How this idea came about.
Figure 1. The first panel.
I've been getting PCBs made up for internal use and customer projects for at least 20 years. It didn't take quite that long to realize that I was continually designing the same subcircuits over and over, but finally I got to thinking that there should be a better way!
The concepts developed fairly quickly once the idea came along. The first panel of boards looked like Figure 1. These were added to a PCB panel for an in-house project. They were “bare” with neither solder mask nor silk screen, but they proved that the mounting method worked. That first order was an array of 5 x 5 boards, 2 off, so I started with 50 boards. The 13 initial designs included both positive and negative 3-terminal regulators, a boost converter, some LED boards, a 555 timer design, a Direct Digital Synthesis design, a 9 V board and most ambitiuously, a version of the FTDI USB to RS232 chip. (That FTDI board used a USB-A jack which was the wrong type to pick, but the board was useable with an A to A cable.)
A number of these early, bare, boards are still in use: some are built in to things and some are employed in temporary test situations as needed.
More panels have been ordered since: so far I've produced more than 700 boards. Later panels have been full double solder mask and double silk screen. Nearly 50 have been farmed out to reviewers, some have gone to customers as part of prototypes (to as far as Australia) but most I've kept. Quite a few have found their way into devices that I use frequently, like my WWVB clock, but I have quite a number still available for in-house use. And yes, a few designs have been duds, but only 2 of the more than 80 unique builds have not been useful. A couple more designs used chips which turned out to be poor choices because of obsolescences, and have had to be re-designed.
I have found that this system is so useful that I want to share it. But…
What to do, what to do?
The biggest problem in getting the concept to market is also the main advantage: some minimum number of designs need to be available for the system to be viable and useful.
The first plan of attack was to identify a minimum set of designs which will be enough to demonstrate the worthiness of the idea. I've been calling these the Minimum Viable Set after the Minimum Viable Product concept of Eric Ries, author of The Lean Startup. Even this set could be called large: over 20 designs made the cut but this left out some of my favourites!
The normal manufacturing process of ordering panels with as many copies of a single board as will fit onto a standard panel would be prohibitively expensive since many different panels would have to be ordered. On the other hand, panels containing a mix of boards would undoubtably have the wrong mix, building up unwanted inventory of some boards, and/or shortages of the most popular ones.
Now, for boards which have either no components installed (generally the uncommitted ones), or which just need a simple through-hole part added (many of the interconnection set which are essentially mounted connectors) this might not be too bad. If an unbalanced inventory situation develops for those then the panel mixes can be adjusted at the next order, though set-up charges could apply to each new mix. But for boards that really need to be built in a pick and place, reflow oven process line, this does not seem realistic.
Another difficulty is the corner holes. The method that I use, bisecting then bisecting the holes again with a shear while cutting apart panels works well enough for me, but it is not compatible with normal manufacturing processes. Nor will my shear remain sharp forever! The obvious alternative is to use V-scoring to partially separate the individual boards. There are a couple of problems with this:
- Some of the designs require copper very near to the edge of the board, such as the ones that can be stacked side to side like the uncommitted DIP design and the 1/10" header. PCB house rules generally require a wider margin between their V-scoring line and copper than these designs have. See Figure 2, below, from a recent panel design where I happened to place some of these designs side-by-side.
- The width of V-scoring is not always well controlled, nor is the scored path necessarily aligned perfectly with the drilled holes, so some panels would likely come back with almost no useable corner quarter-holes at all.
Figure 2. The problem with V-scoring some designs. The red lines are where these boards should be separated, and are much thinner than an actual V-score would be. The smallest distance between pads is 0.54 mm or 0.021", which is smaller than the 30 mil required space I have had to use in the past.
It should be possible to fine tune pad sizes to bring the copper-to-copper distance up to 30 mil or better, but the 2nd problem remains. So, the best plan seems to be, for the uncommitted boards at least, to use routing to create the outlines and the corner holes. However, and perhaps I'm inventing problems, I'd have to require that the PCB house use a smaller than normal router bit. This would be necessary since the corner “holes” have a 0.086" diameter. (I understand that 0.10" bits are normally used here.)
My proposed solution.
A cooperative venture seems an ideal solution: a guiding organization which works to maintain design consistency and tries to ensure good availability should be created. If there were to be a proliferation of sellers, each supplying one or a few boards then this is not optimum at all since a buyer would have to order from mutiple vendors, discouraging purchase. But the proposed system means that they do not have to produce the boards themselves: there are companies like SparkFun that sell both their own designs as well as boards that are made by others. And, as implied above, it would be too expensive to start out producing dozens and dozens of different board designs.
So the following is the mechanism I'm proposing:
- The organization would control the model numbering so that duplicate numbers are not assigned and at least some numbering commonality is maintained.
- The organization would register all makers, sellers and their boards and make the information available to the world.
- The organization would maintain a forum with sections for board makers, sellers, and an open section for users.
- The organization would act as a clearing house for all board designs, in other words it would store and supply all open source documentation for all boards, or at least store and supply pointers to these documents.
- The organization would act as a kind of brokerage to facillitate makers selling to vendors so that vendors might be able to maintain reasonable stocks.
- All makers and sellers would set their own prices. Since any maker could decide to make any given board design then competition could keep prices under control.
- The organization would require a small number of samples of each board produced by a maker. This would allow the organization to maintain consistency.
- There would be nothing prohibiting the organization from being either a maker or seller, and at the outset I imagine both would be necessary.
Initially, there may have to be some means of encouraging makers to make as many different designs as possible: there is no point if all potential makers want to supply the same limited set of boards. Once the level of interest is known for each board, in a general way, then the market can guide production.