- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi everyone,
I use altera fft ip core 9.0, and let my data go through 1024 ifft - GI - GI_remover - 1024 fft, but i did not get the origin data back(not the scaling problem, the data is totally different). Time sequence is correct, I got sop, eop every 1024 cycles. Could anyone give me some advice of it? Thanks!Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
I would recommend that you check your data representation. Ensure that your input data is in the format expected by the FFT component. If your data's time samples span -N/2-N/2, but your fft is configured for natural order, you will have a problem in your output. Since you are recieving the SOP and EOP from the source interface, the issue is probably just a configuration issue.
BTW, what type of signal/data are you analyzing? What is the bandwidth of the signal? Give me a little bit more information about your signal and the configuration and I may be able to provide some additional help. I just spent 2 weeks trying to the FFT in my system working and I finally figured out where all(hopefully) of the pitfalls are.- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi Winder,
Thanks for your reply. My input is just in natural order. And could you tell me how to calculate my data's time samples span? I have real[13:0] and imag[13:0] work under 100MHz. Really appreciate your help- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Sure, ill give you an example. If your signal is lets say, sin(2*pi*t) and you define t = -0.512 : .001 : 0.511, you should configure the FFT's input data as (-N/2 to N/2). However, if you are processing your FFT's out "offline" with MatLab or some other tool, you can do a FFTSHIFT operation that will properly center the output. But if not, you should change the FFT to (-N/2 to N/2).
If that parameter is not set properly, the next component may misinterpret the frequency (or time) values.- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
--- Quote Start --- Sure, ill give you an example. If your signal is lets say, sin(2*pi*t) and you define t = -0.512 : .001 : 0.511, you should configure the FFT's input data as (-N/2 to N/2). However, if you are processing your FFT's out "offline" with MatLab or some other tool, you can do a FFTSHIFT operation that will properly center the output. But if not, you should change the FFT to (-N/2 to N/2). If that parameter is not set properly, the next component may misinterpret the frequency (or time) values. --- Quote End --- Thanks you again, but I do not quite get you... The attached file is my input, two column show I, Q for real and imag. Is there any problem if I use fft ip core to process that?
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
I was unable to upen your attachment. Could you re-link it?
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
I want to know how to use FFT ip freely. if you have licence, can you !give me a one , Thank you !
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
maohuhua123: please don't request licenses on the forum
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page