LC filter or ferrite for input filtering?

[Manx]

I want to measure an external analog signal. It may be (very) dirty so I want to do some robust noise filtering. It's not only about smoothing the signal for measurements, but also about protecting the ADC.

My initial idea was tu use a normal LC filter, with 1 mH and 1 uF. Later I learned that ferrites are all the rage for input noise filtering purposes. Something about filters just reflecting high frequencies when ferrites actually eliminate them.

What does actually happen to filtered-out frequencies? Do they actually bounce back and forth? And what should I use then? Filter seems simple enough and I'd use it without overthinking the matter, if only I haven't learnt about ferrites.

[Siwastaja]

Just use RC.

L instead of R makes sense when you need to conserve power, or need to have 2nd order response, but even in the latter case, it's better to do 2nd order with an active opamp circuit, instead.

You want to dampen the LC with R (in parallel with L, or in series with C, or both), otherwise you are right, energy bounces back and forth, causing resonances at some frequency.

You can model the ferrite bead as an L with R in parallel, so a damped L. This means, the rated impedance is more or less resistive, far from purely inductive.

But, instead of a, say, 600 ohm ferrite bead, just use a 600-ohm resistor, because you don't need to transfer DC power, I guess.

RC filter of course is just -6dB/oct, but if you need more for anti-aliasing for the ADC, then use an additional opamp circuit for a 2nd order response.

[Manx]

Thank you very much.

I think RC is out of the question because the inputs have more complicated function than only connecting to ADC. Also, one of the inputs connects to device's ground and I cannot put a resistor there.

The ideas with dampening resistors sound great. I think I'll do this. It'll be the smallest change to the design and from what you say I think it will work as desired.

Why then do people use ferrites instead of filters? Is it about component count, or something else? Also, from what I noticed, ferrites block only "quite high" frequencies, so filters should be more versatile anyway.

[T3sl4co1l]

The other main thing ferrites are used for, is dampening resonances on cables.

When a wire or cable is grounded to an enclosure, plane, multiple cables, etc., or RF-grounded through a capacitor, that low impedance termination reflects standing waves on the cable(s) in question.  This allows incident radiation to build up somewhat, at resonant frequencies.  To the circuit, this manifests as common mode noise, which affects the circuit directly, or inevitably gets converted to differential mode.

The ferrite bead spoils the resonance, acting as a lossy series impedance, a termination for the waves on the cables.  The improvement is usually modest, maybe 3 to 10dB, but that can make all the difference in a marginal design.

A typical example: say you have digital outputs leaving a device.  These should have low resistance, say so they can drive LEDs if necessary.  Then you can't afford to have series resistors for termination or damping.  And the drivers are lowish resistance as well (logic gates typically 30-70 ohms; bus drivers maybe 1/2-1/3 that), so parallel damping won't do anything (i.e., an R+C from signal to GND).  A ferrite bead in series, will present a modest impedance at high frequencies, softening the risetime and providing series termination, without affecting the DC current capacity of the channel.

Tim