NI-9203 generates noise pulses when acquired

[sherry.etan@gmail.com]

I use the NI DAQ 9203 module and a 4 - 20 mA pressure sensor. I am very new to using Labview and electrical circuit connections. My data output is very noisy, and I have used a digital filter and even a low pass filter in my LabVIEW VI. Though it reduces the noise, I would be more comfortable if my output data had little or no noise. I have tried the RC circuit connection, and it has no impact; there may be something wrong with what I am doing, as I have seen that the RC circuit significantly reduced the noise for some people in this thread. I desperately need help as this has caused months of delay in my research. Any help will be deeply appreciated. Attached are the sensor datasheet, my Labview VI, a screenshot of my data, and the RC circuit I connected. 

[Alex2012]

Hi!

Unfortunately I could not open your VI because I have slightly older LV version on my computer. Seems that your data were collected during a few days. I am wondering what is you sampling rate? Does 0.1 on your VI snapshots (in the doc file) means one measurement per second or 100mHz? If this is the case, I would recommend to to increase sampling rate to say 100..400 Hz (400 is the max for your sensor) and then take an average value. I assume you are from US, so since the electrical power grid is operating at 60Hz there, try to avoid sampling rate which is multiple of 60. So, 100 or 200 Hz sampling rate would produce a single point for you which, I assume, you store in a file. So, you will still get 1 data point per 10 seconds, but this point is well integrated or averaged.

Then, perhaps, you do not need to perform digital filtering. Your filtered/processed data look too smooth, of course I do not know what system you are dealing with, although perhaps you are interested in finding some features (local min/max and bumps) in your pressure curve.

If you suspect that 9203 noise is still affecting your data, try increase sampling rate to a few kHz and then average it out. Even if your sensor is limited to 400Hz bandwidth, it is OK to oversimple because the goal is to overcome noise in 9203.

Let us know how you fixed it in your application by posting some cool plots. Good luck!

[OfirHarari]

Hello,

I'm sorry to hear about the noise issue but it is well known, and we have learned our lesson in this subject.

All our efforts to eliminate the noise were not successful. We have tried to use capacitators, changed wirings etc... with no success.

I have replaced all the 9203 cards with NI 9253 cards, and it is ok now.

Ofir

[Alex2012]

---

@OfirHarari wrote:

Hello,

I'm sorry to hear about the noise issue but it is well known, and we have learned our lesson in this subject.

All our efforts to eliminate the noise were not successful. We have tried to use capacitators, changed wirings etc... with no success.

I have replaced all the 9203 cards with NI 9253 cards, and it is ok now.

Ofir

---

The 9203 had a lame design fault (I skip describing the issue in details). It is even more lame because NI did not bother to test that product before its release. Interestingly, when I opened support ticket back then, the NI support engineers performed "an investigation" for a few weeks and then totally denied the issue, thus, refusing to replace several 9203 modules our organization have purchased. They sent me even a report with totally flat and horizontal plots. That was simply a lie to a big customer on a corporate level.

About the same time I had even worse issue with EtherCat temperature modules REM-11120. Turned out that it was another design fault by the NI's subcontractor company which refused to cure it. It took 1 year for NI to investigate it!

So, having 16 years of experience in application development with LabView I really lost confidence in NI products and nowadays when I'm building a solution or making recommendations for my own customers the NI products are at the bottom of the list, unfortunately.

I am sorry @OfirHarari but I do not easily buy your advise to purchase NI-9253 as a replacement product. I simply do not trust it because it may have similar issue(s). I, and I guess people in this thread, would expect a good work around solution from you (NI) allowing to deal with the problem with existing 9203 which we purchased. Can you advise a proper band pass filter schematics developed by NI engineers? Or a free of charge VI with suitable and flexible digital filter? Or something else like an add-on board, analogue or digital etc? That would be highly appreciated.

[sherry.etan@gmail.com]

Thanks so much for your detailed response. I am sampling at a rate of 0.1 Hz because I didn't want a large amount of output data and didn't know I could output a mean value for every 100 or 200 samples. I am modifying my VI as you advised to see if this will help. It's hard for me because I am not conversant with Labview. Regarding the digital filter I already did, it oversmoothens the data, and I get significantly different values of Diffusion coefficients (which are obtained numerically from the pressure vs. time plot) at different levels of smoothening. My supervisor says he does not want to see the noise in the output data. I will post the results once I know how to do the average.

[sherry.etan@gmail.com]

I have emailed them to ask if I could exchange the 9203 module for one without noise. I am still waiting for a response.

[Alex2012]

---

sherry.etan@gmail.com wrote:

Thanks so much for your detailed response. I am sampling at a rate of 0.1 Hz because I didn't want a large amount of output data and didn't know I could output a mean value for every 100 or 200 samples. I am modifying my VI as you advised to see if this will help. It's hard for me because I am not conversant with Labview. Regarding the digital filter I already did, it oversmoothens the data, and I get significantly different values of Diffusion coefficients (which are obtained numerically from the pressure vs. time plot) at different levels of smoothening. My supervisor says he does not want to see the noise in the output data. I will post the results once I know how to do the average.

---

Another thing which can work instead or in addition to averaging is to apply median filter. The noise in 9203 comes as a quasi-periodic pulses. In theory, such noise can be filtered out more efficiently with median filter. There are already such functions in LabView (pint by point and other). Still a higher sampling rate is required.

[sherry.etan@gmail.com]

All through yesterday, I tried to apply the median filter as you advised, and I did not get any output readings. I know it is due to a misunderstanding of how to apply the sampling rate, seconds per data timing, and the median filter, but I don't know exactly what's wrong because I am very new to using Labview. Since you are unable to open my VI, I have attached a screenshot below for your further help. Thank you so much.

[Alex2012]

---

sherry.etan@gmail.com wrote:

All through yesterday, I tried to apply the median filter as you advised, and I did not get any output readings. I know it is due to a misunderstanding of how to apply the sampling rate, seconds per data timing, and the median filter, but I don't know exactly what's wrong because I am very new to using Labview. Since you are unable to open my VI, I have attached a screenshot below for your further help. Thank you so much.

---

Please look at the example VIs supplied in LabView so you can see how to use it. I guess it is an "iterative" function which you need to use in the loop. Alternatively, you can use median filter function from other toolbox which can be applied directly on the sampled array (no iterative loop).