NI9219 rtd noise

[Greg_Reynolds]

I'm seeing between 0.3F and 0.7F noise on a 3-wire RTD (unshielded) using the 9219 Universal AI.  The noise appears to be coming via the laptop pc power-charger connection and the USB (I have a single slot cDAQ 9171). When I unplug the laptop charger cable, the noise fluctuation (rms) drops to 0.07F.  Repeating this exercise on a NI9217 dedicated RTD module in the same single slot crate, I'm seeing 0.03F rms.  So am I seeing an inherent limitation on noise isolation on the 9219?

[Henrik_Volkers]

I think it should expexted that a dedicated device performes better than a universal one.... all detail in the spec 😉

Without looking into the specs, I assume that the bandwidth of the universal device is higher, so is the expected noise...

You didn't wrote the detailed setup (sample times) you where using.

The range of universal device 505Ohm where the dedicated one is 400Ohm   ...  (OK I had a look in the specs.... )   

Do you use stirred liquid baths for the sensors?  Usual configurations (>80%) a rarely useful for uncertaincies better than 0.1 deg C....  Sorry, as a metrologist I refuse to use non SI units 😉

[Greg_Reynolds]

Hi Henrik,

I was sampling at 100Hz and averaging over 0.5s to get the rms values posted.  The NMRR spec for the 9219/9217 modules is 90/85dB respectively.  I don't know if this is the appropriate spec to consider, given that the noise seems to be introduced via the USB, not the RTD wiring leads, but my measurements correspond to: (single-slot cDAQ connected to laptop via USB)

• 9219 NMRR = 40 to 50 with laptop power chord attached
• 9219 NMRR = 60 with laptop power chord disconnected
• 9217 NMRR = 70

These measurements were not taken at "High-resolution" sample times.  So I'll do that next and see if the modules will reject more of the noise.

Thanks,

[Greg_Reynolds]

The NI 9219 manual identifies "High Resolution" or "Best 60Hz rejection" conversion times for optimal NMRR values, which are around 80-90dB. I had been sampling at 100S/s for 1-2s just to get a solid average for a mean-value measurement, but this turns out to be inherently bad for noise rejection.  So I made the following comparision at several slower sample rates to see if lower noise was realized.  The rms values, T',  are taken from a running average of 50 samples at the noted sample rate.  The RTD is in a insulated water bath (but not stirred) with T average = 70F.

1. 100S/s (high speed) T' = 1.68F

2. 2S/s (high resolution) T' = 0.0026F (NMRR = 89)

3. 10S/s (best 60Hz rejection) T' = 0.049F

With the lower speed sample rates, the noise is now negligible, and the presence of the laptop charger cable (connected/unconnected) is of no measurable consequence.

[Henrik_Volkers]

Great you have a solution now, and provided that information (you should mark your post as a solution).

One thing that can try with the power supply noise:  Get some clamp-on ferrites (or a ferrite ring core)  and wrap an many as possible turn of the DC output of the supply through it.

Or get a supply with a better DC output filter 😉

this pics a from the web, the lower give an idea