Gain error with shunt calibration

[emilg]

Hello,

I have a problem with shunt calibration of bridge based displacement transducers. The shunt calibration routine sets gain factors that give errors around 2% in the measured displacement. I have also found that the shunt calibration gain factors run out of bounds without a preceding offset null.

The hardware is a cDAQ-9188 with 9237 modules. The gauges have four wires, two input wires that I have connected to EX+ with jumper to RS+ and EX- with jumper to RS-/SC-, and two output wires that I have connected to AI+ with jumper to SC+ and AI-. LabVIEW is used with a VI shown in the attached figure.

Does anyone know why the shunt calibration behaves like that or what I do wrong?

/Emil

[antsundq]

Hi Emil, 

Is the bridge you are using calibrated, so that you are able to tell that there is an error in the calibration you are performing in LabVIEW? The calibration only affects the measurement side and can not correct for errors in the sensor. The 2% error you are refering to; is it a constant 2% offset or is it related to noise? You are using a resistance of 350 Ohm for the bridge and I assume that this corresponds to the transducer you are using.

Your connections seems to be in order, except I am not sure how you have connected the RS+ and RS-. They should be connected close to the bridge to correct for resistances in the cables used, is this how you have set it up and is the calibration error stil present if you remove the two RS cables? 

Regarding the offset nulling; this is normal behaviour since, before the zero-level for the signal is set, the signal may go out of bounds since there is a constant background voltage present adding to the total signal. 

//Anton

[emilg]

Hi Anton,

Thank you for your reply!

The gauges are shipped with calibration papers including calibration constants, e.g. 0.002001 mm/uV/V, a rated nonlinearity of 0.01%, and bridge resistances, e.g. 350.2 ohm input and 350.0 ohm output. All ten gauges have bridge resistances between 349.2 ohm and 350.3 ohm.

The 2% error is neither noise nor a constant offset, but a linear gain error so that the absolute error increases proportional to the signal.

Since the bridge is internal to the gauge and the gauge only has two input and two output wires, I have connected RS+ and RS- to EX+ and EX- at the 9237 module. I have tried both with and without jumpers to RS+ and RS-, but it makes no difference. I know that this is the expected behaviour without an extra wire from the bridge to measure excitation voltage drop, but I thought it was worth a try.

/Emil

[antsundq]

Hi Emil, 

What gauges is it you are useing and do you have acces to a whitepaper on them that you could share? To get a better picture of what you are doing, could you briefly explain the application?  

From the LabVIEW code I am assuming that you are using a full bridge configuration. Is the gauges you are using in a full bridge configuration?

And just to clarify; how are you able to measure the 2% error in the measured displacement? 

/Anton

[emilg]

Hi Anton,

I have attached a data sheet for the displacement gauges. When I measured the error I mounted the displacement transducer against a fixed surface and used known gauge blocks to move the head of the transducer.

I forgot to mention that the error is almost an order of magnitude lower with the shunt calibration turned off.

/Emil

[antsundq]

Okay, so you are saying that the calibration actually makes the measurement worse? 

There is very little information in the data sheet, but it says that the trasnducers adopt a strain gage for the sensing. Is there any particular reason that you are useing a the DAQmc VI in bridge mode? Also, in your example code you are not handling any errors, have you checked for errors when running the code?

To continue the troubleshooting I would recomend you to try using the DAQ Assistant Express VI and see if the the calibration works useing this. You could try both in bridge and in strain mode.

[emilg]

Hi Anton,

Yes, the measurements are worse with shunt calibration turned on.

As I understand it the strain gauges inside the displacement transducer are connected in a full Wheatstone bridge, hence the two input and two output wires. This theory is supported by an uncalibrated output close to the correct displacement.

I have now tried adding the “Simple Error Handler VI” as in the attached minimal working example and it reports no errors. I have also tried using the DAQ Assistant Express as you suggested. The result from a bridge calibration is still the same, i.e. gain adjustment values typically around 1.03.

During the latest measurements I noticed something interesting. The typical gain adjustment values for the transducers connected according to previous posts are 1.03 whereas the typical values for gauges connected without jumpers from EX+ to RS+ and from EX- to RS- are 0.97. In other words, the errors in measured displacement are equally large but with different signs. I have four transducers connected with jumpers and six without and the observation applies to all of them. Can that be a coincidence?

/Emil

[antsundq]

Okey, so using the polymorphic Create Channel VI in for measuring a strain can also be setup for a bridge type measurement, and in fact this is how it usually is done. Using the strain gage instance of the VI will allow you to specify information about the actual strain gage, such as gage factor and Poisson's ration. As far as I understand using the bridge instance of the VI should generally be used when you are using a more general type of bridge and you want to use custom scales for obtaining physical units. 

Here is a really useful link for strain measurements:

http://www.ni.com/white-paper/14327/en/#UnderstandingStrainChoosingStrainGages

When using the strain instance I suppose that the calibration factor you were mentioning earlier should be used as gage factor i.e. 0.002001 mm/uV/V and this should be put in units of m/V/V giving 2.001 m/V/V and the Poisson's ratio is usually close to 0.3. Could you try using this instance of the VI or in the DAQ Assist express VI? 

If you actually know how much the calibration gage offset is going to be you could allways set the Gain using a property node, but this is merely a workaround as the calibration procedure should work. 

Generally I dont believe in coincidents, and especially not when repeated. This sounds like there is something fishy with the EX wiring, and it might be usefull to look into it. The gain factor introduced by the leads is given as 2*R_lead/R_bridge, and leads typically have a resistance of 1/10 Ohm/m so I can not directly see that this is due to the remote sensing. 

/Anton

[emilg]

Hi Anton,

Setting the gain manually from a known displacement is not an option as some of the transducers are mounted so that they cannot be tested with gauge blocks.

Setting the DAQ Assistant Express VI or Create Channel VI to strain mode gives the same shunt factors as the general bridge mode.

I have done additional measurements to outrule any chance of coincidence with the sensing wires. With jumpers from EX+ to RS+ and from EX- to RS- on the first four of ten transducers the shunt factors are:

1.0046; 1.0246; 1.0347; 1.0246; 0.9649; 0.9811; 0.9600; 0.9720; 0.9591; 0.9793

with jumpers only on the first two transducers the shunt factors are:

1.0044; 1.0247; 0.9867; 0.9729; 0.9648; 0.9811; 0.9599; 0.9721; 0.9589; 0.9791

and finally for all transducers without jumpers the factors are:

0.9624; 0.9715; 0.9868; 0.9728; 0.9647; 0.9809; 0.9600; 0.9721; 0.9592; 0.9792

The conclusion is that the shunt factors are wrong but stable if the wiring is unchanged. Jumpers between EX and RS terminals effects the shunt factors. The shunt factors are between 0.04 and 0.06 larger with jumpers.

Should the RS terminals be connected for gauges without an extra sensing wire? Should there be a difference with and without the jumpers? Do you suggest anything special to look into regarding the EX wiring?

/Emil

[antsundq]

Hi again Emil, 

Great that you checked the strain mode of the VI. 

You have made a pretty elaborate check and I would take it as conclusive evidence that the calibration gain factor is related to the remote sense cables. The remote sense cables should be connected as close as possible to the bridge, but I understood that you can not do this? I am still not sure how the wire resistance could introduce an error of this magnitude, since the resistance should be quite low. Could you try measuring the resistance between both the two Ex+ and Ex- wires on a gage and between the Ex+ and AI+ wires? They should both be very close to 350 Ohm. 

The remote sensing should generally not be used if you can not connect at the bride. The reason why you use RS is to compensate for the resistance in the wires, and this can not be done if you connect the RS with jumpers at the terminal side. The voltage drop over the wires is compensated for when using the remote sense and a change in the excitation voltage will be indistiguisahable from a change in the gain factor. So what I am saying is that this could be caused by wire resistance, but the magnitude should not be this large. 

One last thing that comes to mind at the moment is that you could check if there is any difference if you change the loaction of the shunt resistance in your code from R3 to e.g. R4. Due to the symmetry there should be no difference, but I am not sure on the configuration of the bridge in your gages. And this probably goes without saying, but I assume that the gages are not biased when you are doing the calibration. 

/Anton