Pico board manufacturing progress

Posted on
  • Good to hear! The boards look great.

    Is the board design finalized now?

  • Is the board design finalized now?

    I hope so! I've got 135 panels of Pico PCBs sitting around at Jaltek :)

    The part outline is in an eagle library at https://github.com/espruino/EspruinoBoar­d/tree/master/Pico

    I've ordered a bunch of the adaptors using that part outline, so in a few days I'll be able to tell you 100% for sure if it's correct or not :)

  • @gordon I'm interested in the process of getting a run of 4000 boards shipped, once the last PICO is out the door can you do us a little write up ? Costs, kickstarter .. etc.

  • Yes, absolutely! I'd like to keep everyone updated a bit more about what's happening right now too, but I just don't have time at the moment!

    I'm currently trying to get an automated test harness working - after that's done I can get Jaltek to run it on all the panels, and hopefully I'll have a bit more time to write everything up.

  • ...thought so... may be not the exact number... but for sure finalized... otherwise @Gordon would not be Gordon ;-)

  • Very nice! How do the pcb board manufacturers create a panel like that where the designer has the ability to "snap" the individual boards off the larger board? PCB manufacturing seems like magic to me.

  • CNC mill, probably the same thing they use to drill the holes.

  • You can either let the manufacturer do the 'panelizing' for you, or you can do it yourself.

    More info from the great EEVBlog
    PART1: https://www.youtube.com/watch?v=VXE_dh38­HjU

    PART2: https://www.youtube.com/watch?v=Uemr8xax­cw0

    PART3: https://www.youtube.com/watch?v=2zGisPMN­stI

    Most use the same drill as they do for through holes and vias etc. That way, you only align the panel once.

    EDIT: Didn't know this forum embedded videos, sorry for wasting so much vertical real estate with this

    Love the PCB for the Pico @Gordon, lovely solder mask too <3

  • Actually it's handy to have the videos in the page anyway. EEVBlog is great, but I don't get how he talks so much - I'm sure he could make his videos 1/4 as long!

    The Pico panel was put together by the PCB manufacturer - it was a bit more tricky because of the castellations... Every castellation needs to be drilled a second time, and then routed (to avoid tearing the copper out of the hole).

    As you can see from the initial image there are 3 'rat bites' - little tabs with small drilled holes - that hold the Pico boards in place. They've got to be strong enough so the board doesn't wobble when running through the pick and place but easy enough to snap out.

    As it happens we got that bit wrong, so each board will have to be cut out manually as snapping it would damage the pads on the end. There's a tool for that though, so it won't take too long.

    I didn't put an ATE connector on the board (described in that last video) though, and I'm regretting it now. I've had to make a test harness that physically moves around the board, which is taking a bit longer than expected!

    We'll see how reliable it is, but it may be that I have to program and test each board manually!

  • but it may be that I have to program and test each board manually!>
    Maybe you could make some kind of JTAG panel with nails that you drop on top, and then use JTAG for boundary scan and even some firmware tests if needed?
    You could then at least test for shorts and such with relatively few operations per panel. Not sure what kind of things you're testing so this suggestion could be stupid...

    Can I ask which manufacturer you used? It looks very high quality.

  • It's Jaltek - they usually do aerospace stuff but with this kind of thing they can keep the production line busy and also get good publicity :)

    The problem with all this stuff is getting the connection - if I could create a panel with 180 perfectly positioned Pogo Pins in, and could then position and press it down reliably with the 18kg of force required (100g per pin) then I might stand a chance.

    Trying to test 3 boards at a time (only 18 pins) seemed like a good plan, but having now tried this it seems like it's not going to be reliable enough, as on the boot pad especially positioning needs to be within at least 0.05".

    It's looking like I'm going to have to go for testing boards individually, which is a bit of a shame. On the plus side, the fact that it's a language interpreter means that you just have to plug it into USB. The firmware is flashed, the device reboots, runs Espruino, and can then check every pin for shorts itself (using the built-in pullups and pulldowns).

  • High quality manufacturer, beauty! (EEVBlog says it all the time 'bjuti', though - as you already pointed out - he says a lot of things during an episode...)

    About testing manually, sounds like this could be a true 'stress test' of the espruino-tools?
    I would love to see a video of how you test the Pico when you get a workflow up and running. Send me a paypal account and I'll give you some beers for it :)

    While on that topic, do you have some sort of setup for donations? I accidentally missed out on the Kickstarter - but I want to be able to contribute to the project.

  • and can then check every pin for shorts itself (using the built-in pullups and pulldowns).

    Very interesting... if this is one of the most common issues, this may be good for the 80/20 rule... (but I don't know the test error profile...) and 80 is not good enough to be called 'quality' and the numbers are too low to do a random sample testing.

    I'm sure Jaltek has means to say how 'perfect' the 'empty' boards (the panels) are before they move on and place / solder the components...

    Since there are now castellations in place, could there not be a different way of testing? ...like a type of socket like in an RJE socket? The positioning is then not that difficult... it is almost self-positioning... On the other hand, for zero-force 'insertion', it would need more mechanics than a plate with pogo-pins requires...

    How many (additional) test points are beyond the castellations?

  • @Gordon do you take into account that there possibly might be some defective boards manufactured; and does the manufacturer refund the amount of defective boards or are there more boards manufactured to make up for the possible loss?

  • Is it possible to flash several boards simultaneously? Make USB fork cable with DO from host to all board, DI from one and hope fo the same timing.
    I think no need to perfom total testing. You can check let say one panel from beginning, middle and end of batch. The probability of short circuit in such production is negligible.

  • @allObjects I think it's going to be extremely rare there are any issues - no board I have tested yet has had any. The bare PCBs are tested for shorts by the manufacturer, and the AOI (automatic visual inspection) will notice any solder bridges. On top of the Jaltek are actually xRaying random STM32s in the batch to check there are no solder bridges under the chip.

    Using the castellations is an interesting idea - although to check for shorts I don't need them at all - literally just 5 wires - gnd, power, USBx2, and BOOT0. BOOT0 is a pain because it's a relatively small area to make contact on (you can see the gold teardrop shape in the picture above).

    @d0773d No, there's no refund or anything. Jaltek will rework any assembly issues, but if the board is broken I just have to write it off. I'm getting around 500 more boards manufactured than I need anyway, so hopefully the yields won't be that bad!

    @user51876 Unfortunately I do have to pre-program every board with Espruino, so each one still has to get connected up - while I'm doing that I might as well run the tests :) I'm not sure if it's possible to connect all the DPs and DMs together - I've just been putting them into a USB hub, but yes - I can flash multiple boards at once. If I could actually connect to all 30 boards on the panel at once I should be able to do them all in one go, but right now I'm sticking with just 3 at a time for the automated tester :)

  • Hi @Gordon. I have a few questions:

    How is the testing going?

    When will you approximately start shipment?

    Will the backers receive an email when their package has been mailed along with the tracking #?

    Does anyone know how long it usually takes to receive a package in the US from the UK?

  • Hi,

    It's going ok - there have been some issues getting the pinned versions of the boards out of the panels though, so I'm getting them much more slowly than the unpinned ones. I've still got around 1000 boards here and we've been making up kits ready to send out.

    The Royal Mail account for postage should be up and running early next week, and when that is done I'll be able to start sending them out (roughly in the order that people backed on KickStarter).

    I'll try and send an e-mail when the items are shipped - whether that contains a tracking number or not depends on whether I can get the tracking number off Royal Mail in a digital format. Also the less valuable packages won't be sent tracked. It's more trouble than it's worth, as many lost/returned packages last time seem to be because the postman couldn't get a signature and took it back to the post office where it wasn't collected.

    Postage to the US should take just over a week I reckon, so you're still well on track to get the boards in April.

  • I have visited PCB manufacturing factory, it needs over forty processes to get a kind of pcb manufactured, i was amazed!!

  • Yes I agree -- very interesting! :)

    I recently picked up this book:


    which was recommended on Hack A Day -- slightly tangential, but focused on China based manufacturing. As someone who spends a fair amount of time in Asia each year, I'm curious to see how easy it would be to setup a (small - say 1000s) production run of a particular board there (say for example an MCU and a WS2812 unit, in an all in one package).

  • IMO - much with any country - the trick is finding a good company to work with.

    I've been using Seeed so far, who do seem pretty competent. They are however not that cheap, and I wouldn't be surprised if when everything is added up I could get boards produced for very similar prices in England.

    I think it says a lot about Seeed that I've been able to get all the Espruino boards produced there without visiting China once. Contrast that to the one company I tried in the UK where I ended up visiting loads, and yet there were still a lot of problems :)

  • These guys?


    Wow, they look serious :) I'll check them out. BTW, that is what alot of the "Hardware Hacker" is about, the search for the right manufacturer in China, and then subsequent quality control....

  • Those are the ones, yes. There have been a few mistakes but they've tried hard to solve them, which is the important thing.

    I could totally recommend trying to use their OPL too - if you can design with the components and part outlines they have, they can assemble the device from their own stock and it's much faster and cheaper.

  • Post a reply
    • Bold
    • Italics
    • Link
    • Image
    • List
    • Quote
    • code
    • Preview

Pico board manufacturing progress

Posted by Avatar for Gordon @Gordon