Below Rails Switch Ringing

[J_M_B]

Hi all

I'm just looking at a simple tactile switch with 10K pullup to 3V3 and 100nF cap to ground.
The switch doesn't bounce which is fine but it does ring.
Im not worried I just want to understand this.
I measured with very short gnd lead across cap so its not a bad probing issue.

Question is
- Can this not cause latch up? If a signal dips more than 0.7V below a supply rail?
- What causes the signal to dip below GND. Mismatch in drive strength and line characteristic impedance?




Any thoughts would be much appreciated
Jamie

 

[betwixt]

It can, and does cause problems sometimes. Consider a perfect system, the capacitor voltage would rise as it charged through the resistor and fall as it discharged through the switch, there is nothing to create a negative voltage so everyone is happy. Now look at a practical system, there will be inductance in the wiring and the components will have inperfections from pure resistance and capacitance. These are what causes the ringing. Almost all of it will come from the inductance of the wiring. As the switch closes the only thing limiting the discharge current is the wiring resistance and inductance, the current spike will be relatively large. The release of magnetic energy from the wiring resistance is what causes the negative overshoot. The simple fix is to slow down the capacitor discharge by wiring a resistor in series with the switch, obviously use a value small enough to ensure the signal is still recognised as 'low' but large enough to limit the current. Probably something in the range 47 Ohms to 100 Ohms would work well and as a bonus, it protects the switch contacts too!

Brian.

 

[FvM]

Discharging the 100 nF capacitor through the switch without a series resistor also reduces switch lifetime.

 

[D.A.(Tony)Stewart]

It was good that you used a short probe ground. I wonder how long the signal path and ground return is for the 0.1uF cap to supply.

Let's look at the ringing on the signal closely.
The overshoot is ~33% at a 1/2 division of 400ns/div= 200ns or 5MHz .
We know there must be a 2nd order LC effect.
At 5MHz , 0.1uF is 300m
The impedance at 3.3MHz for 0.1uF is approx 0.3 Ohm and the required L must be 10nH. This is a series resonance with the switch Rs, Cap and loop L which can be solved for the dampening factor or % overshoot in a 2nd order approximation. We could make it critically damped by making the switch resistance equal the resonant impedance at 5MHz by adding 0.3Ohms in series.

So what can cause 10nH? Trace impedance loop from Cap to supply and return, + vias if they exist or perhaps long leads on the cap with a low SRF. ( unlikely with SMT)

However to debounce the switch , like a pullup RC time constant must be longer than the bounce time (10k* 0.1uF= 1ms ) and if edge sensitive, either debounce in hardware (Schmitt Trigger) or software.

Increasing C would extend the RC rise time which is fast but also lower the LC resonant frequency and the impedance at resonance. Since the switch resistance might increase over time, you could increase C or add a series R or reduce the loop L to reduce overshoot. The protection clamp diodes have an ESR > 1 Ohm so your ringing source impedance is much lower than the clamp diodes.

So increase the source impedance with a series R between the switch and Cap would be the ideal solution to prevent latchup on the order of 10 Ohms for a 1us rise time. For metallic switches , quality contacts are gold plated and don't need a wetting current which can erode the plating. But non gold plated contacts can benefit from the surge current of 10% of the rated current.

So it depends also on the switch plating and the threshold for latchup. IF you exceed the device specs, it should be corrected.

 

[J_M_B]

Thank you all for your great answers.
I kept working after posting yesterday and blundered my way to the same answer. My thinking was it was a reflections issue so looking at circuit thought the only source (+V or GND connection point) that didnt have an impedance was the point the switch connected to GND.

So added 5Ohms between switch and GND to achieve results in the picture below.
So turns out my thinking was totally wrong but I fluked a good solution.

Really appreciate the detailed explanations, its really filled in a blind spot in my knowledge.

Thanks
Jamie