Constantly decreasing signal

[1176789890856754634563452134512356423454]

Hello,

I am using a pressure transducer with a 100mV maximum output. I connected that pressure transducer to a SCB-68 which is connected to a PXI-6289. The transducer is connected in differential mode (ai0 & ai8). I am supposed to measure a constant zero balance output between -5mV and 5mV when I power up the transducer (10V DC) with no differential pressure applied to the transducer.

The problem is :

I measure a zero balance output but the value is constantly decreasing (even after 50 minutes). If the zero balance output is negative, the value still decreases (it does not lean toward 0).

If I connect two transducers on different channels, their zero balance output (which is not the same, depending on the transducer serial number) will decrease at the same rate.

What I tried :

I tried another power supply

I tried to measure in RSE and NRSE

I tried with another PXI station
I tried with BNC-2090
I tried with PXI-6224

I tried with other serial number transducers (3 different with all the same problem)

I tried with other cabling

I tried to measure in a different emplacement using a different power source

I tried to reduce drastically (from 10k to 1 sample per second) the sampling rate

ALL these attempt failed to resolve the problem. 

I contacted my transducer supplier. He guaranteed that these transducers should keep their zero balance output constant for 4 years. Also, the fact that the decreasing rate is almost the same when i connect different transducer would point toward a cause other then the transducer itself.

Here is an example of a test (arround 35 minutes). The voltage is the zero balance output. An unknown event caused a rapid drop at the beginning. 

Hoping to get a fast answer,

Elliot
 

[1176789890856754634563452134512356423454]

Also, I don't know if it can be linked, but when I have nothing connected to my PXI-6289, I get a constantly rising signal (until maximum range at 10V) in RSE mode, a lot of noise with random signal variations in Diff mode and the signal decreases from around 5 V to minus 0.0006V in NRSE mode.

[johnsold]

You may need bias resistors.  Read the manual notes on connecting differential signals.  

The other possibility is thermoelectric effects. Something in the system acts like a thermocouple. 600 microvolts in half an hour is not unusual for warmup drift in many instruments.  What do you get if you substitute a metal film or wirewound resistor for your transducer (keeping as much of the cabling and other stuff the same)?

Lynn

[1176789890856754634563452134512356423454]

Thanks for the fast response. 

My transducer is internally connected in a full Wheatstone bridge. Would that still be seen as a floating source even if it is connected to DC power? That would explain why I would need a bias resistor. 

I will try the bias resistor as well as the resistor test to see if it is a thermocouple effect today.  

[johnsold]

Your bridge excitation source ground should be connected to AI ground on the DAQ device.  Connected that way you probably do not need bias resistors.  A DC path for the bias currents should exist through the bridge and the excitataion source.

You said the excitation voltage is 10 V. That suggests that the common mode voltage will be 5 V. It is not at all clear from the specifications whether the 11 V common mode specification applies to small inputs.  The specifications document does not inclued any input circuit diagrams. You may need to be using the +/-10 V range to make sure that no input is out of range.  If you can use a balanced +5/-5 V excitation, then the common mode voltage would be near zero and you could use the 0.1 or 0.2 V ranges without worrying about the common mode voltages.

The specifications recommend a 15 minute warm up time for the PXI version.

The total temperature coefficient from all sources in the PXI device adds up to about 600 uV/K on the 20 V range and about 15 uV/K on the 0.2 V range.  The drift you are seeing could represent only about one degree on the 20 V range.  Depending on what else is in the chassis and how the room temperature varies, this may be within specification.

How much error in millivolts or microvolts do you calculate for the transducer itself for a realistic temperature change. Unless the manufacturer specifically compnsated the transducer at the temperature you are using, there will almost certainly be some temperature related error.  Remember that the transducer bridge resistors will dissipate some heat due to the excitation. The manufacturer almost never specifies the self-heating effects. It could easily take 30 minutes for the transducer to stabilize thermally. Does the fluid being measured vary in temperature? It is often in good thermal contact with the strain gauge.

Similarly, the excitation source probably changes with temperature.  The common mode rejection of the DAQ device is pretty good, but this will still contribute something to the total error.

Lynn

[1176789890856754634563452134512356423454]

First of all, thanks again for the complete response. 

I tried to measure the signal in RSE mode in order to determine if the bias resistor could solve the problem. I had the same decrease in RSE mode.

Also, if there would be an out of range input, I would see some bar at the top of my graph, which is not the case.

I specified in the DAQ assistant that my range was -150mV to 150mV. I guess it used the 0,2V range. If not, a one degree variation could explain my situation. I will try to use the maximum speed for the fan and I will try to confirm that the device really use the 0,2V range.

For the transducer temperature; it is compensated from 25 Celcius to 80 Celcius. If you want, here is it's specification : http://pdf.directindustry.com/pdf/kulite/miniature-is-pressure-transducer-xcl-072/7414-98687.html Mine are the smallest pressure range model, Gage version. For the moment, the transducers are sitting on the table next to me with minimum air movement arround them to test their zero balance input. The air is room temperature (which would be arround 22 Celcius).

I did not have the time to test the Thermocouple effect. In fact, I will only work on these transducer again on Monday. I will give news then. (It is 18h00 here now...)

Thank you a lot, 

Elliot

[1176789890856754634563452134512356423454]

I connected 4 different transducers on 4 different channels and I let it run for 50 minutes. After a 20-30 minute pause, I ran it again for 30 more minutes. Here are the result. The first graph represents the first test and the second graph represent the second test.

We can see that the rate of decrease is not constant and is not the same for each transducer. Also, we can see that the rate is a lot smaller when I run the test for the second time. This makes me think that it is a thermocouple effect or at least that it is linked with the temperature. 

Now, how could I cancel that effect? 

[johnsold]

It does appear to be a thermal effect.  I notice that the starting values of the second set of data are close to the ending values of the first set.  This suggests that the temperature change during the standby period is smaller than when it is running.  In the second image at about 3360 all four traces show an abrupt drop. Did the room temperature change suddenly or did something else occur at that time?

Without the actual data to check it is hard to tell but it looks like the changes may be proportional to the starting values.  This suggests a gain change rather than an offset change. A thermocouple effect would likely be an additive change which would shift all values by about the same amount.

Can you run a test with one or more of the transducers on some channels and just a resistor connected to another channel?  This would help isolate whether the drift is in the transducer or the DAQ device. Since you have been running 4 channels, try this: Connect the first channel to a transducer.  Use the one with the 6-7 mV output since that had the largest change. Channel 2: Connect to a precision resistor with a value similar to the bridge resistance. Channel 3: Short to ground. Channel 4: Connect a voltage divider consisting of two precision resistors to the transducer excitation power supply. Use values which result in an output of a few millivolts.  The results of such a test should help isolate the source of the drift.

Once you know what is drifting, then you can determine how to compensate for it.

Lynn

[1176789890856754634563452134512356423454]

Okay, I'll do that. One question thought, if my Wheatstone bridge has a 1000 Ohm resistor on each branch, what would be it's total resistance? Also, I'm not sure what is a voltage divider...

I'll post as soon as I have results,

Elliot 

[johnsold]

Elliot,

The equivalent resistance of a Wheatstone bridge of equal resistors is the same as the value of one resistor, 1000 ohms in your case.  You have two resistors in series on each side of the bridge.  Those values add. You have two of those series branches in parallel. Two equal resistances in parallel have an equivalent resistance equal to half the individual values. So 1000*2/2 = 1000.

A voltage divider is two resistors in series across a voltage source.  The voltage at the junction is the input voltage divided by a factor realted to the resistance values.

For your 10 V excitation you could make R1 = 10000 ohms and R2 = 1 ohm.  This will give you about 10 mV at TP1.

Lynn