NI-9229 Sensitivity

[Emad_NRIAG]

Hi,

I'm working on developing seismic digitizer so, we trying to configure the NI-9229 to match the required specifications. 

Please, I wanna to know the NI-9229 sensitivity (what's the relation between counts per volt) in order to calculate the system sensitivity after connecting seismometer ( sensor for measuring seismic wave).

Kindly, download the attached picture for the current configuration. Moreover, what is the corresponding sensitivity in case of changing the voltage input range (V_peak to peak).

Great Thanks 

[croohcifer]

Sensitivity is a loaded term that doesn't have a universally accepted definition.  There are two actual quantities that might help answer your question.  The ADC resolution, aka "quantization error", is (your_selected_input_range)/([2^adc_bitness] - 1), as with any other ADC.  The 9229 is a 24-bit ADC, and note that there is only one input range, +/- 60 V, on the 9229.  So, per that equation, (120 V)/([2^24] - 1) = ~7.2 uV.

As with every ADC I've ever seen, the thermal/electrical noise will be larger than the quantization error.  The 9229 has an "input noise" of 320 uVrms.  This value, and NOT the quantization error, defines your noise floor, and this is the number you should probably care about.

Sensitivity is usually defined as something like "the smallest DC step that can be meaningfully be detected by an ADC" or something equally nebulous and unhelpful (and subject to test instrumentation marketers trying to make their instruments look better than they are).  If your input made a 7.2 uV DC step but you still had 320 uVrms of noise sitting on that signal, I think we'd both agree that you probably wouldn't "meaningfully" register that DC step change unless you averaged the signal for an eternity.  So, the natural follow up is, "How much would the input have to change for you to "meaningfully" detect it?"  Great question.  We could sit here and argue about that very thing for another eternity because it's a qualitative measure, which is why "sensitivity" is a pretty terrible spec.  We could define "sensitivity" as a function of the input noise or something arbitrary like that, but then it would be a fairly redundant metric.  (This is the same reason why a device usually won't define both SNR and ENOB.  One is readily calculable from the other.) 

TL;DR:  Throw "sensitivity" out of your vocabulary and replace it with something more readily quantifiable like input noise.

[Emad_NRIAG]

Great Thanks, Mr. Croohcifer for your organized and professional explanation.

I really understand the sensitivity meaning like the sun rays 🙂