When the accuracy is reported as 12.5mS +/- 12.95uS, 12.5 mS (12.5 milli strain) is the reading value (which for a full bridge is the upper bound of strain) of strain and the accuracy is +/- 12.95micro strain. So, if your strain reading is lower, the corresponding accuracy can be calculated from the equation
absolute accuracy = (gain error * reading) + (offset error * range) + (noise)
You can read more about it on this page : http://zone.ni.com/devzone/cda/tut/p/id/8909
So to answer your question-
12.5 mS is NOT the precision. 12.5 milliS is strain reading corresponding to the 25 mV/V span (which is the maximum for the 9237). The center of the range means, say you got a reading of 10 mV/V (assume a full bridge) which corresponds to a strain of 5 mS, the accuracy in this case would be 10.9 micro strain. As the Off Set and Gain Error are given in terms of % of the reading, the smaller strain you measure the better precison you have -- ( Feel free to correct me if I am wrong)
I havent started using the 9237 yet. So I can't help you on this.
Half Bridge is: 25 mV/V ± 1.45 mV/V. The DevZone article actually has an error in it and kinda helped create this confusion. (I'm working on fixing it now).
The first values it reports as accuracy for the different configurations are correct.
Absolute Accuracy for Full Bridge is: 25 mV/V ± 25.9 µV/V
Absolute Accuracy for Half Bridge is: 25 mV/V ± 1.45 mV/V
The error occurs when it converts those values to strain. It uses the correct formulas, but it switches the values calculated above.
The half bridge tolerence is a spec of the internal completion circuit in the 9237. It's listed on page 25 of the 9237 manual and probably is calculated using the resistor tolerances in the completion cirucuit.
I do not know how to calculate the quarter bridge error, but I know arrangement has the smallest output, and it the configuration where noise is most prominent.
You are correct in the way you describe the accuracy. The first number is what you are reading and the second is the range where the actual value is guaranteed to be.
I cannot understand how accuracy can be the range limit +/- the noise?!? For me accuracy could be the smallest step that can be sensed. So if Vex is 2.5V, and givven the 25mV/V what is the delta-strain that can be sensed (which is usually the noise level above which it is the signal)?
Are you saying that for HALF bridge accuracy is 1450 uS? If YES then I am in trouble. Can you confirm or hopefully prove me wrong? Our calculations givven the 24-bit ADC and noice produced numbers below 1 uS ...?!?
Accuracy is speced as measured reading (A) +/- uncertainty (B). This spec says that if your device is measuring value A, the actual value of what is being measured is guaranteed to be within +/- B of the measured value.
If you look at the formula: absolute accuracy = (gain error * reading) + (offset error * range) + (noise)
The gain error is speced as a percentage. So, the smaller the reading, the smaller the error. Typically, you chose to calculate the gain error portion of absolute accuracy based on a hypothetical measurement at the top of the measurement range to account for the maximum gain error you could possibly see.
The smallest measurable step is typically referred to as resolution. Strain gauges are analog devices, so they "theoretically" have infinite resolution, and the 9237 is a device with 24bit resolution. The noise error while using these devices is a lot greater than the smallest step the 9234 can detect, which is why it is more useful if we look at accuracy based on the noise.
Yes, half bridge accuracy is guaranteed to be within 1450 uS of the reading. Again, because of the way this is calculated, this will be slightly smaller for readings at the middle of the range. You will need to go to a full bridge to reduce your noise and increase your accuracy.
Thank you for clarifying the topic. Do I udnerstand correctly that if I wire additional 2x 120 ohm resistors (1%) to complete the brdige outside of the NI-9237 I will get the 25.9 uS resolution? Any specific requirements?
For a full bridge, you will need four active strain gauge elements. The 9237 has internal half bridge completion circuitry, so you do not need to wire any external resistors. Just make sure the bridge configuration is set to full bridge when you take measurements and you will get the accuracy of a full bridge.
Here is a good document for your reference.