Accuracy of small strain measurements

[bpmcc]

Hi,

I'm having problems quantifying  the systematic errors in a set of strain gauge readings. The uniaxial gauge (gauge factor of 2) is in a quarter-bridge configuration (2.5V excitation), and is connected to a SCXI-1314 terminal block, SCXI-1520 strain gauge module and PXI-6052E DAQ.

The strain levels range from about zero to +/-1000ue (microstrain),  and according to the NI 'Measuring Strain with Strain Gauges' tutorial this corresponds to a maximum gauge output voltage of -/+1.25mV. The SCXI-1314 has no error terms, and according to the SCXI-1520's Data Sheet (+/-10mV Nominal Range, Single Point Measurement at room temperature) the absolute voltage accuracy is given by:

+/-[(1.25mV x 0.001) + 50uV + 20uV] = +/-71.25uV

Note that the offset and noise terms dominate the accuracy. Null compensation was used during the tests, which should (I presume) remove the 50uV offset error, thus improving the accuracy to +/-21.25uV. This gives an RTI error for 1000ue of 21.25uV/1.25mV =+/-1.7%, which seems reasonable. The RTI error of the PXI-6052E has similarly been calculated as 0.056%, which is small enough to be neglected here.

The problem occurs when the errors for small strains are considered. The procedure given is to convert the nominal strain reading to its output voltage, use the RTI error to get the bounds of the voltage, and finally use these bounds to find the strain tolerances. However, as the absolute accuracy is roughly +/-20uV for all of the measured strains, the RTI error will increase as the gauge's output voltage decreases, ultimately going to infinity for zero output voltage. This clearly doesn't make sense.

How should I treat the errors in this case? Using just the absolute accuracy of +/-20uV gives a strain accuracy of +/-16ue across the entire strain range; I don't know if this is reasonable.

Any help or suggestions would be appreciated, sorry about the long post.

Regards,

Brian  

[JonBowers]

Hi Brian,

I am not quite sure I understand the problem you are having. The accuracy should be for the given range that the DAQ-card/SCXI module is set up for. Obviously the smaller the signal from the strain gauge the more it will be affected by the error or noise. Are you trying to determine the error as a percentage of your measurement, because as the measurement approaches zero mathematically this becomes unfeasible.

Regards

Jon B
Applciations Engineer
NI UK & Ireland

[bpmcc]

Hi Jon,

The problem is that NI seems to recommend (in tutorials etc.) the use of RTI voltage errors to calculate measurand errors (strain, temperature etc.), and as you mentioned this becomes unfeasible as the output voltage from the sensor goes to zero. It would make more sense if absolute errors were used instead. For example, if no strain was applied to the strain gauge in my application and thus its output voltage was 0V (ideally), the absolute voltage accuracy of +/-20uV could mean that the strain is reported as +/-16ue: is this the correct interpretation? I understand that the quoted voltage accuracies are usually conservative, but are possible. The absolute voltage accuracy in my case doesn't vary much over the measurement range as the gauge's output voltage is small, so the absolute strain accuracy is fairly consistent. Therefore the strain error as a percentage of the measurement will decrease for increasing strains.

Thanks for your continuing help.

Regards,

Brian

[xseadog]

Hi Brian,

It is very unusal to have a zero output with out any strain applied. If you are not using a 3 wire connection to  you strain gauge  then the cable length has an effect!

 Standard 2 wire connection has a volt drop over the cable length thus the effective resistance of R1,R2 does not equal R3,R4 in your wheat stone bridge circuit.

Using a quarter bridge configuration  without temperature compensation is not a recommended setup. Are you using self temp compensated strain gauge?

With your present set up have you performed a shunt calibration ?  This would enable you to check your accuracy.

Also perfoming  a series of calibrations would enable you to see if  the set up is linear as one expects.  Use precision resistors better than 0.1% if you do not have
a calibrator.


Regards
Xseadog