How do I multiplex strain gauge signals?

[egret2008]

I have an application where I need to measure around 160 strain gauges.

I currently have five NI 9237 modules which will allow me to measure a total of 20 gauges.

I only need to preform static testing, so there is scope to multiplex all the signals into my existing hardware.

I will use a full bridge configuration for each measurement point i.e. one active gauge and three passive gauges. Multiplexing will occur outside of the bridge so that any change in resistance at the multipler switch will have a mininal affect on the measured signal.

On the NI 9237 there are a maximum of ten lines per measurement point.  The two TEDS lines will not be required and the two lines for excitation voltage do not need to be multiplexed. This leaves six lines to be multipled per meaurement point. Therefore I need to multiplex a total of 960 lines down to 120 lines.

Can anybody recommend a suitable analog multiplexer for this application?

One concern I have is how to deal with the offset nulling and shunt calibration. Is it possible to apply, in software, the calibration data for each measurement point each time the multiplexer is cycled?

Regards, Ben

[Chris-G]

Hi egret2008,

Is it a requirement that you also switch the remote sense and shunt calibration channels? The reason I ask is that if you can reduce the required lines to just the AI+ and AI- lines, you will obviously not need as large a switch system. If it's possible to boil the system down to 2 wires per strain gauge, you would only need a switch system that takes 360 channels down to 40, which is a significant reduction. If these lines are required, the switching system will have to be really extensive.

In general, NI offers switches in two form factors, SCXI and PXI. Because this is a new system, I would recommend using a PXI system because it is a newer platform and allows for more future flexibility.

Based on the specs you laid out, each 9237 AI channel would have to be able to switch to 8 strain gauges (160/20 = 😎 and each strain gauge accounts for 6 lines. You could use the PXI-2536 8x68 matrix switch. The 2536 is an FET switch which offers unlimited relay lifetime (as long as the signals are within specs) and extremely fast switching speeds. FET switches are only good for low-voltage, low-power signals which is fine since you will be measuring strain gauge signals. This switch could switch up to 11 strain gauges to one input channel (68/6 =~ 11). With a total of 160 strain gauges, you would require 15 of the modules. All 15 could then fit in one 18-slot PXI chassis (such as the PXI-1045). Consider that you would also need a computer or embedded PXI controller to control the PXI chassis.

Also, you certainly can apply calibration data in the DAQmx software. This will offer just a software level of calibration to the incoming voltage data. However, if your application requires very high accuracy it is recommended that you perform shunt calibration everytime a new strain gauge is wired to a channel. This adds a slight amount of additional complexity and a longer wait time between measurements, but ensures that the channel is very accurately calibrated for each strain gauge prior to taking a measurement.

Hopefully this helps,

Chris G

[egret2008]

Hi Chris,

Thanks for your reply. I would really like to reduce the number of multiplexed wires down to two per channel i.e. AI+ and AI-. But I have a number of concerns regarding accuracy.

Without multiplexing the RS+ and RS- channels I will not be able to measure the actual voltage applied to each bridge. One solution would be to connect the RS+ and RS- lines to the excitation voltage (external power supply). I could then use wires for the excitation voltage with a suitable cross sectional area to ensure the voltage drop from the power supply to the bridge is negligible. This will ensure voltage fluctuations from the power supply are compensated for. An even better solution would be to measure the excitation voltage once and route this signal to all channels (rather than hard wiring each channel). Is this possible?

I think I can eliminate the need to multiplex the two SC lines because I can manually shunt each gauge with a precision resistor at the time of installation to calibrate each channel. You mentioned that for a high degree of accuracy a shunt calibration should be performed each time a gauge is connected to a channel. I assume this applies to each indexing cycle of the multiplexer. Is this to compensate for variations in the resistance of the multiplexer contacts? If I chose not to perform a shunt calibration each index of the multiplexer is it possible to estimate the maximum error this will introduce?

Regards, Ben

[Chris-G]

Hi Ben,

Your concerns about accuracy are certainly all valid. The remote sense and shunt calibration functionality on the 9237 help to increase accuracy of the measurements, but depending on a system's requirements they are not always needed. There's really no solid method of telling whether or not this functionality will be needed for your system. Really, the best way is to to take initial measurements with and without RS and SC to see whether or not they are really required to meet your accuracy goals.

In terms of your question about the excitation signal, you will not be able to simply measure the excitation voltage initially and just use that value for subsequent measurements. The reason for this is that the 9237 internally references the excitation voltage when converting the signal in the ADC. The 9237 does this because this eliminates any inaccuracies due to fluctuations in the excitation voltage source. This is also why you must route an excitation source through the 9237. The good news is that you still don't have to actually multiplex/switch the excitation source. You can just wire the excitation from one of the channels on the 9237 to multiple strain gauges in parallel.

In terms of RS and SC, you can estimate the error introduced by not using RS by measuring the wire resistance of the leads you are using. Page 14 of the NI 9237 User Manual and Specs gives the simple equation for doing so. Estimating error due to not using SC is not as straight forward. There isn't exactly a quantitative way of evaluating the error eliminated by using SC.

The reason for doing a shunt calibration everytime you connect to a new strain guage is each strain guage is slightly different. These slight variations add an additional source of error when switching between guages. However, you could just do the shunt calibration once for each strain guage. Then you could manually enter in the different shunt calibration constant for each channel everytime you switch to a new strain gauge. This wouldn't be as simple as using the DAQmx Shunt Calibration VI though. You would have to manually switch the shunt resistor on, then take a measurement and compare this to what you expect based on the shunt resistor value, then calculate a shunt calibration constant for each strain gauge. You would then have to scale each measurement based on this calculated constant.

The path length through the multiplexer will be similar for every channel through the switch module, so changes in path resistance between different channels should be negligible.

Hopefully this helps.

Chris G