Virtual ground when exceeding common mode voltage

[phasetrans]

Analogs dogs, hoping for an education:

Have a unity gain inverting opamp stage using a modern FET opamp that is rail to rail on the output, but not the input. The opamp is AC coupled on both input and output, non-inverting to ground. I am intentionally exceeding the common mode input voltage range at times as part of the sound.

What happens at the virtual ground point under these conditions with respect to the input current and bias current? Should I add a resistive path from the virtual ground to physical ground to help with good behavior?

Thanks in advance!

[merlinb]

Assuming you mean you're exceeding the CM range enough to actually clip the output then the virtual ground will start to track the input voltage. My guess is the internal FETs can handle voltages up to the rails, but if you're worried then you can add anti-parallel diodes between the two inputs terminals.

[phasetrans]

Assuming you mean you're exceeding the CM range enough to actually clip the output then the virtual ground will start to track the input voltage. My guess is the internal FETs can handle voltages up to the rails, but if you're worried then you can add anti-parallel diodes between the two inputs terminals.

Clipping the output voltage, yes. The opamp (ST TS27X) is rated to the rails for input voltage, but common mode voltage range is Vcc+ -1.5V.

The bias current is super low (pico A), but I was wondering if the input current would ever cause a problem if outside the common mode range for an extended period of time.

[merlinb]

The bias current is super low (pico A), but I was wondering if the input current would ever cause a problem if outside the common mode range for an extended period of time.

If you somehow manage to forward bias the JFET input diode then current will still be limited by the input resistor, so I don't think it would be a problem...

[R.G.]

Really, it depends on the opamp and virtual ground.

Opamps do a variety of things when you exceed common mode voltage range. Exactly what they do has been the subject of opamp improvements for decades, as memorialized in the advertising and specifications of the opamps where they claim freedom from various ugly things their predecessors did. There's probably not a single behavior for "modern" JFET opamp. Which opamp, and whether the maker specifies its behavior under excess common mode will matter.

You're putting the opamp in a situation where the output can't supply/sink the current at the inverting input to keep the differential voltage from inverting to non-inverting inputs at nearly zero. So the inverting input is pulled away from the non-inverting input, increasing the differential input voltage. Opamps vary a lot in how they respond to high differential inputs. Some can't stand it, and do odd things. Some phase invert, some have catch diodes to prevent this happening, some let current flow to the rails. Depends on the input structure. "Modern" JFET opamps are more polite in what they do about this, by design.

So point one - check the differential input mode range on your opamps. That will matter. Some opamps won't care at all, some will do funny things, and some will shunt current between the two inputs through catch diodes or some such. I think this last is what you were referring to.

If you're pushing/pulling current into/out of the virtual ground, it reacts... well, like it reacts. I did a paper on resistive virtual grounds early in geofex. I can look it up for you if you need it. The bottom line is that if you're using a resistive virtual ground like a resistor divider, the current pulls the virtual ground away from center by an amount determined by how much current you're push/pulling.  If you have an active virtual ground like an opamp output set to the desired voltage, it will hold much steadier. Ohm's Law tells you how much your virtual ground moves when pulled off-base. The series resistor on your inverting input limits how much current is pushed/pulled IF your input is one of those that shunts current from inverting to non-inverting.

That's important if you have other opamps using the same virtual ground. If you only have one, what's happening on it is sent out to the other opamps and such that use that virtual ground as an input. They all pick up and amplify the funniness caused by pulling the "ground" around. Using multiple "virtual grounds" is necessary in some situations, down to the place where each amplifier gets its own, minimizing interference between them.

You don't mention whether your input voltage is big enough to pull the inverting input out of the input differential mode range (which is what I think you're exceeding first) and beyond the input common mode range. Input common mode range really only exists in a follower that can pull both inputs close to the power supplies. Inverting amp circuits have both inputs at essentially the same voltage until the output clips, and only then can one or both inputs be pulled off the virtual ground/bias voltage.

[phasetrans]

So point one - check the differential input mode range on your opamps. That will matter. Some opamps won't care at all, some will do funny things, and some will shunt current between the two inputs through catch diodes or some such. I think this last is what you were referring to.

Yes, I was wondering what the opamp would do with the current internally when above the common mode voltage input range, when the output isn't source / sinking current to hold the virtual ground.

The opamp in question is the ST TS27XCDT https://www.mouser.com/datasheet/2/389/ts274-957304.pdf. The absolute max rating for input voltage is the same as the max supply voltage (18V). The datasheet footnote says "the magnitude of the input and output voltages must never exceed the magnitude of the positive supply voltage." The rated common mode input range is 0V to (Vcc+ - 1.5V)

I've attached an image of the stage in question. The driving voltage is from the same type of opamp on the same voltage rails.

[anotherjim]

Inverting input amps are easier to deal with, even parts with small input safety ranges can be made to work.
The trick is to drop the excessive voltage across the input resistor using the op-amp output feedback. Rf and Rin are a voltage divider between the input and the op-amp output. Rf and Rin are sized so that the middle of the divider (the -in pin) must always result in a safe voltage on the -in pin. So if both Rin and Rf are 47k, -in will see half the voltage between the opamp output (can't be larger than the supply rails) and the applied input voltage. The input voltage can be safely limited by using clamping diodes on the input before Rin to the supply rails - then the maximum overvoltage will only result in the -in pin being somewhere about half of total supply voltage. To prevent the input clamping diodes blowing the value of Rin can be split into 2 resistors with one part on the input side so there is some current limiting when a clamp diode conducts.

In the scheme you posted, a negative going input will result in the output via D3 clamping the -in pin to the reference and all the negative input voltage will be dropped across R1.
When the input goes positive, the op-amp output drives negative. Until it reaches the -V rail the -in pin sees the Rf/Rin divider voltage and close to the +in pin voltage even though the input should now be close to the +V supply. At this point, the chip is in no danger whatsoever and the input can carry on going positive to almost x2 of +V.

[PRR]

> what the opamp would do with the current internally when above the common mode voltage input range, when the output isn't source / sinking current to hold the virtual ground.

This ST TS27 is MOS input. Simplisitcally, the inputs are connected to glass insulators. There can be NO current in/out of the input pins. When the amplifier is overwhelmed, the "virtual ground" is not connected to anything in the amplifier, and will follow the usual rules for passive resistor networks.

If you are REALLY over-driving: the glass would break-down around 20V. Almost certainly there are protection diodes to the rails, which being 18V max tends to limit the risk of glass punch-through. If you are taking strange signals from outside sources (600W bass amp speaker?) then you want to consider the current rating of the protection diodes. If you are driving it with a similar amp on the same supply rails, you can't kick-up past the rails (or not much) so no damage likely.

I am assuming you "like" what this abuse does to the sound. If so, I don't see any point in worrying about violating the theory of simple (unabused) opamps and virtual grounds.

[phasetrans]

> what the opamp would do with the current internally when above the common mode voltage input range, when the output isn't source / sinking current to hold the virtual ground.

...There can be NO current in/out of the input pins. When the amplifier is overwhelmed, the "virtual ground" is not connected to anything in the amplifier, and will follow the usual rules for passive resistor networks...

A very lucid summary, thanks. I asked specifically because it was a MOS device. I have educational background, and (ancient) work experience building CMOS logic devices (but not analog stuff).

...Almost certainly there are protection diodes to the rails, which being 18V max tends to limit the risk of glass punch-through. If you are taking strange signals from outside sources (600W bass amp speaker?) then you want to consider the current rating of the protection diodes...

I guess this is the heart of what I should have articulated. "How are protection mechanisms actually implemented inside opamps, especially these MOS devices that have fragile gate dielectric but still have some ESD rating."

I am assuming you "like" what this abuse does to the sound. If so, I don't see any point in worrying about violating the theory of simple (unabused) opamps and virtual grounds.

Whether I like the sound is TBD. I liked the vibe from a nominally similar TI MOS-based opamp, and now I'm giving this ST part a try. But it all got me wondering if that TI "sound" is just from an internal zener (or similar) protection circuit as much as anything from the opamp output.

[composition4]

Hey R.G., did you manage to dig up the link to the resistive grounds article you wrote? Sounds like it might be worth a read for me.

Thanks
Jonathan

[highwater]

Probably this one: http://www.geofex.com/circuits/Biasnet.htm

[composition4]

Oh, that's not what I was expecting, I was thinking it was a paper on common mode considerations... I should probably read comments a bit more closely. Thanks for the link anyway!