Protecting Op Amp

[cyberfish]

I have an op amp set up as a summer with inputs A, B, as shown in attachment.

B is a controlled input, I know it won't go out of range.

However, A can be anything (an oscilloscope's input, if you are wondering).

The Op Amp is powered by +-5V.

How would I go about protecting the op amp?

Adding a clamp at the - input doesn't make much sense to me, because it's a virtual ground, regardless of the inputs. Should I do it at the output then? How would I go about doing that?

BTW, the capacitor is for compensation. Probably doesn't matter. Including it in case it does.

Thanks

[waltr]

A common method is diodes to the power rails to clamp the input. Since there is a 1Meg input resistor you may want a smaller value resistor closer to the op-amp input and connect the diodes between the two series resistors.

Diodes to the rails will allow the input to go a diode drop beyond the rails and current will flow to the op-amp input pin. Check the op-amp data sheet for the max current allowed and size the smaller input resistor accordingly.

Or if the normal input signal is small (less than a diode drop) then back to back diodes to ground could work.

[cyberfish]

Thanks.

After some more thinking about it, I think I will just add 2 diodes to ground at the input.

My reasoning is that, under normal operation, the Vin- will always be a virtual ground.

But when input goes too high, and the output saturates to negative rail, the op amp will no longer be able to maintain the Vin- at 0 through the negative feedback, because the output cannot go any lower to sink more current, and Vin- will go up with A.

Therefore, adding the 2 diodes at the Vin- will be like an "exhaust", allowing the op amp to maintain it at 0 easily (and hence won't be damaged no matter how high A goes).

I think it's the simplest way. Makes capacitance easier to calculate, too (each 1N4148 has 4pF to ground). Schottkeys would be better protection, but the reverse leak is way too high for 1 mega ohm.

Did I miss something?

[waltr]

I think you got it.
Are you planning two back-to-back diodes or two in series to ground for a higher voltage before conducting?

The idea is to keep the Op-amp from getting hurt. If the input voltage is too high then the output is invalid anyway so who cares what the diodes do to the signal.

You did consider the capacitance of the diodes. This will limit the highest frequency but may not be much of a problem.

There is a way eliminate the affect of the diode capacitance by 'bootstraping' the diodes from the Op-amp output but this only works on a follower.  Though I'd mention it in case you need a lower C input sometime.

[cyberfish]

Back-to-back. I'm actually trying to go for a lower conducting voltage, but that's not important.

I'm not sure about the effect of capacitance. I'm aiming for an analog bandwidth of ~30MHz.

The feedback resistor and the diode capacitances can form a low-pass filter (and with an 1M feedback resistor, which I can't change because I need to keep the input resistance at 1M, it doesn't take much capacitance to kill the signal).

[Soeren]

Hi,

Adding a clamp at the - input doesn't make much sense to me, because it's a virtual ground, regardless of the inputs. Should I do it at the output then? How would I go about doing that?

Virtual ground won't do it.
Let's say the op-amps are driven from eg. 12V and imagine that you feed 1kV into "A", your Point B would then need to be at roughly 1kV as well to keep the op-amp able to keep the virtual ground (you only have virtual ground as long as the op-amp has output potential to cope).

You could split the 1M input and clamp at the split, but whichever way you choose, you have to measure and fine tune the input impedance if you wanna make sure it is 1M (and you want that).

If using regular diodes is a problem, a couple of JFETs can be used for the clamping, as they should exhibit a lower leakage and capacity.

[cyberfish]

Let's say the op-amps are driven from eg. 12V and imagine that you feed 1kV into "A", your Point B would then need to be at roughly 1kV as well to keep the op-amp able to keep the virtual ground (you only have virtual ground as long as the op-amp has output potential to cope).

Yeah that I figured, see my second post .

Right now I'm only worried about the capacitance. Not sure if it will affect the BW of the amplifier.

Attached is what I have in mind.

[cyberfish]

Just simulated my design for the first time, and turned out it won't work at all...

All high frequency (I'm looking at >10MHz) op amps have very low input resistance (~1MOhms, some as low as 300kOhms!).

Need to think of something else.

[waltr]

Oops, at least you find out in simulation.
Try out in simulation a three Op-amp instrumentation amp. For >10MHz and high impedance you will need some good op-amps and a very good PCB layout.

[cyberfish]

Yeah, definitely better than finding out $50 and 2 weeks later. I'm going to simulate everything from now on :p.

I'm looking into Op amps with JFET input stage. They have input impedances in the range of teraOhms, and very low (1pF) input capacitance.

They are quite a bit more expensive, but I'm wondering if they have other disadvantages. Otherwise they just sound like exactly what I want.

[Soeren]

Hi,

Just simulated my design for the first time, and turned out it won't work at all...

Don't take a sim as the truth.

All high frequency (I'm looking at >10MHz) op amps have very low input resistance (~1MOhms, some as low as 300kOhms!).

LF357 is a good old 20 MHz op-amp (actually, I think it's more or less out of production, but plenty of shops have them).
OPA2889  goes to 60 MHz. Remember, that the spec'd frequency is the the G/BW product so this is where Av=1, but this should still make the OPA2889 OK for app.

[cyberfish]

The LF357 looks great, except the slew rate may be too slow (12V/us). The OPA2889 has pretty low input impedance (3.5Mohm).

I'm looking at the OPA659. Massive overkill with gain bandwidth product of 650MHz and JFET input, 2550V/us slew rate, but it's not TOO expensive, and I hope I can get free samples (TI has sent me samples before).

[Soeren]

Hi,

The OPA2889 has pretty low input impedance (3.5Mohm).

Thought you said you needed 1MOhm?

I'm looking at the OPA659. Massive overkill with gain bandwidth product of 650MHz and JFET input, 2550V/us slew rate,

And then you'll have to figure out compensation. It's quite common thinking to just go for max. oomph, but it's a bit like using a fighter plane to go to the local mall - you have to land before you have taken off and you'll still go to far.
A 650MHz op-amp might give you all sorts of headaches - I'll still recommend the 2889... At least give it a spin.

[cyberfish]

Oh no, I was only complaining that the one I had only has 1MOhm, I want much more.

I can order some 2889 with my parts order, but I'll have to design the circuit around one (since I'm assuming they have different pinout), and I think the 659 has a higher chance of working (due to the higher impedance). I'm getting a few free samples from TI, too.

The op amp datasheet says it's unity gain stable, and I'll only use it as a buffer (voltage follower). I'm assuming it will be fine?

[cyberfish]

It has been done.

http://cyberfish.wecheer.com/blog/wp-content/uploads/2010/06/scope.png

All comments/suggestions welcome! (before I start layout)