problem in frequency analysis of data from 9237

[Zhe]

I used the NI 9237 and NI cDAQ-9172 module to measure the strain change of a micro-force sensor (AE-801). The sensor is beam type with strain gauge on each side. There is some problem when I did freqency analysis of the data from the NI 9237. When there was no force applied on the sensor, I recorded the measurment from the 9237 and did FFT. There is always high freqency reading in 100, 200, 300 ,400 ,500, 600Hz. Is there anyone can help to explain this problem? 

The sensor was placed on a vibration-isolation table. The sensor beam is free. In 9237 setting, I used the half bridge configuration. The setting of gage factor is 60. Gage resistance 1.12k. Vex value 5. Continuous samples, 5000 to read with 10K Hz rate.

I tried another type of sensor (also strain gauge on each side of beam). This FFT time the reading is 50, 150, 250, 350, 450... Hz.

[hoffiz]

Hi Zhe,

Before taking a look at the frequency reading on the FFT, I will like to know in the time domain: what is the average reading when a force is applied?  What happens if nothing is connected to the 9237 do you still read 100, 200… on the FFT? When you do the FFT what is the value (dB or Volts) of those frequencies? How fast are you expecting your signal to change? Can you explain a little bit more why are you taking the FFT of your signal I would like to know the information are you trying to get from it?

Since the sensor is placed in a vibration isolation table; like you said, those frequencies are more likely to be noise either in the cable/connections or in your Lab. What we need to do is first identify if it is noise and the source of this noise. Some suggestion I have for you are changing the slot in the cDAQ-9172, if it’s possible move the cDAQ to a different place on your lab or a different room, change the sampling rate to make sure we are not oversampling, disconnect any other module from the 9172 when doing this troubleshooting.  While I’m waiting for you to answer those questions I will suggest taking a look at: Measuring Strain with Strain Gauges.

I hope this helps

[Zhe]

Hi Jaime:

 Thanks for your help. I tried your methods but it still did not work. Please look my response between your questions. 

    What is the average reading when a force is applied? 

>>In my application, the average strain reading is around 100e-7 when a force is applied. When there is no force, the average reading is around 0.1e-7. 

What happens if nothing is connected to the 9237 do you still read 100, 200… on the FFT?

>>I can not do the measurment of 9237 in LabVIEW when nothing is connected. In DAQassitant, the error shows "device is not no longer presented". When I tried in "Measuremet and automation explore", the reading is around -0.25.

When you do the FFT what is the value (dB or Volts) of those frequencies?

>>I used the "fft" function in Matlab to do the FFT analysis. The frequency value is represented using power spectrum. A typical value is listed: (when there is no force on the sensor) 

Frequency (power spectrum) at 50Hz(1e+6), 100Hz(1e+7), 200Hz(1e+6), 300Hz(1e+5), 400Hz(5e+4) 

How fast are you expecting your signal to change? Can you explain a little bit more why are you taking the FFT of your signal I would like to know the information are you trying to get from it?

>> I tried to use a piezoresistive micro-force sensor to measred the vibration of a golden wire. When I analyzied the signal, there are always 100, 200, 300 Hz peak in the FFT results. I want to know where are these signals come from?

Some suggestion I have for you are changing the slot in the cDAQ-9172, if it’s possible move the cDAQ to a different place on your lab or a different room, change the sampling rate to make sure we are not oversampling, disconnect any other module from the 9172 when doing this troubleshooting.

>> I tried following: 1)change the NI 9237 to different slot positions in 9172, 2) change different channels in NI9237, 3)change sampling rate (50k,25k,10k,5k), 4) change the cDAQ position in the lab. But all these methods didnot work.

[Neal M]

Hi Zhe,

Since the frequencies are so regular, it appears that it may be due to harmonic frequencies of a signal source.  I have a couple more suggestions of things to try to narrow down the problem or remove the noise:

- Is the noise still there if you replace the strain guages with resistors (precision perferably)?  This would eliminate the SGs as the cause of the noise.

- Have you run a calibration in Measurement and Automation Explorer? This would allow you to remove any offset and do a shunt calibration, which would also remove error when you apply strain. You can do this in the Device tab of a task created in MAX.

- Are you using Remote Sensing?  Depending on the length of your leads, remote sensing would remove any error caused by lead resistances. Making the leads as short as possible would also reduce the chance for noise to be introduced into the system.

- When you answered Jaime's last question about moving the chassis around, etc, you say that "these methods did not work."  Do you mean that the noise does not dissappear completely or it does not change the noise at all?  If the noise changes magnitudes but does not go away, then that is indicative of some external cause (e.g a fan, lights, or other EMF source). I am not sure how your power spectrum units would scale if the source changed, if there is a way to convert it to dB or volts, that would be easiest.

I realize that most of these do not directly attack the noise you are seeing, but reducing total error should help narrow down the cause.

Here is a link to some low level measurement articles.  They mostly deal with multimeters and switches, but there is some good information that could be useful to you.

[Zhe]

Hi Neal M.:

  Thanks for your advice. I think I found the source of the noise. But there is still one question. Please look my response between your questions.

- Is the noise still there if you replace the strain guages with resistors (precision perferably)?  This would eliminate the SGs as the cause of the noise.

  I used two 1k resistors to replace the micro-force sensor (resistors on double sides of silicon beam) . Now there is only 50Hz left in the frequency domain. But when I connected the micro-force sensor back (without changing the 9237 setup and the placement of 9237 and sensor), the harmonics frequencies appear again (100, 200, 300 HZ..) and are very strong. So it seems that the harmonics frequencies come from the micro-force sensor. But I do not know why this happened? The sensor is only a piezoresisitve force sensor. Please advise.  

- Have you run a calibration in Measurement and Automation Explorer? This would allow you to remove any offset and do a shunt calibration, which would also remove error when you apply strain. You can do this in the Device tab of a task created in MAX.

  I did the calibration before I started the measurment. The calibration has no effect to remove the harmonics frequencies in the measured signal.

  - Are you using Remote Sensing?  Depending on the length of your leads, remote sensing would remove any error caused by lead resistances. Making the leads as short as possible would also reduce the chance for noise to be introduced into the system.

  No. The sensor output pins were directly connect to the NI 9237.   

- When you answered Jaime's last question about moving the chassis around, etc, you say that "these methods did not work."  Do you mean that the noise does not dissappear completely or it does not change the noise at all?

  The noise does not change at all.

[Neal M]

Zhe,

From your answer to the first question, I agree that the noise appears to be coming from the micro-force sensor.  I am not terribly familiar with piezoresistance, but from what I do know, I have a couple theories as to where the noise may be coming from.  First, there is a chance that the harmonics we are seeing are the natural frequencies of the material in the sensor being induced by the excitation or something similar.  It also seems to me that sensors of this type would be extremely vulnerable to external noise, so if the sensors are near a noise source (e.g a computer monitor), that could cause the noise as well. 

The 50Hz noise you see when you replace your sensors with resistors is almost always caused by flourescent lights or some other wall powered device. I would reccomend looking in the manual for the sensor or possibly contacting the manufacturer to see how to best get around this noise.