Your custom keyboard is probably illegal - a long post on radio interference testing regulations and keyboards

This has been something I’ve been looking into over the course of designing my 120% keyboard, and something that I’m generally pretty concerned about for the keyboard community. I’ve brought this up in various Discord servers at various times, but would like to write a longform post about this.

Disclaimer - I am not a lawyer, and this is not legal advice. I am not an expert in this field - this is merely a summary of my research, and there may be inaccuracies.


  • Digital devices such as keyboards can generate unintended radio interference
  • Government agencies (such as the FCC in the US) regulate the airwaves, and generally require consumer electronics products to be tested before they can be built, sold,or operated, to ensure that they don’t exceed legal limits for unintended radio interference, and don’t cause issues for other devices that use our airwaves.
  • These tests require very expensive calibrated equipment, and have to be done at a testing lab
  • A test can run north of $2,000 for a given design for a given jurisdiction
  • Most custom keyboards out there do not perform these tests, and are potentially illegal to either sell, build, or operate as a result
  • The fact that custom keebs are built from kits does not seem to help much:
    • In the US - although whether designers and vendors can sell kits without testing is a gray area, current regulations make it illegal for end users to build and operate an untested design from a kit, or to resell it
    • EU does not appear to exempt kits at all
    • I have not looked into other jurisdictions
  • There’s an exemption for DIY projects in the US (and likely the EU), but keyboards built from kits are not exempt, and it appears to be illegal to sell untested keebs built as DIY projects
  • There are potentially ways around this that do not require expensive testing, but may require a change in how custom keyboards are sold

Flip over any prebuilt keyboard, and you’ll find a label on the underside. Usually it’ll list the model name and a serial number, and almost always, you’ll find some regulatory symbols.

For example, here’s the underside of an iKBC CD108:

To the right of the serial numbers are three logos - from left to right, the arrow on the left represents Taiwan’s Bureau of Standards, Metrology and Inspection, the middle logo is for the US’s FCC (Federal Communications Commission), and the CE logo on the right is for the European Union. The logos attest that the keyboard complies with the various regulations in those jurisdictions that cover keyboards and other electronics, and as a result, can be legally imported and sold there. On other keyboards you’ll often find other logos that cover other jurisdictions (for example, CCC for China, or RCM for Australia and New Zealand).

One of the main regulations that electronics have to comply with is what’s called electromagnetic compatibility (EMC). This can also be called EMI (Electromagnetic Interference) or RFI (Radiofrequency Interference). These regulations exist to ensure that consumer electronics and other goods don’t emit excess amounts of unintended radio interference into our shared public airwaves, which can cause problems for other devices (radios, phones,TVs, etc.) that rely on it. There are legal limits for unintended radio emissions, and generally most electronic devices sold on the market are required by law to be tested to ensure that their emissions do not exceed legal limits. In the US, this testing and verification procedure is called “Equipment Authorization”, and is regulated under 47 CFR Part 15. In the EU, keyboards fall under the Electromagnetic Compatibility Directive, in addition to other regulations like RoHS.

Do wired keyboards have to do this too?
Yes - a popular misconception is that these regulations only apply to wireless keyboards. Virtually all keyboards, wired and wireless, are digital devices that operate above 9KHz (your typical keyboard MCU operates between 10-100MHz), and have substantial ability to emit radio waves as a result. In the US regulatory scheme, wired keyboards are categorized as “unintentional radiators”, and are regulated under 47 CFR Part 15 Part B. This is why prebuilt wired keyboards still have regulatory markings on them.

The DIY exemption and kits
However, unlike prebuilts, custom keyboards are usually sold unassembled as kits, to be assembled by the end user.

In the US, personal DIY projects are provided a limited exemption from testing in 47 CFR § 15.23. Unfortunately, to qualify for the exemption:

  • The device must be built for personal use
  • The device may not be marketed (i.e. offered for sale, or sold)
  • Devices built from kits do not qualify

As a result, if the vendor or designer did not test the design, it is technically illegal for people to operate the resulting keyboard without having conducted the tests.

A kit is defined in 47 CFR § 15.3(p), and the fact that switches or keycaps are not included with custom keyboard kits doesn’t change that they are still considered kits.

Whether vendors and designers can sell kits in the US without testing is a gray area due to a lack of explicit regulations on most kits and contradictory (but non-binding) FCC statements on the matter. Sparkfun and EMCfastpass discuss this in depth and reach opposite conclusions. In addition, for computer peripherals specifically, 47 CFR § 15.101(d)(1) says that “No authorization is required for a peripheral device or a subassembly that is sold to an equipment manufacturer for further fabrication; that manufacturer is responsible for obtaining the necessary authorization prior to further marketing to a vendor or to a user." . It can be argued that the end user is an “equipment manufacturer“ that conducts “further fabrication” since they build the kit into a complete keyboard, though I would say this is a stretch.

However, if sellers were not required to test, customers who built the keyboard would end up with the legal responsibility to test, as kits are not given the DIY exemption. It would be unreasonable to expect customers to pay thousands of dollars in testing to legally operate their keyboard, so it really should be up to the vendors and designers to do this.

The EU does not appear to exempt keyboard kits whatsoever. The EU likely exempts personal DIY projects, as the EU EMC Directive specifically only covers products that are “placed on the market (i.e. available for sale). This would probably exclude personal DIY projects as they are not “placed on the market”, but would include keyboard kits bought from a vendor or during a GB. There are also exemptions for custom built R&D evaluation kits and amateur radio equipment kits, neither of which would cover keyboards. For other kits, the flowcharts in the EMC directive guidance document seems to state that electronic components or sub-assemblies “intended for incorporation into an apparatus” “by the end user” (i.e. a kit) falls under the definition of “Apparatus” and therefore falls under the regulation.

The Testing

To test a keyboard, you generally have to bring it to an accredited testing lab, as the equipment required for testing (a calibrated antenna,spectrum analyzer, and a shielded anechoic chamber) are extremely expensive and must be calibrated. The FCC also requires the test labs you use to be accredited. So unfortunately, DIY testing is not an option for fulfilling the legal requirements.

EMCfastpass goes into some details about the required tests. The FCC requires radiated and conducted emissions tests, while the EU requires additional immunity and ESD testing.

A part of this blog post from the Ultimate Hacking Keyboard’s developers also goes into the process a bit (scroll down to “EMC Testing”).

As far as cost,I received a preliminary quote from a lab to get my keyboard design tested, and it was:

  • $2200 for FCC Part 15 Testing & Certification (SDoC, Class B, Part 15.109, 15.107), for the US
  • $900 for Industry Canada Testing & Certification (ICES-003, Issue 6, Class B, CISPR 22)
  • $3500 for CE Testing & Certification (CE EMC | EN 55035:2017, EN 55032:2015, with reference to EN 61000-3- 3:2013, EN 61000-3-2:2014), for the EU

So $6,600 for testing from this particular lab - pretty steep. This also wouldn’t cover other jurisdictions like Mexico, Singapore, Japan, Australia, etc., as those countries have their own standards and their own test requirements, and their own costs.

In the US wired keyboards are allowed to self certify under the Supplier’s Declaration of Conformity process - no submission to the FCC is required, but testing must be done, and the test results must be kept and available to the FCC upon request. Certain regulatory language is also required to accompany the product. I am less familiar with the EU requirements but I believe it is similar.


In the US, the FCC has the power to levy pretty substantial fines (up to $75,000/day) for failing to comply with their regulations. However, there currently doesn’t seem to be a lot of enforcement going on - you can see a list of past enforcement actions on the FCC website under Equipment Marketing Violations, which includes failure to conduct testing. In addition, wired keyboards generally are unlikely to be huge sources of interference, as they don’t use much power and often have metal cases that act as shielding. However the enforcement that has happened (none of them keyboard-related) generally resulted in pretty large fines.

I am not familiar with the EU or other jurisdiction’s enforcement and penalties, but they probably also involve fines, and potentially mandatory recalls and criminal penalties.

Boards that do testing

Some custom keyboards do go through the testing process, and display the FCC and CE logos. Notably all of these boards are either hotswap or offer a prebuilt option.

  • CapsUnlocked tests all their boards
  • Drop appears to test their boards (the Planck was notably delayed due to FCC testing)
  • Mike from Novelkeys said in an email that the NK-series boards have passed testing, though they don’t display the markings. These markings may be added to VIA in the future, as in the US devices that require screens to operate (such as phones and keyboards) are allowed to display their regulatory markings digitally.
  • The Ultimate Hacking Keyboard
  • Boards from larger companies like Glorious and System76 generally go through testing

However, most customs show no evidence of having conducted these tests.

What would a legal custom keyboard scene potentially look like?

As mentioned, there isn’t much enforcement in this space now, but if the scene were to comply with these regulations, this is potentially what the options are.

Option 1: Just do the testing

For higher volume boards, the designer or vendor could simply do the testing. Testing is effectively a one time cost, and several thousands of dollars in testing costs can be easily diffused across thousands of boards sold. For example, a $6,600 testing fee across 1500 boards in a large GB (such as a Satisfaction75) amounts to only $4.40 per board, which is definitely affordable. Large GBs are unlikely to see much issue here.

Some designers and vendors may choose to limit themselves to certain jurisdictions to cut down on testing costs - for example, doing only FCC testing for the US may only cost ⅓ that of doing the tests for the US, EU, and Canada. If it’s unlikely for the EU sales to make back the $3500+ in testing fees to be able to legally sell there, designers and vendors may simply choose to not sell there. This would be quite unfortunate for the community, and smaller markets with their own testing laws (such as Canada and Australia) may be especially screwed over here.

Unfortunately, this is likely not that viable for lower volume boards, which have fewer units to spread the testing costs out to, and would potentially result in substantial price increases. . If you’re a first time designer, or running a highly unusual or experimental board, and only running say, 50 units, a $6,600 testing fee add $132 in costs to each board sold.

Option 2: Separating the PCB and case, and only selling the case to take advantage of the subassembly and DIY exemptions.

Another option might be to only sell the case rather than a complete kit- since the case is just a hunk of metal or plastic and does not contain electronic components, it may not fall under the EMC regulations. The customer would have to buy their own PCB. Some case-only GBs already do this, though for other reasons.

PCBs that are sold separately and not intended for any particular product likely fall under the subassembly exemption in the US, which is discussed in Sparkfun’s and EMCfastpass’s articles. The resulting PCB/case combo might avoid being classified as a kit, and therefore have the DIY exemption available for end users, which would remove the need for testing.

However, this likely only works for boards that use common layouts (such as 60% and TKL) for which several different PCBs by several different sellers are available - if you just sell the case and PCB separately in the same store, and the PCB is intended only to work with your particular case design, it seems a lot more of a stretch to try to exempt the PCB as a subassembly.

The EU regulations also do not appear to cover bare cases, and contain a subassembly exemption that might be similar, so this also might be an option for other jurisdictions.

Option 3: Open Source

The third option would be to take advantage of the DIY exemption, and instead of selling the board, the designer simply publishes the design and the manufacturing files online, so that the designer or vendor aren’t selling anything. The end user would be expected to buy all of the parts and assemble it as a DIY project (not from a kit).

This could potentially be done in conjunction with (2), where the designer may choose to only sell cases, and tell the end users to order their own PCBs directly from the PCB fab.

Unfortunately, this ends up being very uneconomical - there are substantial economies of scale and cost savings from mass manufacture, and everyone making a board will have to pay the cost of a one-off. In addition, options for the designer to be compensated for their work are somewhat limited. However, this may end up being the best option for smaller design that cannot afford the testing costs and for which no off-the-shelf PCBs are available for.

Ignoring the law
Of course, this has been essentially what’s been going on (mostly out of ignorance). No one has gotten in trouble so far, there isn’t a ton of enforcement going on, and with keyboards unlikely to be a huge source of interference, we haven’t seen much of the negative consequences of interference that the regulations were meant to prevent. In addition, in the US, the legal responsibility is possibly on the end user rather than the designers and vendors, and it is unlikely the FCC goes after every single individual typing on an untested custom keyboard. Nonetheless - the community is growing, and no longer able to skirt laws that it formerly got away with ignoring (most notably around intellectual property and keysets). In addition, vendors, especially those that sell outside the US, potentially face substantial legal liability. Finally, although unlikely due to massive potential backlash, larger players are moving into the customs scene, and may be able to report these violations to the government to try to eliminate competition from smaller vendors and designers. So moving forward, this may no longer be a viable option.

I haven’t really thought much on how to conclude this long and technical post - I just wanted to raise this issue as a problem with huge potentially significant impacts on how the customs scene operates, and something that definitely merits discussion. Hopefully, we as a community can find ways to deal with these regulations moving forward.


(I apologize for any typos - I read over this once but am generally bad at catching typos and missed words. I’ve also been working on this post for 2 weeks and just wanted to get it done with)


Excellent post! And I’m probably one of the few that have never given any this a single thought.

Although, with that said, has this ever happened before?

If not, I can feel that it will happen in the near future…and possibly more often as more and more players get involved.

It’s really an issue of enforcement and while selling an untested product is itself a violation if the product itself doesn’t violate any of their other regulations (the things they are testing for) then you are probably pretty safe from getting dinged by them.

You still certainly don’t want to advertise that the product is untested or sleep with the s/o of someone who works at the FCC.

You are 100% correct. This topic was brought up in the community before maybe 5-6 years ago. Same examples, same explanation.

On the ikbc example or other boards that have such stickers, they might reuse existing certs or literally just use the logo without actually ever certifying (check the manual if it has a cert reference/id etc.). It’s sometimes done via the PCB manufacturer, they have established certs for certain “circuits” and they are reused by customers. Hence ikbc etc. might not actually ever even spoken to anyone related to the stamps but their factory might have once validated a circuit for another client. There’s no real good way to validate that. (as far as I know)

overall the 1-2 posts in the past didn’t change anything nor will this one. some shops sell a lot of boards but there always needs to be a cause for a report or ill will and then the agencies have to come after them… question is if anyone really cares, after all those pcbs are nowhere near powerful enough to cause any real world damage to anything. I’d be more concerned about USB cable makers not properly testing cables - a short in a household sounds a lot more dangerous than incredible low powered radio waves in a closed room…

EDIT: did a bit of googling - I found that iKBC definitely certified some products - for example:
Most entries I find are for Korea and Taiwan, not the US FCC.
“National Communications Commission Taiwan”
Not entirely sure how this works with the certificates but maybe they can use those to apply to the FCC? Either way, seems like since they are selling international to big outlets into retail, they probably did all that. There is also a German Mister GMBH invovled and two more US and Chinese companies. Definitely a bit confusing who did what certificate :wink:


This is a good and thorough post. It’s something I’ve thought about a number of times but I not done the research you have.
Seems to me like the worst case scenario is that cases could still be sold but PCB design would have to be open sourced and people would have to get their own PCBs made.

1 Like

Thank you very much for this detailed analysis, very interesting reading.

I may have another solution between Option 2 and 3 (I personnally like Option 3).
Let’s call it Option 2.5:

It would be to use a computer on module.
This module would contain all active components that would be responsible for any radio interference, like the MCU.
This is something already quite used by the DIY community with all the Adafruit and Sparkfun boards out there, beeing Arduino compitible or not.
Also designers wanting to have BLE wireless connectivity commonly use this approach in order to have BLE certification easily (because the computer on module has already passed all the tests).

The module could be slotted on the keyboard PCB with some sort of connector like an NGFF (PCI express standard) one.
The keyboard PCB would not contain any active components that would emit radio interference and would not have be tested.
The module would have to be tested, but would be used on many different projects and testing cost would be easily amortized.

Now the question is:
It would work on a simple PCB design.
But how about adding fancy things such as RGB underglow or RGB lighting ?
You’d need to add on your PCB additional active components like a LED driver that may fall in the test requirement.
And maybe just have LEDS that are PWMed is sufficient to fall under the test requirement.


I think this is a great idea/solution, something like the Unified Daughterboard project but for the entire MCU and associated circuitry. Like the Giga40 but without the retro game console aesthetic.

Not only would this allow for the controller board to be certified but it also gives the end user as much PCB flexibility when i comes to layout, as they currently have with the cutting of customised plates.


In fact I started working on that a few months back, but stopped to concentrate on other projects.

The idea is not new really.
Matrix Lab has already developped NGFF modules for their PCB.
And there is another 60% open source PCB that also uses this NGFF connector but I don’t remember the name.

Unfortunately I’m a little skeptical that this might work to get around the regulations. Doing case-only possibly gets around the regulation because a bare case likely doesn’t count as an electronics component (though it may count as an electronics enclosure, so I’m not too certain about that either). Doing a PCB, which still has other electronics parts on it (most notably diodes, and as you mentioned, possibly underglow LEDs or backlight LEDs), and telling the end user to buy an MCU module to make it work likely results in it being considered an electronics kit rather than a subassembly. The MCU module’s emissions will also definitely be affected by the PCB and case it’s installed in, so unfortunately using a certified MCU module does not necessarily mean the resulting keyboard will pass testing either.

The analogue here is FCC pre-certified Bluetooth and wireless modules. Normally, if you design a wireless device you have to go through the FCC’s intentional radiator testing and go through formal certification, which is much more rigorous and expensive than unintentional radiators. However, if you use a FCC certified wireless module, you can skip the rigorous intentional radiator testing, but you still have to go through the less rigorous unintentional radiator testing for the complete device, because the certification only covers the module you use, and you still have to get the rest of your device tested. It’s not a perfect analogy, but it’s possible the FCC will treat an MCU module for a wired keyboard similarly. Good idea though.


You know I only just now realized, I don’t think pro-micro and pro-micro-alikes are certified. So that would also greatly affect even many ‘open sourced’ designs as they still rely on an mcu assembly to function.


Ok, I was wrongly making the analogy with FCC certification and it looks to be two different things.
So the solution is option 3 (and I like this one).

Now if you want to be very picky, could a new certification be mandatory if you change the type of PCB substrate panel ?
If thinking of going from 1.6mm FR4 to 1.2mm FR4, or even different brands of PCB substrate panel (Rogers, …).

Why borrow trouble?

It is not much about borrowing trouble, more about knowing the potential future troubles.
And it is most likely that nobody will be in trouble for a small GB anyway…

This is certainly an interesting point to raise, I am no expert but here are my two cents:

What does the law say about an individual making custom electronics? If this is more about certifying things that are being sold commercially, then a reconsideration of the way these items are sold would be welcome.

On another note, if this were a legitimate issue, i would like a company like arduino to weigh in, since I expect there is some overlap in hobbies, and they would likely be aware of the litigation surrounding the topic.


Thank you for being clear on this! As someone heavily involved in wireless keyboards, this seems to be a major point of confusion. Several folks have mentioned concerns about selling wireless keyboards in RegionX, and when I ask what they’re doing about EMC testing for their current offerings, the answer is typically “nothing”. I completely understand why that might be their current testing approach, but it also seems a bit backwards to suddenly care about it (especially since I always design w/ pre-certified modules, that have had the intentional radiator testing done, and what’s left is the same unintentional radiator testing you’d do for a non-wireless keyboard.

Appreciate the detail that went into capturing the current state of things, and at least some brainstorming on how things might be different. I do think there’s likely two different sets of folks w/ different solutions, as you note: large vendors, and smaller/niche ones, with different risks, and possible solutions (or non-solutions).


As mentioned in the post, there’s a personal DIY exemption in the US under 47 CFR § 15.23 - Home-built devices, but it does not apply to devices built from kits. So if you’re designing and making your own keyboards from your own design (or building them from subassemblies, but not kits), as long as you’re not selling them, and as long as you’re building less than 5 you should be in the clear. For the EU I believe the EMC Directive does not apply to devices that are not placed on the market, so you’re probably also in the clear there. Not sure about other jurisdictions.

Does a group buy count as selling? The idea in its original sense was that one person would collect money from others to be able to order enough quantity of an item (or components) to bring the price down, the items would then be delivered to each contributor.

The difference between this and selling is that it is a group of individuals coordinating to get a better price on an item or collection of components (as opposed to setting up a company, shop front, profit etc. to sell)

Someone would have to confirm if this is enough to bypass the requirement for testing, but it would seem compatible with the spirit of individual exemption at a surface level.

I wonder if this thread will eventually upheave the established way of doing things in the community, taking us back to smaller, localised, actual group buys again. may also be a relevant company to seek opinions from. They are a pillar of the community (like it or not), and they might have the kind of resources to investigate the matter further - they might even have the answers already.

Anybody care to reach out?



I am glad this topic gets considered in this community :-).

On the technical aspect of it, as mentioned by others, using pre certified modules is only useful for wireless applications. For wired keyboards there are no advantage regarding emission certifications.

Unattended emissions are a system design issue. Ie. they can only be evaluated on complete, boxed up, ready to be used as-is assembled piece of hardware. In some (most?) regulations, even the firmware is part of the certified assembly, with only minor changes authorised.
And it is not because regulators want to be a pain on purpose. It is really because any small change can have a big impact.

Do not regard the MCU as the main offender in unattended emissions. It is not. One of the reason there are exceptions on sub-assemblies is because just the way you wire it in a box can drastically change the profile of unattended emissions. The final integration matters. The main PCB routing matters. The firmware matters.

For the PCB, note that emissions do not come from an active component. Fundamentally to make an emission you need a signal swing in a conductor (ie. an antenna).
Practical application in keyboards: the matrix scan. Yup, just that. Some varying signals from the MCU going into copper tracks.
In addition, could be that all these long columns/rows tracks with diodes makes for some nice rectified signals at various antenna length.

4 layers stack up with proper ground planes would be a low hanging fruit to beginning with.
Some interesting entry level info on that could be to search interviews or talks on youtube with Eric Bogatin or Rick Hartley. They are more about signal integrity and susceptibility if I remember well. But that’s kind of two sides of the same coin as the fundamentals are the same.

On firmware side, one of the easiest trick applicable here is to reduce the drive strength of the outputs that are beings toggled during a matrix scan. Ie., make it such that it goes from low to high to low over longer progressive ramps. It could be some MCU allows such configuration (usually through configurable pull resistances).
Note: it is not the frequency at which these transitions happen that matter (which in keyboards is just going to be some puny kHz). But the speed at which they toggle, which can easily push you into unintentionally emitting in the 100’s Mhz bands with “modern” microcontrollers (as opposed to old keyboards that did not had such issues).


Thank you very much for those very interesting informations.

If I understood well testing the PCB alone does not solve the problem.
So it would be the role of the PCB designer to try to minimize the emissions, but the charge of conducting the tests would be to the people that sell the keyboard (with everything assembled and with a near final firmware).

Going 4 layers for PCBs is a neat way to minimize emissions, but manufacturing is still significantly more expensive than 2 layers ones (although the difference is getting lower with time).
Keyboard PCBs are rather big, so final price can be a concern here.

I am not too much concerned about big GBs, as the big number of units will more or less hide the testing cost.
I am more with small scale ones, low number of units, usually done by a one man operation were passion hobby prevails over capitalistic goals.
I’d like those small scale projects to continue to live as they often drive innovation forward.

We know more or less the cost thanks to @Pylon , but how about the time it takes to make the tests ?
And also what can be the potential failure rate?

1 Like