1200-1640MHz bandpass filter (flat within 0.2dB between 1200 and 1299MHz) The goal is to build a flat filter between 1200 and 1299MHz using the microstrip technique ensuring low losses, good shape factor and ability to cut frequencies around 900MHz. With the microstrip technique, the narrower the passband the higher the in-band attenuation; as a result, a “wide” filter was preferred – providing low in-band attenuation – but shifted upwards, so that the undesired band ends up on the side of the response curve ensuring the highest possible attenuation (a realistic figure is >35 dB). The goal is to obtain an in-band insertion loss of 1dB. The project parameters are the following: Zin, Zout: 50ohm Type: bandpass 1150-1650MHz @-3dB Sub-type: hairpin Distribution of poles: Chebyshev In-band ripple: max 0.5 dB In-band RL: >18dB Attenuation at 900MHz: >35dB The filter is built on PTFE, such as ROGERS RT/Duroid 20mils (0.5mm) 1oz/sqft (35 microns). Another possibility is to use FR4, however the dielectric constant (4.5) is very erratic as well as not well-defined during manufacturing. In addition, dielectric losses are quite significant. When entering the project parameters into one of the several software applications for filter synthesis, this is the resulting geometry: Fig. 1 – Geometry of a 5-section microstrip filter. NOT IN SCALE The same software can also produce Gerber (.gbr) format files, which can be used to build the filter in the right size and with the desired electric features. A Gerber file is attached to this article. By printing it on a regular acetate sheet in two copies with a good laser printer and overlaying the two sheets – in order to obtain blacks completely opaque to UV light – a mask can be obtained (either positive, to be used with sprayable photoresist, or negative, to be used with film photoresist). Below is the expected frequency response based on the synthesis: Fig. 2 – The filter’s simulated S21 frequency response. In order for the actual values to be very close to the theoretical ones, the filter must be built as accurately as possible. The pictures below show the extremely small size of the filter as well as the high degree of precision and accuracy required. Fig. 3 – The filter lines (still is photoresist in the picture) are 0.86mm wide and 0.178mm apart This is the finished filter: Fig. 4 – Filter on ROGERS RT/Duroid 20mils 35 microns (PTFE) A couple of details: Fig. 5 – The lines after the acid attack These are the real values as measured: Fig. 6 – Actual response of the S21 filter Fig. 7 – Actual response of the S21 filter across the whole band Fig. 8 – The filter’s in- and out-of-band RL; notice the values at 1295MHz Fig. 9 – The filter’s insertion loss between 1200 and 1299MHz, <1dB Interestingly, the simulation and synthesis produced by the program and the real-life results overlap nicely – the -3dB passband, the in-band attenuation, the flat in-band response, the RL and the good symmetry of the filter sides.