FFT - Science topic

I get a peak at 240 Hz in FFT plot of tool vibration. The accelerometr is placed at 45 mm from tool tip during machining. Turning of nickel steel of dia 60 mm and length 300 mm at spindle speed 256 RPM. PCLNR2020K tool holder. Panther conventional precision lathe it is.

Our research team met one question on calculating EEG relative power and absolute power at this stage.

When we integrated all negative and positive amplitude/power data in five EEG bands (delta, theta, alpha, beta, gamma), a few relative power results became huge (i.e., 440%(44.44) or even over 1000%). We thought these values were abnormal results. The reason is that the integration result of five EEG bands with negative and positive power values could be 1 or 2 as the denominator, but the numerator could be very large for the integration of one specific band(i.e., delta). The relative power calculation is (sum of spectral power in the band)/(sum of spectral in all bands)

The attached image showed some negative and positive spectral power values.

Therefore, we would like to ask whether we need first to transfer negative value to absolute value to consider relative power or absolute power. Normally, the relative power should be around 0-100%.

Can experts help us? Could experts please share some references with us?

We conducted a 32-channel EEG and used FFT to calculate each frequency's Power-Value. Some of them are negative and I don't know how to interprete those im comparison to positive Power-values.

If you take FFT and then inverse FFT of a HR TEM image. What information is obtained from FFT and IFFT? The corresponding images are attached.

Answers will be highly appreciated

Nodes used for this job:

------------------------

node02

node02

node02

node02

node02

node02

node02

node02

------------------------

Working directory is /data_hp/home/Rohith/2_AuAgCuPt/2_Convergence-test/1_Kp-test/7_kp

running on 8 total cores

distrk: each k-point on 8 cores, 1 groups

distr: one band on 1 cores, 8 groups

using from now: INCAR

vasp.5.3.5 31Mar14 (build Aug 31 2021 16:15:13) complex

POSCAR found type information on POSCAR Au Ag Cu Pt

POSCAR found : 4 types and 32 ions

-----------------------------------------------------------------------------

| |

| W W AA RRRRR N N II N N GGGG !!! |

| W W A A R R NN N II NN N G G !!! |

| W W A A R R N N N II N N N G !!! |

| W WW W AAAAAA RRRRR N N N II N N N G GGG ! |

| WW WW A A R R N NN II N NN G G |

| W W A A R R N N II N N GGGG !!! |

| |

| For optimal performance we recommend to set |

| NCORE= 4 - approx SQRT( number of cores) |

| NCORE specifies how many cores store one orbital (NPAR=cpu/NCORE). |

| This setting can greatly improve the performance of VASP for DFT. |

| The default, NPAR=number of cores might be grossly inefficient |

| on modern multi-core architectures or massively parallel machines. |

| Do your own testing !!!! |

| Unfortunately you need to use the default for GW and RPA calculations. |

| (for HF NCORE is supported but not extensively tested yet) |

| |

-----------------------------------------------------------------------------

-----------------------------------------------------------------------------

| |

| ADVICE TO THIS USER RUNNING 'VASP/VAMP' (HEAR YOUR MASTER'S VOICE ...): |

| |

| You have a (more or less) 'large supercell' and for larger cells |

| it might be more efficient to use real space projection opertators |

| So try LREAL= Auto in the INCAR file. |

| Mind: At the moment your POTCAR file does not contain real space |

| projectors, and has to be modified, BUT if you |

| want to do an extremely accurate calculation you might also keep the |

| reciprocal projection scheme (i.e. LREAL=.FALSE.) |

| |

-----------------------------------------------------------------------------

LDA part: xc-table for Pade appr. of Perdew

POSCAR, INCAR and KPOINTS ok, starting setup

FFT: planning ...

Suppose we should do calculations in frequency domain, so we should use FT of a continuous-time signal that is zero outside a boundary: X(w) = F{x(t)}. We know that FFT can be used for DFT on computers, but it assumes that the signal is periodically repeated outside the boundary.

Then we obtain Y(w) after our desired frequency-domain calculations. Now we want to estimate the time-domain signal, y(t) by applying the inverse FFT: y(t) = IFFT{Y(w)}. But this inversed transform is also assuming that the signal is periodic and is defined from -inf to inf.

Is there any way (i.e. numeric calculations on computers) to obtain the best estimate of out nonperiodic time-domain signal, y(t)?

Actually I have two interferogram ( containing Young's fringes). One interferogram is recorded without any perturbation in the path of light rays. While recording the second interferogram, I have kept a glass plate in the path of one light beam. Now I want to get information about refractive index of the glass plate by analysing the interferogram. Both Interferogram have been recorded usin a CCD camera. Can anyone help me?

I have calculated the gear meshing frequency of planetary gearbox to be 786 Hz. However, when a FFT is performed on the data acquired for the same planetary gearbox I could see peak around 645 Hz and not at 786 Hz.

The calculated mesh frequency was done based on the speed and number of teeth. But the signals acquired during operation was under loaded condition.

Does external load change the natural frequency and meshing frequency of gear?

Is there any reference to calculate the theoretical gear mesh frequency in relationship with load.

Attached FFT plot.

Thanks in advance for sharing you knowledge.

I need to perform Sparse FFT on 1D signal , where can i found the corresponding code in MATLAB?

I need to decompose a signal on Matlab but I ended up having IMFs that with FFT instead of having one peak it shows several. Does anyone know how to decompose a signal with this method?

I'd be appreciated someone who could help me out

Hello I am trying to reconstruct the far field pattern of a patch antenna at 28 GHz (lambda = 10.71 mm ). I am using a planar scanner to sample the near field with a probe antenna. The distance between patch and probe is 5 cm. The resolution of the scan is 1.53 x 1.53 mm². The total scanned surface is 195x195 mm. The NF patterns are shown in the NF_raw file.

The complex near field is then transformed using the 2D IFFT to compute the modal spectrum of the plane waves constructing the scanned near field. (See C. A. Balanis (17-6 a and 17-7b) for this). The modal components are shown in the IFFT file. The problem is that is observe an oscillation in the phase of those modal components that reminds me of aliasing effects in digital images (Moiré pattern).

This effects also procreate when I resample the modal spectrum in spherical coordinates, as seen in the Sampling file. The transformed phase changes therefore too fast per radian. The absolute value of the pattern looks reasonable.

Could someone explain why these effects occur and what steps I can implement to prevent them? Thank you for any helpful input.

The background is, we are trying to calculate an index relying on high frequency band over 100Hz with only 128Hz signal. The assumption is that: Say we have a 128Hz signal, while using fft to convert it into frequency spectrum which will get information from 0-64Hz according to Nyquist. Then, if we have the original signal subtracting ifft of the 0-64Hz spectrum, will it produce some information of 64-18Hz band?

i am trying to plot FFT spectra of bootstrap switch Sample and hold circuit, i have got the fundamental frequency and harmonics distortion components in the graph. i want my graph like 2nd image but i am not getting noise. can anyone please help me out where am i making mistake?

We tried to identify the secondary phase by comparing the d-spacing from the lattice fringes and secondly by taking localized FFT.

a) Is FFT the localized variant of the SAED pattern?

b) How to differentiate between the two different phases and if the same phase has two different orientations in FFT?

c) How to make sure that our crystallites are well oriented along the zone axis?

if anyone can suggest any document/video/paper which shows/tells how to perform 2D-FFT analysis to determine the size of magntic domains in MFM images

I want to detect anomaly from a streaming data. Using FFT or DWT history is it possible to detect anomaly on the fly (online) . It will help a lot if anybody could suggest some related resources.

Thanks.

How to combine multiple frequency response function (FRF) files into a single FRF file ? If I have multiple FRF data files from any FFT analyzer then how could it be possible to combine them into a single FRF files within a certain frequency range ??

Hi all,

I am doing EEG preprocessing using FFT. The sampling rate is 128 Hz, epoch length is 2s, 256 data points. After I applied fft(), there are still 256 points. What is the frequency resolution and frequency range of the result?

The Hough transform is a feature extraction technique used in image analysis, computer vision, and digital image processing. The purpose of the technique is to find imperfect instances of objects within a certain class of shapes by a voting procedure. This voting procedure is carried out in a parameter space, from which object candidates are obtained as local maxima in a so-called accumulator space that is explicitly constructed by the algorithm for computing the Hough transform.

The image below shows transforming Hough from a cell hologram. How do I get image edge pixels?

Dear Researchers,

suppose I have an wideband is signal with F_low=200MHz and F_high 400MHz. I want to decompose this wideband signals into 26 narrowband signals using 256-points FFT.

What I understood is that each FFT point will represents a frequency bin. does it mean we will end up with 256 narrowband signals? if yes, why published works said that we apply 256-points FFT to decompose this wideband into 26 narrowband signals?

I am just confused with these terms (FFT points, Frequency bins and narrowband signals).

I appreciate your clarification

Thank you for your responses

Researcher

Bakhtiar

BICGSTAB-FFT can be used in DDA because the special form of its interaction matrix, which makes the time complexity reduce from O(N^2) to O(NlogN). Iterative method such as bicgstab can also be used to solve the matrix in finite element method, but it seems FFT(fast fourier transform) can not be implemented in this case.

Does this mean, suppose with the same number of elements (or matrix size), finite element method will principally be more time consuming than DDA simply because the more time consuming matrix vector multiplication in bicgstab (suppose bicgstab runs the same number of iterations).

I try to transmit LFM signal using SDR platform (usrp-2932) and receive The Signal using another usrp of The same version , The TX and RX ports are connected through RF Cable With certain length .

MY question is : it Must appear single bin at certain frequency after applying FFT, but The received pin is very fast , what is The factors that affect this phenomena?

In courses about DSP that I did at university we only covered theoretical material, I am looking for a good book that covers practical implementation of DSP in MATLAB like designing filters and DFT or FFT.

also looking for good books on signal processing with MATLAB in general.

Thanks.

Dear all, I have some real data (about 32 equidistant points), and I fitted it a Fourier transform function using the FFT method. Indeed I get 32 complex Fourier coefficients, which correspond to the obtained 16 positive frequencies. I want to apply a low pass filter to smooth the obtained fitted function. Actually I take the fifth frequency as a low pass threshold (so I take only the first five frequencies which correspond to 30% of the total frequencies). I have chosen this threshold, basing on a visual interpretation of the fitted curve. Can anyone suggest a more robust or efficient method to choose the threshold frequency for low pass filter?

Good day all,

the pictures is one of the background noise I captured in anechoic box, in FFT and unit is dB.

Can any body provide me a suggestion on why there is a peak at 12 kHz, 13.5kHz, 15kHz, and 18kHz?

I believe that it is not originated from any structural issues, justification is that I tried excite my chamber with diffuse pink noise, but dB value at these peak stay the same.

The complex signal may be real-imaginary or magnitude-angle form.

I have torques and angular positions data (p) to model a second-order linear model T=Is2p+Bsp+kp(s=j*2*pi*f). So first I converted my data( torque, angular position ) from the time domain into the frequency domain. next, frequency domain derivative is done from angular positions to obtain velocity and acceleration data. finally, a least square command lsqminnorm(MATLAB) used to predict its coefficients, I expect to have a linear relation but the results showed very low R2 (<30%), and my coefficient not positive always!

filtering data :

angular displacements: moving average

torques: low pass Butterworth cutoff frequency(4 HZ) sampling (130 Hz )

velocities and accelerations: only pass frequency between [-5 5] to decrease noise

Could anyone help me out with this?

what Can I do to get a better estimation?

here is part of my codes

%%

angle_Data_p = movmean(angle_Data,5);

%% derivative

N=2^nextpow2(length(angle_Data_p ));

df = 1/(N*dt); %Fs/K

Nyq = 1/(2*dt); %Fs/2

A = fft(angle_Data_p );

A = fftshift(A);

f=-Nyq : df : Nyq-df;

A(f>5)=0+0i;

A(f<-5)=0+0i;

iomega_array = 1i*2*pi*(-Nyq : df : Nyq-df); %-FS/2:Fs/N:FS/2

iomega_exp =1 % 1 for velocity and 2 for acceleration

for j = 1 : N

if iomega_array(j) ~= 0

A(j) = A(j) * (iomega_array(j) ^ iomega_exp); % *iw or *-w2

else

A(j) = complex(0.0,0.0);

end

end

A = ifftshift(A);

velocity_freq_p=A; %% including both part (real + imaginary ) in least square

Velocity_time=real( ifft(A));

%%

[b2,a2] = butter(4,fc/(Fs/2));

torque=filter(b2,a2,S(5).data.torque);

T = fft(torque);

T = fftshift(T);

f=-Nyq : df : Nyq-df;

A(f>7)=0+0i;

A(f<-7)=0+0i;

torque_freq=ifftshift(T);

% same procedure for fft of angular frequency data --> angle_freqData_p

phi_P=[accele_freq_p(1:end) velocity_freq_p(1:end) angle_freqData_p(1:end)];

TorqueP_freqData=(torque_freq(1:end));

Theta = lsqminnorm((phi_P),(TorqueP_freqData))

stimatedT2=phi_P*Theta ;

Rsq2_S = 1 - sum((TorqueP_freqData - stimatedT2).^2)/sum((TorqueP_freqData - mean(TorqueP_freqData)).^2)

Hello every one

I Recently started working on POD Galerkin's method for Approximating the PDE's (N-S and Energy Equation), i stuck at the calculating Laplacian operator on POD modes (These are orthonormal basis ) , earlier I used FFT but its not working because i have non periodic data, So is their any another way to compute this .

In the image T0 will be the ensembled average of Temperature and Phi will be the POD modes ...

I want to decompose some sensor data using wavelet. The FFT shows that my data does not contain higher frequency component(It has an influential DC component). Therefore which mother wavelet would be appropriate for decomposition?

Dears.

I have a system where the input was a constant value, I plotted the output in function of time.

Later I used FFT to get amplitude per frequency.

The total sampling time was 1.3 second.

I know that for the power spectrum plot it give the frequency range where the strong variation occur.

but how to interpret it in my case since when you look at the frequency it is clearly a very slow one (10^-5)

Dear I would like to plot the FFT while I am having only a table of two vectors times and amplitude!

is that possible? could someone guide me!

Hello,

I am studying the Webrtc AEC3 algorithm from "https://topic.alibabacloud.com/a/an-introduction-to-webrtcs-echo-cancellation-aec-aecm-algorithm_8_8_10267058.html".

Here, I read:

"Considering the convenience of the actual processing, in the c code of webrtc, the frequency-domain power spectrum after the FFT transformation is divided into 32 subbands, so that the value of each specific subband Xw(p, q) can be 1 bit. It means that a total of 32 bits are needed, which can be represented by only one 32-bit data type."

Now, so far my understanding is that:

When you take FFT (e.g real fft (rfft) that will result in Fs/2) the number of frequency bins will be dependent upon NFFT size. E.g If I set the NFFT size to 512, the total number of frequency bins will be 257 with a frequency step size = 31.25 Hz at sampling rate = 16000 and obviously will vary with respect to the sampling rate.

Regards,

Khubaib Ahmad

Hi all,

I have done a CFD for a cylinder and obtained a transient pressure data at a point using DES. I have put time step size 1e-04 seconds. I got a frequency (ater FFT of the data) of 540 Hz. Now when I performed the same in experiment with 5000 Hz sample frequency, I obtained various frequencies at 100, 200, 500 and 900 Hz. while the 500 Hz is nearer to my CFD it should be the largest mode since analytically also the same frequency is obtained using a/2L formula. Please help !

i am recording the vibrations from accelerometer while vehicles are passing from the bridge. i recorded the vibrations produced from vehicles. now the problem is how can i address the effect of vehicle mass on the fundamental frequency. how can i find the solution of this problem. the frequency is increasing when vehicle is present. but its agaist the physics because when mass increase the frequency decrease because they are inversely propotional to each other. the frequency is decreasing when vehicle is not present at the bridge. so this is the actual problem i am facing right now for my research. is it possible to find natural frequency while vehicle is present on the bridge or is there any way to find natural frequency. i read your suggestion about find the frequency when vehicel leave the bridge it sounds valid but is there any other method to address this problem

Good afternoon, I would like to ask you the method to get acceleration PSD profile from time-acceleration data which was obtained through vibration test.

First of all, I though that the time-acceleration data should be changed to frequency-acceleration data. Therefore, I used FFT.

And then, I though that transformed data to frequency-acceleration data should be squared and divided by their own frequency. Because as seeing Acceleration graph, the parameter of X-axis is frequency(Hz) and that of Y-axis is (g^2/Hz).

So, I wrote matlab code like below;

(Here, THM is the time-acceleration data)

time1=THM(:,1);

t_leng1=length(time1);

dt1=time1(2)-time1(1);

Freq1=(0:t_leng1-1)/dt1/t_leng1;

x=THM(:,2);

xft=fft(x);

xft=xft(1:t_leng1/2+1);

psdx = (abs(xft).^2/Freq1');

psdx = 2*psdx(2:end-1);

figure(1)

plot(time1, x);

hold

xlabel('Time(sec)');

ylabel('Acclearation(g)');

title('Time-domain Accelaration of X axis');

figure(2)

plot(Freq1(2:t_leng1),abs(psdx(2:end-1)))

hold

xlim([2 200])

xlabel('Frequency (Hz)');

ylabel('Accelaration (g)');

title('Frequency-domain Accelaration of X axis');

Freqq=Freq1(2:t_leng1)';

Xresopons=abs(xft(2:t_leng1))/t_leng1*2;

However, when running this code, the errors appear like below;

Error using plot

Vectors must be the same length.

Error in FFT_PSd (line 59)

plot(Freq1(2:t_leng1),abs(psdx(2:end-1)))

Here are my questions.

1. Is the correct method to obtain the acceleration PSD graph?

2. if it is correct, how can I solve this matlab error.

3. If it is incorrect, please let me know the method.

Thank you.

I have many samples of a arrays representing signals in the time domain that I want to classify using artificial neural networks. Do I need to use feature extraction methods (such as the FFT or WT) before applying them to the ANN ? or do I apply the signal directly as it is (600 dim array) to the NN.

Thanks in advance.

I have developed matlab code of OFDM index modulation based on look up table but not getting the desired curve...i want curve as shown in link given below on page 5544....somebody plz help me with the coding part...there are many mistakes in my code..plz help me to rectify the code...

clear all;

close all;

clc;

nsym=10^2;

%nbitspersym=12

bits=(1:12)

n=4

g=2;

%g1=bits(1:6) % division of 16 FFT size into 8 FFT subcarriers.

%g2=bits(7:12) % division of 16 FFT size into 8 FFT subcarriers

ipbits = randi([0,1], 1, 600) % generation of random bits

M=4;

Z=reshape(ipbits,100,6)

k=log2(M)

%n=nFFT/g;

EbN0=1:10

for i1=1:10

EbN0dB=i1

BER=[];

for i=1:100

p1=Z(i,:)

% p1=ipbits(1:6) % it will take first 8 bits

%p1=[1 1 0 1 1 0];

%p2 = ipbits(7:12) ;

s1 = p1(1:2)

s2= p1(3:6)

% s3=p1(5:6);

% s4= p1(7:8);

% s33=s2

s22=reshape(s2,2,2).'

s222=bi2de(s22, 'left-msb')

ipMod = (1/sqrt(2))*qammod(s222 , M)

F=zeros(4,1);

if s1==[0 0]

F(1,1)=ipMod(1)

F(2,1)=ipMod(2)

F(3,1)=0

F(4,1)=0

elseif s1==[0 1]

F(1,1)=0

F(2,1)=ipMod(1)

F(3,1)=ipMod(2)

F(4,1)=0

elseif s1==[1 0]

F(1,1)=0

F(2,1)=ipMod(1)

F(3,1)=0

F(4,1)=ipMod(2)

elseif s1==[1 1]

F(1,1)=ipMod(1)

F(2,1)=0

F(3,1)=0

F(4,1)=ipMod(2)

end

F.'

%=ifft(Sq1,N)*sqrt(N)%

xt=ifft(F).'

%xt=xt.'

xt=[xt(4) xt]

%xt=(nFFT/sqrt(nDSC))*ifft(fftshift(xF.')).';

%x_with_cp = [x(4) x]; % cp addition

% adding noise

nt = (randn + j*randn)./sqrt(2)

yt = xt +( (10.^(-EbN0dB/10)).*nt)

yyt=yt(2:5)

yyt1=fft(yyt)

[temp, idx] = sort( yyt1, 'descend' )

n = temp(1:2)

idx = idx(1:2)

%zzt=(abs(yyt1))

zzt1=(abs(yyt1).^2)

[max1, ind1] = max(zzt1)

zzt1(ind1) = -Inf;

[max2, ind2] = max(zzt1)

zzt1(ind2) = -Inf;

rec=[(ind1) (ind2)]

a1=1;

while(a1==1)

if rec== [1 2]|rec== [2 1]

bitts=[0 0]

a1=0;

elseif rec== [2 3]|rec== [3 2]

bitts=[0 1]

a1=0;

elseif rec== [2 4]| rec== [4 2]

bitts =[1 0]

a1=0;

elseif rec==[4 1]|rec==[1 4]

bitts=[1 1]

a1=0;

elseif rec==[3 1]|rec==[1 3]

BER12=0

break

% bitts=[1 1]

elseif rec==[4 3]|rec==[3 4]

BER12=0

break

% bitts=[0 0]

end

% finding the minimum distance between received vector and original symbols

% original symbols

orig=[0 0]

conv=bi2de(orig, 'left-msb')

ipMod = (1/sqrt(2))*qammod(conv , M)

orig1=[0 1]

conv1=bi2de(orig1, 'left-msb')

ipMod1 = (1/sqrt(2))*qammod(conv1 , M)

orig2=[1 0]

conv2=bi2de(orig2, 'left-msb')

ipMod2 = (1/sqrt(2))*qammod(conv2 , M)

orig3=[1 1]

conv3=bi2de(orig3, 'left-msb')

ipMod3 = (1/sqrt(2))*qammod(conv3 , M)

xxt=[ipMod ipMod1 ipMod2 ipMod3]

D1 =norm(n(1) - xxt(1))

D2 = norm(n(1) - xxt(2))

D3 = norm(n(1) - xxt(3))

D4 = norm(n(1) - xxt(4))

D=[D1 D2 D3 D4]

D5=min(D)

result = find(D==min(D))

DD1 = norm(n(2) - xxt(1))

DD2 = norm(n(2) - xxt(2))

DD3 = norm(n(2) - xxt(3))

DD4 = norm(n(2) - xxt(4))

DD=[DD1 DD2 DD3 DD4]

DD5=min(DD)

result1 = find(DD==min(DD))

G=[DD5 D5]

[tempp, idxx] = sort( G, 'descend' )

nn = tempp(1:2)

idxx = idxx(1:2)

idx1=sort(idxx) % sort according to tempp;

% idx1=[result result1]

%

% reeesult=[result result1 result2 result3]

% [temp1, idx1] = sort( F, 'ascend' )

% n = temp1(1:2)

% idx1 = idx1(1:2)

% [min1, ind1] = min(F)

% F(ind1) = Inf;

% [min2, ind2] = min(F)

% F(ind2) = Inf;

% rec1=[(ind1) (ind2)]

if idx1== [1 2]

biitts=[0 0 0 1]

elseif idx1==[2 1]

message('i m here')

biitts=[ 0 1 0 0]

elseif idx1== [2 3]

message('i m here')

biitts=[0 1 1 0 ]

elseif idx1==[3 2]

message('i m here')

biitts=[ 1 0 0 1 ]

elseif idx1== [2 4]

message('i m here')

biitts =[0 1 1 1]

elseif idx1==[4 2]

message('i m here')

biitts =[1 1 0 1]

elseif idx1==[4 1]

message('i m here')

biitts=[1 1 0 0]

elseif idx1==[1 4]

message('i m here')

biitts=[ 0 0 1 1]

elseif idx1==[3 1]

message('i m here')

biitts=[1 0 0 0]

elseif idx1==[1 3]

message('i m here')

biitts=[0 0 1 0]

elseif G1==[1 1]

biitts=[0 0 0 0]

elseif G1==[2 2]

biitts=[0 1 0 1]

elseif G1==[3 3]

biitts=[1 0 1 0]

elseif G1==[4 4]

biitts=[1 1 1 1]

end

S=[bitts biitts]

% A2 = sort(F(:))

% out=A2(1)

% out1 = A2(2)

% result33 = find(DDDD==min(DDDD))

BER12=sum(abs(p1-S))/6

end

BER=[BER BER12];

end

T_BER(i1)=sum(BER)/100

end

semilogy(EbN0,smooth(abs(T_BER)),'*-');

xlabel('SNR');

ylabel('EbN0dB');

legend('BER curve for OFDM index modulation');

Hello, I need to find the amplitude of the FFT of a real signal in Matlab. I would like to get the same amplitude in the frequency domain (with fft) and in the time domain.

I've read about some methods:

1) Division by N: amplitude = abs(fft (signal)/N), where "N" is the signal length;

2) Multiplication by 2: amplitude = 2*abs(fft(signal)/N;

3) Division by N/2: amplitude: abs(fft (signal)./N/2);

4) Others follow Parseval's theorem: amplitude = abs(fft(signal)./factor), being "factor" equal 1/fs (fs - sampling frequency).

Can anybody help me?

Thank you

Dear Community,

I am measuring heart beats through a PPG signal in realtime. So far I processed the raw PPG with a bandpass filter in the range [0.4, 4] Hz. Then, a peak detection algorithm was applied. However, it is sensitive to different noise sources, leading to errors in the heart beat detection.

I though about another simple and robust method for heart beat detection, e.g. through correlation. The FFT is too computationally expensive for the ESP32 system, where we implement the algorithm.

Thanks, blessings!

Fernando

During parameter study for a project (signal processing applied in mechanical vibration), when I decrease the frequency to a very low value (less than 200 mHz, for a interested range of frequency of a few hundred Herz), the frequency response I have gets "squishy" (instead of a single line going up or down, I have a waveform going up or down).

Checking on some forums online yield me the connection to "dirac delta function", but I do not understand it fully. So I'd like to ask if there is any connection between dirac delta function and frequency resolution, with regards to FFT in vibration analysis. Thanks for your help.

How should I exactly process the vertical acceleration signal (with respect to time) to be able to adapt it on the ISO2631 curves (as attached). As far as I know, First I should find rms of the acceleration, then FFT and after that 1/3 octave filter. is there any matlab code for this process? or any pdf that help with using iso2631 figure?

another question is, what if the peak point of the fft curve of the acceleration signal does not meet one of the ISO2631 curves (as shown in the attached picture)? Is there any interpolation method to find the exact amount of time that the human body can endure the vibration in this case?

After taking the frequency spectrum of an impulse force (from a shaker, sampled from a vibration analysis test), there are noticeable peaks at 50 and 150 Hz (presumably due to some electrical/electronic artefacts or some mismatch impedance between the shaker's coil and its structure). This specific experiment aims to study the effect of frequency resolution on the spectrum.

When the frequency resolution is smaller (same bandwidth, or F_max - F_min, is used, and more spectral/FFT lines are used), these sudden peaks start to appear. It should be noted that the original input signal is the same. The question is why this happens?

Thanks for your help

I have Frequency Weighted RMS Acceleration Data for ISO-2631-1 ride comfort analysis in time domain.

I want to compare it with ISO-2631-1 Time Exposure Limit plot by superimposing the available FWRMS acceleration above mentioned (As shown in attached Figure). So can I directly use FFT to above data to obtain it Frequency domain from Time domain and superimpose on Time Exposure Limit graph available in frequency domain?

Dear All,

Hope you are fine and doing well in this pandemic.

I am working on a FORTRAN program, where I need to sum the Gaussian functions over time. I have observed that my computing time is scaling linearly with the number of iterations. I found FFT is useful for such tasks while reading about techniques to reduce the iterations. So, my question is if FFT can efficient for my task, how can I implement it?.

Pseudocode or suggestions are appreciated.

Cheers!

Anji

I did the FFT of a periodic current profile from the axis of a machine tool. The FFT amplitude profile is also periodic, i.e. the profile has a peak at 7, 12, 17, 22, 27 Hz, i.e. every 5 Hz I do have a new peak. I really do not understand if this behaviour is a consequence of some physical phenomenon or it is some kind of numerical error. Anyone can help?

I am working on a project related to sEMG classification. In order to process the signal, I am trying to filter the signal, especially power interference. The signal is acquired from Delsys Bagnoli 16, and the sample rate was 4000. I collected 6 signals simultaneously, and some signals showed this abnormal behaviour in fft (see the figure, zoomed to 0-700Hz). Therefore, I tried notch filters to remove the spikes, but it seems like it has other spikes around 50Hz. For example, see the next figure. I need to know the reason for this behaviour. Is it due to fft calculation? (see the last figure, Matlab code. it was recommended by the Matlab (https://www.youtube.com/watch?v=VFt3UVw7VrE , at 5:17)) or due to a filtering problem at the amplifier?

Does anyone know any book that contains general descriptions and basic mathematical formulation of frequency analysis methods like PSD, FFT or Allan variance?

Hi,

I am keen to know that If I have two signals say :

Farend_signal (y[n]) and Nearend_signal (x[n])

and x[n] also contains an echo of y[n] with some delay.

Is there any technique to align these two signals temporally is Spectrogram? For example, any correlation technique to find the temporal displacement between signals after making a spectrogram(STFT) of two signals?

I have tried to measure the third harmonic of discharge current performing FFT of the discharge current signal. I found out that the amplitude of the third harmonic is not prominent enough to detect.

Can anyone suggest me some methods to measure the same distinctly?

Currently, I have a series of human motion data recorded by accelerometer sensor at a sampling rate 100HZ. Based on literature review, many papers extract the feature called “First 5 FFT coefficients” from the sliding windows. They claimed that the first 5 of the fast-Fourier transform coefficients are taken since they capture the main frequency components, and the use of additional coefficients did not improve the accuracies [1][2].

I googled a lot and it seems Numpy.fft module could solve my problem. However, I am not sure this module is appropriate for my purpose. From the document, the numpy.fft.fft() function accepts either a real or complex array as an input argument, and returns a complex array of the same size that contains the Fourier coefficients. The real part contains the coefficients for the cosine terms and the imaginary part contains the negative of the coefficients for the sine terms [3].

I am not sure about the relationship between First 5 FFT(fast-Fourier transform) coefficients and output from fft module in Numpy. Sorry for my poor signal processing background. Any hints or suggestions will be much appreciated.

Reference

[1] Incel, Ozlem Durmaz. "Analysis of movement, orientation and rotation-based sensing for phone placement recognition." Sensors 15.10 (2015): 25474-25506.

[2] Lee, Jin, and Jungsun Kim. "Energy-efficient real-time human activity recognition on smart mobile devices." Mobile Information Systems 2016 (2016).

[3] http://snowball.millersville.edu/~adecaria/ESCI386P/esci386-lesson17-Fourier-Transforms.pdf

This project is related to transient states.

I want to study the behaviour of a circuit (like a current source) by suddenly increasing the resistance from one value (5 ohms) to another (10 ohms).

At first, I used Brick resistance, which I could not achieve to my goal.

Now I use Carbon resistance which has not worked.

Which resistance are used for high frequencies (100ns; 10 Mhz)?

If I use carbon resistors, is there a proposed circuit that I can increase their frequency?

I have a question based on application of the FFT. It has been 30 years since I worked with FFT’s so, please explain as if explaining to a neophyte as there are many cobwebs. I will explain the image as basically as I can since I do not know if anyone will have seen this type of medical image before.

The image below shows the Doppler spectrum produced by an FFT. The FFT bins correspond to velocity through the Doppler equation (note the scale to the right of the Doppler spectrum). There is now a “raging” debate in the medical field about the lighter envelope portion of the spectrum (red dotted line) that basically duplicates the darker envelope (modal velocity traced in light green) below the baseline. Some are arguing that this higher velocity (red envelope) is artifactual based on the FFT processing and some are arguing that this is the result of a hemodynamic situation which results in an increased velocity.

I want to discuss one more general point about Doppler FFT processing before I can really described the situation and pose the question. The Doppler signal is separated into an I and Q channel so as to detect flow direction (represented as flow signals above or below the baseline). I am well aware of the “mirroring” that can occur in the Doppler spectrum when some of the signal “crosstalks” between the I and Q channel. If there is not perfect separation between the two channels, we see the result as replication of the real signal “mirrored” across the baseline.

I will briefly describe the situation. This picture represents the imaging of a prosthetic mitral valve (top image with color) and a continuous wave Doppler, spectrum below. Continuous wave implies that the flow velocity is detected along the entire Doppler line (dotted white line down the center of reference image above spectrum). In this case, the valve has been replaced with a mechanical prosthetic valve with metal discs. A well understood artifact that occurs sometimes in ultrasound (and frequently with specular reflectors like metal which acts like a mirror to sound waves) is a reverberation artifact. In essence, the sound, instead of making a single path down and back from each part of the image, makes two or more paths (reverberates) between the strong reflectors. The result is that the specular reflectors (and of course every structure between the specular reflectors, is duplicated a second time (and possibly more times) below the location of the actual structure within the ultrasound image. The fact that this reverberation artifact is happening in this image is confirmed with other views I am not including.

So some people are theorizing that just as the image is being replicated, the Doppler shift is being replicated (this I completely believe since the sound beams used to detect Doppler shifts behave in the same manner as the sound beams used to create the 2D image). However, the velocity presented is related to the Doppler frequency shift, not if the shift is detected twice, so detecting the same Doppler shift again does not explain the increase in velocity shown. In other words, detecting the same frequency twice should result in energy in the same frequency bins, not higher order bins corresponding to higher frequencies( velocities). The question is whether it is possible that since the “same” shift is being replicated and detected twice instead of once can result in an explainable result in the FFT the produces signal in higher order FFT bins (the signal in the fainter envelope traced in red). In other words, is there a mathematical explanation (like there is for the “mirroring” artifact for this “double envelope” artifact?

For completeness: others are theorizing that this is not related to the FFT processing at all, but is the result of the vena contracta and complex flow acceleration that results from impingement of this flow on the septal wall. I will not go into more detail on this theory as this relates to fluid dynamics and I know this is a forum for answering questions about Fourier Transforms.

I have a prototype tested results of a 4-bit ADC for which the FFT spectrum is attached.  2.5V p-p input was used with an input frequency of 100 MHz sampled at 800 MS/s

2.5V p-p input was used with frequency of 100 MHz sampled at 800 MS/s

I need to calculate the dynamic range of this converter. How can I do that?

Thanks in advance

Anush

I am new to fpga. Can anyone help how to use inbuilt fft module of fpga

what is the reason for the difference in complexity

any answers or suggestions from the scholars will be highly helpful.

Thanks

Brij Mohan Kumar

M-Tech (Communication systems)

IIT, Patna

I am reproducing the results reported by the author in a research paper. In which he takes the STFT of EEG signals. The input data has a shape of (1 X500) and I used Scipy library built-in function to calculate STFT and the output has shape (257 x32) I need to extract the data between two frequency bands (6-13 Hz) and (17-30 Hz). The author reports the extracted bands to be of the size of (16 X 32) and (23 X 32) but in none of the setting, I am getting this frequency resolution. I tried contacting the author but no response so far. I hope some of you could direct me in the right direction. Thank you very much.

I need this one to write about cepstral analysis. It doesn't seem to be available anywhere.

Full reference: Bogert, B.P., Healy, M.J. and Tukey, J.W., 1963, June. The quefrency alanysis of time series for echoes: Cepstrum, pseudo-autocovariance, cross-cepstrum and saphe cracking. In Proceedings of the symposium on time series analysis (Vol. 15, pp. 209-243). chapter.

Dear All Fellows,

I hope my question is clear, I need to know which plane of my materials showing me interplanar (d-spacing), which I have calculated through FFT image. I need a litrature link, where I can check the materials planes (001, 100, 002, etc.) by puting my known d-spacing (nm).

Regards

Dear friends

I am designing a underwater optical acoustic sensor Using COMSOL Multiphysics software. I would like to analyze the received signal signal by sensor in both the time domain and frequency domain. I have converted the time domain to frequency domain using FFT node which is available in the COMSOL itself but the amplitude is different. Is it correct or wrong?. Amplitude should be same or not if not same then why?

For understanding the question I have attached the figures below please find it.

Thanks & regards

Nilakanta

It's how to to convert spectral radiance from W/cm^2/sr/nm to W/cm^2/sr/cm-1. Fisrt one is the radiance represented by wavelength and second one is represented by wavenumber.

Consider solving the nonlinear Schrödinger equation (NLSE) with 2D transverse dimensions. Or more simply, consider the beam propagation method (BPM).

We know that the original formulations of Feit&Fleck relied on the FFT to compute the derivatives.

In the 90s, many works about finite differences (FD) were proposed. The advantage is that the grid is more flexible and the boundary conditions are better understood and controllable (PML, TBC...). Prof. Fibich in his book suggests FD is a better approach to tackle collapses.

With the advent of GPU, FFTs with lots of points seem to be almost costless. Do you think it is not worth anymore to study finite differences? What about non-paraxial propagators?

Thank you for your suggestions.

Generally, FFT is used to obtain the spectrum in the frequency domain. What are the informations we can get from the FFT analysis of a spectrum?

I need to classify an EEG signal (most likely using an artificila neural network) but for the inputs I would like to know the differences between using the result of the FFT or using the signal with different band-filteres for example butterworth.

The advantages of the FFT is that is much smaller (a 1 second window it will have ~60 inputs) and that the signal is still there. On the other hand using band-filters I would have as input all the signal (a 1 second window @ 256 Hz it will have 256 times the number of bands that I want to obtain). Also the signal gets distorted.

So my final question is, for classification purpouse, does the FFT contains the same information that the signal filtered?