Hi Steve,
> If one applies a 10V excitation, the 9237 automatically adjusts its input range to cope, i.e. my 20mV/V sensor is now seen by a 250mV/V 9237.
Units correction: If you apply 10 V excitation, a 2 mV/V sensor has a 20 mV output, not 20 mV/V.
2 mV/V * 10 V = 20 mV
Since the NI 9237's range is +/-25 mV/V, using 10 V excitation will make it measure sensor outputs up to +/-250 mV. However, the NI 9237 measures the ratio between the sensor output and the excitation in hardware, so that any variation in the excitation voltage cancels out.
> To cope, it seems some form of sensor amplification is necessary.
Are you sure that it's necessary? The number of bits output from the ADC isn't a guarantee of the measurement's accuracy, dynamic range, or noise characteristics. There are separate specifications for those in the NI 9237 Operating Instructions and Specifications. What are your accuracy and noise requirements?
> An alternative would be to set the excitation to 1V but instead of using it, use an external excitation set at 10V. Then we would have a 20mV/V sensor feeding a 25mV/V module.
If you configure the NI 9237 to use internal 1 V excitation, and you connect the sensor's excitation leads to an external 10 V excitation source instead of connecting them to the NI 9237, the NI 9237 can't measure the ratio between the sensor output and excitation voltage in hardware. Any fluctuation in either the external excitation or the NI 9237's internal excitation will show up in your measurements, instead of canceling out. This is likely to be much worse than what you have now. Also, increasing the excitation voltage increases sensor self-heating effects, which can also reduce accuracy.
To use external excitation with the NI 9237, you connect an external voltage source to the external excitation connector on the NI 9237, and the NI 9237 connects the excitation to the ADC references and the EX+/EX- pins for each channel. This allows the NI 9237 to measure the ratio between the sensor output and the excitation in hardware, and doesn't require the external excitation voltage to be calibrated.
> Another option would be to use an analog 24bit ADC input module, e.g. 9239, and set the input range as close to the sensor range as possible, it would suffer on not having excitation or calibration though.
The NI 9239 has a fixed +/-10 V range. The accuracy of the measurement would depend on the accuracy of your external excitation source. This approach also wouldn't support remote sensing (which can be used to compensate for lead wire resistance, if your sensor leads are long). You could do strain calibration (offset nulling and shunt calibration) yourself, but you would need to do all of the math in your application and devise another way to switch the shunt in/out of the circuit.
Brad
---
Brad Keryan
NI R&D
