[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: [linrad] Finally!



> It seems like the Athlon chip is quite happy processing 4 
> channels of 96khz.
> I need to do some cruel tests, Leif can you offer any suggestions for some
> settings that would be particularly taxing for the processor?

It seems that Athlon and Pentium 4 are similar in performance.
I have made a comparison at:
http://ham.te.hik.se/~sm5bsz/linuxdsp/hware/p4_athl.htm
or
http://nitehawk.com/sm5bsz/linuxdsp/hware/p4_athl.htm
or
http://www.g7rau.co.uk/sm5bsz/linuxdsp/hware/p4_athl.htm

I am not sure about the memory speed for the Athlon but
I think it was 266MHz while the Pentium used 333MHz.
The big difference is that Athlon does not have efficient
SIMD instructions but on the other hand it seems to have
more efficient MMX instructions. The smaller cache size of
the Athlon may have a significant influence if an fft size
is selected that will fit to the Pentium4 cache but not to the
Athlon one.

Cruel tests:
1) Graphics speed. Deselect second fft, set fft1 bandwidth 
to 200Hz with a N=4 window. Set the waterfall average number to 1
(you have to click in the box in the lower right corner of the
main spectrum to make it show 1. Averaging is a multiple of this 
number). Resize the windows to make the waterfall occupy 75%
of the entire screen. If your computer graphics is not fast enough
the waterfall will loose data. Try it by connecting a variable 
frequency source and sweep across the frequency range. If you
see frequency jumps the screen is not fast enough. 

If your computer passes this test (my Pentium 3 at 600 MHz 
is ok here but with very small margin) try to set the fft1 
bandwidth to 400Hz. My computer is far too slow for that;)

The fast waterfall can be used for many purposes, not only
for radio signals. Connect an antenna directly to the
soundcard and watch ionospheric phenomena (protect the 
soundcard with a high pass filter at say 1 kHz).
You may use it to look at audio signals, music, speech 
analysis and other things.


2) Processor speed. Deselect second fft. Set window N=9 and
fft1 bandwidth to 0Hz. This will provoke a 65536 size for
the first fft. My 600MHz PIII is not fast enough. If it
works you have a very interesting spectrum analyzer if
you make the main spectrum fill most of the screen. 



73


Leif  /  SM5BSZ
LINRADDARNIL
g