Short, comprehensive, to the point. Loved it. I've been going through tons of resources around the net the past few days, and I couldn't find what I needed till I stumbled upon this video.
The FFT gives you a two-sided spectrum. When you shift the output array elements to place DC (index k=0) in the center you get negative frequencies on the left side and positive frequencies on the right side, and fmax appears at the far right side. The bidirectional bandwidth refers to the *total* spectrum occupied by this two-sided spectrum, and because DC is in the middle, fmax is only half the bidirectional bandwidth.
This could only ve understood if you're working on the hardware and have experienced how acquisition of bandwidth is done when tune a radio at a certain center freq.
@NTS. ....If time(t) starts at zero then time instant of last sample minus that of first sample(0) is (N-1)*Td/N ... which is not equal to the time length of signal (Td)....what the heck you guys are talking......It read more than hundred of paper on FFT and DFT , all of them are using this dumb index system (0 to n-1) and that silly mistake though its has no such significant effect it can;t be neglected.... Correction way 1: If you take N samples ..then it has N discrete time instants ...but indeed (N-1) intervals only....that's make size of interval to be Td/(N-1) in fact indeed..... Correction way 2: If you stick at signal length to be Td then...your index should run from 0 to N.....thereby making total number of samples to be (N+1) .....total number of interval to be N.... and one more thing what you described is DFT... not FFT...not at all hehe
What is bidirectional bandwidth, why does it equal to sampling frequency, and why does maxima of frequency equal to half of bidirectional bandwidth? Thanks for the video btw.
Because the first sample is at time zero its sample index is also zero. For example, suppose you have 8 samples beginning at time zero. The index of the first sample is 0, the next index is 1, and the final sample has index 7. If you started indexing at 1, then the final sample index would be 8.
why the time domain signal time interval, Δt divided by N and not N-1? Since it is the interval. There will be an offset by (100/N)% by the sampling time calculation.
Thank you for your reply, even after some years. I also end up to this when I code the algorithm and it turns more convenient to start from zero. Continue your great job.
short precise and to the point explanation of FFT ....it transform time domain waveform to frequency domain spectrum ......but why there is a need for FFT please explain
I don't understand the concept of bidirectional bandwidth. Shouldn't 0 (DC) be at the left and then the higher frequencies go to the right? I don't understand about or why you put 0 (DC) in the center.
Hi, Thanks for the video. I have a doubt what do you mean by "typically display only lower half of the output array". If there are N samples in => N samples out. So did u mean, its N samples in=> N/2 Samples out?
If you go past N/2 in the frequency array, it is actually is just a mirror image of the previous N/2 frequency bins. See this answer: dsp.stackexchange.com/a/4827/11807
Time starts at 0 and runs until N-1 meaning there are N samples for duration T_d. So you got T_d/N as width of time steps or Delta_T. In other words, you would be right if n started from 1 instead of 0.
@NTS.....If time(t) starts at zero then time instant of last sample minus that of first sample(0) is (N-1)*Td/N ... which is not equal to the time length of signal (Td)....what the heck you guys are talking......It read more than hundred of paper on FFT and DFT , all of them are using this dumb index system (0 to n-1) and that silly mistake though its has no such significant effect it can;t be neglected.... Correction way 1: If you take N samples ..then it has N discrete time instants ...but indeed (N-1) intervals only....that's make size of interval to be Td/(N-1) in fact indeed..... Correction way 2: If you stick at signal length to be Td then...your index should run from 0 to N.....thereby making total number of samples to be (N+1) .....total number of interval to be N
Short, comprehensive, to the point. Loved it. I've been going through tons of resources around the net the past few days, and I couldn't find what I needed till I stumbled upon this video.
Hh5 GT6
This is one of the best explanations of the FFT results that I have found on the Internet.
wrong, this tells nothing about FFT, they're just DFT principles
The FFT gives you a two-sided spectrum. When you shift the output array elements to place DC (index k=0) in the center you get negative frequencies on the left side and positive frequencies on the right side, and fmax appears at the far right side. The bidirectional bandwidth refers to the *total* spectrum occupied by this two-sided spectrum, and because DC is in the middle, fmax is only half the bidirectional bandwidth.
Could you explain with an example
This could only ve understood if you're working on the hardware and have experienced how acquisition of bandwidth is done when tune a radio at a certain center freq.
@NTS. ....If time(t) starts at zero then time instant of last sample minus that of first sample(0) is (N-1)*Td/N ... which is not equal to the time length of signal (Td)....what the heck you guys are talking......It read more than hundred of paper on FFT and DFT , all of them are using this dumb index system (0 to n-1) and that silly mistake though its has no such significant effect it can;t be neglected....
Correction way 1: If you take N samples ..then it has N discrete time instants ...but indeed (N-1) intervals only....that's make size of interval to be Td/(N-1) in fact indeed.....
Correction way 2: If you stick at signal length to be Td then...your index should run from 0 to N.....thereby making total number of samples to be (N+1) .....total number of interval to be N....
and one more thing what you described is DFT... not FFT...not at all hehe
Thank you so much for the video. Best explanation so far I found in the internet
Thank you. This gives an overall image on FFT. Great video.
I needed to remember the FFT concepts. Your video was a very useful way to do it, so thanks you !!
Thank you so much for this video. May Allah give you the best in the life and the next :).
It is clear but sadly it tells nothing about the FFT, just some basic definitions of DFT.
thank you!! no formulas, no actual plots of time and frequency domains ...a bit disappointed here tbh
Equal about fft is very many on the internet. This video is exactly what i need.
this is exactly what i needed! gracias!
What is bidirectional bandwidth, why does it equal to sampling frequency, and why does maxima of frequency equal to half of bidirectional bandwidth? Thanks for the video btw.
Because the first sample is at time zero its sample index is also zero. For example, suppose you have 8 samples beginning at time zero. The index of the first sample is 0, the next index is 1, and the final sample has index 7. If you started indexing at 1, then the final sample index would be 8.
This is perfect explain on dft. It clear up the basic concept. I am wondering if got similar topic for 2d fft, tks a lot
thank you very much, it is useful and informative video
why the time domain signal time interval, Δt divided by N and not N-1? Since it is the interval. There will be an offset by (100/N)% by the sampling time calculation.
in 2:39 you say that the sampling interval Dt is Td/N, normally Dt is equal to Td/(N-1) so in the end total time Td=(N-1)*[Td/(N-1)]=Td
Thank you for your reply, even after some years. I also end up to this when I code the algorithm and it turns more convenient to start from zero. Continue your great job.
Very usefull for me. I don't care how the FFT is calculated, I just want to know whats the input and whats the output
Very clear. Thank you so much!!
Thank you so much. brief and clear.
Excellent explanation!!!
Question, why Bb=fs ? Thank you.
Excellent video!! thank u
short precise and to the point explanation of FFT ....it transform time domain waveform to frequency domain spectrum ......but why there is a need for FFT please explain
I don't understand the concept of bidirectional bandwidth. Shouldn't 0 (DC) be at the left and then the higher frequencies go to the right? I don't understand about or why you put 0 (DC) in the center.
exactly what I needed
After going through so much bullshit over internet and youtube, this is what I wanted.
Hi, Thanks for the video. I have a doubt what do you mean by "typically display only lower half of the output array". If there are N samples in => N samples out. So did u mean, its N samples in=> N/2 Samples out?
If you go past N/2 in the frequency array, it is actually is just a mirror image of the previous N/2 frequency bins. See this answer: dsp.stackexchange.com/a/4827/11807
Am I wrong or Delta_t should be equal to T_d/(N-1)?
Because t_i=i*Delta_t so t_N-1=(N-1)*Delta_t=T_d
Time starts at 0 and runs until N-1 meaning there are N samples for duration T_d. So you got T_d/N as width of time steps or Delta_T. In other words, you would be right if n started from 1 instead of 0.
how number of input samples is equal to out samples?
Thanks a lot , very helpful
Can someone explain what is the bidirectional bandwidth?
was very usefull, thanks a lot!
why is the time of the final sample [(N-1)/N]*td?
isn't it just td?
good explained fft
Great video.
Thanks,
Anthony
Well explained thank you
Very clear ty!
Good explanation
how Bb = Fs?
thanks alot, it is helpful
Thanks a lot for clear explanations on FFT.
By
Kz
God bless you, sir.
jdyehynshd
jsksj
Thank you so much sir
cool cool cool, but where is the algorithm????
This video is very well done. Saving my ass!
can somone explain this? f(t)=sin(at)+sin(bt)
Thank you sooooooooooooooooooooooooo much
Thanks a lot
Thanks!!
many thanks
Thanks
thank u so much
i don't understand monkeyshit, not from this video, not from any other i've watched about the FFT or any explanation i've read on the internet.
@NTS.....If time(t) starts at zero then time instant of last sample minus that of first sample(0) is (N-1)*Td/N ... which is not equal to the time length of signal (Td)....what the heck you guys are talking......It read more than hundred of paper on FFT and DFT , all of them are using this dumb index system (0 to n-1) and that silly mistake though its has no such significant effect it can;t be neglected....
Correction way 1: If you take N samples ..then it has N discrete time instants ...but indeed (N-1) intervals only....that's make size of interval to be Td/(N-1) in fact indeed.....
Correction way 2: If you stick at signal length to be Td then...your index should run from 0 to N.....thereby making total number of samples to be (N+1) .....total number of interval to be N
Nimda
ww