Validity of fourier spectrum obtained

[Water]

Hi,

I have some query on the validity of the Fourier Spectrum I have obtained.

I am currently using Labview 7.1 and S-series DAQ card PCI-6133 which have an analog resolution of 14 bits.  I am trying to sample a signal with a certain frequencies of 500Hz. I have used the sampling frequency of 10000Hz and a sampling size of 100-200 samples.

I have used the Express Vi for Spectrum measurement with the following configuration

Selected Measurement: Magnitude (Peak)

View Phase: Wrapped and in radians

Windowing: Hamming

Averaging: Hamming

Weighting mode: Hanning

Number of averages: 10

Return a signal only when averaging complete

Setting the out put to be in dB

The output which I have output to a graph suggested a presence of a signal at its frequency of 500Hz but the magnitude written is of around -60dB...which seem to suggest the reading is at uV...  Even though I expect the signal reading to be weak... but I do not think it would be that weak. Also is it possible for such a reading for this S-series card or the reading is just some mathematical computation result?

I am wondering whether such a signal reading is valid or to what extent is this plausible?

Thanks for your help!

[GerdW]

Hi water,

to check the integrity of your signal it's maybe easiest to plot it BEFORE doing FFT. Do you really get a sine at ~500Hz or just white noise?
How can we help you, if we don't know anything about your signal? You can attach vi's or pics from front panel/block diagram...

[Water]

Hi,

I think I have verified that the signal is due to some electrical coupling signal from my func generator... Anyway, is there any good way we can reduce such noise?

Thanks for your help!

[GerdW]

Hello water,

you speak of noise in the order of -60dB. You use an ADC of 14 bits resolution, giving you a ideal maximum of ~80dB dynamic range (when my calculation is right).
So at the moment you loose about 4bits of accuracy due to noise.
What do the specs of the DAQ card say about accuracy of the ADC measurement? What kind of setup do you use?
The first (good way 🙂 is always to use better shielding...

[johnsold]

water,

With 10000 samples per second and 200 sample data sets you get about 10 cycles of your 500 Hz signal. With the windowing only a few cycles will be present for the FFT algorithm. Usually FFTs work better with longer sample sets. The other issue is that the VI is set to do 10 averages. I have not looked into the internals of the VI, but if it is averaging the datasets and the sampling is not synchronous with the 500 Hz signal, then the averaging will tend to cancel it out rather than enhancing it.

Try running a dataset of 4096 or 8192 samples (assuming the signal frequency is stable over that time frame) and turn off the averaging. See if that makes a difference in what you see.

Lynn