudes and setup/1-off costs. More so from a new customer they don't know.
If you use one company a lot, and they know you will give them data etc. exactly as they want it, you may find setup costs reduce significantly over time.
It is frustrating that UK places still tend to charge £150-200 for stencils when they are available so cheaply from China. An issue here can be their printers need large fixed sizes.
Youtube channel:Taking wierd stuff apart. Very apart.
Mike's Electric Stuff: High voltage, vintage electronics etc.
Day Job: Mostly LEDs
 

Offline SMTech

  • Frequent Contributor
  • **
  • Posts: 845
  • Country: gb
Re: Questions on small batch PCB assembly - esp in UK
« Reply #6 on: August 30, 2018, 12:56:23 pm »
Quote
It is frustrating that UK places still tend to charge £150-200 for stencils when they are available so cheaply from China. An issue here can be their printers need large fixed sizes.

I don't think you should find it so, industry statistics put the primary process to cause issues for surface mounting is printing, something IRO 80% of all issues happen here. Yes you can get cheap stencils from China, but they are far from excellent, merely "adequate" , it doesn't take many PCBs to justify the premium which isn't exactly a huge number.  Certainly in most cases a pretty irrelevant sum compared to the development time  put into the PCB or even the other costs of parts and labour. Plenty of people spend far more on tooling for enclosures, labels, keypads, packaging and all the other bits and bobs that go into making a finished product, why cut corners on the bit that makes it function? However I think the OP was not really thinking of very small batches or prototypes so moot I guess. However if it does bother you, you can always extract a little value by putting more than one product on it.

Chinese stencil manufacturers tend to use a very basic/low end finish on the foil.
  • They give no feedback whatsoever on your stencil design.
  • They implement no modifications based on a set of customer rules.
  • They don't get approval before making it (because they simply cut to your drawing).
  • Are seemingly obsessed with putting apertures where fiducials are.
  • Only do laser cut & single level - which might not suit the design.
  • In some cases the apertures seem "generous"...
  • The frames are flimsy, often warped and dented (and this matters more if you are manual printing)
  • The frames can be interesting sizes
  • The foils are glued into the frame untidily this can mean there is glue/epoxy in the print area, or the the glue zone on the underside intrudes on the area the PCB will sit making for a poor gasket between the board and foil, so too much paste on the board, shorts on fine pitch, mid-chip solder balls on your passives etc etc.
  • Take 5+ days, our European ones take 24 hours
  • Don't clean as easily (probably related to finish and cutting quality)


 A sub-contract house works to a standard, to reach that standard they stick to certain procedures and trusted suppliers and they have to do that to be able to have things like ISO or VOSN approval. That might not matter to you but it will matter to them and their other clients, sticking to these rules helps keep quality up and failures down, ask them to bypass them and you should expect & accept a corresponding reduction in quality.

As to printers, different models have different capabilities, some might expect the image to be centered, some require fiducial marks on the stencil underside. Some accept smaller frames using rails clamps that move (with limits), some achieve that with an adaptor frame requiring one for every weird size that came your way and I get the impression some only accept the size they were built for which might be 29"..

In our case we implement a basic set of rules, but a proper european sourced foil is always the best option.

V Small batches/prototypes typically 5 or less preferably small - A frameless foil will suffice if costs require it
A larger PCB/batch or a design we expect to repeat either occasionally or already anticipate revisions - Framed Chinese will suffice if costs require it.
A volume production stencil, fine pitch, stable design , meet any of these criteria and its a real stencil foil.
« Last Edit: August 30, 2018, 01:42:06 pm by SMTech »
 

Offline mrpackethead

  • Super Contributor
  • ***
  • Posts: 2845
  • Country: nz
  • D Size Cell
Re: Questions on small batch PCB assembly - esp in UK
« Reply #7 on: August 31, 2018, 12:13:15 am »
    I don't think you should find it so, industry statistics put the primary process to cause issues for surface mounting is printing, something IRO 80% of all issues happen here.
    Quote
    In our line, i would put it at over 90% of defects can be atrributed to bad paste application.     You simply can't ignore it.    this includes, your paste, stencil and application.   

    Quote
    Yes you can get cheap stencils from China, but they are far from excellent, merely "adequate"


    The stencils i'm getting from China, have prooved to be excellent.   The are using a LPKF machine, and electropolishing them.       I am *very* specific about them making them exactly to my specification. No changes to the gerbers! Zip, Nudda, nothing.       They really want to 'help' you, but i have through experience, and experitmentation, worked out apatures sizes that work really well for us.    This means that i can use $30-50 stencils and get awesome results.   

    Quote
    Chinese stencil manufacturers tend to use a very basic/low end finish on the foil.


    Some do. Some dont. Like everthing from china, you can get awesome and crap.    I would'nt be so bold as to say all of them are crap.

    • They give no feedback whatsoever on your stencil design.

    This has not been my expereince.  Some will want to offer lots of feedback.

    • They implement no modifications based on a set of customer rules.

    Thats good.  Send them the stencil design, exactly how you want it made!

    • They don't get approval before making it (because they simply cut to your drawing).
    • Are seemingly obsessed with putting apertures where fiducials are.
    • Only do laser cut & single level - which might not suit the design.

    Sounds like your experiences have been with a manufactuer who doe'snt care, or you've not taken the time to understand what you need to do so, they udnerstand what you need.

    • In some cases the apertures seem "generous"...

    Send them the gerbers EXACTLY how you want them.

    • The frames are flimsy, often warped and dented (and this matters more if you are manual printing)

    I've never experienced this.

    • The frames can be interesting sizes

    Yes, they can be.  A range of frames to suit your application. Is that bad?


    • The foils are glued into the frame untidily this can mean there is glue/epoxy in the print area, or the the glue zone on the underside intrudes on the area the PCB will sit making for a poor gasket between the board and foil, so too much paste on the board, shorts on fine pitch, mid-chip solder balls on your passives etc etc.

    This is just poor manufacturing. Find another manuactuer. Theres plenty of them in china. that do a good job.

    • Take 5+ days, our European ones take 24 hours

      I order pcbs + stencils, ( 4 layers typical ) and often they arrive in 5-6 days.   No point in having a stencil in 24 hours if you dont' have boards to go with it!


    • Don't clean as easily (probably related to finish and cutting quality)

    Again. sounds like you have just used a poor manufactuer.


    [/list]


    > A sub-contract house works to a standard, to reach that standard they stick to certain procedures and trusted suppliers and they have to do that to be able to have things like ISO or VOSN approval. That >might not matter to you but it will matter to them and their other clients, sticking to these rules helps keep quality up and failures down, ask them to bypass them and you should expect & accept a >corresponding reduction in quality.

    I run our own line, for our own products.  I've done a huge amoutn of work to make it work really easily for us.   There is no way in teh world i'd take on contract work, because people just dont' get what it takes to design pcb's for manufacture, or are prepared to pay to get it done.        Apature design makes  a massive difference to overall relaiblity.   All the little things that i do, ( fids, tooling strips,  index holes, alignment markers ( " simple text like.. " This side to front " ).. 


    >As to printers, different models have different capabilities, some might expect the image to be centered, some require fiducial marks on the stencil underside. Some accept smaller frames using rails >clamps that move (with limits), some achieve that with an adaptor frame requiring one for every weird size that came your way and I get the impression some only accept the size they were built for ?>which might be 29"..

    FOr what i do, i tend to use smaller stencils,  and try to keep my panels at a resonable size.      Theres some gains in panelizing, but when you get too big, everythign gets harder.    An A4 page sized panel works well for me as a maximum, but i'll often do somethign smaller as well..

    On a quest to find increasingly complicated ways to blink things
     

    Offline mrpackethead

    • Super Contributor
    • ***
    • Posts: 2845
    • Country: nz
    • D Size Cell
    Re: Questions on small batch PCB assembly - esp in UK
    « Reply #8 on: August 31, 2018, 12:23:47 am »
    • Is is possible to PnP onto Vee-Cut panels? If not, would I typically need to supply a panel with routed/tabbed separation, or can I supply individual boards?

    Absolutely.  Often a combination fo V-cutting and routing works well.  Mouse bites can be a pita to clean up.       


    • Presumably a multi-panel PCB needs a multi-panel stencil - at what point (# boards) would this typically be cheaper than individual boards? (eg size for 1 board: 5x5cm). Would it be say around 2000 PCBs or something similar?

    Yes. you need a stencil that matches the panel.  I never leave panelisation to the PCB manfuactuer.  I always do it myself..  And send them panelised data for both the PCB's and stencils. Theres thigns that the just dont' know about your design that they cnat' design for.     Overhanging connectors are a classic.    You need to support them when they are being reflowed and they cant' sit on top of other parts. A PCB house wont' know this. 




    • Would a multi-panel PCB need repeated PnP data, i.e. per board? Or could the machine re-use the same PnP data for each board within the panel?

    It is normal to just use the PNP data for a single board. You set up the panel as an array of these boards.    When PCBs are being manfuactuered,  it is possible that a single pcb in the panel will fail test.  This is called an 'X-out'.   Accpeting 'Xouts' in your panels saves you lots on PCB costs.  ( if you panel has 16 boards and one fails, they dont' have to chuck out the entire panel ), they just mark the failed panel with a big X.  You need to tell your PNP machine not to load that particular board.     On my line this is actually automated, i have fiducudials that we black out if its a DNP place panel.   If i ordered 100 panels i'd expect to see 7-8 panels with X-outs in them.  Normally just one board.   Its a very quick task for the operator to deal to, well worth the cost savings.




    • Would the PCB assembly house insist on their own stencil - eg to fit their equipment?

    Sometimes.    its such an important part of the process.  You have to work with them.

    On a quest to find increasingly complicated ways to blink things
     

    Offline SMTech

    • Frequent Contributor
    • **
    • Posts: 845
    • Country: gb
    Re: Questions on small batch PCB assembly - esp in UK
    « Reply #9 on: August 31, 2018, 09:20:48 am »
    • Is is possible to PnP onto Vee-Cut panels? If not, would I typically need to supply a panel with routed/tabbed separation, or can I supply individual boards?

    Absolutely.  Often a combination fo V-cutting and routing works well.  Mouse bites can be a pita to clean up.       


    • Presumably a multi-panel PCB needs a multi-panel stencil - at what point (# boards) would this typically be cheaper than individual boards? (eg size for 1 board: 5x5cm). Would it be say around 2000 PCBs or something similar?


    Missed properly answering this bit, noboody really wants to build a 5x5cm panel, in fact on many machines that is as small as the conveyor will go, always panelise small boards. A panel 20CM up about A4 is going to be pretty optimal for most assemblers, at that size it should fit most lines inlcuding superfast dual lane ones that often can't handle larger boards. Mouse bites can be a pain but are often unavoidable, some assembly houses might have fancy de-panlisers that can do all sorts of shapes tidily for you, but you will probably find most have guillotine types and/or a mouse nibbler.

    Every time an automated line pulls in a PCB, it travels to the PCB stop, the machine inspects the fiducials, then it places, it fairly quick but its still time and if all you've got on there is 15 components or something, it could well be more time than it spends placing. On a standalone machine an operator opens the cover physically locates the board on the fixture, closes the door & tells the machine to place, now the machine checks fiducials and places. As you might guess this is now easily in the 30secs-1min bracket when you factor in the operator is standing side the machine like a sprinter on the blocks watching for that exact moment the previous board finishes and the door unlocks.

    Quote
    And send them panelised data for both the PCB's and stencils. There's things that the just dont' know about your design that they cnat' design for.Overhanging connectors are a classic.    You need to support them when they are being reflowed and they cant' sit on top of other parts. A PCB house wont' know this. 


    Well if your assembly house does the paneling for you they should have the gumption to spot this and again this would be a conversation everyone would be involved in, "we've done this is it OK?" etc. Beg to differ on the connector point tho', often they should be routed underneath so that they are not overhanging a neighbouring PCB or border edge, otherwise anyone doing the (naughty) breaking out by snapping along the v-score will break them if they pull upwards. Equally an overhanging device gets in the way of some de-panelling options like guillotines. Something else to bear in mind with connectors is they often don't come in placement friendly shapes and sizes and might need very specific part numbers ordering to suit the manufacturing process and even then some might need a dedicated tool. If you choose an SM connector be sure to have a reason to, or be aware they might be going on with a pair of tweezers.

    There is another clue hidden away in mrpacketheads other reply however, if manufacturing is not something you have direct experience of, take the advice of whoever you choose to works with when it comes to DFM, he is his own manufacturer so has control over everything. From there you should get some feedback on how to tune your later designs. Because we do builds for all sorts of people we see all sorts of things and all sorts of datasets. Tiny clients with perfect data to massive clients who for come reason can't give me Centroid files. BOMs that contain everything we need to know to BOMs that include parts that do not exist e.g "C5 1uf 0805 100V NPO" - good luck with that lolz. Some generate stencil layers, some make us do it, they use different IPC standards (or make up their own) to derive the pad sizes etc. That's why its nice to be able to say to the stencil maker, please crosshatch all the large apertures, apply homebase pads or a 10% reduction, they have tools that do that more quickly and reliably than we can although FAB3000 looks pretty good at that kind of thing and I use that for Gerber to centroid creation. If your builder has no questions or feedback when you first work together, I would be worried.
     

    Offline AndyC_772

    • Super Contributor
    • ***
    • Posts: 4208
    • Country: gb
    • Professional design engineer
      • Cawte Engineering | Reliable Electronics
    Re: Questions on small batch PCB assembly - esp in UK
    « Reply #10 on: August 31, 2018, 09:55:45 am »
    To the OP, all your questions are really best directed at the specific contract manufacturer that you're thinking of using.

    Most can and will get PCBs manufactured in China, though perhaps not at the ridiculously low cost that the sites popular amongst hobbyists somehow manage.

    Nevertheless it's always a good idea to order completed PCB assemblies from one place, simply because you have a single point of blame in the event something goes wrong. If you supply the PCBs, and a completed assembly comes back with connectivity issues because the board was defective, then you have no comeback at all.

    Similarly, if your boards are not usable for whatever reason (panel size, fiducials, V grooves, finish / solderability or whatever), then they're scrap, at your expense.

    Very occasionally I'll consider supplying some free issue components, if there's a reason why I can get hold of them more easily than a CEM. Otherwise I'd always supply a BoM and Gerbers, and let the CEM take care of everything. You may not get the absolute lowest possible price, but you do have someone who is definitively responsible in the event of a problem, and that's worth paying for many times over.

    Online Ice-Tea

    • Super Contributor
    • ***
    • Posts: 3063
    • Country: be
      • Freelance Hardware Engineer
    Re: Questions on small batch PCB assembly - esp in UK
    « Reply #11 on: August 31, 2018, 10:10:44 am »
    Overhanging connectors are a classic.    You need to support them when they are being reflowed and they cant' sit on top of other parts. A PCB house wont' know this. 

    Actually, I've a proto specialist that does know. He catches stuff I miss from time to time. Most of the time I also provide him with a front/back 3D shot which helps but... He knows ;)

    Offline mikeselectricstuff

    • Super Contributor
    • ***
    • Posts: 13693
    • Country: gb
      • Mike's Electric Stuff
    Re: Questions on small batch PCB assembly - esp in UK
    « Reply #12 on: August 31, 2018, 10:22:42 am »
    As regards free-issue vs. contractor procurement, there are a few reasons I tend to free-issue everything :

    Contractor has no motivation to get the best price ( though some may get good pricing anyway due to their volumes).
    They may leave ordering to the last minute for cashflow reasons, risking out-of-stock issues ( especially at the moment)
    Unless given specific digikey etc. order codes there is a risk they will order the wrong part
    They may charge an unreasonable amount for procurement

    This is somewhat contractor-dependent - IME some seem to just not want to get involved with procurement, some prefer to procure themselves. 
    Youtube channel:Taking wierd stuff apart. Very apart.
    Mike's Electric Stuff: High voltage, vintage electronics etc.
    Day Job: Mostly LEDs
     

    Offline mrpackethead

    • Super Contributor
    • ***
    • Posts: 2845
    • Country: nz
    • D Size Cell
    Re: Questions on small batch PCB assembly - esp in UK
    « Reply #13 on: August 31, 2018, 11:31:29 am »
    There is another clue hidden away in mrpacketheads other reply however, if manufacturing is not something you have direct experience of, take the advice of whoever you choose to works with when it comes to DFM, he is his own manufacturer so has control over everything. From there you should get some feedback on how to tune your later designs. Because we do builds for all sorts of people we see all sorts of things and all sorts of datasets. Tiny clients with perfect data to massive clients who for come reason can't give me Centroid files. BOMs that contain everything we need to know to BOMs that include parts that do not exist e.g "C5 1uf 0805 100V NPO" - good luck with that lolz. Some generate stencil layers, some make us do it, they use different IPC standards (or make up their own) to derive the pad sizes etc. That's why its nice to be able to say to the stencil maker, please crosshatch all the large apertures, apply homebase pads or a 10% reduction, they have tools that do that more quickly and reliably than we can although FAB3000 looks pretty good at that kind of thing and I use that for Gerber to centroid creation. If your builder has no questions or feedback when you first work together, I would be worried.

    The best thing i ever did was to stop using Contract Manufacturing.  Its revolutioned how we do things.  We can manage thigns from design to manufacture.   Often a change of product will only involve putting 1 or 2 new feeders on the machines.  and we can do a swap from one product to the next in 10-15 minutes.    Its amazing what you can build from 240 different parts is.   I dont' have to worry about what the CM does or does not need;

    My workflow sees me create two files.  One is a 'Machine_load" file.  It takes the BOM for a board, and tells me, what parts are already on the line,  what parts are in stock ( and were they are to pick them from the shelf ) and what ( if anything ) needs to be ordered to be able to do the job.    The other file that is generated is a native file for the PNP machines.  It does all the top/bttom splits and also because i run across two machines,  the split between the two boards.   

    We are certainly not high volume.  But the money and time i've saved over using CM, has well paid for itself, and its very fesible for me to run just a few panels, if i need / want to.   This means i dont' need to keep the levels of inventory of finished parts we used to, and can get much closer to JIT.


    On a quest to find increasingly complicated ways to blink things
     

    Offline mrpackethead

    • Super Contributor
    • ***
    • Posts: 2845
    • Country: nz
    • D Size Cell
    Re: Questions on small batch PCB assembly - esp in UK
    « Reply #14 on: August 31, 2018, 11:34:41 am »
    With the cost of PNP coming down and down,  its quite fesible for small business's to actually make it work out. ( not hobby level ).   There are projects that realistically are too small for Contract Manufacturing..

    The real cost factor for CM is the time and effort for information transfer to occur ( in both directions ).   A huge amount of  info is needed.
    On a quest to find increasingly complicated ways to blink things
     

    Offline SMTech

    • Frequent Contributor
    • **
    • Posts: 845
    • Country: gb
    Re: Questions on small batch PCB assembly - esp in UK
    « Reply #15 on: August 31, 2018, 12:28:17 pm »
    Regardless of free issue or otherwise I would suggest creating your own internal part numbers for parts you use just as it if you were manufacturing yourself. Then you can keep a list of acceptable/approved alternatives as well, later you can refer to this list to avoid using multiple parts where one might have done e.g using 0.1uF in 3 different but close voltages in the same package. If the line operator is having to cross reference a last minute alternatives you're adding inconvenience and another route to errors.  As an added bonus internal part numbers are often easier to read than some of the gibberish you see on reels of passives.

    If your assembler buys the parts they then have an answer to the exact part you specified not being available  as you can share this list with them, our automotive customers do by default as part of their accreditation.

    If you free issue be sure to understand the requirements the assembler has for kitted parts as a rule they will typically want to be slightly over-issued and tapes should be on reels with headers on, its rare for a machine to be able to pick the first or last component of a strip of tape, at least reliably, some feeders expose multiple pockets at once which also creates wastage.

    If you give an assembler decent data they should be able to purchase correctly, we do the purchasing for most of our customers, partly because even the big ones don't want to buy some of the parts in the manner necessary to ensure stock, or sometime because their system can't understand the concept of them owning components that are in our building not theirs. We always look for best pricing, there's a clear incentive to  do so, we also have automatic discounts of varying scales with RS and Farnell (volume doesn't matter) & credit account with mouser and Digikey, something you may not have. In this current environment I would suggest it is just as likely for a part to suddenly appear back ion stock as it is to disappear.

    If you look at Newbury they have a huge stock of parts you can draw from some of which (last I checked) you can have for free.  To date almost all our clients have been very 0805 focused so for very generic 0805 values it makes sense to use ours. Mixing free issue with assemblers procurement is also an idea, for instance we have a client who specializes in Radio, some of the parts they use come from suppliers we never deal with and are very expensive -they are better placed to source them than we are.

    There preference of an assembler one way or another will depend somewhat on how they work. If they have a dedicated kitting/feeder loading team it is easier for them to have a "kit" come in from a customer, load that onto their feeder stock as a job, possibly on a quick load trolley and when that Job comes around it goes on the line just like that. Meanwhile the previous job can be unloaded at leisure from the trolleys they took off. this is super efficient for machine utilization, less so on people.
    If they are smaller and don't have a huge feeder stock they might instead have fewer feeders and want to minimise componment changeover, no its much easier if all the parts common to sequential jobs stay on the machine/in the feeders and that is harder to do if the client has kitted everything, unless they kit multiple products at once.

    Other things:
    Fiducials
    Fiducials
    oh yeah Fiducials. (not in symmetry at least 2, preferably 3 or more, on the individual pcbs and the panel borders for bad marking as previously mentioned, 5mm+ in from the edge)
    Make pin1/+ve obvious just because your centroid file has a rotation number in it doesn't mean the contractor knows which way round it goes, your CAD, our definition or both may be wrong and not follow the IPC rules (they do exist).
    Put borders on, conveyors need something to hold, so do fingers and fixtures, nothing quite as annoying as a tiny board stuffed edge to edge, or double nightmare, like that and double sided. Equally flex circuits need support and PCB "fingers" should be protected both of which can often  be done with borders otherwise you should be using tooling = $$$.
    Generate an assembly layer if not all parts are on the silkscreen- this can be useful for all sort of reasons including the very strange assembly houses who enter their data without using Centroid files. ( no idea why, but we outsourced something recently and they did their pick and place programming this way, it took days because they "couldn't read that format" [CSV!!???] when it should have taken hours at worst. )

    Use a silkscreen, in low volumes with no golden sample or other data source, building boards with no visual aids at all adds complication, its basically free these days anyway.

    Worth mentioning is the blog from US contractor screaming circuits which discusses the common issues contractors encounter at all stages.
     

    Offline SMTech

    • Frequent Contributor
    • **
    • Posts: 845
    • Country: gb
    Re: Questions on small batch PCB assembly - esp in UK
    « Reply #16 on: August 31, 2018, 12:50:40 pm »
    With the cost of PNP coming down and down,  its quite fesible for small business's to actually make it work out. ( not hobby level ).   There are projects that realistically are too small for Contract Manufacturing..

    The real cost factor for CM is the time and effort for information transfer to occur ( in both directions ).   A huge amount of  info is needed.

    I think this and the previous comment are highly case dependent, almost every large brand subcontracts.

    Some my recent work has been for a client who gets all their SMT done in eastern Europe, they then bring it back to the UK and do the through hole themselves. They free issue all parts, I find it baffling, if I were them I would be building in house the volume would run multiple lines, and yet they don't even have an NPI / prototype assembly line.
    You can be quite a big fish in a niche area and not have the work to justify a line, we produce things for several different niche markets which mean that in the UK the chances are that you have come in close proximity to something made by me on my line, there no way the customers I built them for could justify a line and the staff and the skills to build them in house, they'd be twiddling their thumbs for weeks at a time. It doesn't take much to build 1000's of boards.
    We have our own products in our own niche market, we couldn't build some of those if we subbed out SMT, having the line makes them feasible. Having contract work makes the line feasible..
    When we were shopping for our current placement machine, one of the demo sites we visited had purchased theirs to bring their assembly work fully in house (they already had a little bit so had the skills), they reckoned it would pay for itself in a year, the machine ran all day everyday so it clearly made some sense, however it was also the wrong machine - it was new and at capacity.
    Lastly there is another relationship, where the client just wants a product that does what they want, so in our niche we manufacture bespoke products to spec or our own products with a client label on. The client might have the skills to specify something but not design it, they certainly won't want to be manufacturing too.
     

    Offline mrpackethead

    • Super Contributor
    • ***
    • Posts: 2845
    • Country: nz
    • D Size Cell
    Re: Questions on small batch PCB assembly - esp in UK
    « Reply #17 on: September 01, 2018, 12:32:17 am »
    I think this and the previous comment are highly case dependent, almost every large brand subcontracts.

    almost every large brand subcontracts..  My point was not about large brands.   My point is that there is a huge  information transfer that has to go from design --> manufacture and back again for sucessful manufacturing.    If design and manufacturing are one in the same, then that information transfer does'tn have to happen.    For large projects the transfer gets divided across a large number of devices..  For small projects.. its a big cost item. 

     
    >You can be quite a big fish in a niche area and not have the work to justify a line, we produce things for several different niche markets which mean that in the UK the chances are that you have come in close proximity to something made by me on >my line, there no way the customers I built them for could justify a line and the staff and the skills to build them in house, they'd be twiddling their thumbs for weeks at a time. It doesn't take much to build 1000's of boards.

    That may well have been the case in the past. but things are changing, and the economics of doing it in-house are changing.


    >We have our own products in our own niche market, we couldn't build some of those if we subbed out SMT, having the line makes them feasible. Having contract work makes the line feasible..

    For the same reasons that I became extremely fustrated with contract manufacturing,   I just wou'dnt want to take on CM work from anyone.   The only thing I would potentially look at would be something that we wholely designed, end to end for someone and i guess, that is the same as just making a product and then selling it to somone.
    On a quest to find increasingly complicated ways to blink things
     

    Offline mrpackethead

    • Super Contributor
    • ***
    • Posts: 2845
    • Country: nz
    • D Size Cell
    Re: Questions on small batch PCB assembly - esp in UK
    « Reply #18 on: October 27, 2018, 08:13:11 am »
    1, we do have quality stencil suppliers who check your stencil design, modify your stencil and send it back for you to confirm and even give you some advice on how you can make better stencil, have the ability to make step stencil, and etc. However, due to communication problems or other (sometimes the cost is just too high for detailed communications before making because the customer just didn't pay enough for the service. )  Low cost stands only for one part of the Chinese manufacturers, you can find a lot of cost-effective quality suppliers if you find the right way to source.
    2, It requires a lot of work to make a good stencil especially for those complex boards. For example, our engineering team will study the design of stencil and the critical parts on the boards and decide what stencil it should be and what modification should be made and send a detailed requirement file to the stencil maker, and the stencil maker will also study our requirements and send their advice back on time. For urgent cases, we can get a stencil within 8 hours after confirmed.

    I dont' want *any body* to make *ANY* changes to my stencil designs.  I want them made exactly how i've sent the gerbers.  I dont' need anyone to change anything.  Because i've carefully considered how the stencil should be made,  I know what the parts are, i know what my solder stencil process is like, i know my reflow process.. All things a stencil manufactuer just wont' know when all that is provided is the gerbers.    It does'nt matter if the stencil maker is in Germany, China, USA or Australia,  i just want it made exactly  ( read exactly ) how its made, with NO modificatioins..

    Why do you assume you can do it better than i can, when you dont' have all teh information i have.

    And if its not right, ( but its made how i asked for it ), then i'm happy to accept the problem.
    On a quest to find increasingly complicated ways to blink things
     


    Share me

    Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
    Smf