S E T U P F O R M C H A P T E R 8
Blackman, Blackman-Harris: We use Blackman-Harris to design all filters in the signal
processing chain for all modes. This provides for minimal bleed through as we have described and
the best shape factor for our needs in the overlap-save filtering routines. While it has some good
features, these features are probably not ideal for the CW power spectrum displays. There is a
penalty to pay for the smoothing near the ends afforded by the windows. If you have experience in
the use of FFTs and the display of the power spectra that result, you know that a smaller FFT at a
fixed sample rate causes a wider range of frequencies to be contained in one bin of the power
spectral results. This means that a tone will look more like a large lobe or finger than a tone spike.
The same phenomenon is present in the best windows. While the spreading of the “main lobe” of a
tone is not as bad as taking an FFT of half the size, it is wider than one bin. The formulae for all of
the Blackman filters are of this form:
f(x) = a
0
− a
1
cos(πx) + a
2
cos(2πx) − a
3
cos(3πx)
The individual parameters by type:
o Blackman: a
0
= 0.42, a
1
= −0.5, a
2
= 0.08, a
3
= 0
o Blackman Harris: a
0
= 0.35875, a
1
= 0.48829, a
2
= 0.14128, a
3
= 0.01168
o Blackman Nuttall: a
0
= 0.3635819, a
1
= 0.4891775, a
2
= 0.1365995, a
3
= 0.0106411
Parzen, Bartlett, Exponential, and Riemann: These windows are much less widely used and are
included for completeness. They sometimes perform better for a particular application and some
experimentation on the users part should be undertaken to find the window that gives the most
pleasing display to you.
[The rest of this page has been left blank intentionally]
89 FlexRadio Systems
To Next Page
Loading...
Need help?
General
