Constantly decreasing signal

[1176789890856754634563452134512356423454]

I made the test. I ran it for 84 minutes. See the graphs for the results. On the first Graph: White is the transducer, Red is the 995.1 Ohm Resistor, Green is the short to ground and Blue is the voltage divide

Also, I tested my transducer (Not the same as the preceding test) with a Voltmeter (Not a NI instrument). Using only the source from NI (PXI-4110), I measured some voltage at different time. Here is the big picture:

Start : 5.69 mV

3 minutes :5.57 mV

26 minutes : 5.02 mV
34 minutes :4.89 mV
46 minutes :4.78 mV
50 minutes :4.73 mV 

I trust the source to be constant. I measured it with a Voltmeter and it stayed around the same value all the time. Also, I tested my transducer with another source (Not a NI instrument) and I had the same effect.

Elliot

[1176789890856754634563452134512356423454]

.

[johnsold · Accepted solution]

Elliot,

Excellent!  This clearly demonstrates that the transducers are the source of the drift. Your voltmeter measurements are consistent with the DAQ measurements. And no drift was visible on the measurements which only looked at the DAQ device or the excitation source.

Now you can begin to think about how to compensate the transducers.  One experiment which may be worth considering is to leave the transducers powered continuously (perhaps overnight) and then run the test again to see if they drift forever or eventually stabilize.  If the nature of your usage will not allow this as a solution, it may not make sense to do this.

One expensive compensation technique is to use a spare transducer with no pressure connections to monitor the drift.  The drift pattern of all the transducers seemed to be quite similar.  Since the drift seemed to be a multiplicative effect you could use the ratio of the peak value (near the start) to the current value as a correction to the other transducers.

Other possibilities might be to use a curve fitting process for the compensation.

Any of these methods leaves some uncertainty about their accuracy because you do not have a good model for what is causing the behavior. If you base the correction on time but the temperature is the real issue and the temperature changes differently the second or third (or 407th) time you run, then you are adjusting incorrectly and do not know it.

How big is the temperature signal you will typically be measuring?  How much error can you tolerate?  Do you really need to correc at this level?

Lynn

[1176789890856754634563452134512356423454]

Well, this is the only probable cause. I'll have to speak to Kulite about their transducers. I called them 3 times, and each time, they guaranteed my this effect would not come from their transducers. 

Thank you a lot Lynn, you really helped my narrow down the real cause of this frustrating error.

This error is very frustrating because I will be measuring pressure gradient with these transducers. So, the difference between two adjacent transducers might be 1 - 2 mV or even less. That is why I really need to correct this error. Also, the temperature of my lab will vary with the exterior temperature as I don't have any air conditioning.

Thanks again,

Elliot 

[johnsold]

Elliot,

My work is solving problems like that.

How fast do the pressures change?  Do you have any means of calibrating the transducers? How many pressure points will you be measuring? How long will any test/run last?

You saw that there are small (but significant for your application) differences between transducers.  With calibration and some means of tracking the temperature it might be possible to detect submillivolt differences but it will be ugly.

Lynn

[1176789890856754634563452134512356423454]

Well, I think I will simply power up my transducers the night before my tests. As the signal tend to stabilize, I will have a steady response at when I'll do the tests.