I leave it up to you to guess what percentage of low-volume devices are actually tested, but I'm pretty sure "design it sensibly and we'll take the business risk that we're not going to get found and hosed by the FCC" is wildly more common than non-technical people would assume. For devices released in under quantity of 10K units, I bet the percentage of digital devices without intentional transmitters that are certified rounds to 0%. (Arduino was only FCC certified on the 5th version, IIRC.)
And the answer to "why 9 kHz?" is pretty obvious from: https://www.ntia.doc.gov/files/ntia/publications/2003-alloch...
Or do you think those "RadioNavigation" users magically don't care about interference?
Small volume manufacturers just try to design well and roll the dice because (almost?) no one is spending four figures on an accredited lab to do emissions testing on a device that sells a few dozen or couple hundred units per year.
Some try to find a way to indicate that it's used only in transportation, that it's industrial/commercial test equipment, that it's a peripheral device or sub-assembly and not a finished end-user device, others just realize that the FCC isn't going to bust down anyone's door because a device not emitting harmful RF doesn't have a testing certificate on file.
https://linxtechnologies.com/wp/wp-content/uploads/fcc_resou...
The only thing that really chafes for me with the FCC is their absolute retardation with Internet regulation, and I'm still not sure I'm 100% okay with idea of "spectrum auctions". I get the idea behind it, but I seriously question the wisdom of anything that basically self-selects for corporate ownership over anyone else.
Still a bit of a noob on the RF scene though, so I've probably got nil real insight into the real reason the FCC is actually doing a good job.
If your power is going down a long wire, it tends to generally get better travel on the signal (for accidental transmitters, intentionally sending signals can can use math and design to better results)
So if your Arduino operates on a loop, and sends LOW voltage, and occassionally HIGH voltage, you get a radio. The frequency is not exactly always the loops executed per second, it tends to get mirrored onto many frequencies for some reason I'm not familiar with.. you can force it to a certain frequency by getting you loop to operate at a specific speed, but it will also appear in other frequencies.
Would love a radio guy to expand on my explanation here.
Play with this for a few minutes: http://www.jezzamon.com/fourier/index.html
And watch any of these: https://www.youtube.com/results?search_query=vibrating+reed+...
Okay, so think of a radio receiver as being sensitive to a particular frequency, because it has a filter in front of it. A vibrating reed of a specific length, if you will. (This is almost exactly true for certain types of crystal and SAW filters, actually.)
But when you flip a digital output from 0 to 1 or vice-versa, it changes state very quickly -- the output is essentially a square wave. (Physical limits mean it does have a finite slew rate...) Which means it contains components of every frequency. Think of coming up to one of those vibrating-reed meters and smacking it -- a single impulse excites ALL the reeds.
Smack it regularly (change the pin state at a specific frequency) and the reeds that're way off frequency will just sort of bounce around but not be particularly excited. You're creating noise, but it's just noise. But the reeds that're odd multiples of the smacking frequency will vibrate too -- these are the harmonics. They'll get really excited, almost as excited as the fundamental. (The evens essentially cancel out.)
Here's what it looks like on the radio spectrum: https://www.youtube.com/watch?v=eC36AqL5mw8
Incidentally, this is why so many chips now boast "slew-rate limited" outputs: to intentionally slow down those transitions, to reduce the amount of noise that has to be filtered and shielded in order for the device to pass RF emissions testing.