FFT？Problems when learning precision testing techniques! Please answer in detail, thank you for your

Take a seat, I'm writing a program for you, please wait.

The first question, to explain the problem with the program, first according to your requirements listed**, and then give the answer, the procedure is as follows:

fs = 1000;% sampling frequency 1000hz

f = 50;% Fundamental.

n = 2048;% Sampled points: 2048

n = 2000;

n=1:n;

t=(n-1)/fs;

am=[220 0 10 0 5];% of the amplitude of the signal component contained.

x=zeros(1,n);

m = 5;% Three components: 50, 150, 250, the highest 250 is the 5th harmonic, m takes 5

p=1:m;

ph=p.*;The phase is just a given.

for k=1 : m

x=x+am(k)*cos(2*pi*f*k*t+ph(k));Generate an analog signal.

end Below do the FFT analysis and draw the frame frequency map.

y = fft(x);

yabs = abs(y(1:n/2))*2/n;

fx = (n-1) *fs/n;

stem(fx(1:n/2),yabs);

According to your parameters, the first spectral line with an amplitude of not 0 after FFT is the first, that is, 0Hz, the spectral line with the largest amplitude (because I set the amplitude of 50Hz is the maximum) is about 50Hz, the reason for the inaccuracy is that your points are set to 2048, not the whole period is truncated, resulting in spectrum leakage, if you change the sampling point n to 2000, the signal is a full 2000 (1000 50)=100 cycles, or change the sampling frequency fs to 1024, It is also a full 2048 (1024 50)=100 cycles, at this time, the spectrum leakage w is 0, and the amplitude and frequency map only have three spectral lines, which are 50Hz, 150Hz, and 250Hz, which is exactly consistent with the setting; The above analysis can be verified by myself according to my program.

301hz seems to be specific, you can go to the crescent forum to ask.

x is the signal, and n is the number of transformation points.

y = fft(x) is performed on the signal x.

fast Fourier transform;

y = fft(x,n) is a fast Fourier transform on the first n points of the signal x, if n is greater than the number of points of x, then the first n points are directly taken, if it is less than n, then x is first zeroed and expanded to n point sequence and then find n point fft.

In general, N should be taken to the power of an integer closest to the length of x, which can achieve faster FFT and improve computational efficiency.

The results calculated by fft and dft are exactly the same, both are correct, and fftshift is used to make the results intuitive or for post-processing.

Personal understanding: If you look at the relationship between points from the perspective of 2-dimensional space and deal with it, it is spatial image processing, such as median filtering. In the frequency domain, it must be the 1D and 2D images after the Fourier transform and pass through the spectrogram.

Are you entering too much data. The less data, the faster the computation

First of all, we should clarify the concept, and the horizontal standard is the frequency domain. bai

The data acquisition rate is 250hz, know.

If it is 256 points FFT, then the first 128 points of the array after the DAO calculation is the result, and the last 128 are symmetrical.

The first 128 points correspond to 0 250Hz, and every 250 128 = a point, which point has a higher amplitude, that is, the amount of collected contains the frequency signal corresponding to that point.

1. The source can be intuitively observed in the temporal domain.

Horizontal reading is the frequency domain.

The data acquisition rate is 250Hz, depending on how many points you have when calculating FFT (I'm talking about discrete microcontrollers and the like), if it's 256 point FFT, then the first 128 points of the calculated array are the result, and the last 128 are symmetrical.

Then the first 128 points correspond to 0 250 Hz, and every 250 128 = one point. The amplitude at which point is higher means that the amount to be collected contains the frequency signal corresponding to that point. See.

Sidestep the question. Sidestep the question. Sidestep the question. Sidestep the question. Sidestep the question. Sidestep the question.

Yours is not a simple harmonic motion, but an approximation of simple harmonic motion, which has only one peak in the frequency domain, while non-simple harmonic motion has other frequency peaks, which should be a small low-frequency component.

FFT variation is all about converting your approximate simple harmonic spectrum to multiple simple harmonic spectrums.

You see only one peak on the spectrum, but there should be at least two peaks, so you're looking at the smaller peaks.

The damped vibration can become a lot of simple harmonic vibration, which is the FFT transformation. The peak value is the possible value, and when this part is removed, the energy of the vibration that remains is small.

In the Founier module, the amplitude and phase of the voltage can be determined.

with phase-locked loops. The phase of the 1-phase pll output can be subtracted.

To put it simply, it is displayed in an oscilloscope.

Did you find this module? Can you tell me where this module is?

It seems that everyone's situation is similar, my graduation project is based on F2812, and I also use the FFT algorithm, and it is a hybrid of C and assembly language, and I am still working on it.

Landlord I'm worse than you I'm doing the finishing design, based on 5416DSK for FFT (compilation) How do you learn now Can you communicate

It is more convenient to call the library with TI, that is, it is done with assembly, and it is very easy to use through the TMS320 algorithm standard. Reference books can also be looked at, take your time.