op-amp nV input offset voltage

Hi,
As you know, the *input* offset voltage is the voltage required across the op-amp's input terminals to drive the output voltage to zero.
Although it has been my experience that for most op-amps the input offset voltage is due to the "-" input pin for the *most* part. For example, according to Spice the input offset voltage on the "+" input pin on a LMC660A op-amp for a non-inverting amp circuit is a few nanovolts, disregarding thermoelectric effects mind you, but a few millivolts on the "-" input pin. Although as you know the input signal is not applied to the "-" input pin for a non-inverting amp circuit, which means there's just a few nanovolts on the input of such a circuit if we disregard thermoelectric effects.
I have a INA116PA Instrumentation op-amp where Ib typ = 3fA, Ib max 25fA, and Vos typ = 0.5mV. Now it seems to me in order for there to be 0.5mV on the input of this Instrumentation op-amp circuit with 3fA bias current that the DUT input impedance would have to be 0.50mV / 3.0fA = 170 Gohms. On the other hand, if the DUT input impedance is say 200 Kohms then would the input offset voltage be 3.0fA * 200Kohms = 0.6nV, disregarding thermoelectric effects?
INA116PA datasheet: http://focus.ti.com/lit/ds/symlink/ina116.pdf
Regards, Paul
Add pictures here
โœ–
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
wrote:

The offset voltage is *differential*. You can blame it on either pin, or both pins... it doesn't matter who you blame, the result is the same: offset voltage becomes measurement error.

The offset voltage error is a different thing from the input bias current. They are unrelated [1]. You can of course generate a real, external-to-the-opamp error voltage by dumping the bias current into real external resistance, but that's a different matter entirely.
John
[1] Some opamps have low offsets and high bias currents, and some vice versa. Chopper amps are low on both; cheap bipolars are high on both.
Add pictures here
โœ–
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Jun 21, 8:41 am, John Larkin

Here's my main concern. If I build the INA116PA for DC application, which is an internal Instrumentation op-amp chip (3 op-amps), and the impedance of my DUT is 200 Kohms, then what bias currents could a good EE such as yourself expect? I mean, for a 200K ohm DUT input source we cannot have both 0.5mV offset and 3fA bias on the DUT. I think V=I*R applies, so if the bias current is 3fA then V = 3fA * 200Kohms = 0.6 nV.
Thanks, Paul
Add pictures here
โœ–
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Sat, 21 Jun 2008 08:59:10 -0700 (PDT), the renowned Paul

Think of the "offset voltage" as something like an internal battery (ideal floating voltage source*) of voltage equal to the "input offset voltage" connected in *series* with one of the inputs. End of story.
Now, bias current can be thought of as current sources (or sinks) connected to each of the input pins. The value of the bias currents on each input need not be similar to the other (on bipolar op-amps it often is fairly well matched, on MOSFET input op-amps where it's just leakage, it could be anything). They are unrelated to the "input offset voltage", and voltage drop resulting from input source resistance could add or subtract from Vos.
So, as you can see, the bias current certainly can be xx fA and offset voltage can be xx mV, regardless of source resistance.
In your example, with source impedance of 200K, and if Ib is really 3fA (including PCB leakage) then the bias current effect is negligible.
* of course it will vary from unit to unit, and is a function of temperature, time and (sometimes) previous history of differential voltage, but to keep it simple as a first approximation, it's constant (different from unit-to-unit, and may have a 'hole' in the distribution if you choose to buy a low grade unit), as a better approximation, it is a function only of temperature.
Best regards, Spehro Pefhany
--
"it's the network..." "The Journey is the reward"
snipped-for-privacy@interlog.com Info for manufacturers: http://www.trexon.com
  Click to see the full signature.
Add pictures here
โœ–
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Paul wrote:

Incorrect, one *can* have eve 3mV offset and 3fA bias. Learn what an ideal op-amp is, then learn about each of the various real-life error components.
Add pictures here
โœ–
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

My apologies if this has been covered in the other branch of this thread, which I don't have time to fully read...
As others have pointed out, the input bias current and input offset voltage of the part are characteristics of the part. HOWEVER, the external circuit strongly influences how well you can take advantage of those characteristics. That is, the external circuit can completely wipe out the potential benefits of either a low bias current or a low offset voltage or both, even. For example, I used a chopper-stabilized op amp to amplify the output of a diode RF detector. The op amp has typically a pA of input bias current and about a uV of input offset voltage. First, I had to be very careful to guard the detector traces against currents leaking in from outside, and then I had to be careful that the resistance between the guard trace and the detector output trace (feeding the op amp input) was high enough that a 1uV offset would not result in a current as large or larger than the pA op amp bias current. For this part, that's only a megohm or so, fairly easy to do, but in an earlier design using a non-chopper amp where the offset voltage was up to a millivolt or so, it was a killer. The RF detector diode is shunt between the guard and the amplifier input, with RF fed in through a capacitor, and the zero bias RF detector diode shows considerable current at a millivolt.
Seems like there should be some good references on applying low offset, low bias amplifiers. I know that Bob Pease has had some good articles on the trials and tribulations of testing amplifiers down in the fA region--not trivial! His articles can be found with a search on the web...
Cheers, Tom
Add pictures here
โœ–
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Tom Bruhns wrote:

One good set is Jim Williams' 2001 trio of articles on his 1 ppm accuracy DAC. They're on the EDN web site someplace--search on "1 ppm" and "20 bit DAC", and they'll come right up. He even goes into details like the thermocouple constant of two different vendors' copper wire: +-0.2 uV/K, believe it or not--as big as Cu:Au or Cu:Ag.
Cheers,
Phil Hobbs
Add pictures here
โœ–
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Jun 21, 8:41 am, John Larkin

The LMC660A has a typical voltage offset of 1mV and bias current of 2fA, but that depends what type of op-amp circuit. According to Spice the input voltage offset for an inverting or differential circuit is about what the Vos spec says, but for a non-inverting circuit it's a few nanovolts on the "+" input pin. I'm wondering if the Vos in datasheets is referring to a certain type of op-amp circuit such as the inverting type (http://hyperphysics.phy-astr.gsu.edu/Hbase / Electronic/opampvar.html#c2).
Regards, Paul
Add pictures here
โœ–
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
wrote:

No. Those are characteristics of the chip.
The actual error voltage that a *circuit* generates is made up of several contributors. One is the input offset voltage of the chip itself. An additional error is any voltage drops created in external resistors by the opamp bias currents. The errors are generally assumed to add, because we can maybe know the polarity of the bias current from the datasheet (but often we don't even know that) and we never know the polarity of the offset, unless we measure it on one real opamp.
According to Spice

I don't think that's right. The Spice model shouldn't be that stupid. Sounds like an interpretation issue.
I'm wondering if the Vos in

No, it's a property of the chip, not the circuit it's used in.
John
Add pictures here
โœ–
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Hello Paul, Maybe it helps if you think about the transistor circuit of an opamp.
The first stage of an opamp consists of a differential amplifier made by a pair of two well matched transistors. The difference of the Vgs(Mosfet opamp) or Vbe(bipolar opamp) of these two transistors in the input stage is the main contributor for the offset voltage.
Offset voltage is always measured between the + and - input. What you have measured at the +input is the bias(leakage) current multiplied by the value of the resistor connected to the +pin.
Best regards, Helmut
Add pictures here
โœ–
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
wrote:
Newsbeitrag

I appreciate all of the replies! All of these years I've had this false idea about the datasheets Vos burnt into my head. I've always assumed that if the datasheet said the op-amps Vos was say 50uV then that's the lowest input voltage (by my def: the voltage applied on the input device due to the op-amp) one can expect with a typical op-amp circuit such as an inverter or non-inverter.
So it's true that one could achieve input voltages in the nanovolt region on a 200K ohm DUT from an Instrumentation op-amp chip such as INA116PA even though the datasheet Vos spec is 2mV?
Thanks, Paul
INA116PA datasheet: http://focus.ti.com/lit/ds/symlink/ina116.pdf
Add pictures here
โœ–
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Hello Paul,
Yes you can apply voltages as small as you like. they will be still amplified by the gain G, set with the feedback resistors. The drawback of any Vos is that you will have an output voltage of (Vos+Vin)*G . This menas you have to either adjust the offset voltage already at the input or you have to subtract Vos*G at the output.
Best regards, Helmut
Add pictures here
โœ–
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
wrote:
Newsbeitrag

Thanks! As you said the output offset can always be corrected, but it's great to know that a 2mV op-amp chip such as the INA116PA can apply DC voltages as low as a few nanovolts on the input device without adding shunt resistors. Of course one can always add a shunt resistor to lower the input voltage across the DUT, something I knew about, but of course that has obvious effects of decreasing the DUT's effective input voltage to the op-amp.
I'm wondering if there are any op-amps or perhaps a BiFET amp circuit that could achieve a few nanovolts across say a 200K ohm device while consuming no more than a few microwatts. The idea is that such a microwatt amp would have considerably less input thermoelectric effects. Thermoelectric effects can generate a half dozen or more microvolts on the DUT unless carefully balanced with dummy resistors. I believe Linear Tech has some microwatt op-amps, but nothing near 25fA bias current.
Thanks, Paul
Add pictures here
โœ–
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Hi,
I'll try to clarify:
I am referring to the input voltage on the *DUT* caused by the op-amp, and therefore if the bias current through the DUT is decreased then the offset voltage on the DUT will be less-- ohms law.
The op-amps I am working with have offsets around 0.5uV to a few uV. Therefore thermoelectric effects should be considered. As far as I know instrumentation op-amp appear to have to least thermoelectric effects since both input pins go to the same polarity on both op-amps, the + pin, but there are still thermoelectric effects since both op- amps are not 100% identical. Other circuits such as the inverter require dummy resistors and such to help reduce the thermoelectric voltages on the DUT.
My interest in BiFET's is to design a low power amp circuit with low bias current.
Thanks, Paul
Add pictures here
โœ–
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
wrote:

To prevent thermoelecric voltages, keep all pins at the same temparature. But also all surrounding resistors etc. A cooling airflow gives temperature differences. And resistors and opamps that get warm may give some effects.
Pieter
Add pictures here
โœ–
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Pieter wrote:

This is especially true of temperature gradients across devices that aren't metallic throughout, e.g. metal oxide resistors and semiconductors. The thermocouple constants between silicon and any metal are up in the 700-1500 mV/K range, depending on doping---which dwarfs the slope of any metal/metal thermocouple.
Cheers,
Phil Hobbs
Add pictures here
โœ–
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Paul wrote:

The last time I checked, the offset voltage would be the difference between the (-) and (+) input with the Op-amp in (-) loop back mode.
So, if you were to put an op-amp in (-) loop back and lets say 5 volts on the (+) input, the (-) input should be offset no more than what the spec's state.
Or, I guess if you were using a +/- to common supply, you can simply tie the (+) to common with op-amp in (-) loop back. You should be seeing that offset factor at the (-)/output..
Maybe things have changed but that is what I go by..
http://webpages.charter.net/jamie_5 "
Add pictures here
โœ–
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Polytechforum.com is a website by engineers for engineers. It is not affiliated with any of manufacturers or vendors discussed here. All logos and trade names are the property of their respective owners.