Hi, Greg Richter here. I designed the Revolution Deva light, and as an electronic engineer of 30+ years experience I can help you with your EMI/RFI question. Few things:
CMH ballasts do not put out zero EMI. The boost PFC stage can radiate quite a lot, and all arc-light ballasts have this stage. Interestingly enough, LEDs are often the worst offenders for the same reasons. While LEDs are DC devices, they are run on switching AC power supplies which can make a lot of noise. A good example of an LED driver chip can be seen here:
https://led-driver.power.com/products/lytswitch-family/lytswitch-4/
LED designs are based around switching power supplies and these generate EMI. Quite a lot of effort is expended in minimizing that EMI, and the Power Integrations link above talks about some of what they do in that particular chip to keep things quiet. So, LED does not equal low EMI. Often it means the opposite, as the absolute noisiest lights we ever tested were all LEDs. COB means that the LED chips are supplied on a board, which makes it easy for low-tech manufacturing, but doesn't change EMI spectra. The DRIVER is what generates the noise, and COB, discrete colors or white light LED are all driven from a switch-mode supply plugged into the AC mains. A lot of info out there is just plain wrong on this topic, and I'll do my best to set the records straight with facts, links, and stuff you can actually look up.
With any arc light be it HPS, Halide, CMH or other, the first stage is always a boost PFC (power factor correction) circuit whose job it is to draw power from the AC line in pulses and keep EMI low and power factor high. This stage is typically very noisy, and better designs use two or even three stage filters to keep the noise down.
The output stage of an HID (arc light) ballast usually runs from 35 kHz to 100 kHz, and this can be a source of EMI as well. Name brands like Gavita, PL, Phantom and Galaxy do a lot of filtering to keep this noise down, and they are all much better now than they were two years ago. These particular ballasts did very well in our testing, which is why I mention them. Two years ago there were quite a few ballasts being sold that didn't even pass the minimum FCC limits for "commercial" as opposed to the tighter specs for "consumer" use, and there are quite a few still being sold that don't pass, as in are illegal to sell or operate. Google ARRL GROW LIGHT to read all about THAT big drama...
The Phantom does not run at 32 MHz, it runs at 100 kHz like the Gavita and other similar designs, just FYI to keep the record straight!
Now:
Our Revolution Deva light has a three stage EMI filter, and an output stage that runs 150 Hz. Not kilohertz, 150 cycles per second, which is fast enough for no flicker, but low enough the output stage puts out very, very little noise. I can post a video if you want showing the spectrum analyzer traces with the light at 1150W output generating about as much EMI as a Macbook laptop. It's very, very close to the limit of the analyzer to record it, and low enough to be considered "very good" for professional audio equipment, much less a grow light! We're really very proud of it. It's also digitally controlled, amazingly priced, and ... OK, I'm biased. Take a look.
Then:
The EMI/RFI thing became a problem a while back and I talked with fellow ham Tom Thompson about his experiences in Colorado and then went on to develop an EMI filter for the lamp side of all the noisy ballasts. We sell the thing for close to cost, just to get the problem fixed. Hydrofarm has them in stock, and they work really well, and are mentioned above also. To fix the power line side, if your PFC stage is noisy, usually requires a redesign. If that's not happening, you can throw ferrite at it to dissipate the high frequency noise, likely to solve most problems. You can see the filter and a suggested fix for power line noise at
http://FightEMI.com where I specify a ferrite you can get from Mouser for a couple bucks:
http://media.wix.com/ugd/ae3c18_6a5096a62a744c85bd0ccf24e96452d3.pdf
That paper also shows a well-know ballast spectacularly failing a conducted EMI test. You can pick up this POS at almost a half a mile with an AM radio!
It's why we didn't make a power line filter -- you can just go buy a ferrite core or two and that should help quite a lot. Either you can buy a big ferrite and loop the cord a few times, or a few of them and clip them on. Adding Y capacitors and stuff is dangerous, and can put hazardous voltage on the case of the ballast so we didn't go that way. I can get into that a little if there's an interest, but this is already a freakin' Book of a post ...
Faraday cages, BTW, are really hard to build in practice. The ones we use in the lab has lots of contact surfaces on the doors, screening, welded seams, etc. Better to stop the noise at the source than try and trap it all inside somehow.
I'll stop there. If I can answer anything else, just let me know. Engineers are like musicians, we'll work an eight hour shift for money, then go hang out with our friends and do it all again for fun.
-Greg