Feasibility of a small run/prototype SMT assembly service

Community PCB Design
1

I’m a big fan and long time reader and shopper at sparkfun, but first time poster - I thought this might be a good place to get some feedback and gauge some interest on an idea that I’ve been working with some folks on for a while. Please let me know if this is the wrong forum to post it in, I tried to find the best fit. Also this is cross posted on to http://forums.adafruit.com/viewtopic.php?f=42&t=31465 - as I’m also an avid reader of that forum and hope to get as much feedback as possible. There is already a response over there and some further explanations, so if this post interests you feel free to respond either place. Thanks!

I think anyone making a SMT based prototype has wondered why assembly can be so expensive, ok maybe not everyone, but I hate doing my own SMT soldering, even with all the tricks and good equipment. Obviously automated assembly for a few units is so expensive because of the setup costs - just looking at all the trials and tribulations with pick and place machines on this forum is a good indicator.

This got me to thinking what would it take to provide the SMT assembly service equivalent of OSH Park (formerly dorkbotpdx) or BatchPCB besides a really patient operator?

After some serious thinking, researching, talking to vendors and operators, and general pricing and exploration of all the details here is what we’ve arrived at:

PCB fabrication, solder paste application, SMT assembly (a combination of PnP and manual), and reflow soldering for $10/sq in(plus parts) for 2 boards (so a 1x1 board would be $10 for two copies plus parts).

Obviously this price blows existing prototype assembly companies out of the water so here are limitations we’ve arrived at:

Eagle and KiCad libraries would be provided for download of 50ish+ most common parts (determined by vote?) sorta like Jellybeans - Many resistors, caps, a few voltage regulators, maybe a few ICs - these parts would be preloaded on a pick and place reducing setup time. These would be priced at digikey or better price.

A variety of popular but lesser used parts placed for free. Again these would be priced at digikey or better price.

Custom parts for $1 per part - you send in the parts we don’t start your order or charge you until the part is received.

Any through hole components would be left for the buyer to assemble.

Discounts once the total of your copies exceed around 200 sq inches (we’d have a simple calculator to tell you how many copies of your board this is).

0603+ and no fine pitch/BGA to start - but all equipment would support it.

Here are our feature ideas, you know beyond SMT assembly!

Website that accepts Eagle Board files, or Gerber+Drill+Centroid (OSH Park is our inspiration here of how a PCB order should work)

Website shows preview and totals part costs, allows you to define any custom parts to be sent in.

Live status on website and turn around estimate based on current queue.

All USA based from PCB to assembly (assembly here in Washington state).

All order shipped out priority mail ($5 per order regardless of number).

Open, transparent, documented process and business (we think it’d be great to have competitors!)

Extra raw PCBs (we’d order 3 at least to ensure we have 2 good ones) would be thrown in the order, maybe random rejected parts from the PnP as well

We come from backgrounds in programming, pcb design, and industrial equipment operations. We’ve put a huge amount of research into this, including working up custom options with Pick and Place manufacturers, PCB fabricators, etc and have the capital, space, multiple 220v hookups, and passion to make SMT assembly into a truly accessible service - we’re thinking about using kickstarter to gauge interest/reserve pre-orders.

Please by all means - rip apart our idea, give any feedback you have, or any suggestions!

Thanks!

Erik Kettenburg

2

I think you’ll struggle with the setup and component handling.

Existing SMT assembly companies are already competitive on reasonable volumes, where the setup costs can be amortised over hundreds, thousands or millions of units. The niche you’re looking to fill is for small volumes, where the time each board spends on the machine is tiny compared to the time it takes to set up and program.

SMT machines require components to be supplied on in tubes or on reels, and that’s not going to happen if the customer only supplies two or three parts. For $1 per part you’ll need to have someone load a part into a tube or tray, install it on the machine, and configure the machine with the location and type of the component.

Any placement errors will be costly too. On a run of 1000 boards, the first one can go in the bin once all the errors have been spotted and fixed. Not so on a run of 2 PCBs; that QFP that got placed 90 degrees out has to be recovered, cleaned, legs straightened, and put back in the tube to try again.

How would you apply the solder paste to the board? The cost of a stencil will be prohibitive for a very small run.

For me, the value of a commercial service like this would be to allow me to use components that simply can’t be soldered by hand, ie. BGAs and QFNs. The rest I can solder by hand, and that might be your best bet too: don’t bother with the machine, hire an operator to place the components and solder them manually. By the time you’ve taken all the setup and programming time into account, as well as having worked out what to do with all the parts that arrive loose in bags and boxes, it’s probably quicker to put the chips straight on the board by hand. An operator who can solder is cheaper than a technician who can program and maintain a pick & place machine too.

If you can find a way to make it work then I think it would be a great service, but every manufacturer I’ve ever worked with makes their money on volume production, which is consistent, repeatable and keeps the utilisation of the machine nice and high. Setup and prototyping has high overheads and is therefore costly, and I don’t think that keeping reels of 10k resistors and 100n caps permanently fitted on the machine is going to solve that problem. Sorry :frowning:

3

Andy,

Thanks for the feedback!

The idea is that no components would be placed by the SMT machine besides the standard 50+ components found in the common library. Given this would be hugely limiting but I think it would provide enough basic building blocks for most hobbyists - my boards seem to have about 12 parts common to all of them and those parts cover about 95% of the parts used. The $1 (or whatever the price worked out to be) per part was for hand placement of those parts that the customer needs to send in. The 50+ common parts would have a library (Eagle and maybe KiCad) for the customer to use in their design, this would ensure common footprints. Evolved from the adafruit discussion - I’m wondering if accepting Eagle Board files and using a custom ULP to get the XY data for those parts might alleviate some of the placing errors (plugins for KiCad and others to be developed if it worked).

Solder paste could go stencil (we have a laser cutter and mylar is not too expensive) or paste dispensing (though I fear that isn’t accurate enough).

Placement errors certainly could be the death of a project like this - for it to work we’d have to be able to reasonably accurately calibrate and run the machine off a combo of XY data, vision, and fiducials (which would be at the corners of each panel). That is the big one, admittedly, whether we can write custom software and get an accurate enough machine to accomplish that.

I currently solder all my SMTs by hand, including 0402s and fine pitch, but it sure would speed up my development process if I didn’t have to hand solder all my prototypes, and that’s where this idea comes from.

Thanks again for the feedback, I certainly here that this would be quite a task to accomplish - I’m thinking I might use production of some of my boards to justify the PnP purchase and go from there trying to make something like this work.

4

I honestly have no idea what a pick & place machine costs, or how many boards you could have assembled commercially for that much. I’m sure you’ve done the calculations.

I think the overheads will be the biggest issue you face. For your $1, you need to have someone identify the correct part from a box of bits, open its packaging, remove the part, check its number and orientation, find the corresponding place on the board, and place the component correctly into the solder paste. I think that’s a lot of time spent for a small amount of money in return - and if someone sends you 2 parts and you drop and lose one, that’s a problem. If you were making 100 boards at a time it would be much less of an issue.

I’m not sure what you gain by insisting on common footprints, especially for relatively undemanding parts like SOICs or 0603s. I design products commercially, and my footprints come either from my own library (which I draw myself from manufacturers’ data) or from the library of a PCB design contractor. Maybe it’s less of an issue for hobbyists, but if you can’t place parts on the same footprints I’d be using for volume production, then I couldn’t use your service. I use Orcad PCB Designer (a variant of Allegro), so using your footprints isn’t an option for me.

As you’ve mentioned yourself, you’d need a very good 1st time placement rate to make it work. Every panel you do would be a 1st-off, all the X-Y positions and orientations would have to be right, the stencil alignment would have to be spot on. I wouldn’t envy you.

5

Hey Erik,

Have you guys worked any more on this idea? I think it’s interesting and I can certainly see the value for a few places:

  • Microcontroller placement . . . even experienced folks have trouble placing QFNs.

  • High-volume parts . . . . thinking of some boards that have hundreds of resistors or capacitors on them.

So, any progress on the service?