# Instrumentation Amplifier Saturation to Rails

I am working with a TI INA821 Instrumentation Amplifier to try to amplifier a 0-5mV (delta) analog signal off a strain gauge. To reiterate, every DC analog value between the delta of 5mV represents a specific load (weight).

The delta is the difference between (+) ~2.528 V and (-) ~2.533 V. The circuit is operating off +/-5 V supply.

The circuit I am using is that of the generic recommended configuration in the INA821 Datasheet:

Where:

1. V+ = 5V
2. V- = -5V
3. C1 = C2 = 0.1uF
4. R1 = R2 = 1M
5. Vref = 0V (GND)
6. INO+ = INO- = 2.528V @ no load
7. INO-' = 2.528V @full load
8. INO+' = 2.533V @ full load
9. R3 = Varies (see below)

The only configuration that does not result in Vout being pushed to +5V is with R3 DISCONNECTED which yields a gain of G = 1.

Using two 50mVpp, and 60mVpp test sinusoidal signals (same generator, voltage difference via voltage divider) at IN+ and IN- successfully amplifies the delta and reflects change when a pot is used in place of R3.

My only thought is that perhaps the strain gauge, being a DC type signal is causing an offset large enough to send the output to the positive supply rail? However, in theory I did not think this should be an issue given that the amplifier design is of the differential type?

• Set R3 = 249 for Av~200 to get ~1V out or R3=100 for Av=500 to get ~2.5V out Apr 17, 2019 at 2:04

## 1 Answer

Your question was going along well until the last sentence. There is NEVER a valid reason to make this assumption, even if you paid extra for a balanced bridge. Get a voltmeter out and measure the voltage between the inputs.

• This is not an answer. I have measured the inputs at the bridge which is where the 2.528 and 2.533 values come from. Given that they are subject to temperature changes, and are variable devices, I have observed a fluctuation of perhaps +/- 2mV undisturbed. Though this is no reason for a very-high gain at the output. I have also ruled out faulty components and/or connections using a test signal. Looking for more reasoning here as seems to be some circuit design issue I am not seeing.
– Matt
Apr 17, 2019 at 2:35
• The only difference I am seeing is that my test sinusoidal signal is AC-type, and my real strain gauge is DC. Thus, my reasoning in my last sentence is poking at the idea of something causing fault when ran in a DC-steady-state? I have no idea. Thus, looking for outside opinions or validation.
– Matt
Apr 17, 2019 at 2:40
• I'm sorry. The wording of the question made me think the data given was theoretical. You need to supply more information about the input circuitry. Apr 17, 2019 at 6:18