NI-9203 generates noise pulses when acquired

[Alex2012]

It appears that NI-9203 generates noise when data are acquired from it. See attached PDF for more details how tests were made. Interestingly, noise peaks are ejected by the card into the input channel! Oscilloscope shows them clearly on input. They are coming with the same frequency as sampling rate and they have quite a bit of amplitude. When data are not acquired (from either MAX or LabView) and 4..20mA current continues flow through any of input channels then no noise is detected by oscilloscope.

And there is no way to understand or detect this (that noise is coming from 9203) by analyzing acquired data because user is sampling with the same frequency as noise peaks repetition. User see a random white noise with amplitude of about 10-20 times exceeding digital 16bit step. You need to sample at least twice more frequent but you do not do that. Only connected to the input oscilloscope or another DAQ card having higher sampling rate can show this phenomenon.

As a temporally workaround I am shunting inputs with a capacitor. For slow DC measurements it is OK but for faster applications this solution will reduce bandwidth, the card is specified for 200kS/s!

9203 costs quite a bit of money, it is clearly a hardware design flaw (or NI driver bug?). I have seen few similar posts here but NI experts seem to be diverting users to problems with their measurement setups. Really? 4-20mA measuremnt is an industry standard and should work like a clock! I have mailed this problem to NI dealer here in Sweden. I hope I will get another model for the same money.

[Henrik_Volkers]

Nice documented finding.

All multiplexed cards I ever inspected emmit EMC noise at the inputs...  even single channel  high precision DMM do it to a certain extend (If you look close enougth 😉 ) ...

The question is always: Does it effect my measurements?

[Alex2012]

Hello Henrik,

Yes it does affect measurments. I am measureing distance with laser sensors. Their "spec accuracy" is +/-1mm or ~1.5% (though usually even better) which is acceptable for me. The lasers are read by 9203 digitazing 4..20mA signal. Due to EMC noise the accuracy drops to +/-7mm which is not acceptable for my application.

[Alex2012]

Now NI support from England called me. They will look at the problem. There are several reasons which might be responsible for the noise as we have discussed:

 - ground coupling with internal ADC

 - design flaw (for example too low 20pF input shunting capacitor etc)

 - faulty unit which I've got

Well, I have a hope now 🙂

[renehabel]

Hey Alex,

i've got the same problem with this module and acquiring 4-20mA signals. It leads to some worse peaks in our measurement.

So we decided to oversample this module with a "fast-task" and re-sample the results to our "normal-task"

Normally: 1000Hz (temperature and voltage)

Fast: 25000Hz and then resampling to 1000Hz. (current)

This works well...

Best wishes

René

[Alex2012]

René, thanks for tips! Sure one can do that, but what if you need to sample 100kS/s or 200kS/s (specs max) then you cannot sample more than specs limit. A capacitor on each input solved my issue.

Well I've got answer from NI saying that "... issue is caused by opening the channel in the ADC. I have also confirmed that a similar issue occurs with other models of cards." And they cannot do much about this.

My conclusion is simple: DO NOT PURCHASE NI-9203 CARD.

I have soldered current sensing boxes myself, usually one uses current sensing IC and ADCs chips from Texas Instruments for the cost of few $$. And surely they do have some nV/sqrt(Hz) but it is on the level of the digital noise. Here 9203 has simply terrible design I suspect. Also NI “other models” do not emit noticeable spikes into the input line, I have not tried all, but have not seen anything like that with many other NI cards. What a lame design...

[Sean365]

Hi Alex,

I am experiencing the same issue with NI-9203. What size capacitor did you use and how did you wire it in?

Thanks,

Seán

[Alex2012 · Accepted solution]

---

Sean365 wrote: I am experiencing the same issue with NI-9203. What size capacitor did you use and how did you wire it in?

---

Hi Seán! I used a couple of µF which might be overkill if your sampling rate is some kHz or higher. For my upto 50Hz sampling it was totaly OK. For kHz sampling 100nF might be enough, you have to test it. I connected it in parallel with each NI-9203 input I've used, right near the input because this noise is coming out from 9203 and you don't want to propagate it further into your circuit. So, kill that noise right at the place of origin

[Sean365]

Thanks Alex.

I am sampling at 150Hz.

Before installing the Capacitors, I was experiencing 0.4mA variation per channel which results in a variation of 0.3mm with the disance sensors that I am using. The sensors have a resolution of 1um and I need 10 to 20um for my application.

After I installed a 100uF capacitor across each channel the variation dropped to 3um!!!

This problem has caused me a huge amount of grief and delays in my project. Now hopefully I can make some progress, thanks all.

[Alex2012]

---

@Sean365 wrote:

Thanks Alex.

I am sampling at 150Hz.

Before installing the Capacitors, I was experiencing 0.4mA variation per channel which results in a variation of 0.3mm with the disance sensors that I am using. The sensors have a resolution of 1um and I need 10 to 20um for my application.

After I installed a 100uF capacitor across each channel the variation dropped to 3um!!!

This problem has caused me a huge amount of grief and delays in my project. Now hopefully I can make some progress, thanks all.

---

I am glad it helped! Please click "accept as solution" my reply above. Though note that too high capacitance (100µF is a bit high) will cause slow response of your distance sensors. You will not be able to measure fast changes, instead they will come with a delay (immediate change will be measured as several points approaching final value slowly). I would recommend 2-10µF.