| General > General Technical Chat |
| Is partial SMT board assembly a practical option? |
| << < (3/4) > >> |
| mikeselectricstuff:
--- Quote from: OwO on May 08, 2020, 02:42:32 pm --- It would be nice if JLC left the unsoldered pads tinned already because then all I need to do is add flux and do the reflow. --- End quote --- In that case just leave the paste pads there ( or maybe use smaller ones, to put anly a small layer of solder from the first reflow. ) One trick that can help is to use a hard flat tool to flatten the tops of the pads so parts don't fall to one side |
| KL27x:
--- Quote from: Siwastaja on May 08, 2020, 09:43:17 am ---Otherwise than that, I don't see the point because hand soldering a lot of SMD parts is slow and expensive. If you have only a few units, and want to hand solder some parts, why not just hand solder everything, and save the expensive assembly setup cost, which dominates in small projects? --- End quote --- Maybe you have a lot of units including only one or two expensive microcontrollers that need to be flashed with proprietary code or bootloader. For Americans, if we have that assembled in China, then we have to pay an additional customs fee for importing these microcontrollers from China. We also have to either handle the logistics of getting preflashed chips to china, or we have to flash the boards after assembly, which this is a significant amount of labor involved in itself. Esp if you have your own chip flashing machine and know the easiest way to solder such ICs (can be fairly trivial). If you send the flashed chips from the US and back, you can file some paperwork to avoid the customs. But you spend on the shipping and you extend the leadtime from 5 weeks to w/e your end is going to take... at the minimum it's the 2 or 3 days of emails from random China-time to bounce back and forth to set that up plus your time to get the stuff there. (And if you have a shipping department to pawn that off to, that's great. Some people don't.) This lead time sometimes matters for your project. And that is more time your money is tied up with nothing to show for it. And who likes paperwork? It can be nice to just order the boards with the rest of the jelly beans on there, to get that ball rolling. Then you have 5 weeks to buy and flash the micros, to get those ducks in a row. And the first batch can be ready hours after the boards arrive. |
| OwO:
I've never heard of hand soldering SMT in production, but I guess it's feasible if your volumes are < 30 a month. The main concern would be soldering defects, as hand soldering is generally worse quality than pnp and reflow. Programming should almost always be done after assembly imo, the logistics of dealing with flashed chips is just a mess. You have to build a test jig for functional testing of the final product anyway, so why not combine flashing into the test jig? |
| KL27x:
--- Quote from: OwO on May 08, 2020, 06:31:45 pm ---I've never heard of hand soldering SMT in production, but I guess it's feasible if your volumes are < 30 a month. The main concern would be soldering defects, as hand soldering is generally worse quality than pnp and reflow. Programming should almost always be done after assembly imo, the logistics of dealing with flashed chips is just a mess. You have to build a test jig for functional testing of the final product anyway, so why not combine flashing into the test jig? --- End quote --- What is 30 a month? Are you soldering 1 IC per day? Or are you spending one day a month to solder these 30 ICs? :-DD (I soldered 300 IC's yesterday, and that was only a few hours of my day; I guess I'm done for the year.) Programming after assembly can be in some cases be just as bad as soldering an IC. Every time you have to handle an individual board, that is costly. When flashing through ICSP, no matter how easy, it still means you handle every board and require feedback... in this case to look at and make a connection, then look at a machine to verify good.* The soldering iron can be just as easy and provide instant feedback. Dependent on the package type, this can go faster or at the very least more easily. When something works 99.999% of the time and it is completely obvious when it works or not as you do it, the throughput can be higher than something what is half the time but only works 95% of the time on the first try. You have to also deal with the 5% fail, and you are constantly taking a wrench out of the machine that could otherwise be cranking on minimal oversight (instead of daydreaming, you will have to deal with a boring and repetitive reality what demands your constant attention like a crying baby). *I have used a CNC machine to flash panels, and that helps a little. But it's costly in time to setup between different panels, and it only helps a little. You still have to clear the machine on a bad flash, either fixing it on the spot or marking the bad pcb for later processing. And you have to stand up and walk over to reload the machine every 5 minutes. With some more motors and robots, maybe that could be improved, but the set up time just gets worse the deeper you go that route. I know the cool kids are installing their micros with preflashed bootloader and mousebites. That is not necessarily cost-effective on every project, either, due to board constraints and panel layout or available I/O's (or learning/executing, i.e. work). And that pre-flashed bootloader costs minimum 10 cents per, as far as I can tell, for small customers. --- Quote ---the logistics of dealing with flashed chips is just a mess --- End quote --- The only thing worse is having to deal with the logistics of pre-flashed chips weeks earlier where they can creep into the lead time. Now physically, dealing with chips does not have to be difficult. They are very tiny, relatively sturdy, very dimensionally consistent, oxide-free, and they're flat. They don't come in weird, arbitrary shapes with burred up edges, some of which are not easy to handle in an automated way, esp with the much greater comparative tolerances in manufacturing. And chips come in a variety of nice packaging to help you out. Dealers choice. Dealing with boards/panels is the hard part. Whether flashing through ICSP or slapping a DIL chip on there in about 20-30 seconds per (including chip and board handling and fluxing from the second you sit down and working at a rate you don't mind while listening to some tunes), you handle the boards. True though, you should have it combined with testing, where that is practical. It's not always. And if the flash time is significant, you're adding this time to what might be the worst, most monotonous, most baby-sittingest part of the process. The more of this process you can do while the chip is still a chip, in a standard, easy-to-deal-with package, where the flashing can be essentially 100% successful without creating additional steps, without requiring human hands or much attention/feedback/wrenches-in-the-gears, the better. |
| OwO:
Realistically at these volumes you won't be having dedicated personnel for it, which really means your engineers are going to be doing the manual labor. 30 a month means under 100 per 3 months, and 100 boards can probably be done in a day or afternoon. One day per 3 months of an engineer's time isn't too bad. How do you do functional tests if you don't handle individual panels/boards? I assume you already have a bed-of-nails or plug-in test setup, so it shouldn't be hard to add two more pins for SWD? You do not need a preflashed bootloader if you simply bring the SWD pins to the testing header or test points. Flashing time shouldn't be a problem if you have enough test jigs and the whole process is automated; the operator only needs to load/unload panels onto the test jigs, and can load more panels without waiting for completion. If you ever do burn-in testing you have to deal with long times on the machine anyway. |
| Navigation |
| Message Index |
| Next page |
| Previous page |