Analog differential input - how to make it measure 0V when floating

[esm.]

Hello.
I am using an external ADC (MCP3421), 16-bit resolution, with internal PGA (internal gain) of 8x to read signals from a thermocouple type T. Within my current ADC configuration, the differential input impedance is around 280K.
But when the inputs are floating, the result of ADC conversions behaves incorrectly (negative and positive aleatory values, sometimes with high values are read from ADC).

I want this ADC to output conversion values near to zero when the inputs are floating, to detect a sensor fault, or sensor disconnected, for example.

How could I do a "pull down" to it? because the Vin- input (reference) and Vin+ input (input signal) from the ADC and and the digital ground of the IC are not in the same voltage reference.

 

[ronsimpson]

You can place a resistor across the (-) to (+) input to try to make floating look like 0. This may make the thermocouple read incorrectly.
I have also used a resistor to +5v to pull a input up to a unrealistic value, with no thermocouple. 14,369 degrees=no thermocouple

 

[crutschow]

You might try adding a high value resistor (much larger than your signal source resistance) from each input to signal common.

 

[esm.]

This is the problem, I want to use a high value resistor.
The ADC differential input impedance (between Vin+ and Vin-) is 280K within my current ADC configuration.

If I put a lower value between Vin+ and Vin-, the signal measurement will be affected, because I am measuring multiple thermocouples using an analog switch (with around 200 ohms Ron). There will be voltage losses across the analog switches as much as the resistor value is decreased.

 

[ronsimpson]

This is the problem, I want to use a high value resistor.
The ADC differential input impedance (between Vin+ and Vin-) is 280K within my current ADC configuration.

If I put a lower value between Vin+ and Vin-, the signal measurement will be affected, because I am measuring multiple thermocouples using an analog switch (with around 200 ohms Ron). There will be voltage losses across the analog switches as much as the resistor value is decreased.

YES!!!

Now your 280k is about 1 part in 1000. You could try using a 22k resistor and see how much that moves the data and if you can correct for that in software.

OR

In software look for noisy data and declare noise=no thermocouple.

 

[Reloadron]

Generally when working with a thermocouple you either want Upscale or Downscale Burnout. For example in a heating application you would want to drive the TC input above the max TC mV value if the TC opened and in a cooling application you would want to drive the TC input below the max low. This is generally done placing a small voltage through a resistor to the TC + input to your ADC. Maybe something like this is what you are after?

The positive input of the operational amplifier of the measuring circuit is connected to the positive (or negative) power supply through the high-value resistor R. R resistance is so high that the current through it is as low as 0.1mA. The current has only negligible effect on temperature measurement when the thermocouple circuit is closed, since it is diverted toward the thermocouple circuit whose resistance is low. We call the current a burn-out current. When the circuit opens by burn-out or disconnection of the input wire, the burn-out current changes the amplifier output to drive the transmitter output higher than 100% (or lower than 0%).

I would work burnout from that angle. A Google of Thermocouple Burnout will or should bring up more hits and suggestions.

Ron