Gain error with shunt calibration

[emilg]

Hi Anton,

The resistances between EX+ and EX- are 250 ohm to 251 ohm for the four transducers I measured. Between EX+ and AI+ I measured resistances of 267 ohm to 270 ohm. According to my calculations a balanced bridge should have a resistance of 263 ohm between EX+ and AI+. I guess that the bridge in the displacement transducer is balanced when the transducer is in its middle position. In my set-up the displacement transducers are not at their middle position at rest. Is 7 ohm unbalance realistic?

Changing the shunt resistor position from R3 to R4 gives the attached error message.

/Emil

[antsundq]

Hi again Emil, 

I think the calibration should work as long as the bridge is in a stationary state and at rest, it should not need to be in the middle position as long as you do an offset nulling. 

So the deviation from your calculations, could it be due to cable resistance? Taking a cable resistance of 3.5 Ohm, since you are allways measuring the 7 Ohm over two cables, would give a gain error of 2% according to my calculations (se attachement). Wasn't this exactly what you were experiencing? This would suggest that the gain error is in fact due to the cable resistance.

In order to eliminate this error you would need to get remote sense cables connected at the bridge. You could alternatively make the cables shorter to reduce the effect. If you knew the cable resistance exactly, you could manually compensate for this, but I am afraid that it is hard to determine the cable resistance with high enough precission.  

/Anton

[emilg]

Hi Anton,

I agree that shunt calibration can be applied to unbalanced bridges. However, I do not understand your shunt error calculations. Unbalance is not the same as cable resistance. According to the data sheet, the cables are 2 m long with 0.065 mm^2 area which for copper wires gives a resistance of approximately 0.5 ohm. As derived by Stanford and Box (2001) in the publication “Errors in shunt calibration of strain gauge circuits due to cable resistance”, the shunt calibration error due to cable resistance for a remote full bridge is

D = 1-(Rg+2RL)(Rg+2Rs)/(Rg(Rg+3RL+2Rs))

where Rg is gauge resistance, Rs is the shunt resistance and RL is the wire resistance. In my case the shunt error should be 0.3% due to cable resistance.

A 0.3% error is not negligible in my application, but still much smaller than the error I measure.

/Emil

[antsundq]

Hi Emil,

So my calculations were a little more approximative and do not take the shunt calibration resistance into account, but they give roughly the same answer. Your formula gives a gain error of 1.5 % and my gave 2%. 

What I assumed was that the RL is 3.5 Ohm and not 0.5 Ohm as you approximated it. I thus assumed that the error of 7 Ohm that you were getting in the resistance measurements were due to cable resistance. The cable resistance is depending on the quality. 

As of now I think that the problem is due to the wire and contact resistance, is there any way you could measure this directly?

[emilg]

Hi Anton,

Maybe I was to quick to assume the theoretical wire resistance. I will try to measure the contact resistance.

Can lead resistance also explain the change in sign of the error when jumpers are connected between EX and RS terminals?

/Emil

[antsundq]

Hi Emil, 

To me it sounds plausible that this would be the case, but from the top of my head I can not say directly how these are related. When you use the jumpers you will have a voltage drop over the cables and this is not compensated for, causing the excitation voltage to be somewhat lower than expected. This in turn would have the same effect as a larger gain factor. I think this should be a well known issue and that it should be well documented in the litterature.

/Anton

[emilg]

Hi Anton,

I have measured the contact resistance between the gauge wire and the 9237 module to less than 0.1 ohm. Together with the cable resistance of 0.5 ohm and some additional resistance on the gauge side the total resistance should still be below 1 ohm. Thus, the shunt calibration error should be no more than 0.5%. Also in combination with the attenuation error

Da = 2RL/(Rg+2RL)

that you derived earlier, the total measured error should never exceed 1%.

I realise that the estimated error is now really close to what I measured in the first place, but I am still surprised by the behaviour when I connect the RS terminals. Without RS terminals connected I assume that the internally regulated voltage that is supplied to the EX terminals is taken as reference excitation voltage. Thus, jumpers between EX and RS terminals should give a reference voltage somewhere between the voltage at the regulator and the voltage at the gauge and the error should be smaller than without jumpers and with the same sign, i.e. the shunt factors should never be larger than 1.

Could you provide some information on the internal connection between regulated output voltage and the sensing input?

/Emil

[antsundq]

Hi Emil, 

I am very sorry for the late reply, but to be quite honest I can not really get my head around this. Even if I believe that the cable resistance is a part of the problem I agree with you that this is not the complete picture and that there is something that we have missed. 

When the RS terminals are connected the excitation voltage is measured internally, and it is compensated for in the output of the module. The module senses the open circuit when the RS terminals are unwirred and in this case the excitation voltage is taken as the voltage drop between the two EX terminals. As far as I can see ther should therefore be no difference between using jumpers and not and I can not really understand what this effect depends on. Offcourse there is some current goint in the remote sense circuit, but this should be negligible. 

I went back to look at your code and I would like to understand why you use a 2mV/V range when according to the white paper on the transducer the rated output is at 5mV/V. Are you not using the whole range? 

I also made some considerations about the accuracy of the measurements and a quick and dirty calculation gives an accuracy of about 15 uV/V assuming you use the whole range of 5 mV/V. This is quite large and also an overestimation, but I think the 1-2% error is close to the limits of what you can acchieve wiht that transducer and module. How old is the module you are using? 

There is still two mysteries that I can not understand. First, did I understand you correctly that the error is smaller before shunt calibration? And second, I can not see why the jumper causes the change in the gain. To understand this we would need to get some information of the inner circuits in the module. However you are not supposed to short the Ex to Rs terminals using jumpers. 

/Anton

[emilg]

Hi Anton,

I totally agree that the jumper part of the problem is strange and that jumpers or no jumpers should not make a difference if the excitation voltage is taken as the voltage drop between the two EX terminals when the RS terminals are unwired.

You have understand correctly, the error is smaller for all transducers without shunt calibration. The modules were bought about a year ago. I did a calculation of the accuracy when I built the LabVIEW VI and to increase the resolution I reduced the range of the transducers. I need the 10 mm mechanical range of the transducers, but the displacements I want to measure are much smaller (within a 0.1 mm range). To be on the safe side, I just made some measurements with full range and the shunt factors are still the same.

I hope this information can be useful although it is not much.

/Emil

[antsundq]

Hi Emil, 

Decreasing the range will not actually increase the accuracy using this module since there is no active amplification of the signal, so changing the input range in the software will not influence the meassurement. In general you should use a sensor that can make use of as much as possible of the range and measuring over several orders of magnitude with high precission requires high precission transducer. 

This however does not solve the problem since it can not explain why the calibration makes the measurements worse. I am running out of ideas. You have connected everything correctly and you are using the software properly. The only thing I can see that would lead to a gain error is the Ex lead resistances. 

Do you have access to any other sensors/bridges that you could make a test with to see if the problem is in the transducers? 

/Anton