Prototype fab assembly or hand assembly?

[eecook]

Hi Folks,

I am going through a predicament here. I started doing independent design work and now I am faced with manufacturing 5 prototypes for a client. In terms of both cost and reliability, I am wondering if it would be more convenient to have the boards assembled by a manufacturer or by myself (considering I don't have a reflow oven).
The specs of the board are the followng...
Dimensions: 5.55in x 7.25in
Copper Layers: 4
Copper weight: 2oz.
Number of holes: 1224
Min trace width/spacing: 8/8 (mil)
Total Components: 387
THT Components: 25
Unique components: 135
Board has SMD components on bottom layer as well

My budget for Assembly is around 1k USD, what concerns me the most is reliability, since I've never had assembly done for a prototype with anyone beyond Screaming Circuits (which is completely out of my budget) and don't know anything about these other guys.
I am feeling a little bit insecure about this, any advice from more experienced engineers?

Thanks,

Simon.

[andyturk]

If you have to deliver 5, you'll probably want to build more than that. The extra time could be significant if you're building them by hand.

You could look into an online service like Tempo Automation, MacroFab or PCB:NG do the boards and assemble the SMD parts for you. Through hole is probably harder for them to do cheaply, but you could do that yourself as a second step.

I think a $1k budget is probably too low for professional assembly in the U.S. (if that's where you are).

[eecook]

Thank you Andy. I actually have to deliver 3 boards, so I figured 5 is reasonable. I am still waiting on some qoutes from 7PCB, Screaming Circuits, PCBCart and OurPCB. I could probably stretch the budget to 2k, but not more than that.

[Siwastaja]

I think the complexity and numbers sound like you are quite on the brink here, so the answer is: maybe.

I would consider doing them by hand, but only if you are fairly experienced in doing that. As you have a paying customer and a budget, having them professionally done would leave you more time to do what you do best (need to write some firmware for the thing? That's doable when waiting for the boards.)

I guess the Cheap Chinese assembly houses would be around $1000 for PCB manufacture + assembly. Parts come on the top, and if you let them do the sourcing, it's a bit more than what you pay on Digikey. Expect +30% to +50% on components.

Personally I have used PCBway for assembling small batches and have been happy so far. Not perfect, but no horror stories either - in fact, given the complexity of my designs (quite many BOM items), and low price for the assembly, I'm surprised at the low level of human error happening.

But also do note that building your CAD files so that you can really communicate with the assembly house almost 100% certainly requires some significant extra work, especially if you have never outsourced assembly before. OTOH, this is a useful skill if you ever consider doing that again.

Also, if there is any chance this will be manufactured again, in bigger numbers, starting with an assembly house from the first proto batch on gives you a head start.

Reliability will of course be better than hand-soldered.

[ajb]

You're definitely right on the cusp of where hand assembly becomes a pain.  If you have BGAs or a lot of QFNs or parts that are otherwise challenging to place and solder, I'd err on the side of automated assembly. 

At such low quantities a CM will certainly increase their markup for having to buy extra components and/or deal with short tapes, and if you do a consigned build there's the labor cost of building the kit.

If I were in your position I would probably go with hand assembly, but we have a decent reflow oven and stenciling setup and I have a tech who can reliably hand place anything we design (down to QFNs and 0402 currently).  As a bonus, that tech actually enjoys hand placing, so I don't even feel bad about handing off that work--if I had to place that many parts myself I'd go insane!

[jbb]

OK, the SMT parts could be tricky with 2oz, 4 layer board. For reliability I would strongly recommend use of a paste stencil and reflow oven.

What SMT parts do you have on the bottom?

The TH parts can quite possibly be hand soldered. But many assembly houses are good at that anyway.

If you’re getting quotes out of Asia, the components may well be hand placed with tweezers. Programming a pick and place machine is actually a lot of work, so giving a technician the PCB soca can be quite attractive...

[XYZVector]

You will have to hand assemble these things with a 1K budget. Now don't fret there is a way to do this without going insane. You can get a manual pick and place machine, there is a cheap one
pre built for 500.00 bucks floating out there or you can build one yourself. I built my own manual machine and it does make it a lot easier. Reflow oven they have the cheap Amazon ones for 150.00 which if you go that route make sure you have a fire extinguisher and do the re flow outside. If you spend a bit more you can get a more professional oven on amazon. I built my own out of a toaster oven and it was not to bad. Your biggest problem is going to be the double side components. You might as well forget about those unless you can dab some holding paste on those chips to keep them in place until you can get the board into a reflow oven. USE A SOLDER PASTE STENCIL!!! This is doable but the double sided SMD components will make your life difficult. For the prototype try to design those out if you can, and then this becomes a problem of just building a manual place http://trains.socha.com/2014/07/building-manual-pick-and-place-machine.html machine, and getting a reflow oven. You can go as cheap as a toaster oven, and it works quite well. Or you can go with the cheap amazon re flow ovens that work a smidge better. Make sure you double up on boards, your first two will probably be scrap learning the process. However my first home re flowed board worked a treat. I was nervous but looking back it's actually quite easy. You have a big bom so take your time and remember solder paste gives you about 3 hrs to place before it becomes garbage. Good luck!

[ajb]

Doing top- and bottom-loaded boards isn't that hard.  You place and reflow one side, then place and reflow the other.  The solder on the first side should have enough adhesion to hold all but the heaviest components in place even while upside down while the second side is reflowed.  It's mostly a matter of choosing which side to do first and then having a way of supporting the board properly in the oven.  There are rules of thumb for component weight vs solder pad size that will give you an idea of whether or not a component on the first side is liable to fall off -- 44g/in^2 according to this.  I've actually exceeded that by about 30% in the past with no trouble, but YMMV.  If you have heavy inductors on both sides or something then you may need some glue, but this can often be applied after the first reflow.

[bson]

If it's just a small number of simple components on one side, like a bunch of decoupling caps, then just paste, place, and hot air reflow those manually.  Especially if it's 0402 or bigger where pasting and manual placement is easy.  Although if you get a stencil it will probably have both sides on it.  (Just not worth securing and dealing with when you can quickly paste pads manually.)

[XYZVector]

Your right I didn't even think of this, I produce a grill controller with a ESP32 and once I place it on the board it is a real bear to move it even before re flowing. I have no real need to do a double sided. However you bring up an excellent point. Most components are light in comparison with the surface tension of solder, or solder paste. So if you support your board correctly in the oven parts shouldn't fall off. I'm now curious I think I will try double sided placement on my next grill controller model. I can half the size of it with double sided placement, and it would be way easier to route. It's just three extra steps.

-Thank you

[KL27x]

I have a bit of experience hand-assembling pcb with a lot of small SMD parts. I would think it depends on your time and other workload and what skills and equipment you have. And what you like to do. Some people will not enjoy this process, at all.

If you have the time and desire to slog through it and learn as you go, a $1000 budget could buy a stereo microscope, a pickup tool, a hot air station, an automated reflow oven, years worth of flux and paste, and a couple extra boards to screw up and learn on. Time is the unavoidable factor, because you will need a fair bit of it to learn what you can and can't do (and end up with a working board). Heck, it can take weeks to even get a stereomicroscope workstation setup, comfortably. 

If you have hard deadlines, you could shelve the idea. But maybe buy an extra board and components to try your hand at. You might find, in the process, that some of your component layout/orientation and selection, thermal pad design/strategy, and other random things could be better tweaked to help you assemble your pcbs with your particular methods and equipment in the future.

Honestly, in the end, I personally feel like the stereomicroscope and a DMM are the most important physical tools for this task. And a scope, perhaps, depending on the board. Along with some debugging skills (which you probably are the best at, on your own board), there's pretty much nothing you can't fix as long as you can actually see what you're doing with good lighting and magnification. But you can easily underestimate the amount of sheer space and organizational energy that you will need to dedicate to your workspace in order to assemble and test/debug a pcb of any complexity.

[Siwastaja]

You will have to hand assemble these things with a 1K budget.

Not necessarily true. Many Chinese PCB assemblers will do this for under $1K. Use pcbshopper.com to compare prices. You'll find that assembly kind of "starts" at around $500, but given this is quite a large design, it's closer to the $1000. Note that in reality, the price will be a bit higher due to the need to buy excess components compared to hand soldering, and then there are either extra fees for sourcing, or extra shipping fees if you source the components yourself and ship them to China. That's why it gets near $1K, but it should be enough. Not much leeway though.

(I'm assuming the $1K budget is for PCB&assembly only, and doesn't include the basic component cost, because that's what you need to buy anyway, and we have no idea how expensive the components are.)

Budget is (barely) there, but IMO the decision depends more on:
* Do you have experience on producing the necessary files, instructions, and communicating with PCBA fabs? Do you want to invest your time to learn this for future projects? If no, do it yourself.
* Is there any chance this is going to be manufactured in bigger numbers (even just 10-20) later? Then buy the assembly service now.

If it's really a one-off, and you think you're going to do similar small one-offs, similarly with limited budget (or even more limited), it's probably a better investment to get better reflow equipment and get better at it.

On the other hand, if it seems like you could have small budget increases, and in the future need to do 5-10 pcb batches with $2000 budget instead of $1000, then learn how to go the Chinese assembly route and use your precious time in design work & lower level prototyping - you are going to solder things up anyway!