maddog's Doghouse

From time to time, I hear people talk about "Free" or "Open" Hardware and what it means to them. After all, we have Free and Open Source Software (FOSS) and shouldn't some of the same ideas of collaboration that surround FOSS apply to hardware? In many ways, Free and Open Hardware (FOH) could be like FOSS. Creators of hardware could collaborate with potential users to help determine the characteristics of the hardware. Developers who want to use the hardware to make a product could help test the functionality and interfaces of the hardware to make sure it actually works and meets their needs.

However some of the features of FOSS do not apply equally to hardware. For example, if you give away a copy of your hardware, it is a physical thing you are giving away, and, unlike software, it probably will cost money to replace it.

Costs to prototype and manufacture hardware have dropped over the years (as has the cost of the hardware itself), so it is easier than ever to design hardware outside of a large company. However, there are still significant costs for the development of hardware. Let's look at some of the costs of developing a single board computer (SBC). I will warn you that this will be from a very high-level view.

First, you should decide what features your SBC will have. What architecture should the CPU have, what Graphical Processing Unit (GPU) will it have, how much RAM memory, what I/O peripherals, power supply, and so forth. You might generate a list of features from looking at other SBCs in the marketplace.

Then, if you have the knowledge and skill, you either find or create a circuit diagram that implements the features you desire. Sometimes you might find a very expensive board with lots of features and you decide that you can save a lot of money, heat, power usage, etc. by "depopulating" an existing design to fit your needs.

Eventually, you will have a design that you like, so you may try to create a printed circuit board with your parts on it. This is where the design may start to get expensive. Although many of the parts (particularly the resisters, capacitors, and other "passive components") are very inexpensive, the "active components" (RAM, CPU/GPU, etc.) can be fairly expensive and can also be ruined by static, heat, and other environmental issues. Many components are also "surface mount," meaning that they only sit on pads of connectivity that are very, very small and require a steady hand for placement OR a machine called a Surface Mount Technology (SMT) machine (or its smaller cousin, the pick and place machine). A world class SMT machine can cost well over four million dollars and place 60,000 components per hour. To set up an SMT may cost between $100,000 and $400,000. A "pick and place" machine that sits on a table top could cost several thousand dollars and place fewer components slower, but is more practical for smaller numbers of boards or prototypes.

Another issue is getting access to the documentation for the active components. Often this documentation is not available to the public, and you have to pay the manufacturer to get this information, or perhaps you can get access to the information through a distributor who has a partnership with the vendor.

Perhaps you now have a working board. You test it and it seems to do everything you want. Does it have WiFi? If so, you might need to get the wireless certification from your national certification agency, and this might cost tens of thousands of dollars. Depending on the target market, you might have many more certifications to test, patent royalties to pay, and other regulatory costs.

Also, you might need an operating system, or a specific distribution of an OS, to be ported to your board. In order to port the board (assuming you do not also have those skills), you might have to generate several of your boards to lend to software people to finish the port. Now you have a finished board with some distributions ported to it. Assuming you want to build a lot of the boards to use in products, you have to raise the money to purchase the parts for large-scale manufacturing, as the parts suppliers will want to be paid before they release the parts for manufacture.

After that, you only have to purchase the packaging, decide how to ship the finished product to customers, and determine how to take back boards damaged in manufacturing and shipping.

Development costs have come down, but a new SBC can still cost between three to four million dollars, taking into account everything necessary to release the product to the customer. As you can see, while collaboration is fine, free hardware is not as free as free software.