In a message dated 2007/07/14, gmvoeth@........... writes:

> The power needed for a real-time FFT display of any value just does not 
> exist for a typical home computer. You would need a special card like these high 
> speed video cards.
> 
> I have seen the machines that do real time fft and back in the 1970s they 
> were like 1000X faster than your home computer and would do 500 points and be 
> only a few milli seconds behind real time. You most probably will have to 
> stick with after the fact processing.

Hi Geoff,

       There have been several developments in FFTs since the 1970s!!

       The amount of maths required to produce a FFT has been drastically 
reduced by new algorithms.
       The processors are now thousands of times faster, several may be 
included on one chip and they can use vectored instructions.

       It is largely a question of what you are calling 'real time'. 
       If you use a multitasking processor, the interrupts are maybe 20 to a 
few milliseconds or less. Such time delays are not usually important in 
amateur seismology. The most likely delay is the relatively slow rate of seismic 
sampling. By opting for a 10 Hz low pass filter, you are accepting a 50 milli 
second delay minimum.

       I would not expect there to be a problem in providing a FFT display to 
update at this rate. I note quoted times of less than 1 milli second to 
complete a FFT sample.

       Regards,

       Chris Chapman   
In a me=
ssage dated 2007/07/14, gmvoeth@........... writes:



The power needed for a real-tim=
e FFT display of any value just does not exist for a typical home computer.=20=
You would need a special card like these high speed video cards.



I have seen the machines that do real time fft and back in the 1970s they we=
re like 1000X faster than your home computer and would do 500 points and be=20=
only a few milli seconds behind real time. You most probably will have to st=
ick with after the fact processing.




Hi Geoff,



       There have been several developments in=
 FFTs since the 1970s!!



       The amount of maths required to produce=
 a FFT has been drastically reduced by new algorithms.

       The processors are now thousands of tim=
es faster, several may be included on one chip and they can use vectored ins=
tructions.



       It is largely a question of what you ar=
e calling 'real time'. 

       If you use a multitasking processor, th=
e interrupts are maybe 20 to a few milliseconds or less. Such time delays ar=
e not usually important in amateur seismology. The most likely delay is the=20=
relatively slow rate of seismic sampling. By opting for a 10 Hz low pass fil=
ter, you are accepting a 50 milli second delay minimum.



       I would not expect there to be a proble=
m in providing a FFT display to update at this rate. I note quoted times of=20=
less than 1 milli second to complete a FFT sample.



       Regards,



       Chris Chapman