[SOLVED] Chebyshev filter with real components

[DavidZlavan]

Hello all,

I'm trying to implement a Chebyshev bandpass 3rd order filter at 433MHz.

Values are:

Element 1 , Orientation : series
C = 0.165 pF, L = 820.909 nH
Element 2 , Orientation : shunt
C = 365.229 pF, L = 0.37 nH
Element 3 , Orientation : series
C = 0.165 pF, L = 820.909 nH

Idealy simulated the response is as expected. (I use agilent ADS)

But when trying to put real manufacturer values the response is gone.

For example, Changing the first capacitance from 0.165pF to 0.2pF 12dB of losses are appreciated!

I will now try to sweep values manually to find the best trade off but I can learn from you tips or another way to solve this.

Thank you!

 

[andre_luis]

The higher the order of the filter, the more critical the displacement of each element becomes from calculated bandwidth, which ultimately does have a cascade effect on the final response of the whole filter. What you can do is to perform components association series/parallel in order to try to tune up to the desired value, although tolerance remains the determining factor on deviation from original specs. Monte Carlo analysis is quite useful in determining the components with the greatest impact on performance degrade. Keep in mind that at RF frequencies, layout matters, so that some of the above small values ( 0.2pF ! ) do not even exist to sell, or are in the same magnitude order of the parasitic elements of the components/board.

 

[DavidZlavan]

The higher the order of the filter, the more critical the displacement of each element becomes from calculated bandwidth, which ultimately does have a cascade effect on the final response of the whole filter. What you can do is to perform components association series/parallel in order to try to tune up to the desired value, although tolerance remains the determining factor on deviation from original specs. Monte Carlo analysis is quite useful in determining the components with the greatest impact on performance degrade. Keep in mind that at RF frequencies, layout matters, so that some of the above small values ( 0.2pF ! ) do not even exist to sell, or are in the same magnitude order of the parasitic elements of the components/board.

Thank you for your answer! What we have done is reduce to order 2, increase to BW=150MHz (this has been very helpful because value varation didn't affect so much) and do a ADS optimitzation for S11 and S22.
Now losses are about 0.6dB.
It's a pitty not to have a good filtering at 315MHz but it was prioritary to reduce losses and increase matching.

PS: Murata GRM15 capacitors have values of 0.2 and even 0.1pF and it's true the layout would have these parasites but I don't think it should affect due to lambda/100 of 433MHz is about 7cm and layout is smaller.

 

[BigBoss]

You probably request a "non feasible/tight" response from this filter.I mean extremely high attenuation in ban-dstop region, very narrow-band etc. characteristic. Therefore the filter elements are uncommon for 400MHz.I suggest you to loosen the specs. of the filter in order to find more realizable components.

 

[FvM]

To avoid extreme L/C ratios, a coupled resonators topology might be more appropriate.



But I basically agree with BigBoss that your bandwidth specification is probably beyond a feasible range for lumped filters without tuning components.