TVS with a small breakdown to clamping window?

[ACharnley]

Hi,

I have a circuit which clamps at 33V. The circuit components have an absolute rating of 40V.

Upon clamping there's occasionally ESD spikes, the maximum of which I've recorded is 43V lasting 50ns.

A TVS, which on paper sounds ideal, can have a clamping voltage of 36V however the breakdown voltage is somewhere around 25V. It is likely then the TVS will be conducting well before the clamp engages -which could destroy it.

Two other options are;

1. to lower the clamp voltage into the 25V region (yuk!)
2. to hope and pray the duration of the spike isn't enough to kill the IC's (unfortunately there's FET's in there and in my experience they don't suffer over-voltage stress).
3. use a 36v zener (which likely won't be fast enough).

Thoughts?

Andrew

 

[dknguyen]

If you have a 33V power rail then you can connect non-TVS diodes (i.e. schottkies) between your line and those rails so that they are normally reverse biased and forward bias during clamping. So the clamping mechanism is forward cvonduction rather than reverse breakdown like a TVS.

THe disadvantage compared to TVS is that the 33V power rail has to be powered for these to function, but the benefit is that you have a tighter, more current indepependent conduction voltage.

 

[rjenkinsgb]

How about adding some small ceramic caps to absorb and distribute some of the spike energy?
That should reduce the level and allow a slower device to clamp it, if needed.

A small inductor in the power feed may also reduce the level on any spikes or noise.

 

[ACharnley]

I tried a 36v zener across the line but it wasn't quite fast enough, even with additional capacitance.

Now I'm close, I added 0.1uF to the line (absorption) and a 10k resistor across the input. As it nears the 30-40v it shunts more current, enough to dampen the spikes.

I've ran it for a while under different conditions and spikes are all below 38v, which is promising. I am however wondering if a MOV can be found with characteristics closer to the zener which could replace this resistor and make it more effective. 38V is still too close for my sanity.