How to select appropriate bias current resistor for instrumentation amps

[AlessandroAU]

Hi all, I have a project where I am trying to measure vibrations from beam mounted PZT sensors. I'm using an AD8253 which is a digitally controlled instrumentation amplifier.

In the datasheet it specifies a requirement for input bias resistors in the input stage of the amplifier. Since I am taking measurements from very high input impedance piezo sensors I want to keep the input impedance as high as possible. Since input bias resistors directly effect input impedance I want to keep these values as high as possible.

What determines what values of input bias resistors I can use? the input bias of the ad8253 is 50na.

http://www.analog.com/static/imported-files/data_sheets/AD8253.pdf

[Galaxyrise]

Are you referring to this part of the datasheet?

An external resistor should be used in series with each of the inputs to limit current for voltages greater than 0.5 V beyond either supply rail

Are your inputs outside the opamp supply voltages?

[AlessandroAU]

This is the part I am referring too.  The input bias currents need a path to ground or else the output will gradually float to one of the rails. I want to know how I can work out what resistance I need.

My inputs are in the <100mv range

[tszaboo]

Can't you just ground the negative input? Or connect it to ground with 10K and leave the positivefloating?
INA128 datasheet figure 3.

[mzzj]

Hi all, I have a project where I am trying to measure vibrations from beam mounted PZT sensors. I'm using an AD8253 which is a digitally controlled instrumentation amplifier.

In the datasheet it specifies a requirement for input bias resistors in the input stage of the amplifier. Since I am taking measurements from very high input impedance piezo sensors I want to keep the input impedance as high as possible. Since input bias resistors directly effect input impedance I want to keep these values as high as possible.

What determines what values of input bias resistors I can use? the input bias of the ad8253 is 50na.

http://www.analog.com/static/imported-files/data_sheets/AD8253.pdf

how much gain yuo are using?

Lets say you have gain of 100, max 100mv input signal and 15v supplies:
You have keep the input bias current effects enough small so that your opamp output doesnt saturate.
If you want to measure 100mv signal and the bias current causes 500mV offset at the input things wont go well since at 100x gain your opamp would try to get 50 volts at the output and its outside of the possible output range.
20mv offset might be acceptable, so resistance works out to 20mV/50nA=400kohm

[wraper]

say you have gain of 100, max 100mv input signal and 15v supplies:
You have keep the input bias current effects enough small so that your opamp output doesnt saturate.
If you want to measure 100mv signal and the bias current causes 500mV offset at the input things wont go well since at 100x gain your opamp would try to get 50 volts at the output and its outside of the possible output range.
20mv offset might be acceptable, so resistance works out to 20mV/50nA=400kohm

Do not forget that this is instrumentation amp and both inputs need to be biased. Therefore it's not abut that offset on input but offset difference between inputs.

[dannyf]

Those resistors are there to limit current though the clamping diodes when the inputs are overloaded.

to minimize the impact of input bias current offset, they are usually equal.

[mzzj]

say you have gain of 100, max 100mv input signal and 15v supplies:
You have keep the input bias current effects enough small so that your opamp output doesnt saturate.
If you want to measure 100mv signal and the bias current causes 500mV offset at the input things wont go well since at 100x gain your opamp would try to get 50 volts at the output and its outside of the possible output range.
20mv offset might be acceptable, so resistance works out to 20mV/50nA=400kohm

Do not forget that this is instrumentation amp and both inputs need to be biased. Therefore it's not abut that offset on input but offset difference between inputs.

Yes, this would be more correct way.
Select  same size of resistors for both inputs and hope that input bias currents are roughly same so that offsets would counter-act each other.
The opamp in question has a max 40nA input offset current so the end result for maximum resistor value is roughly same anyways.

[mzzj]

Those resistors are there to limit current though the clamping diodes when the inputs are overloaded.

to minimize the impact of input bias current offset, they are usually equal.

Did you bother to look at the picture on 3. post?

[David Hess]

I usually pick maximum input resistance based on either doubling the voltage noise or doubling the offset voltage.

For broadband noise that would be about 10nV/2pA or 5 kohms (2.5 kohms at each input) which is about right for a good bipolar input stage.

For input offset voltage it will be about 200uV/40nA or again, 5 kohms but since you input is effectively AC coupled, this is not a limitation.

The AD8253 is really the wrong kind of part for this.  You should be using an FET input or maybe picoamp bipolar input stage if you are interested in good low frequency response from a piezoelectric transducer.  If you like Analog Devices then maybe something like an AD8421.

[AlessandroAU]

I usually pick maximum input resistance based on either doubling the voltage noise or doubling the offset voltage.

For broadband noise that would be about 10nV/2pA or 5 kohms (2.5 kohms at each input) which is about right for a good bipolar input stage.

For input offset voltage it will be about 200uV/40nA or again, 5 kohms but since you input is effectively AC coupled, this is not a limitation.

The AD8253 is really the wrong kind of part for this.  You should be using an FET input or maybe picoamp bipolar input stage if you are interested in good low frequency response from a piezoelectric transducer.  If you like Analog Devices then maybe something like an AD8421.

AD8253 looks very good but it has poor THD at upper frequencies. What type of op-amp would be best suited to amplification of frequencies from a piezo at about 20khz to 500khz with a low THD?

If I understand this correctly, using 1Mohm resistor as input bias for example would just shift the input common mode voltage higher but as long as it does not exceed the range of the inst-amp this would be ok? (I only care about differential voltages) Does the common mode input offset also scale with gain?

[mzzj]

I usually pick maximum input resistance based on either doubling the voltage noise or doubling the offset voltage.

For broadband noise that would be about 10nV/2pA or 5 kohms (2.5 kohms at each input) which is about right for a good bipolar input stage.

For input offset voltage it will be about 200uV/40nA or again, 5 kohms but since you input is effectively AC coupled, this is not a limitation.

The AD8253 is really the wrong kind of part for this.  You should be using an FET input or maybe picoamp bipolar input stage if you are interested in good low frequency response from a piezoelectric transducer.  If you like Analog Devices then maybe something like an AD8421.

AD8253 looks very good but it has poor THD at upper frequencies. What type of op-amp would be best suited to amplification of frequencies from a piezo at about 20khz to 500khz with a low THD?

If I understand this correctly, using 1Mohm resistor as input bias for example would just shift the input common mode voltage higher but as long as it does not exceed the range of the inst-amp this would be ok? (I only care about differential voltages) Does the common mode input offset also scale with gain?

OPA637 rings a bell for your requirements.
or CLC5509, too lazy to work out the specs right now.

[David Hess]

I usually pick maximum input resistance based on either doubling the voltage noise or doubling the offset voltage.

For broadband noise that would be about 10nV/2pA or 5 kohms (2.5 kohms at each input) which is about right for a good bipolar input stage.

For input offset voltage it will be about 200uV/40nA or again, 5 kohms but since you input is effectively AC coupled, this is not a limitation.

The AD8253 is really the wrong kind of part for this.  You should be using an FET input or maybe picoamp bipolar input stage if you are interested in good low frequency response from a piezoelectric transducer.  If you like Analog Devices then maybe something like an AD8421

AD8253 looks very good but it has poor THD at upper frequencies. What type of op-amp would be best suited to amplification of frequencies from a piezo at about 20khz to 500khz with a low THD?

How much gain do you need?

I do not understand why you need an instrumentation amplifier or low THD.  Can't the piezoelectric transducers be used with a single ended gain stage?  Maybe noise pickup in the wiring is a problem.

As for parts, from Linear Technology I would be looking at the LT1102 and LT1169.  Given your requirements, I would consider using a low input current stage with JFET or MOSFET inputs configured as either a buffer or low gain differential amplifier to buffer a fast and low noise bipolar input instrumentation amplifier.

If I understand this correctly, using 1Mohm resistor as input bias for example would just shift the input common mode voltage higher but as long as it does not exceed the range of the inst-amp this would be ok? (I only care about differential voltages) Does the common mode input offset also scale with gain?

The 1 Mohm resistor will also raise differential low frequency current noise to possibly incredible levels and also raise the differential input offset voltage.  Any common mode input offset is divided by the common mode rejection so gain should not be a factor.

[AlessandroAU]

I am doing research on characterising damage in composites by looking at non-linearity of acoustic waves. Basically I pass two frequencies, a high and low frequency through a composite sample and based on the amount of non-linear wave interactions observed determine the severity of damage in the sample. Specifically, I am interested in the detection of side bands around the high frequency carrier caused by MECHANICAL wave mixing and intermodulation with the low frequency wave. The goal is to then relate this to the amount of damage present in the specimen.

I am recording data at 24bits, 2.5MSPS. The high frequency carrier can be anywhere between 10-500khz.

Since I an interested in detecting mechanical non-linearity the front end to my piezo transducers must be of high fidelity so that I can be sure any non-linearity is a result of mechanical rather than electrical interactions. Since the piezo's are mounted to conductive substrate (one terminal of the piezo electrically connected) quite alot of common mode noise is picked up hence the need for differential measurement.

In some testing configurations a gain of up 500x is required to measure very small excitations. I would be very happy with 100x however.

I am partial to using the AD825x series because I am using labview and a micro to remotely set gain levels. At the time I did not realise that these were unsuitable for use with a high impedance source.

[David Hess]

I am doing research on characterising damage in composites by looking at non-linearity of acoustic waves. Basically I pass two frequencies, a high and low frequency through a composite sample and based on the amount of non-linear wave interactions observed determine the severity of damage in the sample. Specifically, I am interested in the detection of side bands around the high frequency carrier caused by MECHANICAL wave mixing and intermodulation with the low frequency wave. The goal is to then relate this to the amount of damage present in the specimen.

This makes sense to me.

Since I an interested in detecting mechanical non-linearity the front end to my piezo transducers must be of high fidelity so that I can be sure any non-linearity is a result of mechanical rather than electrical interactions. Since the piezo's are mounted to conductive substrate (one terminal of the piezo electrically connected) quite alot of common mode noise is picked up hence the need for differential measurement.

I do not know how much non-linearity to expect from mechanical deformation but I would expect it to be higher than non-linearity from the electronics.

In some testing configurations a gain of up 500x is required to measure very small excitations. I would be very happy with 100x however.

I am partial to using the AD825x series because I am using labview and a micro to remotely set gain levels. At the time I did not realise that these were unsuitable for use with a high impedance source.

500 kHz is not all that fast but imbalance in instrumentation amplifiers can limit common mode rejection at high frequencies.

I would look at buffering the low impedance instrumentation amplifier inputs which will be required for good AC performance with a high input impedance low gain differential preamplifier.  Higher performance could be achieved with discrete JFET inputs but I doubt that will be necessary.

Low noise is going to be important so look for fast JFET or MOSFET input amplifiers with a low voltage noise specification.  These will also be the ones with high bandwidth and low distortion.

Distortion to support 16 bits at 500 kHz requires cutting edge circuit design.  12 bits is more realistic if you want to keep things simple.

[AlessandroAU]

Thank you very much for your input David

What do you think of using this example from the OPA637 datasheet as a front-end buffer for my existing AD825x circuit? How much gain should I preform in the first stage?

[David Hess]

The OPA637 is one of the better options.  It is a decompensated amplifier however so it has to operate with a minimum gain of 5 or higher which may be an advantage in this case.

Given your bandwidth and gain requirements, I would distribute the gain evenly between stages based on the gain-bandwidth product of each stage for maximum total bandwidth.

An INA217 would be the type of instrumentation amplifier I would consider.

If more gain is needed, then I would either add another differential preamplifier or an amplifier at the output of the instrumentation amplifier.

[dannyf]

That IS the typical 3-opamp topology for instrumentation amplifiers.

You can do that, or use a real instrumentation amplifier. Bunch of them in the INA family.

[nctnico]

Unfortunately the GBW of most instrumentation amplifiers is very limited. Gettting 500kHz at a gain of 100x let alone 500x is going to be tough. I think several stages are required.

[AlessandroAU]

The OPA637 is one of the better options.  It is a decompensated amplifier however so it has to operate with a minimum gain of 5 or higher which may be an advantage in this case.

Given your bandwidth and gain requirements, I would distribute the gain evenly between stages based on the gain-bandwidth product of each stage for maximum total bandwidth.

An INA217 would be the type of instrumentation amplifier I would consider.

If more gain is needed, then I would either add another differential preamplifier or an amplifier at the output of the instrumentation amplifier.

Hi David, I took a look at the INA217 it has all the right specs but doesn't it have an input bias current that is far too high?, its 2 ua according to the datasheet. (Edit: Ah I see you meant using that as the inst-amp in the chain, not as the front end buffer)

http://www.ti.com/lit/ds/symlink/ina217.pdf

I some more poking around, what do you think of this op-amp? OPA657

http://www.ti.com/lit/ds/symlink/opa657.pdf

It seems to be all around better than the INA217 OPA637 if I can interpret it correctly. Is there some reason this wouldn't work or is worse choice?

That IS the typical 3-opamp topology for instrumentation amplifiers.

You can do that, or use a real instrumentation amplifier. Bunch of them in the INA family.

The issue I'm seeing is finding one that has both low input bias, low noise and sufficient bandwidth all in one.

[David Hess]

Hi David, I took a look at the INA217 it has all the right specs but doesn't it have an input bias current that is far too high?, its 2 ua according to the datasheet. (Edit: Ah I see you meant using that as the inst-amp in the chain, not as the front end buffer)

Yes, the INA217 would need some type of buffer.

I some more poking around, what do you think of this op-amp? OPA657

http://www.ti.com/lit/ds/symlink/opa657.pdf

It seems to be all around better than the INA217 OPA637 if I can interpret it correctly. Is there some reason this wouldn't work or is worse choice?

It has low open loop gain and low common mode rejection ratio which shows up at higher distortion at lower frequencies than the OPA637.  On the other hand it will have lower distortion at higher frequencies because of its higher gain-bandwidth product.  Comparing the two is difficult at 500 kHz because they do not have comparable specifications.  The OPA657 will never deliver 16 bits of performance at any frequency.

If you can live with lower supply voltages, TI or someone else may have a better substitute for the OPA637.