High BW Input Protection for OpAmp (Ins-Amp) Operational Amplifier

[Mechatrommer]

currently working on differential signal using instrumentation amplifier. i'm seeking advice on how to input protect the ins-amp and subsequent circuitry (output). this amp will take ±330V (direct mains) attenuated to ±3.3V by using simple smd resistors divider. to accomodate for high volt spike (assuming something bad happens in the mains line), voltage suppression diodes will be used (probably).

1) is using resistors divider alone as in schema (1/100X) enough? (ie i mean say 1KV spike will get divided to 100V 10V, i suppose ins-amp internal have protection to this "attenuated" level? or just simply saturated output)
2) is voltage suppression diodes in schema sensible? (spike will be routed to psu which is the one will probably be dead? well?)

i'm aware there's other more bulletproof method such as isolation transformer, or optoisolator (is there?), but i'm anticipating high BW type to the range somewhere greater than 10MHz (preferably up to 50MHz and beyond?). any advice on this?

(ps: ok before you complain, the 14148 will break down easily on ±330V input, i just quick sketch it drawing updated to remove confusing 4148. and the INA115 ins-amp is 100KHz BW only, but still i need input protection to 10MHz++ )

edit: ok now i see why putting suppression diodes on ±330V input to ±12V power supply is a joke, that is a stupid mistake. they should be put on lower/attenuated volt input (ins-amp) side

[jahonen]

I'd think twice before using this kind of circuit (but that said, I have done it myself in the past, although amplifier was just ordinary opamp connected as difference amplifier), but since you are going to do it anyway, put at least several, say, 5, 1M resistors in series, normal resistors can only take 250 volts or so. That also ensures that no single resistor failure will cause a direct hazard. That would not probably meet IEC61010 requirements but it is a start.

For functional protection, you can put  those 1N4148 clamping diodes to supply rails directly at amplifier input (where voltage is low), that should work ok.

Regards,
Janne

[Mechatrommer]

5x1M in series, thats the idea thanks. so are you suggesting 5Mohm input impedance or higher? isnt that bad for input noise/leakage? sorry i'm not a master in noise suppression and cant afford too much on very good pcb to test this more advance "black art" thing but i'll keep/take your advice in the next revision. i did put the diodes on the high impedance side (attenuated ins-amp input), but they roll-off the BW sooo badly in sim (around 1MHz on 50MHz circuit), i should get better faster respond and lower leakage diode, any suggestion?

although amplifier was just ordinary opamp connected as difference amplifier

bingo!

[jahonen]

I think the noise is related to what resistance opamp sees, it is not the same than those resistors in series. Of course, your equivalent input noise voltage will be quite high due to that you have quite a lot of attenuation anyway.

The idea of those series resistors is to keep user safe.

I think that the diode protection will work just fine, at least for the configuration I used in the past. Well, some peaking around 1 MHz. One may tweak those two capacitors to get rid of that. They should be equal valued. See attached simulations.

Regards,
Janne

[Mechatrommer]

ok thanx Janne, you gave me some idea and confirmation in security

[SeanB]

Using multiple resistors is good, but using 1N4148 diodes is not good for protection. You need low reverse current and low capacitance diodes specially made for high frequency work and meant for input protection. Better to make the input resistors leaded units, and use 3 or 4 low TC metal film resistors in series as input.  Try to get input impedance to around 1M total so use 3 330k resistors, which will be low enough current that the input diodes that you choose can handle it ( they will have a very low max current rating) and remember that they too will be light sensitive, you might need to place them under a shield to keep light out.