Reply by Fred Marshall February 26, 20082008-02-26
"Steve Underwood" <steveu@dis.org> wrote in message 
news:fq129n$gfl$1@nnews.pacific.net.hk...
> Fred Marshall wrote: >> In general, the wider the main lobe, the lower the sidelobes - for >> well-designed window functions. You have been looking at some of the >> best. >> >> Sidelobe decay is another measure. In general, the faster the decay, the >> higher the first sidelobe. Again, you've been looking at some of the >> best. >> >> I believe the final selection is subjective and only can be objective if >> you have very specific requirements to relate back to the functional >> characteristics. > > Huh? If its subjective, why bother weighting at all? Surely you always > have a desgin goal, though you may not always end up choosing the optimal > window. >> >> So, "better" is in the eye of the beholder at the level of selection that >> you have already before you. > > Better ought to be the quality of fit against the system goals. > > Steve
Steve, I did say: "and only can be objective if you have very specific requirements to relate back to the functional characteristics" I find that most people have rather fuzzy objectives such as: "narrow main lobe" and not something like "I need the main lobe to be no greater than 10% wider than a sinc at the 3dB points" and "low sidelobes" or "-40dB sidelobes" rather than "I need the sidelobes to be 40dB down within 10% of the main lobe center frequency and 80dB down outside that" It's only when the criteria are more specific that one can make choices between what are already pretty good window functions. fred harris' paper of course is a great review of such things but one needs to take that informatin and apply it to their own situation. In my experience one can be very picky when designing filter/window design algorithms and will be a bit more arm-waving (subjective at some level) when designing a system. Usually what you "need" is an educated guess that you will compare with what's possible. And unless you're doing simulations it will be a fuzzy choice. Thus my comment about subjectivity. Then, if you are doing simulations it's a good bet that there will be multiple solutions that are "good enough". My favorite quote that I often offer to engineers comes from Voltaire and translates roughly into: "Better is the enemy of good enough". And, that, it seems to me, is the apex of subjectivity. How "good" is "good enough"? And that becomes a very interesting circular discussion about angels and pins and ..... Fred
Reply by Steve Underwood February 26, 20082008-02-26
Fred Marshall wrote:
> "boon chun" <boon-chun_lim@agilent.com> wrote in message > news:ZMKdnS3jJOP9fF_anZ2dnUVZ_jCdnZ2d@giganews.com... >> Hi, >> I'm looking for a good window which has narrower main lobe and lower >> side lobes. I try to use matlab to compare Blackman-Harris, Kaiser, and >> Chebyshev windows. And I found the Blackman-Harris has the best main lobe >> width and lower side lobes. Does any one has any comments on this? Is >> there any other type of windows better than this available? >> Thanks. >> Regards, >> BC
Only those windows? Why, you've hardly started. :-) You missed the commonest - Hamming - and there are some only people from particular problem domains ever seem to have met, like Taylor.
> In general, the wider the main lobe, the lower the sidelobes - for > well-designed window functions. You have been looking at some of the best. > > Sidelobe decay is another measure. In general, the faster the decay, the > higher the first sidelobe. Again, you've been looking at some of the best. > > I believe the final selection is subjective and only can be objective if you > have very specific requirements to relate back to the functional > characteristics.
Huh? If its subjective, why bother weighting at all? Surely you always have a desgin goal, though you may not always end up choosing the optimal window.
> > So, "better" is in the eye of the beholder at the level of selection that > you have already before you.
Better ought to be the quality of fit against the system goals. Steve
Reply by Fred Marshall February 26, 20082008-02-26
"Fred Marshall" <fmarshallx@remove_the_x.acm.org> wrote in message 
news:afKdnZCnb8hdL17anZ2dnUVZ_tijnZ2d@centurytel.net...
> > "boon chun" <boon-chun_lim@agilent.com> wrote in message > news:ZMKdnS3jJOP9fF_anZ2dnUVZ_jCdnZ2d@giganews.com... >> Hi, >> I'm looking for a good window which has narrower main lobe and lower >> side lobes. I try to use matlab to compare Blackman-Harris, Kaiser, and >> Chebyshev windows. And I found the Blackman-Harris has the best main lobe >> width and lower side lobes. Does any one has any comments on this? Is >> there any other type of windows better than this available? >> Thanks. >> Regards, >> BC >> > > In general, the wider the main lobe, the lower the sidelobes - for > well-designed window functions. You have been looking at some of the > best. > > Sidelobe decay is another measure. In general, the faster the decay, the > higher the first sidelobe. Again, you've been looking at some of the > best. > > I believe the final selection is subjective and only can be objective if > you have very specific requirements to relate back to the functional > characteristics. > > So, "better" is in the eye of the beholder at the level of selection that > you have already before you. > > Fred
I also should have mentioned that improvements in sidelobe levels generally come at very small increases in mainlobe width. It's a matter of changing energy between them. So a small amount of relative energy from a very large main lobe can result in large changes in sidelobe levels. Fred
Reply by Fred Marshall February 26, 20082008-02-26
"boon chun" <boon-chun_lim@agilent.com> wrote in message 
news:ZMKdnS3jJOP9fF_anZ2dnUVZ_jCdnZ2d@giganews.com...
> Hi, > I'm looking for a good window which has narrower main lobe and lower > side lobes. I try to use matlab to compare Blackman-Harris, Kaiser, and > Chebyshev windows. And I found the Blackman-Harris has the best main lobe > width and lower side lobes. Does any one has any comments on this? Is > there any other type of windows better than this available? > Thanks. > Regards, > BC >
In general, the wider the main lobe, the lower the sidelobes - for well-designed window functions. You have been looking at some of the best. Sidelobe decay is another measure. In general, the faster the decay, the higher the first sidelobe. Again, you've been looking at some of the best. I believe the final selection is subjective and only can be objective if you have very specific requirements to relate back to the functional characteristics. So, "better" is in the eye of the beholder at the level of selection that you have already before you. Fred
Reply by bharat pathak February 25, 20082008-02-25
Hello,

    You may also try dolph-chebyshev window. It has a programmable
    attenuation factor.

    win = chebwin(N, As);

    As is sidelobe attenuation in dB. 

    This window is slightly better than blackman-harris 11 Term. Meaning
    for the same attenuation factor (270 db), chebwin will have lesser
    main lobe width.

    Also you may check wikipedia. Good information on various
    windows and relative comparison is given.

    Happy Window Shopping!

Regards
Bharat Pathak

Arithos Designs
www.Arithos.com


>Hi, > I'm looking for a good window which has narrower main lobe and lower >side lobes. I try to use matlab to compare Blackman-Harris, Kaiser, and >Chebyshev windows. And I found the Blackman-Harris has the best main
lobe
>width and lower side lobes. Does any one has any comments on this? Is >there any other type of windows better than this available? >Thanks. >Regards, >BC > > >
Reply by boon chun February 25, 20082008-02-25
Hi,
    I'm looking for a good window which has narrower main lobe and lower
side lobes. I try to use matlab to compare Blackman-Harris, Kaiser, and
Chebyshev windows. And I found the Blackman-Harris has the best main lobe
width and lower side lobes. Does any one has any comments on this? Is
there any other type of windows better than this available?
Thanks.
Regards,
BC