I've a problem with an AD7190 (or my layout) having a huge settling time after powering up. When a new measurement is started, the signal needs about ten minutes to settle. (See PowerOnSettlement.png. For comparison, the steady state noise is shown in Noise.png.)
The ADC configuration is: sampling rate 12.5Hz, gain 128, chop enabled, sinc4. But also changes to these parameters don't avoid the settling process. For the recorded data, a "measurement" bridge consisting of four normal resistors is used, to avoid any problems which may be caused by the actual measurement bridge. If the normal measurement bridge of strain gauges is connected, the settling process is the same. The measured data reacts instantaneously if the beam on which the strain gauges are clued is touched or bended. Thus, the measurement device appears to work normally, apart from the settling.
The schematic of the ADC part of the circuit in question follows the example provided by AD in CN0118 (weigh scale system).
The crucial part seems to be the switching on of the measurement bridge. It doesn't make a difference if the ADC is powered all the time or if it is shut down between measurements. The only way to avoid the (repeating) settling process is that the bridge is powered all the time.
If the bridge voltage is cut only for about a second, a clear drop in the measured voltage can be observed, and the settling curve is continued from this new start. If the bridge voltage is cut for somewhat longer, the whole settling process is repeated when the new measurement is started.
I'm wondering what causes this settling process. And how to avoid it.
In the FAQs concerning the AD7190 it is mentioned that the solder joints act as thermocouples and therefore affect the overall thermal behavior. But it is realistic that these effects are caused by the 30mA-heating of the bridge current and that they are that strong? The layout of the routing of the connections of the measurement (AIN1/2) as well as the reference voltage (REFIN1+/-) is nearly symmetric.