Low voltage amplifier choice

[mbless]

I need to design a low voltage amplifier and wanted to get some feedback on the amplifier choice. It will mostly be used with thermocouples but can be other low voltage sensors that produce 0-40mV. The amplifier output will be 0-5V, so a gain of approximately 120, and the bandwidth needs to be >100 kHz. I'm thinking of powering it on 4x AA (6V) to eliminate potential noise sources, so the opamps likely need to be rail-to-rail.

My current direction is to use an instrumentation amplifier (e.g. AD8421, INA821), but I'm wondering if there is a better route to go. Also, is it recommended to make your own InAmp, or is it too difficult to match the resistors and individual amps like integrated InAmps?

[Wimberleytech]

AD8241 looks like a good solution.  I would not roll my own. 
I have used the AD624 in several projects where I had dual supplies.

[Kleinstein]

The AD8421 is a nice really high performance amplifier, but with 5 V min. supply . So the common mode range is rather limited at a low supply voltage. The output is also far from Rail to Rail.

For just thermocouples the question is if one really needs an instrumentation amplifier. A simpler zero drift OP may be sufficient, though problematic with high BW. Thermocouple and 100 kHz BW does not match well.

[Wimberleytech]

The AD8421 is a nice really high performance amplifier, but with 5 V min. supply . So the common mode range is rather limited at a low supply voltage. The output is also far from Rail to Rail.

For just thermocouples the question is if one really needs an instrumentation amplifier. A simpler zero drift OP may be sufficient, though problematic with high BW. Thermocouple and 100 kHz BW does not match well.

Good eye.

Analog Devices has a nice web ap to generate diamond plots for their instrumentation amplifiers
https://tinyurl.com/qlzewzd

I would suggest more voltage and configure into a split supply.  In the projects I did, I used two 9V batteries (+/- 9V)

[mbless]

The AD8421 is a nice really high performance amplifier, but with 5 V min. supply . So the common mode range is rather limited at a low supply voltage. The output is also far from Rail to Rail.

Oops, I did not see that... I thought of solving the common mode issue by using pull-up and pull-down on the 2 inputs to center the common mode voltage. It would also serve as open connection detection. The INA821 has rail-to-rail output, but the minimum output of 150mV is still not good enough. It looks like a dual supply is the best option.

For just thermocouples the question is if one really needs an instrumentation amplifier. A simpler zero drift OP may be sufficient, though problematic with high BW. Thermocouple and 100 kHz BW does not match well.

An InAmp may not be necessary, but I don't know what environments the device will be exposed to. Since cost isn't an issue, I thought that using a higher quality device would prevent some headaches and troubleshooting later on.

Can you expand on "Thermocouple and 100 kHz BW does not match well."? If you mean thermocouples don't need that much bandwidth, the requirement was provided by the client, so necessary or not I'll provide it if I can.

Analog Devices has a nice web ap to generate diamond plots for their instrumentation amplifiers
https://tinyurl.com/qlzewzd

I would suggest more voltage and configure into a split supply.  In the projects I did, I used two 9V batteries (+/- 9V)

Thanks for the link and idea. A dual supply would solve some problems.

[David Hess]

Adding a separate rail-to-rail output operational amplifier as a second gain stage will considerably relax the requirements on the input amplifier and dividing the gain between two stages according to their gain-bandwidth product will yield much better bandwidth.  The relatively inexpensive AD8231 has that all built in and might be ideal for what you want.

[floobydust]

Thermocouples and 100kHz bandwidth is a bad requirement for many reasons.
You don't want to amplify mains AC hum and SMPS noise. It is customary to LPF thermocouples as early as possible in the chain. Your client may need to explain the need for a response 100,000x faster than the sensor lol.
Second, your amplifier's gain-bandwidth requirement is getting up there to do Av=120 at 0.1MHz is beyond many low voltage single op-amps of lower cost. It's adding a speed requirement that would have to be done with two op-amps.

[mbless]

Adding a separate rail-to-rail output operational amplifier as a second gain stage will considerably relax the requirements on the input amplifier and dividing the gain between two stages according to their gain-bandwidth product will yield much better bandwidth.  The relatively inexpensive AD8231 has that all built in and might be ideal for what you want.

That is another good option. With the AD8231 I would be able to split the gain between the two amps get the required bandwidth. I'll have to dig into the details more of that one.

Thermocouples and 100kHz bandwidth is a bad requirement for many reasons.
You don't want to amplify mains AC hum and SMPS noise. It is customary to LPF thermocouples as early as possible in the chain. Your client may need to explain the need for a response 100,000x faster than the sensor lol.
Second, your amplifier's gain-bandwidth requirement is getting up there to do Av=120 at 0.1MHz is beyond many low voltage single op-amps of lower cost. It's adding a speed requirement that would have to be done with two op-amps.

Thanks for the potential pitfalls, but I don't think they quite apply here. It will be battery powered so no AC or SMPS noise to deal with. I have a general idea of its application, and the thermocouple response will be much faster than 1 Hz. They probably have decided on Nyquist level resolution plus some additional factor 

The AD8421 and INA821 discussed earlier can both do 120 gain with BW > 200 kHz. Yes they are costly, but cost is not an issue.

[David Hess]

Thermocouples and 100kHz bandwidth is a bad requirement for many reasons.

They are but it is not uncommon to want a "universal" input which can accept thermocouple signals or other types of faster signals.