AD8226 instrumentation amplifier+bridge design

[ChristofferB]

Hi all! Some of you might possibly remember one of my million billion threads on my DIY gas chromatograph.

I'm currently designing the next detector, which is a 4-element wheatstone bridge of tungsten filaments (3mm ground-open light bulbs).

I've never done much design work with instrumentation amplifiers, or wheatstone bridges for that matter, but i've settled on the AD8226 ins. amp.

I'd be thrilled if you'd look at my design so far and see if it's reasonable. See attached schematic.

I have a current source, a small potentiometer in the bridge to balance it out for a zero reading, an attenuator, and then the amp with variable gain. I'm leaving gain as a potentiometer for now as well, as I don't know the sensitivity of my sensor filaments yet, in the future that might be fixed, and the (stepped) attenuator used to reduce the signal if very concentrated samples are analyzed.

Is there any reason not to go single-rail on the bridge and amp?  I want to input the amp output into a 5V analog atmega pin, is it better to have my amplifier at 5V supply or the full 24V and then put a diode clamp on the output?

Thanks for your interest!

--Chris

[RandallMcRee]

Much missing information, there.

Have you used AD's diamond plot tool? You ought to give that a try and see where that gets you.

https://www.analog.com/designtools/en/diamond/#difL=-0.003&difR=0.003&difSl=-0.003&gain=1500&l=0&pr=AD8429&r=10&sl=10&tab=2&ty=1&vn=-15&vp=15&vr=0

At the very least, it forces some important decisions on you. Ones you can't ignore.

[SiliconWizard]

If you want to be able to deal with negative differences, but also very small differences, you should not connect Vref to ground if you're on a single supply. Set it half-way (2.5V for instance). This will become your zero at the output. Don't bother with switches.

As RandallMcRee suggested, use this: https://www.analog.com/designtools/en/diamond/pr=AD8226
and re-read the datasheet several times. A rail-to-rail output instrumentation amp. always has a limited operating range regarding its inputs. A common trap. Just check that you will be within its operating range for any usable input values you're going to get. As a first step, as said above, I suggest setting the Vref pin at 2.5V and go from there.

As to protecting the 5V analog input, you can use a limiting resistor and a clamping diode. Do not omit the series resistor and set it to a value that will not impact the accuracy of the ADC. You could also add a capacitor to ground after this resistor so you'd get a low-pass filter as a bonus. I don't recommend powering the AD8226 at 5V. It will not be happy with such a common mode voltage if powered at 5V.

[ChristofferB]

Thanks for the advice! The diamond chart is handy, I'm just unsure what boundaries to set. My Vcm would be 12V when the bridge is in balance, right? and the difference probably wont be more than +/- a few mV (that's a total guess). Since I have a balance potentiometer that can potentially skew the balance, do I need some kind of input protection to make sure the voltage doesn't go out of range?

I've added some details to the schematic, and implemented the output clamping and filtering.

I haven't set the ref midways between the rails, I'd really like to keep the input switch idea, as the software that's going to be handling the data expects a low-voltage baseline with positive peaks. I could flip it on the computer but I'd rather not have to.

Again thanks!

 

[ChristofferB]

Okay followup question, I'm gonna build a quick prototype of above circuit. If the maximal 1000x gain isn't enough for my signals, can I cascade another ad8226 after the first one, or should I just use a normal op amp?

Thanks again!