We get asked this question a lot so this post is to cover the issues with onboard radio frequency (RF) emitters aka wireless and our company’s general guidelines and perspective on this important topic.
Short Answer
We do not include intentional RF emitters on any board due to (in no particular order) cost, interference, design complexity, regulatory compliance, signal integrity/performance, shortages, obsolescence, and validation. We may release certain products restricted to certain markets with onboard RF on the rare condition that the RF component is a validated part of the vendor platform design. See recommended dongles.
Design
One of our key design considerations is composibility. We want our hardware to be a re-usable component in a final design. Incorporating RF emitters forces cost and product design decisions on customers who may not want/need it. In order to deliver high quality validated products, we need to have scale in our supply chain, manufacturing, testing, and distribution pipelines. This is very hard to do when there are too many product permutations. It is better and simpler for customers to build two layer PCB with SPI/SDIO/USB/PCI-E RF.
Performance
Onboard RF emitters would have to use either PCB antenna or external antenna. Both impose a set of limitations on the enclosure and signal integrity to the outside. Onboard antennas within an enclosure will experience a significant drop in signal to noise ratio and directional limitations depending on pole design and case material. External antennas must be properly routed and create complex extra installation steps for end-users with potential for short circuit. The board would also have to be upgraded as new and better RF solutions become available.
Interference
RF emitters create numerous types of interference. High speed CPU clocks, DRAM busses, PCI-E clocks, USB PHY clocks all have the potential to interfere with RF and vice versa. It is very complex and costly to get the hardware right in a small form factor computer. You can mitigate some issues in software so that the operating system knows which frequencies are being generated by what and take remedial measures to avoid interference. This makes composibility very difficult and makes design a cross domain discipline which not many companies or people have to skills or will to tackle.
Regulatory Compliance
Unintentional RF emitters are covered by relatively simple governing laws in different countries of the world. FCC and CE are two major regulatory bodies for the United States and Europe respectively. Depending on the target field, use or sale of a product minimally require internal testing and self-declaration of compliance. Intentional RF emitters are covered by extensive governing laws that vary on a country-by-country basis with more regulated certification process with specific bodies, agencies, and processes. This complexity can be avoided by using pre-certified hardware available in the market.
Supply Chain
There are only a few mainstream RF IC manufacturers and even fewer high volume module makers. Component shortages and lifecycles are real risks when spec’ing a product. Minimizing specialized components in the bill of materials ensures that products will be supported for years to come.
Validation
Separately validating RF components makes it easier to validate the board as the processes are very different and specialized. Test suites and setups designed for SBCs are very different from those for testing RF. By combining the testing, complexity, throughput, and quality become multiplicative towards cost.
How do other companies do it?
It is not absolute but many of them cut corners and/or skirt legal requirements. If you search for SBC interference, you’ll arrive at many such instances where the functionality or performance is compromised. We want to do things correctly as much as possible.