fft matlab, scaling amplitude problem

Hi I've faced a problem in my homework coding. please help me to solve it.
I wrote a code for a problem, which i have its results. but after taking fft function from input, the amplitude of output is twice more than expected result.
what are the reasons that might be caused this?
I really appreciate it.

1 Comment

I wrote a part of my code related to fft :
clc; clear all;
fm=40; T=10; a=0; N=32768; %2^15 Cm=1.5; h=(T-a)/N; omega=2*pi*fm/T; hh=1/h; Iamp=2.25e3; V(1)=-65;
mhf(1)=mhfinf(V(1));
mhs(1)=mhsinf(V(1));
mNap(1)=mNapinf(V(1));
EL=leak(V(1),mhf(1),mhs(1),mNap(1));
t(1)=a;
Iinp(1)=a;
for j=1:N-1
Iinp(j+1)=Iamp*sin(omega*(t(j)^2));
%%%%%%%%% 4 step method of Runge-Kutta
K1=h*fftNapH(t(j),[V(j); mhf(j); mhs(j); mNap(j)],EL,Iinp(j+1)); deffv1=K1(1,1);mhf1=K1(2,1);mhs1=K1(3,1);mNap1=K1(4,1);
K2=h*fftNapH(t(j)+(h/2),[V(j)+(deffv1/2);mhf(j)+(mhf1/2);mhs(j)+(mhs1/2);mNap(j)+(mNap1/2)],EL,Iinp(j+1)); deffv2=K2(1,1);mhf2=K2(2,1);mhs2=K2(3,1);mNap2=K2(4,1);
K3=h*fftNapH(t(j)+(h/2),[V(j)+(deffv2/2);mhf(j)+(mhf2/2);mhs(j)+(mhs2/2);mNap(j)+(mNap2/2)],EL,Iinp(j+1)); deffv3=K3(1,1);mhf3=K3(2,1);mhs3=K3(3,1);mNap3=K3(4,1);
K4=h*fftNapH(t(j)+h,[V(j)+deffv3;mhf(j)+mhf3;mhs(j)+mhs3;mNap(j)+mNap3],EL,Iinp(j+1)); deffv4=K4(1,1);mhf4=K4(2,1);mhs4=K4(3,1);mNap4=K4(4,1);
V(j+1)=V(j)+1/6*(deffv1+(2*deffv2)+(2*deffv3)+deffv4); mhf(j+1)=mhf(j)+1/6*(mhf1+2*mhf2+2*mhf3+mhf4); mhs(j+1)=mhs(j)+1/6*(mhs1+2*mhs2+2*mhs3+mhs4); mNap(j+1)=mNap(j)+1/6*(mNap1+2*mNap2+2*mNap3+mNap4);
t(j+1)=t(j)+h;
end
V2=fft(V1,N); V3=V2(1:N/2); amp=abs(V3); ampb=amp(2:length(amp)); % remove DC part f= (0:N/2-1)*hh/N;
figure(2),plot(f,ampb) xlabel('Freq') ylabel('Amplitude') axis([0 50 0 2000])

Sign in to comment.

Answers (2)

You should always show your code:
fs = 1000;
t = 0:1/fs:1-1/fs;
L = length(x);
xdft = fft(x)/L;
plot(abs(xdft))
Exactly as I expect two peaks with amplitude 0.5
Or
xdft = 2*fft(x)/L;
xdft = xdft(1:length(x)/2+1);
plot(abs(xdft))

4 Comments

Thanks for your reply. i tried it, if i apply (/L) the result is absolutely different with expected result, but if i remove (/L) the result is same as mine.
Again, you're not showing your code (not sure why). If you are using the fft() for amplitude estimation, you want to divide by the length of the input signal.
thanks for your reply Wayne King. i wrote a relevant part of my code
I wrote a part of my code related to fft :
clc; clear all;
fm=40; T=10; a=0; N=32768; %2^15 Tfft=10; Cm=1.5; h=(T-a)/N; omega=2*pi*fm/T; hh=1/h; Iamp=2.25e3; V(1)=-65;
mhf(1)=mhfinf(V(1));
mhs(1)=mhsinf(V(1));
mNap(1)=mNapinf(V(1));
EL=leak(V(1),mhf(1),mhs(1),mNap(1));
t(1)=a;
Iinp(1)=a;
for j=1:N-1
Iinp(j+1)=Iamp*sin(omega*(t(j)^2));
%%%%%%%%% 4 step method of Runge-Kutta
K1=h*fftNapH(t(j),[V(j); mhf(j); mhs(j); mNap(j)],EL,Iinp(j+1)); deffv1=K1(1,1);mhf1=K1(2,1);mhs1=K1(3,1);mNap1=K1(4,1);
K2=h*fftNapH(t(j)+(h/2),[V(j)+(deffv1/2);mhf(j)+(mhf1/2);mhs(j)+(mhs1/2);mNap(j)+(mNap1/2)],EL,Iinp(j+1)); deffv2=K2(1,1);mhf2=K2(2,1);mhs2=K2(3,1);mNap2=K2(4,1);
K3=h*fftNapH(t(j)+(h/2),[V(j)+(deffv2/2);mhf(j)+(mhf2/2);mhs(j)+(mhs2/2);mNap(j)+(mNap2/2)],EL,Iinp(j+1)); deffv3=K3(1,1);mhf3=K3(2,1);mhs3=K3(3,1);mNap3=K3(4,1);
K4=h*fftNapH(t(j)+h,[V(j)+deffv3;mhf(j)+mhf3;mhs(j)+mhs3;mNap(j)+mNap3],EL,Iinp(j+1)); deffv4=K4(1,1);mhf4=K4(2,1);mhs4=K4(3,1);mNap4=K4(4,1);
V(j+1)=V(j)+1/6*(deffv1+(2*deffv2)+(2*deffv3)+deffv4); mhf(j+1)=mhf(j)+1/6*(mhf1+2*mhf2+2*mhf3+mhf4); mhs(j+1)=mhs(j)+1/6*(mhs1+2*mhs2+2*mhs3+mhs4); mNap(j+1)=mNap(j)+1/6*(mNap1+2*mNap2+2*mNap3+mNap4);
t(j+1)=t(j)+h;
end
V2=fft(V1,N); V3=V2(1:N/2); amp=abs(V3); ampb=amp(2:length(amp)); % remove DC part f= (0:N/2-1)*hh/N;
figure(2),plot(f,ampb) xlabel('Freq') ylabel('Amplitude') axis([0 50 0 2000])

Sign in to comment.

That means that, before calculating the fft, you've made some errors, which we can't find, because you have not posted the code.

2 Comments

thanks for your reply. actually, It has 10++ line of code and i do not know which part do you need? is there any distinguished reason for my problem? because, my result is exactly twice of expected result.
Wayne King
Wayne King on 30 May 2013
Edited: Wayne King on 30 May 2013
Does 10++ mean 12 lines? If it is a reasonable number, please post it all. Did you look at the code I posted below?

Sign in to comment.

Categories

Find more on Fourier Analysis and Filtering in Help Center and File Exchange

Asked:

on 30 May 2013

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!