QOne customer used AD7718 to sample the thermocouple's output. The signal chain
is: Thermocouple + RC Filter + ADC.
She tested AD7718, configuring the input as differential input, when she
connected the two input to the GND, the output is OK,the output noise is
several uV. Once she used the K Type thermocouple as the input, the output
become worse,the noise can be 100 uV maximum. Also she tested
cs5523+10k+0.1uF(Cutoff Frequency is 1kHz),CS5523 is better than AD7718.
So she felt CS5523 is better than AD7718.
Whether you have ever seen similar application? Maybe you can share some points
AWhen the AD7718 is configured for differential inputs, the analog inputs are
buffered. The voltage on each analog input pin needs to be between GND + 100mV
and AVDD - 100mV. The customer has the thermocouple biased about GND. With both
inputs shorted to GND, the customer is actually outside spec (even though they
are measuring good noise). The thermocouple voltage should be biased up so that
the voltage on AIN+ or AIN- is above 100mV. The customer could use the
reference voltage to bias up the thermocouple voltage.
The AD7793 and AD7794 have a bias voltage generator on chip which simplifies
the design. So, another option is to select the AD7793/4 rather than the AD7718.