Expert Advice Responses Page (Febuary)

Published Tuesday, Febuary 19, 2008 | Last Update 3.04.08
Here are your replies to the Febuary Expert Advice column

This month, Rafi Hershtig, K&L Microwave talks About the Band-Reject Filter (BRF), or Band-Stop, is the solution of choice in an increasing number of cases. Exhibiting low loss and good power-handling, this device offers high attenuation of unwanted signals in a localized area, while impacting the entire remaining spectrum very little.

Reference Guide Winner

Comments: Dear Rafi, thanks for this great description. I have implemented a similar approach but have not yet understood the theory enough to find out how to tune the filters. Due to your explanation I do know how to now. One questions remains: What is the temperature stability for the industrial temperature range? Chris

Christian Honegger, Siemens Building Technologies
2/26/2008 3:50:34 PM

Response from the Author: Hi, Christian. Thank you for the taking the time to read the article. I’m happy it shed some new light. If I understand your question correctly, then it is a matter of standards. Typically, the “commercial” applications are from 0 to +50 C, although I have seen cases of -10 to +60 C. I hope I have answered your question correctly. Please don’t hesitate to clarify more if needed.

Reference Guide Winner

Comments: Given the challenge of converting an optimized netlist (equivalent circuit) into a physical design, I find it easier to start with a physical design that can be optimized for electrical performance according to physical attributes. Viewing the netlist topology and values may be better for understanding how component values (i.e. shunt capacitance or series inductance) affect filter performance, but I think it may be better to obtain first pass values from a lumped element version but quickly move over to a physical version before investing alot of time optimizing. I also agree on the importance of EM design validation before fabricating any prototypes. Of course, my work is mostly in microstrip, so this approach may not be for everyone. Thanks for the info, Rafi. Mike

Michael Fallon, BAE
2/29/2008 3:46:20 PM

Response from the Author: Mike, that’s a good point and very much valid for printed filters. The gaps between the resonators, for instance, carry too much information that is needed up front and during the optimization and not at the very end. As for cavity filters, they do offer a larger percentage bandwidth then printed filters and you know from the early beginning if you are against a very tight spacing between resonators or not. In addition, while printed filters are covered by linear simulators and 2.5D, the cavity filters in the mechanical sense are covered only by 3D modeling, which is more time-consuming in handling the bandpass band-stop diplexer as one entity. Many thanks.

Reference Guide Winner

Comments:Dear Rafi, I was glad I caught your presentation in the MicroApps in IMS2007, and now this article which provides some valuable insight to the design of diplexers. We're trying to design some diplexers and multiplexers using LTCC technology, as part of a integrated microwave module development. For the narrowband signals (e.g. point frequencies) we can do it reasonably fast using loop directional filters, but we're facing some problems with those requiring more than 30% fractional bandwidth for the passband (some about up to 50%), as loop filters will not be appropriate for such bandwidth. One approach we are considering is to use 3D planar LC implementation for the wider passband requirements (centered at, say, X-band). Can you can offer us some advice on the potential pitfalls for this intended approach, and advice on how to tune the bandpass & bandstop separately? Another approach we're considering is to use a pair of wideband 3-dB couplers (maybe using re-entrant coupling for the mid coupling section) with a pair of conventional "reflective" filters to form a directional filter. We have noticed from past measurements that we can't implement bandpass filters with high rejection (i.e. beyond 40 dB rejection)at high frequencies (X-band and beyond) unless we have proper chassis for the housing of the filter (i.e. implementing the side "PEC" walls). If we wish to implement this approach completely in LTCC (i.e. having the pair of conventional filters and 3 dB couplers desiged as embedded passive components in the stackup and without any mechanical chassis surrounding the passive components), and we want the filters to have up to, say, 65 to 80 dB of rejection, will we still face the same problem as mentioned above (i.e. not able to obtain high rejection without the side "PEC" walls), or will this problem be minimised since the pair of filters are now part of a directional filter, which is an absorptive filter (i.e. should not be facing the same problems as a more commonly found reflective filter)? Your advice is greatly appreciated. Thank you, and best regards. Kevin

Kevin Ho, DSO National Laboratories
3/4/2008 11:58:24 AM

Response from the Author: Dear Kevin, The situation you are describing encompasses a wide angle of the design spectrum and it is hard for me to reply intelligently without having the specs in front of me. The design strategy of the planar elements and the LTCC technology is solely based upon 2.5D or 3D simulators right from the start, which is different then the cases I described in my article. The closest example I have is the 15GHz design where the BS was done using Suspended-Substrate-Stripline and the BP was a TEM filter, but I’m not sure if this is the answer to your case. I can say that in this particular example, the combined network and optimization was done using Touchstone linear simulator and the S-par file of the BS was created by Sonnet. I hope this helps you a little bit. Cheers, Rafi Hershtig

Reference Guide Winner

Comments: Dear Rafi, your presentation is very interesting and well focused on to highlight interference/spurious mitigation aspects about Mixer/Up Down Converter chains. In my experience, I`ve found that in order to BRF the "Holy Bible" of the Filter Design (Matthei, Young) is not as reach of pratical design aspects as BPF. So, facing the need to design two tunables Band Reject cavity filters (Frequency Band 5 GHz and 8 GHz), and giving the difficult to full model and optimize a consistent equivalent circuit which can match the "real" performance of the BRF, I`ve implemented a solution using a three port device (circulator). The incoming signal feeds the port One of the circulator, the port 2 is connected to the input port of a BPF(1) (the output port of the filter is connected to a 50 Ohm Wide Band Industrial Load). The output port of the signal is the third port of the circulator and is connected to a BPF(2). The BPF(1) is optimized from the point of view of the S11 (input matching) into the band where unwanted signals are present, providing Band Stop effect onto incoming signal, as needed, and low loss and good power handling. Up to 35% of the fractional Bandwidth is achievable, in addition it is possible separatly tuning of the BPF(1) and (2). The jointly effect of BPF(1) and BPF(2) allows to maximize attenuation of unwanted signals in a localized Band. Moreover this approach enables to focus the simulation and optimization on a BPF model, which has a great coverage in the literature. My work is mostly focused in cavity filters, but I think this approach may be extended to the others technologies. What do you think about this? Many thanks for your attention and cooperation Andrea Costa Selex Communications

Andrea Costa, Selex Communications
3/6/2008 10:52:13 AM


Response from the Author: Dear Andrea, Thank you so much for reading the article and providing another possible angle. If I understand your layout correctly, your output is through BPF2, which is a wide band filter that covers BPF1. But the main reason why you chose to use a circulator/ reflection BPF arrangement is because it is easier to tune the BPF over a wide range then a BRF. This layout will always be limited by the amount of isolation that the circulator can provide and over wider bands, they tend to provide a lower isolation level. The fact that you can “tweak” BPF2 to achieve a deeper notch depth (Is this why you have the BPF2 in the first place?) is not a repeatable measure that you can count on if you have to produce many modules like this. The reason is that the isolation level of the circulator dominates the final performance of the module. In short, the circulator/reflection type BPF is best used over narrow band application at lower frequencies (<2GHz), where >25dB isolation levels are available and the equivalent BRF size is too large. Since not much information was provided about the tuning range of the notch, I can’t comment if your design is the only choice you might had. Thanks, Rafi