The following was posted on the BattleBots Forum by John Hoffman:

I've been involved in some of this lately at work; here are some guidelines that apply to us. This is mostly for the 'standard' 72-75mhz band; 900 mhz is a different beast.

The main source of RF noise is the brushes of the motors. These are intermittent spark gap sources, and produce a very wideband spectrum. Being at high current just means the energy in the noise is greater. Another source is due to the switching of high currents by the motor controller; at each transition, the ringing produces many harmonics, which can reach up into the RF band.

With noise, it's always best to try to eliminate it at the source. Once the genie is out of the bottle, everything becomes an antenna, and you're stuck with trying to shield everything. Try to find where the noise is generated, and let that energy make the smallest loop possible; let it 'go home' as quick as it can.

For the motors, the typical solution is to include bypass capacitors across the brushes. It can be done at the input leads, but is probably better done inside the case. A single cap across the leads may be sufficient, but as long as I'm in a motor, I like to also include a cap from each lead to the case. You should use non-polarized caps with a voltage rating of at least twice the operating voltage of the motor, and values between .01 and 1 microfarad. Ceramic capacitors fit the bill well and are commonly available at places like radio shack. Dan has a good application note at Teamdelta.com

Next most important are the power leads, starting with the run from the motor to the speed controller, but including all the high current wiring. You should keep the length to a minimum, and twist the wires together. This minimizes the antenna effect.

Distance is important; keep all power leads as far away from all signal leads as you can. Pay extra attention to your receiver's antenna, but this also applies to all those RC servo leads. If a power and signal wire must be close together, try to have them cross at right angles; parallel wires tend to couple the noise.

Ferrite beads are commonly used to reduce noise in this frequency band. If you use them, here are some guidelines:

Brushless DC motors are better on one hand because they don't have the sparking of brushes - but on the other hand, there is much more switching noise, as the current is turned on and off many times per revolution. In this case, extra attention should be paid to the leads between the controller and the motor.

Also by John:

Some more things to try:
1 Twist the motor leads together
2 Also twist the power leads to the motor controllers
3 Make a shield for the motor leads; An overbraid is best, but just wrapping the twisted pair with a grounded wire may help.
4 Watch out for ground loops. Ideally, all cables should be grounded at only one end. It's a little tough if the motor case is grounded.
5 Add another decoupling cap at the outlet of the motor controller
6 Consider adding decoupling caps from each brush to ground
7 Work on location of the wires;
-Keep the power stuff as far away from the signal stuff as possible
-Never bundle the two together.
-Try to prevent parallel runs of them - if they cross, do it at right angles.
-Try to get some metal between the power and signal; i.e.: run the power on one side of an internal brace, and the signal on the other.
-Maybe even consider rerouting through your frame (assuming tube construction); with a couple of small holes, you can run wires down the inside of the tubes that make up the frame. Works for either power or signal.
8. Do a little voodo dance and make a sacrifice to the EMI Gods. Don't ask me how, though cause the only one I know ain't working. (on a non-robotic project)

-----

Not all ferrites are created equal thou must find the frequency of the interference and choice the correct material. Different manufactures have different numbering schemes for listing what type of material the ferrite is made of. If the frequency is over 450MHz ferrite will not be effective. Other suppressive and shielding techniques will have to used.

Find out if it is a conducted immunity issue or a radiated immunity issue, there are different techniques for each one. Reply back when you know more or if you need help in diagnosing which one is causing your issue.

Yes I am writing this on a PC that has now lockup on me twice while writing this email. @#$%@#

Remember both common mode and differential mode techniques maybe used.

It may be black magic but Maxwell’s equations still apply they’re normally just to complex or long to solve easily. I normally joke that my whole job is to find the perfect spot to place a cap or ferrite bead.

If you look around the cubes where I work you can find a plastic skull with a mouse crawling out of the eye socket, a black and red feather duster and a squeaky rubber chicken for use when all else fails. It really disturbs the management types when you start dancing around their product with them. Especially if the issue could have been solved with more attention to layout early in the product development cycle.

Posted by a
NARTE Certified EMC Engineer
 

1) All ferrites convert RF energy to heat.
2) Ferrites are used to decrease the energy from source, change the coupling path impedance or increase the receiver system immunity

Ferrite beads that are placed into individual lines are working on differential mode currents; they have low DC resistance but high impedance to RF.

Ferrites are spec’ed at a impedance at one frequency, “600ohms at 100MHz” but its impedance at 40MHz maybe 10ohms or 1000ohms you need to look at the impedance vs. frequencies to get an idea of what the ferrite is really like.

Watch out that you do not saturate the ferrite bead by having too much DC current going through the ferrite. MY rule of thumb is take the spec given and divide it by 2 or 4 and use that number as the working DC current.

Ferrite Cable clamps work on Common Mode currents.

Common Mode Chokes work on Common Mode currents

Capacitors work in both Common mode and Differential mode currents depending on how they’re placed.

Consider a loop antenna of size X, adding caps to the system decreases the loop size thereby reducing the loop area and decreasing its radiating efficiency.

Placing a ferrite or cap in a system may cause the frequency of interest to go down but others to go up.

And thread 7420 has this from Joe Richards, responding to Andy Miller:

One thing you could try Andy is to make a "sniffer" from an old car radio. Hook up the radio to a power source and one speaker.Set it on AM. Add a shielded antenna lead to the ant. input. On the other end of the cable strip 1/2" back so that just the core wire sticks out, that's the sniffer. You can trace down noise in a circuit or from a motor. leave the volume on the radio set at the same level. Then you can tell any improvements you've made by the level of noise in the speaker.

Team Delta has a lot of information about supressing motor noise on their website. Application Note #1 - Reducing Motor Noise has this to say about the type of capacitors you should use:

A good value capicitor to use is 0.1 microfarad monolithic ceramic rated for at least three times the motor voltage.