PSD calculation

hi
iam trying to calculate the PSD for the acceleration data.But it is giving
unexpected result. Im collecting data for every 2msec., and im giving that
data directly to PSD function in MATLAB, is this correct way or is there
any other way to calculate PSD, My main aim is to find the gain and key
frequencies involved in the data, so i approached to PSD, is any other
method is there to find my above parameters. My data length is minimum 5000
bytes, each byte represents a seperate acceleration data. When im
calculating PSD and FFT with 512,1024 and for entire length,frequency
remains same but magnitude incresing as number of points increases. Is
there any relation ship between gain and this N-point FFT, please suggest a
correct method to resolve my problem.

Thank you

On 23 Aug, 19:18, "sudheer_raja" <sud...@efftronics.net> wrote:
> hi
> iam trying to calculate the PSD for the acceleration data.But it is giving
> unexpected result. Im collecting data for every 2msec., and im giving that
> data directly to PSD function in MATLAB, is this correct way or is there
> any other way to calculate PSD,

There are several ways to estimate the PSD. Do note that
few of the classical PSD estimatods go well with problems
where calibrated data are involved, since PSD estimators
uses all sorts of averaging techniques and give rather
crude results.

> My main aim is to find the gain

You need calibration data for that.

> and key
> frequencies involved in the data,

What *might* work here depends on what kind of information you
want and what kind of signals you have. Estimating parameters
of near-ideal sinusoidals might not be problematic, but once
the signal starts deviating from the ideal sum-of-sines
model you might find yourself in trouble.

Rune