Very low frequency caused by sample frequency in FFT analog input?

[BartL]

I'm measuring a very low frequency on my analog input, this frequency is in connection with the sample frequency of the Analog Input. At a sample frequency of 1000Hz I see a frequency of 0,05Hz in my FFT, at a sample frequency of 500Hz I see a frequency of 0,02Hz.

Attached is a screenshot of an example how I see this very low frequency.

My hardware: NI USB 6008 --> measuring on AI-0 (in this example the input is unwired). But in my real measurement I see the same FFT + signals I want to see (about 2 Hz).

In my real measurement I windowed the FFT (1-3Hz) so I see only the FFT I want to see. But I suspect that my complete signal moves along with this very low frequency of 0,05Hz. I saw this in my measerement.

What did I do wrong?

[mikeporter]

Assuming a 1kHz sample freq, if you expand the freq scale to 500 Hz are there any other strange peaks in your data? Also what antialiasing filters are you using?

Mike...

[BartL]

Attached you can see a screenshot FFT 0-500Hz. There are no other "relevant" peaks, when I zoom there's only a very small peak at 50Hz.

I didn't consciously use an antialiasing filter. The only two things I setup are the buffer of 20000 samples and a flattop window (antialiasing filter?).

[mikeporter]

First, do you live in Europe? If so, that 50Hz could be power-line pick up.

Antialias filtering must be done in hardware before the DAQ. Because of the way aliasing works if you have sampled the signal it's already too late, you're hosed and no amount of digital filtering can remove the aliased signal. In terms of filter specifications, the filter cutoff needs to be at twice the highest frequency you are interested in seeing. For example, if you are looking for signals in the 2- to 5-Hz range, your antialiasing filter should cutoff at around 10Hz.

Obviously good signal management is also needed: shielding, appropriate signal termination, proper lead dress and spacing from known noise sources, etc...

Mike...

PS: There were no attachments to your last post.

[BartL]

- I abviously forgot the attachement. --> now it is attached

- Yes I live in europe so 50Hz is not that strange, but also a very small signal.

- With the sample frequency of 1000Hz I mentioned the sample frequency I can change in Labview. I don't sample in my hardware. The first screenshot show's the software, and my hardware is an unwired Analog input on the USB6008 NI card.

[Hossein128]

Dear Bart,

I think that this low frequency is 1/f noise that you have in your application. By adding some AI/O with higher frequency you can see the influence of that in your signal. I think that you can ignore that if you will work in the higher range of frequency.

Regards,

[ejkaiser]

It can also be a DC offset. It would would show up in the very first bin of your FFT.

Cheers

Edgar

[johnsold]

I would suspect DC offset also.  1/f noise is not likely very significant at 50 Hz.  What does that spectrum look like when the mapping is set to Logrithmic?

Lynn

[Vali.G]

Hello,

Did you find any solution to this problem? It seems I have a simmilar issue with a sound card.

Cheers,

Vali

[johnsold]

Vali,

It might be better for you to start a new thread.  You could link to this one if you feel it is relevant.

Please be as specific as possible in describing your issue. Posting your VI with typical data sved as default can be very helpful. Note that most sound cards do not process very low frequencies (~<20 Hz).

Lynn