Can ferrite beads be used on digital PWM output?

Question

I am thinking how to reduce the EMI noise of the large LED indicators driven by tlc5947. This chip has digital outputs that limit/regulate current through light emitting diodes by switching on and off. The internal oscillator runs at about 4 MHZ and it looks like quite a powerful circuitry, 12 V, 600 mA for all assembly.

One of the ideas coming to mind is to use ferrite beads sequentially with LED diodes in the output. The concerns against are that be these inductors, they may generate the voltage spikes when current is turned off. From the other side, some sources say that ferrites do no accumulate as much energy as a coil of relay that would absolutely require a diode.

There are 24 channels per chip, and there are multiple chips, so this means lots of ferrite beads. But they are quite small and rather cheap so probably can be mounted if the bead is enough. With such amount of components, adding extra capacitors, resistors per every channel looks like no fun.

Could anybody with the knowledge say how good is the idea to put ferrite beads on the digital PWM output of the LED control chip running at 4 MHZ? I understand that the best would be to build the two prototypes and measure with near field probe (that would show the difference between the versions but not 'good enough' - 'not good enough') or at least just directly the waveform. There is a job and cost to build however. Maybe there is some generic knowledge on how good the idea is.

The goal is just to reduce the possible EMI noise. The indicators work very well, they do not even thing to flicker or the like.

P.S. Answers of the kind I should not be designing the circuit just because I am asking this question are out of scope.

Answer 1

Big picture: good signal integrity equals less EMI, even at ‘only’ 4MHz. Those harmonics can haunt if you don’t deal with them.

So look at several things in the SI area first.

• Signal rise time. The ferrites could actually help here. You could also accomplish the same thing with an RC network.
• Impedance matching. Reduce signal reflections by using controlled impedance an appropriate termination strategy.
• Loop area. Make sure your signals and returns are as close to each other as they can be. The smaller the loop area, the less antenna effect.

If that isn’t enough, there’s a couple more tricks in the bag to further suppress EMI.

• Common-mode filtering. If your signals and returns can be routed together in the same harness, put a ferrite around all of them to suppress common-mode noise.
• Shielding. Have an overall shield that does not serve as a ground return. Look at a USB or HDMI cable for example.
• Lossy materials (ferrite). Convert the EMI to heat.

Answer 2

From specs 8.3.4 they describe the 6x4 mux delay to reduce the PWM current spikes by 4 by design.

The rise times will be in and above the 10MHz band but may still cause EMI crosstalk modulated with the 1kHz PWM rate on high impedance mic’s , etc.

The best solution is to minimize impedance of long lines using twisted pairs as long wires alone are 1uH/m. Having a small area for the current return path will go a long way to reducing the EMI. Keep this mind in your power distribution and decouple with a good e-cap & plastic cap near the driver source and feed twisted pair from the supply. Choosing a thin insulation will lower the impedance which is best to prevent ringing as you will have a 50V/1.6A =+31 Ohm load and low impedance driver. So something <=50 ohms is OK ? Loose wire would result in those really noisy scope signals you have ever seen and std twisted pair 220 Ohm would work but <=50 Ohms is better to reduce reflection ringing.