Decoupling capacitor vs snubber

[hazuki]

Hi all. A basic question for you. I have an unwanted high frequency on a pin that I want to filter out. I can either put a decoupling cap on the pin to ground or I can install a snubber (RC circuit). Why would I want to do one vs the other?


Thank you!

[c4757p]

I have the usual basic question for people asking these basic questions: what's the application? Where does this HF come from, what kind of device is this pin attached to, and what kind of device is it?

That or we can just roll the solution dice.

[digsys]

I'm assuming you're coming in from the "outside world" to a MPU pin. You use straight decoupling when you need to preserve SPEED and
risetime etc. You use a snubber when don't care about delay (say 1-20mS) and want protection from over-voltage or severe contact bounce.
Usually you'd add a zener or clamp diodes as well in this case.

[c4757p]

Usually you'd add a zener or clamp diodes as well in this case.

Depending on the MCU and the application, a large series resistor could be enough to allow the ESD diodes to give protection.

[hazuki]

I have an MCU pin that is picking up noise from a switching converter on the board. I see about 400mv ptp of 100MHz in noise on that pin. My understanding is that an rc snubber will actually dissipate that HF but a decoupling cap will just couple the noise onto the ground plane.

Thanks

[c4757p]

The noise is relative to the ground plane, coupling it there is the same as getting rid of it. (Edit: This is only if you've grounded everything very well. Otherwise you're just coupling the noise into an inductor.)

Is this a signal pin or a power pin? And if it's a signal pin - what data rate?

[digsys]

My understanding is that an rc snubber will actually dissipate that HF but a decoupling cap will just couple the noise onto the ground plane. 

That's what the ground plane is for, as long as you've routed all the analogue / digital returns correctly. The RC will still return "noise" current to the
ground plane, albeit not as much. Both do the job, and as I said, depends on the SPEED of the signal you want to protect at that pin.

[gregariz]

The key difference between these two options are that the decoupling capacitor is a RF/AC shunt and the RC is a filter (ie low pass filter).

We use a shunt for the power pin (or any DC pin) because we want every bit of AC to ground.

We use an RC if we have a signal on the pin and don't want to lose it. Ie. if you have serial data, signal data or anything similar on the pin... use the RC.

[c4757p]

I'd argue that that isn't really a "key" difference, or really, not a difference at all. A shunt doesn't work unless it can pull against some impedance, in which case you've just made a LPF. A decoupling capacitor is just a LPF formed between the capacitor and the parasitic inductance of the trace.

You use a "shunt" for the power pin because a resistor would give a DC voltage drop. If it's just a DC signal you'll get better and more predictable performance with a proper RC filter.

[gregariz]

Sure.. you can use a LPF anywhere you use a shunt (including the power rail if you have the supply).

However, as a response to a beginner... a shunt is a reasonable approach for DC pins, and one you will see used all over industry.

I recommend the OP look at a few digital chip application notes. You will quickly find the type of decoupling commonly used for various pin types. Have a poke around the spec sheets of something like a serial ADC.. you'll see DC reference pins, signal pins and data pins.

[c4757p]

I don't think it's a good response to a beginner, especially if you don't define what you mean by "DC pins". If you have, for example, an op amp driving an ADC input, that could easily be considered DC if it's reading a slow value like a thermocouple temperature, but you most certainly do not want to stick a capacitor on it, no matter how tempting it might look because of the lack of voltage drop. Don't give misconceptions in the name of simplicity.

[madshaman]

How about filtering the output of that dc-dc converter so it's not a problem in the first place?

A quick question, I *always* decouple power pins on ICs as a matter of course, and use good NP0 caps (in parallel groups of increasing capacitance) on any supply to a precision analog device (like an instrumentation opamp).  Not because I'm experienced or anything, but so many application notes go into detail as to why this can be vital and I've always decoupled power pins by default like this.

Is there a cost savings to not decoupling be default like this?

[digsys]

  How about filtering the output of that dc-dc converter so it's not a problem in the first place?   

Often, you don't have any choice. You could waste hours or days trying to be "perfect", then at another load, there's a new resonance.
Naturally, you aim for best figures at reasonable time / effort.

   Is there a cost savings to not decoupling be default like this?   

In the scheme of things, cost of a few caps isn't what you should be striving for. IF you can't evaluate a noise tolerance range,
go for best practice, as you seem to be doing. Eventually, experience will tell you how much you can get away with.

[cthree]

Why would you not filter the dc-dc output with a ferrite bead?

[digsys]

  Why would you not filter the dc-dc output with a ferrite bead?   

IF the DC-DC supply is a separate unit, then that is an option, if needed. When the P/S is part of the same PCB, it's VERY tricky !
I have seen / put embedded ferrite IN a PCB in critical applications, but it's rare. Usually, you'd go for a common-mode filter.
Plus, it's not a guarantee that it will clean up ALL the noise anyway. You need to decide what part of the spectrum is important
to you. All the "tricks" have their pros and sometimes, cons.

[cthree]

I'm currently working on a project using an Atmel ATmega128RFA1 SoC which is very sensitive to noise. All the app notes and eval board schematics I've read all specifically say to use a ferrite bead on the power rail output to prevent injecting noise into the power rail. The app notes for similar Ti low power RF parts specify doing the same. The concept seems entirely sensible to me. Clean the noise where it originates and keep the power and ground planes clean (to a non-engineer like myself).

According to Atmel's AVR040 Appnote "A very important general rule is that all types of noise should be handled as close to the source as possible, and as far away from from the sensitive parts of a circuit as possible." Figure 4.3in that document describes decoupling with a series inductor and a cap. It's worth a read and can be had here: http://www.atmel.com/Images/doc1619.pdf

Both Ti and Atmel reference designs spec a Murata BLM15HG102SN1D but they have a whole family of EMI filters they call EMIFIL. You can look them up and see if anything there is appropriate.

HTH

[Kremmen]

Also the reference mentioned in Atmel's paper, i.e. Tim Williams' "EMC for Product Designers" is to be heartily recommended.
It really should be required reading for all who design circuitry for a living. Very useful to everybody else as well.