06-08-2015 02:14 AM - edited 06-08-2015 02:18 AM
Hi everyone,
I'm trying to measure the noise of an single axis accelerometer with a PCI-4461 card.
First, I have manually measured this noise using an HP35665 signal analyser. I obtain something like this:
And now with LabView and the PCI-4461 card, I get this:
My question is: where does this part come from? And how to remove it ? (Since it doesn't exist with the signal analyser)
I'm using LabView 8.5.
This is how I configured the DAQ acquisition:
and this is my VI:
Thank you in advance for helping
Arthur
Solved! Go to Solution.
06-08-2015 08:03 AM
My first guess: Backfolded aliasing effects.
Try a higher samplerate, (highest possible) and have a look at the complete spectrum.
06-08-2015 02:02 PM
And if you were sweeping the HP source while you took the measurement, the transitions between discrete frequencies would result in high-frequency components that alias across the spectrum.
06-08-2015 10:47 PM
By the way, can you post that LVM file?
06-09-2015 04:18 AM
Hello,
First thank you for your answers
According to the datasheet, I thought the anti-aliasing filter was integrated right into the card. Is it enabled by software? (because I did not see this setting into the DAQ Assistant)
This is the LVM file: Sample rate: 2k and 400k samples
I have done other measurements with these parameters :Sample rate: 200k and 2M samples, but the results were the same (I tried to upload the LVM, but it was too heavy)
06-09-2015 05:42 PM
Antialiasing filters are built in to all the NI-DSA devices. You still have to be careful, because it can be easy to decimate in software and still introduce aliasing.
Loading your time-domain data, it is easy to see why the spectrum shows spikes at multiples of (0.94 Hz) - the time-domain shows odd signal artifacts every ~1+ s.
Can you verify the acquisition with a known input (output a sinewave on one of the analog output channels and wrap it to an input)? What more can you share about your sensor and setup?
06-11-2015 01:58 AM
Hello
Yes indeed, the time signal seems to have two paterns (in green below), once negative and positive once about 50 ms later. This pattern is repeated with a frequency of 0.94 Hz (for simplicity let's say 1Hz).
I have reproduced this pattern with matlab using a decreasing exponential:
I have obtained this spectral signal:
That closely looks like what disturbs the spectrum, indeed..
Then I have tried to output a sinewave and acquire it:
I didn't see any problem.
So I made the measurment with other sensors.
The sensors are ICP accelerometers.The sensors that I am using are in page 7 of these document:
https://docs.google.com/viewer?url=http%3A%2F%2Fvibrasens.com%2Fimages%2Fpdf%2FVS_Catalog.pdf
(I'm currently doing an intership in this company)
And finally I found that the problem came from the cable: when I use the shielded cables, the problem is present; and when I use "normal" cable, there is not the problem, as you can see :
(In red, the sensor is connected with a non shielded cable, and in white, with a shielded one.)
06-11-2015 05:10 AM
One more question:
How can I have de noise power spectral density (in Volt/rtHz or g/hertz ?) and the rms noise from those measurements?
Because I assume that the unit of the temporal graph is in "g/s" since it is a parameter of the DAQ assistant
but I don't undertsant in which unit is the power spectrum.
Thank you in advance for helping
06-11-2015 08:33 AM
The power spectrum in your code is in g^2 rms. If you want power spectral density, search (in the palette search box) for "PSD" or "Spectral Density". The Spectral Measurements Express VI (in LV 2014) also has an option to output the PSD by choosing "Power spectral density" for the Selected Measurement.
06-11-2015 09:18 AM
Thank you
Is there any difference between FFT power spectrum.vi and FFT power spctral density.vi ? Because both give me the same result :
Is it the same function?