Keeping Voltage on Op-Amp Input(s) When Unpowered

[bostonman]

I have a question regarding an op-amp (TLV9144).

Below is a section from the datasheet that I believe explains my question, but not sure I fully understand. In a circuit I'm laying out, a condition will occur where the inputs of the op-amp has voltage but Vcc is low or 0V.

Maybe I'm wrong, but believe one of the LM324 datasheets stated this condition is safe (although I can't find it now) whereas the datasheet for this op-amp doesn't seem to be so specific. My thought is to add a 10k series resistor on the inputs to protect the internal diodes, but, if I understand the datasheet, providing I have a high impedance Vcc line, the op-amp can have voltage on the inputs without Vcc. One reason the paragraphs from the datasheet don't make sense: would all power circuits for Vcc be high impedance?

On a side note, the reason I went from the LM324 to this is due to needing a rail-to-rail.


Another common question involves what happens to the amplifier if an input signal is applied to the input while the power supplies +VS or –VS are at 0V. Again, this question depends on the supply characteristic while at 0V, or at a level below the input-signal amplitude. If the supplies appear as high impedance, then the input source supplies the operational amplifier current through the current-steering diodes. This state is not a normal bias condition; most likely, the amplifier will not operate normally. If the supplies are low impedance, then the current through the steering diodes can become quite high. The current level depends on the ability of the input source to deliver current, and any resistance in the input path.

If there is any uncertainty about the ability of the supply to absorb this current, add external Zener diodes to the supply terminals; see Figure 6-6. Select the Zener voltage so that the diode does not turn on during normal operation. However, the Zener voltage must be low enough so that the Zener diode conducts if the supply terminal begins to rise above the safe-operating, supply-voltage level.

The TLV914x input terminals are protected from excessive differential voltage with back-to-back diodes; see Figure 6-6. In most circuit applications, the input protection circuitry has no effect. However, in low-gain or G = 1 circuits, fast-ramping input signals can forward-bias these diodes because the output of the amplifier cannot respond rapidly enough to the input ramp. If the input signal is fast enough to create this forward-bias condition, limit the input signal current to 10mA or less. If the input signal current is not inherently limited, an input series resistor can be used to limit the input signal current. This input series resistor degrades the low-noise performance of the TLV914x. Figure 6-6 shows an example configuration that implements a current-limiting feedback resistor.

[magic]

My thought is to add a 10k series resistor on the inputs to protect the internal diodes, but, if I understand the datasheet, providing I have a high impedance Vcc line, the op-amp can have voltage on the inputs without Vcc. One reason the paragraphs from the datasheet don't make sense: would all power circuits for Vcc be high impedance?

Same question asked differently: would voltage on your Vcc rail increase a lot if a small current is injected into it through the chip's internal diodes?

Probably not. A resistor in series with the input will limit current even if the rail is heavily loaded, packed with capacitors and whatnot.

[johansen]

the vcc is going to be 1 diode drop lower than the voltages at the inputs because the esd protection diodes dump it right into the rails. (neglecting the internal resistance of the esd diodes)

they suggest limiting this current to 10ma, so your 10k resistor can probably be reduced to 2k ohms, assuming you have 18 volts available.

no, not all VCC rails for opamps are high impedance. usually they are low impedance, but if the power supply is turned off to save power, its impedance may be very high, hence the concern about over-volting the powersupply or opamp.

[Berni]

Most opamps have internal diodes going to GND and Vcc on their input. So you just need to limit current enough to not damage it.

However there are exceptions! Some opamps are missing one or both of the ESD diodes for the purpose of improving input performance. On these opamps it is perfectly fine to place a few volts on the input with the opamp unpowered. The inputs usually stay high impedance too and draw near 0 bias current (tho it might not necessarily still meet the fancy picoamp specs in the tables). This is very useful for selecting powering down like that. However it also means it is easier to blow up the input, so extra care must be taken in some cases to protect it externally.

[bostonman]

In my setup, I will have 12V Vcc and near 12V at the inputs (hence why I want a rail-to-rail).

Due to RC circuits on the input, Vcc should fall to 0V sooner than the inputs. Worst case, the full 11V or so volts on the input may be present when Vcc is 0V.

The datasheet wasn't explicit about this being acceptable. Obviously it mentions limiting current but I wanted to double check that my design of a 10k series resistor was fine; or lower it to 2k as it was suggested.

[Berni]

Well the power supply also has capacitors on it that slowly drain, so the capacitor in the RC just makes the supply stay up slightly longer by discharging gradually trough the ESD diodes.

But as long as you are not trying to milk all the performance possible from an input on a chip it is always a good idea to just add a series resistor for protection. By limiting the current it lets the chip survive way more abuse in the input. Useful for any IO pins that go outside a product and could become shorted to higher voltages.

[bostonman]

But as long as you are not trying to milk all the performance possible from an input on a chip it is always a good idea to just add a series resistor for protection. By limiting the current it lets the chip survive way more abuse in the input. Useful for any IO pins that go outside a product and could become shorted to higher voltages.

Good to know.

Until recently, I wasn't aware about damaging the inputs in such manner. Once I saw a datasheet for (I believe) the LM324 that stated it's safe to have voltage on the inputs without Vcc, I began looking explicitly for similar notes on this datasheet.

Seems it's not necessarily a matter of whether the datasheet states this, but just a matter of reducing current through ESD (and any other) diodes by using series resistors.

In my circuit, Vcc should drop fairly quickly but the inputs have an RC of >2s starting at near the rail voltage.

[peter-h]

I recall some designs having a diode in series with VCC, so that if there is reverse current in VCC (as a result of the upper input protection diode conducting) it does not meet with a significant load.

But then such an input excursion will end up powering the entire op-amp so you have to ask what is its output pin driving?

[magic]

Until recently, I wasn't aware about damaging the inputs in such manner. Once I saw a datasheet for (I believe) the LM324 that stated it's safe to have voltage on the inputs without Vcc, I began looking explicitly for similar notes on this datasheet.

Seems it's not necessarily a matter of whether the datasheet states this, but just a matter of reducing current through ESD (and any other) diodes by using series resistors.

Some chips like LM324 and LM393 can accept input voltages up to 30V completely unpowered because no current flows into their pins under such conditions. In digital jargon, you would call that "30V tolerant inputs".

Majority of chips have diodes from inputs to VCC and absolute maximum ratings on "input current", which limit how much current can be forced into the pins.

[ArdWar]

LTC (ADI) have a whole line of "Over The Top" opamps (e.g. LT6016, ADA4097 etc) that are specifically designed to do just that, working with cm>>Vdd and surviving any Vin>>Vdd.

Many current sensing amplifier designed for highside may also able to do that, although some of their gain and level shifting configuration may be a bit too specific to be easily "misused" as normal opamp.

[SCSKITS]

Most inverting circuits with op-amps have a resistor in series with the input so those would limit any current.
If using a non inverting circuit, a small resistor (maybe 1K) in series with the input is all the protection you should need.

[peter-h]

I looked up the LT6016 and cannot see how it can work as an op-amp above the +ve rail



I can see it can withstand high voltages if you do something in e.g. the I1 current source.

AFAICT, solutions for Vcm above the +ve rail have to involve resistors e.g. the LTC2862; no schematic published but it can easily achieve it with a resistive divider on the input(s).

[Ian.M]

The LT6016 works above the V+ rail because as the PNP long tailed pair  cuts off, its tail current source transfers via Q9 and is reflected by the current mirror Q11,Q12  to bias the dual current mirrors Q4,Q5 and Q3,Q6.  The input voltages on their emitters differentially steer the reflected bias current between the two mirror  collector outputs, which re-joins thje signal path from the long tailed pair collectors.  As Vin drops below V+ by a bit under a volt, Q9 cuts off, disabling the differential mirrors and the long tailed pair takes over again.

The main limitation is the reverse Vbe breakdown voltage of the long tailed pair.  Its obviously custom engineered to have a much higher reverse breakdown voltage than any normal discrete transistor.

[peter-h]

OK; thank you. Very clever!

I too could not see how they did the Vbe issue which is normally about -5V max. I did a dig around and could not find much info that Vbe (rev) can be achieved above about 10-15V. But a trivial way is a diode in series with the base...

[bostonman]

The LT6016 works above the V+ rail because as the PNP long tailed pair  cuts off, its tail current source transfers via Q9 and is reflected by the current mirror Q11,Q12  to bias the dual current mirrors Q4,Q5 and Q3,Q6.  The input voltages on their emitters differentially steer the reflected bias current between the two mirror  collector outputs, which re-joins thje signal path from the long tailed pair collectors.  As Vin drops below V+ by a bit under a volt, Q9 cuts off, disabling the differential mirrors and the long tailed pair takes over again.

I am always amazed at the knowledge people have on this site. While I'm an "engineer", sometimes I'd like to know half the stuff some of you have forgotten.

[Salitronic]

Most modern rail to rail op amps including the TLV914x have input to rail ESD diodes, so with VCC at 0 V the input will back power the device unless the current is limited. The datasheet warnings are about current, not voltage. A small series resistor on each input is the correct fix and should be sized so the injected current stays well below the 10 mA limit, where 2 kΩ to 10 kΩ is fine at 12 V. LM324 type parts are an exception because their inputs are effectively tolerant when unpowered.

[shapirus]

Protecting outputs of unpowered op amps should be considered too, if it is possible that a potential may be applied there when the op amp is unpowered. I'm not sure if there are (or if there are always) builtin protection diodes between outputs and VCC, so it may be desirable to add an external one.

[bson]

This can happen if the there are multiple rails which on power-on become stable at different times.  Buck and boost converters always take some time to start up.