Hi All,

I would like to choose an op amp that will be configured as a trans-impedance amplifier. A 5V source is driving a variable resistance, from 20KOhm down to 500Ohm, at up to 100kHz (voltage frequency, not resistance changes). Furthermore, it is very important that the voltage at the noninverting input is as close to zero volts as possible, say in the nanovolts range (the virtual ground is essential for circuit stability).

To this end I have been looking at op amps with an inherently low input offset voltage, and also perhaps those with the ability to trim said voltage. Will manually trimming the offset voltage load the input in any way (its very important that I can precisely calculate the variable resistors resistance)?

In the end I need some advice on choosing an amplifier because there are countless op amps to choose from and I'm not particularly experienced!

Cheers

EDIT: It's for a PCB implementation, hand soldered and hot air gun, so a surface mount where the pins can be accessed is ideal

 

Advise you look carefully at noise analysis for this design, that
will overwhelm your nanovolt requirement and drive the design.

Google "nanovolt input offset opamp", many hits on this topic.

An ap note of considerations, attached.

Regards, Dana.

 

I based a transimpeadance amplifier design on the obsolete OP-41; This http://www.analog.com/en/products/a...age-op-amps-greaterthanequalto-12v/ad549.html is it's suggested replacement.

It's not an SMT part and not low voltage. I did use a teflon insulated post for the inverting input.

 

Choose Rf to get the full dynamic range.
Your input current is 10mA to 25uA, if amplifier runs from 12V, then 1V/mA is achievable and the feed back resistor (Rf) will be 1K. At 25uA you get 25mV.
Lets say you want 10bit accuracy. You need less than 24uV (25mV/1024) noise across the 200kHz bandwith, an amp with less than 6nV/sqrtHz is needed.And if you want to measure 25mV @ 10bits, the offset voltage needs to be under 24uV. Bias current will need to be 25nA maximum.
There will be some shunt capacitance (Cin) at your input. For stability:

Now you need an amplifier that can easily manage to output 10mA into 1K @ 200kHz. An LT1007A might be a good starting point.
It might not be possible to get all the features you want from a single stage amplifier. Composite amplifiers can help by putting a DC precise amp and a fast slewing amp together:

If you are digitising your signal you can deal with offset and gain errors in software.

 

Advise you look carefully at noise analysis for this design, that
will overwhelm your nanovolt requirement and drive the design.

Input voltage noise (and current noise, too) is just one of several problems that are going to overwhelm the nanovolt requirement:

• Even the very best chopper-stabilized op amps (one that I've used is the MAX44241) can only achieve input offset voltages of around a microvolt; I know of none that achieve nanovolt-level input offsets.
• Thermoelectric effects, both on-board and off-board, can produce error voltages of many microvolts which vary with temperature. Remember, EVERY place two dissimilar metals join (solder joints, switch contacts, IC sockets, etc.) is a thermocouple, and in any design dealing with low-level signals these effects can be a major source of error unless extreme care is taken.
• Finite op amp open-loop gain means that the summing node of a transimpedance amplifier will NOT remain exactly at zero volts at all times; it will be equal to the amplifier's output voltage divided by the gain. For example, an op amp with an open-loop gain of a million, but which is otherwise ideal, will have an input voltage of 1 μV for every volt of output.
• Finite op amp gain-bandwidth product makes the above problem much, MUCH worse at 100 kHz. Above DC, the gain of an op amp varies inversely proportional to frequency. An op amp which has a gain-bandwidth product of 10 MHz (which is a fairly fast amplifier) has an open-loop gain of only 100 at 100 kHz; thus, if the op amp's output swings 10 volts peak-to-peak, the summing junction will be seen to swing 100 mV peak-to-peak, a far cry from the nanovolts the TS is seeking.
• Resistors, like all other components, have tolerances (i.e., uncertainty as to their exact value) and these will limit the accuracy of any circuit. 5% and 1% resistors are common and very inexpensive; 0.1% resistors are readily available but can cost several cents; 0.01% resistors can cost several dollars; and 0.001% resistors can run up to $50 or even more.
• Resistors have temperature coefficients, meaning that their resistance varies with temperature. Most common, inexpensive 1% metal-film resistors have specified tempcos of ±100 ppm/°C (±0.01%/°C). Resistors with zero temperature coefficient are available, but cost upwards of $50.

There are probably other "gotchas" that I've overlooked, but these are the considerations I've had to deal with in precision, low-level stuff.

 

You might consider the ICL7650 chopper stabilized amplifier, but you will probably need a negative power supply.



You can add an offset adjustment if you like in your case of a transconductance amplifier by summing a tiny offset current into the summing node.

 

Cheers guys, I really appreciate the design resources, I think i can probably figure it out now, the device is for a medical grade device, so I'm lucky in that 'cost is no issue' also.

I do have another question, regarding data acquisition. I have about 5 of these variable piezoresistive and op amp setups, and I am uncertain as to whether I should use the onboard ADC of the MCU, or if I should utilize 5 external ADC's.

As far as I can tell, the issue with external ADC's would be the bottleneck in communication with the MPU, and if performance gains are not significant, is there any improvements worth noting?

(The design is going to be scaled up in the future, so it could be that an external number of ADC's, up to 20, may not really be beneficial).

Cheers