01-07-2019 04:23 PM
I read in an NI whitepaper that the accuracy of my strain measurement is calculated by summing gain error, offset error, and noise uncertainty. I'd like to understand how these factors impact information I'm monitoring in real time. Noise seems like an error source that would vary with time. Do the gain error and offset error vary with time or are they constant run to run or point to point.
01-08-2019 02:41 AM
@Testr00 wrote:
I read in an NI whitepaper that the accuracy of my strain measurement is calculated by summing gain error, offset error, and noise uncertainty. I'd like to understand how these factors impact information I'm monitoring in real time. Noise seems like an error source that would vary with time. Do the gain error and offset error vary with time or are they constant run to run or point to point.
Wellcome in the world of metrology 🙂
What that (bold part of your text) tell you is the expected uncertainty of your bridge voltage aquisition part of the measurement.
There are many more parts in a real uncertainty budget for a $unitofinterest measurement.
Just to name some: temperature influence, nonlinearity, drift , creeping, ... (more noise, due to cable EMI, ..)
Many books and papers .... one source with a more detailed uncertainty budget for your task should be noted in the calibration standard(s) for your $unitofinterest sensors.
01-08-2019 03:27 AM
The strain gauge manufacturer HBM features a few good white papers and information about strain measurement and uncertainty/noise. As far as I remember, they can be found under the menu "Services & Support".
01-08-2019 08:03 AM
Hi Testr00,
To add to Henrik's great answer, the noise specs in NI device manuals are RMS values, meaning they're a sort of average over time. It's possible (though very unlikely) to see a spike beyond that range, but over time the RMS noise value will be within the specifications in the device manual/datasheet.
@Testr00 wrote:
Noise seems like an error source that would vary with time. Do the gain error and offset error vary with time or are they constant run to run or point to point.
Yes, the noise varies with time, but we can't predict how it will do so.
As such, we just slap some error bars onto the measurement knowing that the value measured will be within those error bars. Sometimes the actual signal-minus-noise value will be right in the center, sometimes right on the edge, mostly somewhere in between, but if we can't accurately predict it, all we can do is report a range of possible values.
01-10-2019 11:59 AM
Thanks for all the great replies. Would it be correct to say that the gain and offset errors have little impact on variation in the data in the minute to minute time range?
01-10-2019 12:29 PM - edited 01-10-2019 12:34 PM
Well, gain and offset errors have a direct impact on you measurement uncertainty.
Depending on your measurment task, changes due to drifts in gain and offset and maybe temperatur (again only the drift) maybe have a minor impact.
(Asuming no correlation (usually at least questionable 😉 ) a simple error budget calculation end up in a squareroot of sum of squares of rel. errors)
EDIT:
Hard to say, unless you provide a complete description of your measurement task.
01-10-2019 04:30 PM
An example would be measuring up to 1000ustrain at room temperature on using a cDAQ 9188 with NI9236 modules.
01-11-2019 01:12 AM
The aquisition part is already done ... well known, easy investigated 😉 .... major part is gain,offset (error/drift) + drift +tempcos
what about the physical unit (usually measured in SI units 😉 ) you want to measure?*
the transducer (sensor) (again gain,offset,drift,noise,tempcos...
it's mounting
the operator
...
*)voltages and currents are booring unless quantities of these units don't tell us something else 😉 (well, apart from some physics ... but then these quantities are often even harder to measure .... :D).
Voltmetering (and time) is the hammer (you onlyeasy have) in measurements, so all(most) your measurement problems are now a nails voltage(time) measurements.