Re: Q Factor
Hi, all. Talking about Filter Q, I think Randy Rhea from Eagleware had a very clear explanation. He give the expression of I.L for Cheby BPF, which is independent on topology. It is related to Qloaded, Qunloaded, and gs constant (from conventional filter cookbook).
His forum is also very informative, especially on filter design(printed, ceramic resonator type etc). I just love it.
www.eagleware.com
aaqiao
Joined: 22 Oct 2003
Posts: 3
Posted: Mon Dec 01, 2003 8:47 am Post subject: Insertion loss of the narrow band hairpin bandpass filter
--------------------------------------------------------------------------------
I am a graduate from China, and I am trying to design a hairpin bandpass filter with the following parameters:
2856MHz center frequency
60MHz bandwidth
30 dB stopband attenuation and no special limits for the transition bands
I designed it with M/FILTER and finally got a 7 degree hairpin microstrip filter. Its stopband attenuation seemed pretty good but its insertion loss of the passband was very bad, about 7db. I know that the insertion loss of the passband is inversely proportional to the unloaded Q of the resonator, but I have no ideas what to do next. Shall I select other kinds of transmission lines, such as coaxial line or waveguide? Waveguide bandpass filter seems attractive but I don't know how to design it. And can GENESYS do the work? I am eager for some suggestions.
I will be very thankful to your kindness.
Geng Zheqiao
Back to top
Randy Rhea
Eagleware Staff
Joined: 12 Jan 2003
Posts: 91
Posted: Mon Dec 01, 2003 1:42 pm Post subject:
--------------------------------------------------------------------------------
Yes, you are correct, insertion loss is inversely proportional to resonator unloaded Q. More specifically, the IL at the passband center is given approximately by
IL = 4.34*Qloaded/Qunloaded*(summation of the G values)
Qloaded is the filter center frequency divided by the filter bandwidth, in your case, 2856/60 = 47.6. This is a high loaded Q for lumped element and printed filters. Notice that insertion loss is directly proportional to loaded Q.
4.34 is one half the neper constant. The (sumation of the G values) is the sum of the reactive G values for the lowpass prototype used to design the filter. These do not include the termination G values G(0) and G(N+1). The G values are given in the G Values tab of the M/FILTER property box. The summation of the G values is given in the Summary tab once you have entered the passband ripple and the filter order.
The best way to reduce loss is to use resonators with higher unloaded Q. In microstrip, this means as thick a substarte as possible, because unloaded Q in both lumped and distributed resonators increases with increasing physical volume. If loss is a primary concern, I would use a substrate thickness up to 0.125" or 3.2 mm at 2.8 GHz. At higher frequencies thinner substrates must be used or radiation and additional operating modes become a problem. With thick substrates, evanescent modes exist which are not simulated by circuit theory simulators. Therefore, for your application I highly recommend using the EMPOWER electromagnetic simulator to improve accuarcy. I also recommend considering an interdigital filter with the thick substrate. It is degraded less by evanescent modes.
Thick substrates are expensive. That is why machined filters are often used for applications where the ground plane spacing must be large to reduce loss. An example filter type is slabline which is round rods between two flat ground planes with air as the dielectric. These are often done in interdigital or combline. There is an example slabline combline with photographs in my book HF Filter Design and Computer Simulation.
You might find that it helps a little to use a wider bandwidth and then increase the order to achieve the stopbands. It all depends on whether the increased summation of the G values or the wider bandwidth has the most influence. You can use your GENESYS simulator to quickly compare the loss of different bandwidths. This is a little easier in the latest versions of the software which allow you to interactively use synthesis and simulation.
Waveguide filters would have low loss but they would be extremely large. GENESYS will simulate coaxial processes, but M/FILTER synthesizes only coaxial in the stepped impedance bandpass which is probably not suitable for your application. M/FILTER handles slabline very well, but 2% bandwidth with large ground plane spacing will suffer greatly from evanescent modes.
My final recommendation: thick microstrip or suspended microstrip as an interdigital with M/FILTER to start the design and EMPOWER to refine the design.
Clear skies and high Q