Measuring Spectrum value at particular frequency

[Water]

Hi,

I am trying to measure the spectrum amplitude at a frequency of 5KHz. The current method which I am doing is by reading off the  spectrum value which is obtained through the FFT of the captured 2000 samples at 100kHz. I am using PCI 6133. The main problem I have is that the speed at which I can read off is pretty slow. Time for acquisition of samples took about 0.02s (2000 X 1/100 000 =0.02). Is there any good way to improve on the time taken for signal readout? I would need to read numerous spectrum amplitude at varying location.

Some other question I have

1) Is it true that a large sample size help to improve the noise floor?

2) Can we use interpolation to reduce the amount of samples we would need to take. For example, can we interpolate for the samples between two acquired samples and hence increase the acquired sample size by 2 in this manner?

3) Could we try signal mixing, mixing the signal with a 5Khz digital generated signal and retrieved the value required?

Thank you!

Message Edited by Water on 08-26-2007 11:10 PM

[Alex A.]

1) A larger sample size will improve the resolution of the noise floor and allow a separation of the signal peaks in FFT.

2) Interpolating the samples would not do anything to help your measurements.  It would actually slow processing down due to calculations and would not increase the  sample size effectively.

3) For this question I would recommend our  NI Spectral Measurements Toolkit. 

In all, the Spectral Measurements Toolkit would be the best suited for your application because as your simple equation stated earlier, the only way you'll be able to speed up the time is to change your number of samples or sampling rate.  The toolkit would allow you to optimize the time and seeing how it affects the accuracy of  your measurements more clearly.

[ChrisR.]

Even without the Spectral Measurement Toolkit, there's the "Extract Single Tone Information" and "Extract Multiple Tone Information" vi's in LabVIEW that should provide the info you need pretty efficiently. You say your measurements are slow: are you saying the 20msec of signal acquisition is too long, or that it is taking too long to process the data? How long is too long?

Chris