help with instrumentational amp

[rudie]

Hello guys,

I am using the MAX4194 to measure the voltage across a resistor (R2). I chose the MAX4194 because it has a voltage error of 450 uV at a gain of 1. It also has a low gain error. To test if it works i build up a simple circuit (see atachment). As stated in the datasheet i connected a Rload of 25K (R5) to Vcc/2. The Vcc/2 is realized by the voltage divider of R3 and R4. I did not connect a resistor to select the gain because if you leave the pins open the gain is one. The Vcc is 5V. If i measure the output pin it has an inaccuracy of 4mV!!!! and not 450 uV. How is this possible, what did i do wrong?

Please help

greets Rudie

 

[FvM]

It's not suggested in the datasheet to connect a load resistor to Vcc/2. That's just a defined load to specify minimum and maximum output voltage. Minimum output voltage is probably what you see in your circuit, although 4 mV is much lower than the value specified in the datasheet. Rail-to-Rail capability of the amplifier doesn't mean that it can reach exactly zero output voltage related to negative supply. That's physically impossible cause the output saturates with a finite resiudual voltage. From the datasheet values, it should be actually able of operating at a common mode input voltage equal to negative supply, althoug with decreased performance below Vee + 0.2 V. But the output is intended to operate with a virtual ground respectively a Vref above Vee, then it can achieve also the offset specification with zero input voltage.

It depends on your application, if it makes any sense to have a ground referenced precise, low offset measurement, cause most high resolution data converters wouldn't be able to interface it at that level (without an additional negative supply, that could also supply the amplifier if present). A differential input converter could have it's inverting input connected to Vref, avoiding the output saturation and allowing also bipolar measurements.