+ Post New Thread
Results 1 to 14 of 14
  1. #1
    Advanced Member level 1
    Points: 6,805, Level: 19
    boy's Avatar
    Join Date
    Sep 2002
    Location
    Thailand
    Posts
    481
    Helped
    98 / 98
    Points
    6,805
    Level
    19

    Microwave Filter Design

    I have a question on microwave filter design. Can the filter prototypes such as Eliptical, Bessel, Gaussain be used in couple line struture or hair pin structure? In @DS design guide, it seems that those prototypes are not avaliable in those structure. any reason why? I am also curious how can I design a very flat group delay filter without sacrifice impulse response. Anyone know that someone do an optimization for group delay in filter design in stread of passband/stopband ripple? thank you

  2. #2
    Advanced Member level 5
    Points: 36,491, Level: 46
    flatulent's Avatar
    Join Date
    Jul 2002
    Location
    Middle Earth
    Posts
    4,629
    Helped
    494 / 494
    Points
    36,491
    Level
    46

    try Genesys Eagleware

    In the Genesys Eagleware MFilter program you can select bessel filters where every you can select butterworth and chebychev. It is all a matter of using the right G value in the coupling. Filters with zeros have to use other topologies.



    •   AltAdvertisement

        
       

  3. #3
    Full Member level 4
    Points: 4,049, Level: 14
    nandopg's Avatar
    Join Date
    Apr 2001
    Location
    Over the Rainbow
    Posts
    205
    Helped
    8 / 8
    Points
    4,049
    Level
    14

    No reason for flat delay in microwave frequencies

    Dear friend,
    As Flatulent said above you can implement any transfer function through coupled lines. However Butterworth or Gaussian filters are seldom used in microwave frequencies.
    What happens is the bandwidth of the microwave (channel) filter should be wider than the occupied bandwidth of the modulated signal so that the delay introduced by the filter doesn't disturb the modulated carrier and the undesirable spurs be attenuated properly.
    If for any reason you need a microwave (channel) filter with a bandwidth just about the same as the modulated signal you should introduce some group delay pre-distortion at IF and this is done most of the time associating 2nd order all pass networks. The IF pre-distortion should be complementary to the distortion introduced by the channel filter.
    If you want more help, please post a block diagram of the system you are studying.

    NandoPG



  4. #4
    Advanced Member level 1
    Points: 6,805, Level: 19
    boy's Avatar
    Join Date
    Sep 2002
    Location
    Thailand
    Posts
    481
    Helped
    98 / 98
    Points
    6,805
    Level
    19
    thank you. that is a good idea to put pre-distrotion in IF stage. I only think about pre-distortion in RF :). any reason why Guassian filter is seldomly used in Microwave??



    •   AltAdvertisement

        
       

  5. #5
    Full Member level 4
    Points: 4,049, Level: 14
    nandopg's Avatar
    Join Date
    Apr 2001
    Location
    Over the Rainbow
    Posts
    205
    Helped
    8 / 8
    Points
    4,049
    Level
    14
    The Gaussian filter is seldom used because of its poor stopband attenuation characteristic.
    You see, why are you placing a filter at channel level? For sure to attenuate unwanted spurs and to do this effectively the filter can not have a poor stopband, consequently Gaussian filters are not a choice. Also you always have the spurs frequencies far from the channel, what allows to use a channel filter broad enough so that its group delay doesn't disturb the modulated carrier.

    NandoPG



  6. #6
    Advanced Member level 5
    Points: 36,491, Level: 46
    flatulent's Avatar
    Join Date
    Jul 2002
    Location
    Middle Earth
    Posts
    4,629
    Helped
    494 / 494
    Points
    36,491
    Level
    46

    transitional filters

    There are transitional filters that are gaussian to -x dB and high cutoff beyond. Before the days of computer programs to optimize equalizes, these were very popular. The Filter Solutions program at http://www.nuhertz.com/ has these. You should download it and take advantage of the 20 day free trial to do some experimenting with different parameters on the pulse and frequency response.

    Attached are examples of a 8 pole gaussian to 3 dB and a 6 pole 1 dB ripple chebychev (with and without equalization) Notice that the amplitude responses are very close at the 2x and 10x frequency points. Also notice the precursor ripples on the equalization.



    •   AltAdvertisement

        
       

  7. #7
    Full Member level 4
    Points: 4,049, Level: 14
    nandopg's Avatar
    Join Date
    Apr 2001
    Location
    Over the Rainbow
    Posts
    205
    Helped
    8 / 8
    Points
    4,049
    Level
    14

    Transitional Filters are also very seldom used

    Transitional filters have the stopband a little better than the Gaussians but even so poor to be used as a channel filter.
    For example for a lowpass prototype with 5 sections at frequency 2*Wc, a transitional filter Gaussian to 6dB has an attenuation equal to 20dB. In the same situation a Chebyshev 0.1dB ripple has 41dB. Attached are the charts that allows to compare both filters.
    Also for the transitional filters, the Gaussian passband characteristic is a problem for several kind of carriers, while the Chebyshev filter has a flat passband rippled by a known (and small) value, what is very desirable in the most of the cases.
    Using the attached curves one can see the passband for transitional and Chebyshev lowpass filters.

    NandoPG



  8. #8
    Full Member level 3
    Points: 2,788, Level: 12

    Join Date
    Dec 2001
    Location
    China
    Posts
    166
    Helped
    3 / 3
    Points
    2,788
    Level
    12
    Suggest to use the GENESYS.



  9. #9
    Member level 1
    Points: 1,881, Level: 10

    Join Date
    Aug 2003
    Posts
    35
    Helped
    0 / 0
    Points
    1,881
    Level
    10
    IICCEE
    Can you tell me more about designing microwave filter with genesys?



  10. #10
    Advanced Member level 5
    Points: 36,491, Level: 46
    flatulent's Avatar
    Join Date
    Jul 2002
    Location
    Middle Earth
    Posts
    4,629
    Helped
    494 / 494
    Points
    36,491
    Level
    46

    step by step

    They have a m/filter module. You put in the physical method (stripline, microstrip, etc) and then in the next frame you put the general type (stepped impedance, coupled resonator, etc.) and then in the next frame you put the frequency response and then presto, they give you the physical dimensions. These can be imported into the analysis proram for optimizing.



  11. #11
    Newbie level 4
    Points: 1,722, Level: 9

    Join Date
    Jul 2003
    Posts
    7
    Helped
    0 / 0
    Points
    1,722
    Level
    9
    You can use "bridge" coupling. But you cannot obtain high selectivity and flat group delay simultaneously. This is a conclusion.



  12. #12
    Newbie level 5
    Points: 2,173, Level: 10

    Join Date
    Sep 2003
    Posts
    9
    Helped
    0 / 0
    Points
    2,173
    Level
    10

    Re: Microwave Filter Design

    you can use genesys2003,you can find couple-line-struture .



  13. #13
    Full Member level 2
    Points: 2,828, Level: 12

    Join Date
    Oct 2004
    Posts
    138
    Helped
    4 / 4
    Points
    2,828
    Level
    12

    Re: Microwave Filter Design

    as to group delay filter please refer to :

    tables for nonminimum-phase even-degree low pass prototype networks for the design of microwave linear-phase filters.
    J.H.cloete, IEEE MTT-27 no2.1979



    •   AltAdvertisement

        
       

  14. #14
    Full Member level 4
    Points: 4,903, Level: 16
    sinatra's Avatar
    Join Date
    Mar 2002
    Location
    Neverland
    Posts
    214
    Helped
    19 / 19
    Points
    4,903
    Level
    16

    Re: Microwave Filter Design

    Hello boy and all others.
    I would like to raise some more points that may help.

    1-No eliptical filters, why?
    The coupling between lines can be normaly modeled as a capacitor that will just be an imitance converter in the modern filter design theory. With a single capacitor in series we can only realize transmission zeroes in the infinity. Because of this, filters based in coupled line structures normally cannot have finite transmission zeroes. Since the eliptical filters have finite transmission zeroes it is normaly not possible to realize them using coupled line structures. That is why you don't find examples of eliptical coupled line filters in the literature......

    2-Other polinomial functions that have no finite transmission zeroes, like butherworth, chebishev, gaussian, etc. can be realized.

    3-What analog filter to use to have the best group delay reponse?
    The best filter concerning group delay response is always the eliptic filter. The reason is, because of its very abrupt transition from pass-band to stop-band (low pass example) all the poles are higher Q and are near the transition. This means that the group delay will be very much disturbed near the cut-off frequency comparing to other types of filters, but much less in the pass-band, why?

    Now, assuming that we have all filters adjusted to satisfy our filter requirements (filter orders of each type of filter adjusted to satisfy the same specification) you will see that the eliptic filter (the lower order one) will have the worse group delay around the cut-off frequency, but it will have the best group delay performance in the pass-band comparing to all others (assuming the same filter specs). The eliptic filter is then the best choice concerning group delay in the pass-band. The chebyshev filter will have a similar pass-band group delay performance as the eliptic one.

    4-In your case, for a microwave filter based in coupled line structures, the choice is then made. Chebyshev. You will just have to put the cut-off frequency far enough so that the pass-band group delay performance satisfies your requirements.

    6-Another remark: If your process is not very precise, i. e., may have variations like , substrate thickness, epsylon, track width (due to under eaching for example), etc.. than I would sugest that you try a butherworh filter . It will be much more tolerant to process variations. Unfortunately you will have to put the cut-off frequency further than in the chebyshev case, etc.. etc..

    Hope it helps.
    Greetings to all
    S.



--[[ ]]--