| Re: FFT phase? [message #82902] |
Fri, 25 January 2013 13:48 |
Craig Markwardt
Messages: 1869
Registered: November 1996
|
Senior Member |
|
|
On Friday, January 25, 2013 8:18:28 AM UTC-5, Yngvar Larsen wrote:
> On Friday, 25 January 2013 07:49:00 UTC+1, xqin...@gmail.com wrote:
> OR, if you know the frequency of your signal exactly a priori, you have to truncate your data to a periodic signal, i.e. an integer number of cycles, in order to extract A and B reliably. This is what Craig did in his reply, using only a single cycle.
Well, 2 cycles, but yes. :-)
Craig
|
|
|
|
| Re: FFT phase? [message #82909 is a reply to message #82902] |
Fri, 25 January 2013 05:18  |
Yngvar Larsen
Messages: 134
Registered: January 2010
|
Senior Member |
|
|
On Friday, 25 January 2013 07:49:00 UTC+1, xqin...@gmail.com wrote:
> Thanks. I have known the reason. What I use is like x= 2*!dpi*2*dindgen(25)/16, so the amplitude and phase are not so accurate.
As a rule of thumb for harmonic analysis, you need at least 10 cycles in your data to get a reliable phase/magnitude estimate. Your example using around 1.5 cycles is probably the worst case scenario.
OR, if you know the frequency of your signal exactly a priori, you have to truncate your data to a periodic signal, i.e. an integer number of cycles, in order to extract A and B reliably. This is what Craig did in his reply, using only a single cycle.
--
Yngvar
|
|
|
|
| Re: FFT phase? [message #82911 is a reply to message #82909] |
Thu, 24 January 2013 22:49 |
xqinshan
Messages: 21
Registered: November 2008
|
Junior Member |
|
|
Thanks. I have known the reason. What I use is like x= 2*!dpi*2*dindgen(25)/16, so the amplitude and phase are not so accurate.
在 2013年1月25日星期五UTC+8上午10时11分30秒，Craig Markwardt写道：
> On Thursday, January 24, 2013 12:28:02 PM UTC-5, xqin...@gmail.com wrote:
>
>> Hi,
>
>>
>
>>
>
>>
>
>> just use FFT(y). For example, y=A*cos(x+B), C=fft(y). I think atan(C,/phase) should equal to B, but the return reslut is not. How to obtain A and B from complex C?
>
>
>
> It is correct. Example:
>
> x = 2*!dpi*2*dindgen(16)/16 ;; Angle in radians
>
> y = 0.7*cos(x + 1.6000)
>
> c = fft(y,-1)
>
> print, atan(c[2],/phase)
>
> ==> 1.6000
>
>
>
> Craig
|
|
|
|
| Re: FFT phase? [message #82915 is a reply to message #82911] |
Thu, 24 January 2013 18:11 |
Craig Markwardt
Messages: 1869
Registered: November 1996
|
Senior Member |
|
|
On Thursday, January 24, 2013 12:28:02 PM UTC-5, xqin...@gmail.com wrote:
> Hi,
>
>
>
> just use FFT(y). For example, y=A*cos(x+B), C=fft(y). I think atan(C,/phase) should equal to B, but the return reslut is not. How to obtain A and B from complex C?
It is correct. Example:
x = 2*!dpi*2*dindgen(16)/16 ;; Angle in radians
y = 0.7*cos(x + 1.6000)
c = fft(y,-1)
print, atan(c[2],/phase)
==> 1.6000
Craig
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
