Low-pass Filter Configuration

The commonly used low-pass filter is shown in Fig. 3.10(a), and this is a fifth order low-pass filter. After adding an inductor (0.82 nH) to this conventional low-pass filter (shown in Fig. 3.10(b)), the return loss becomes bad and the pass-band response becomes uneven, thus the pass-band insertion loss is too much. The simulation results are illustrated in Fig. 3.10(c), the dash lines and the solid lines denote the scattering parameters of Fig. 3.10(a) and Fig. 3.10(b), respectively. Consequently, it is obvious to see how much the bond wire will affect the input impedance matching. In order to solve this problem, we have to design other structures of low-pass filter that can utilize the bond wire to compensate the input matching, as discussed in Section 3.2.

The schema of the proposed low-pass filter incorporating the bond wire effect for this module integration is shown in Fig. 3.11. Compared with the filter in Fig. 3.10(b), the proposed low-pass filter has only 6 elements rather than 8 elements. Moreover, the inductor can be implemented by utilizing external bond wire which is used to connect the low-pass filter and the top-surface bare dies (the T/R switch). In Fig. 3.11, the resonator formed by a capacitor and a stripline can provide two finite transmission zeros at the second and the third harmonics of the operating frequencies. By changing the length of the stripline or the value of the capacitor in the resonator, we can control the frequency locations of the two finite transmission zeros. The proposed low-pass filter is simpler and has better performance than those of the conventional low-pass filter. The ideal simulation results are shown in Fig. 3.12, and it’s obvious to see that this low-pass filter has low insertion loss at pass band and high suppression at the 2^nd and 3^rd harmonics.

(a)

(b)

(c)

Fig. 3.10 (a) Schema of conventionally used low-pass filter (b) Schema with a bond wire (c) Simulation results of (a) and (b).

Fig. 3.11 Schema of the proposed low-pass filter for module integration.

Fig. 3.12 Ideal simulation results of the proposed low-pass filter.

3.3.2 EM Simulation

In this section, the design flow is as the same as Section 3.2.2. We use the EM simulator (HFSS) to draw the 3-D layouts of the low-pass filter and to simulate the scattering parameters. The 3-D layouts including oblique view, side view, and top view, are shown in Fig. 3.13. Because this filter is simple and has only 5 buried elements, we use only 3 layers for this layout. In order to avoid the fabrication error affecting the filter too much, we use thick layer for this layout to reduce the sensitivity. The interdigital capacitor will shift a bit due to the inaccuracy of fabrication process, thus may change the value of the capacitor. As a result, we can enlarge the distance between the two parallel plates and make one of the two parallel plates larger than the other, and the sensitivity to the fabrication error can be reduced. On the contrary, once we enlarge the distance between the two parallel plates, the area of the two plates also has to be increased to remain the value of demand. Consequently, that is a compromise among size, cost, and performance. Fig. 3.14 shows the EM simulation results of the low-pass filter. From Fig.

3.14, the insertion loss at pass band is less than -0.44 dB, and the suppression at the second and third harmonics are higher than 30 dB. Significantly, the finite transmission zero at 10.55 GHz can provide a suppression more than 60 dB.

(a) Oblique view

(b) Side view

(c) Top view Fig. 3.13 3-D layouts of the low-pass filter.

Fig. 3.14 EM simulation results of the low-pass filter.

3.3.3 Experimental Results

After the analysis and EM simulation, the designed low-pass filter was fabricated using the CT2000 LTCC process with the dielectric constant of 9.1 (at 2.5 GHz), the loss tangent of 0.002 (at 2.5 GHz), and the thickness of the silver alloy of 0.012 mm. The commonly used printed-circuit board FR4 with the dielectric constant of 4.4, the loss tangent of 0.02, and the thickness of 0.4 mm was applied as the evaluation board to measure the performance of the fabricated LTCC filter. This filter was measured by probes which were connected to the network analyzer HP8510C. The first filter was measured without the bond wire, and its result is shown in Fig. 3.15. Note that the pass-band response was not flat due to bad input matching. The insertion loss at pass band are a little higher, the losses at 4.9 and 5.85 GHz are -0.41 and -0.733 dB, respectively. The scattering parameter S21 at higher frequencies about third harmonics

can not provide enough suppression, and it likes that the S21 will ascend as the frequency increases. After bonding a wire to the I/O port of the filter, the measured result demonstrated in Fig. 3.16 shows that the pass-band response is much better than that in Fig. 3.15, and the input matching became deeper. Consequently, one can observe that we have to consider the effect of bond wires before integrating active components and LTCC buried circuits into a package. In Fig. 3.16, although the measured results showed a bit shift to higher frequencies due to fabrication process inaccuracy, the two results were found to agree well with each other. The measured insertion losses from 4.9 to 5.85 GHz were less than 0.34 dB, with a minimum value of 0.235 dB at 4.9 GHz. Moreover, with the transmission zero providing a suppression of 54.4 dB at 10.76 GHz, the second harmonics could be suppressed more than 26 dB. In addition, the suppression at the third harmonics was higher than 25 dB.

Fig. 3.15 Measured results of the low-pass filter without a bond wire.

Fig. 3.16 Measured results of the low-pass filter with a bond wire.