Design Procedure

Fig. 5.13 is the design procedure for the proposed A-CMF, which is expected to characterize all of the physical parameters when designing. The details of the flow are listed as follow:

Step 1: From the given (or desired) mode impedance ZDM and ZCM, the substrate height h and relative permittivity εr, and the metal thickness tM, the line width w1 and spacing s1 of the microstrip coupled lines can be synthesized.

Step 2: For the desired center frequency of common-mode absorption f0, calculate the corresponding CM wavelength of the coupled microstrip lines, and let the center-to-center distance d1 between slots equal a quarter of the CM wavelength.

Step 3: Select a suitable value for the width w2 of the thinner slots (the smaller the better, and always limited by the design rule), and extract the characteristic admittance Y2 and guided wavelength λg2 of the thinner slotline. Then plot the trace of yin2 in the Smith Y chart (normalized to Y2) based on the length of slot II (0.5s) at frequency f0, as point C to D’ in Fig. 5.10(c).

Step 4: Select a suitable value for the width of the wider slot (w3). For an ideal circuit, it is better to have larger w3 for wider DM bandwidth. But in order to avoid unwanted parasitic effects in the real structure, there are some constraints for w3, which will be introduced in Step 5 and Step 7. Then extract the characteristic admittance Y3 and guided wavelength λg3 of the wider slotline.

Step 5: Select a suitable value for the length of the wider slot (l3). l3 should be equal to or larger than w3; otherwise, the wider slotline will act as a y-directional transmission line rather than an x-directional one, and may contribute some

unwanted parasitic effects. Then plot the trace of yin0 in the Smith Y chart (normalized to Y3), as point A to E’ in Fig. 5.10(c).

Step 7: Check if the following condition in (5.7) is satisfied. In other words, whether point F’ is at the right-hand side of the conjugate of point D’ in Fig. 5.10(c). If yes, continue the procedure to Step 8; otherwise, go back to Step 4 and choose a new value for w3.

Fig. 5.13. The design procedure of the proposed A-CMF.

5.3 Results and Validation

5.3.1 Frequency-Domain Response

Both the unidirectional and bidirectional A-CMF using step-impedance slots are fabricated based on the parameters designed in the previous sections. The photographs of the samples shown in Fig. 5.14(a) and Fig. 5.14(b) are the bidirectional and unidirectional A-CMF, respectively. Since the differential coupled lines on the top layer are too close to mount the SMA connector directly, feed lines with a length of about 35 mm are added at all of the ends of the signal lines. The S-parameters are measured with a four-port vector network analyzer (Agilent N5230A). To remove the effects of feed lines, thru-reflect-line (TRL) calibration is applied here. However, the available measuring frequency range is limited by the length of feedlines, and is around 7.5 GHz.

The measured S-parameters of bidirectional design are plotted in Fig. 5.15. In Fig.

5.15(a), the magnitude of CM transmission coefficient (|Scc21|) is lower than -10 dB from 1.7 GHz to over 7.5 GHz, which provides a broadband noise rejection with a fractional bandwidth of 126%. The center frequency of CM absorption is located at 2.3 GHz, with a slight shift from the original design (2.4 GHz). This may be caused by the fabrication error and the variation of dielectric constant of the substrate. The simulated

|Scc11| and |Scc21| by full-wave approach are also shown in Fig. 5.15(a). The agreement between the measurement and simulation is reasonably good. As to the DM insertion loss as shown in Fig. 5.15(b), it is less than 3 dB from dc to 7.4 GHz, which matches well with the simulation result. In addition, the measured DM group delay shown in Fig.

5.16 is flat and agreed well to the simulated one below 6.5 GHz. With these results, it can be expected that the proposed A-CMF can still transmit the DM signal well even in the high-speed system, which will be validated in the next section.

(a)

(b)

Fig. 5.14. Photograph of the samples for measurement. Feed lines with SMA connector are added at the ends of each signal line. (a) The bidectional A-CMF. (b) The unidirectional A-CMF.

(a)

(b)

Fig. 5.15. The measured S-parameters of the proposed bidirectional A-CMF using three step-impedance slots and two resistors. (a) The CM responses. (b) The DM responses.

Fig. 5.17 shows the CM responses of the unidirectional A-CMF. The CM reflection coefficient at Port 1 (|Scc11|) has a TZ at the same frequency as the CM transmission coefficient (|Scc21|), while the CM reflection coefficient at Port 2 (|Scc22|) is with a transmission pole there (around 0 dB). As expected, the absorption is only unidirectional for the CM noise from Port 1 in this design, but half of the size can be reduced. The DM responses of the unidirectional design are almost the same as the bidirectional ones in Fig. 5.15(b), and therefore are not plotted. To conclude, the concept of CM absorption in the proposed circuit are proved both by the full-wave simulation and experiment.

Fig. 5.16. The measured DM group delay of the proposed bidirectional A-CMF.

5.3.2 Validation of Signal Integrity in Time-Domain

To check the signal integrity of the proposed A-CMF, a reference board is also fabricated and measured. The reference board has the same differential line structure as the bidirectional A-CMF except that the patterned ground plane is replaced with a solid ground plane. Eye diagrams are used to determine the DM signaling performance. Both the reference board and bidirectional A-CMF are excited by a pulse pattern generator (Anritsu MU183020A) with a differential (2¹⁵-1)-bit pseudorandom binary sequence (PRBS) whose peak-to-peak voltage is 1 V, a bit rate is 7.5 Gbps, and rising/falling time is 40 ps. The output eye diagrams including the effects of the device under test (DUT), feedlines, and cables are measured by a digital storage oscilloscope (Agilent DSA-X 91604A). The result of reference board is shown in Fig. 5.18(a) and the one of the Fig. 5.17. The measured CM S-parameters of the proposed unidirectional A-CMF using two step-impedance slots and one resistor. CM noise from Port 1 is absorbed and CM noise from Port 2 is reflected at 2.4 GHz.

proposed bidirectional A-CMF is shown in Fig. 5.18 (b). In these two figures, the vertical scale is 200 mV/div and the horizontal one is 20 ps/div. As listed in Table 5.1, the eye height (defined by 40%-60%), eye width (defined at 50%), and the root-mean-square (RMS) jitter of the proposed A-CMF are 600 mV, 114 ps, and 2.67 ps, respectively. These values are very close to the ones of reference boards, which are 603 mV, 113 ps, and 2.74 ps, respectively. This indicates that the proposed A-CMF will not degrade the SI even under this high bit rate.

Reference board Proposed A-CMF

Eye Height (40%-60%) 603 mV 600 mV

Eye Width (50%) 113 ps 114 ps

RMS jitter 2.74 ps 2.67 ps

Table 5.1. Parameters of the measured DM eyes.

(a)

(b)

Fig. 5.18. The measured DM eye diagrams of the reference board and proposed bidirectional A-CMF under 7.5 Gbps PRBS. (a) Reference board. (b) Proposed A-CMF.

5.3.3 CM Noise Absorption in Time Domain

The simulated time-domain CM suppression using measured S-parameters is shown in Fig. 5.19. To compare the suppressing mechanism, a reflective CMF with a similar stopband of CM transmission coefficient (|Scc21|) as the proposed A-CMF is selected to be compared with. The structures (reference board, reflective CMF, and A-CMF) are all excited with the same PRBS source as in the previous section except there is a time skew of 13 ps between the positive and negative channels. The skew is usually unavoidable in the practical circuit and will induce some CM noise. As the green dashed curve in Fig. 5.19(a), if no CMF is applied, CM noise with a peak-to-peak voltage of 96 mV exists at the output port. By using either kind of CMFs, the noise can be reduced significantly, to 36 mV with reflective CMF (red curve) and to 11.7 mV with proposed A-CMF (blue curve). But as to the reflected CM noise seen at the input port as shown in Fig. 5.19(b), the result of the case using reflective CMF (red curve, 37 mV) is about twice as large as the reference one (green curve, 18 mV). If the proposed A-CMF is applied, the reflected CM noise will become much lower (blue curve, 12 mV). This is because the proposed A-CMF absorbs the CM power while the reflective CMF reflects the noise to the input port. Again, the advantage of the A-CMF can be seen and proved.

(a)

(b)

Fig. 5.19. The CM suppression in the time domain. (a) The transmitted CM noise.

(b) The reflected CM noise.

5.4 Summary

Based on the proposed equivalent circuit model and an analytic design method, an absorptive common-mode filter (A-CMF) in 2-layer PCB is implemented in this paper.

Using the transformation between short and open circuit, the proposed A-CMF has a deep transmission zero for |Scc21| at the Wi-Fi 2.4 GHz band and prevents the CM noise from propagating in the frequency range from 1.7 GHz to over 7.5 GHz. Besides, it can also provide a perfect match to the CM wave and make it reflectionless. Hence, the CM noise can be absorbed well. The absorption can be observed in both time- and frequency-domain simulation and measurement. With the help of step-impedance slotlines, the structure can be miniaturized and the available bandwidth of DM signal can be enhanced. The DM signal can transmit through the proposed A-CMF without degrading even when the data rate is as high as 7.5 Gbps.

As listed in Table 5.2, the proposed A-CMFs are compared with other state-of-the-art CMFs, including different fabrication process, DM cut-off frequency (insertion loss of 3 dB), available CM stopband (|Scc21| < -10 dB), fractional bandwidth of CM stopband (FBW of |Scc21|), CM absorbing band (both |Scc21| and |Scc11| are less than -10 dB), cutoff frequency ratio of DM (-3 dB) to CM (lower bound of -10dB), and electrical size (normalized to λg which is the wavelength in dielectric at the frequency of lower bound of the CM stopband). Most of the conventional CMFs can only reflect the CM noise [32]-[34], [36], [42], and only one of them [41] can absorb and eliminate the CM power. However, the idea in [41] is realized in the process of the integrated passive device (IPD) with higher cost. Although the proposed A-CMFs have the relatively large size, they can be implemented in cost-efficient 2-layer PCB process. In addition, compared with [41], the proposed bidirectional A-CMF has a wider CM stopband with

an FBW larger than 126% and a higher DM-to-CM cutoff frequency ratio of 4.4. The proposed circuit is the first A-CMF that can be implemented in PCB and with a good response in both DM transmission and CM suppression.

Table 5.2. Performance comparison with state-of-the-art CMFs

Chapter 6 pter 6 Conclusion

6.1 Conclusions of this Dissertation

High-speed digital differential signaling system can transmit a large amount of data and have high immunity to the noise interference at the same time, which has been widely used in the electrical devices and can provide higher performance and better experience to the users. Under this scenario, CM noise is in known as one of the most serious sources that cause the RFI/EMI problems, which may degrade the throughput of the wireless circuits or even make the devices harmful to other products and human bodies. Although this phenomenon is well known and the mechanism has been studied, it’s still uncertain why the radiation is mainly caused by CM but not DM power, and to what degree the radiation will be caused by a discontinuity.

In this dissertation, an analysis method and three CMF designs are proposed to characterize these interference problems. In Chapter 2, a modal method is developed to deal with the discontinuities in the MTL systems, and can be applied in analyses of both SI and EMI issues. The proposed method provides a more accurate prediction compared with the conventional circuit simulation tool, and takes much less simulation computing resource and time compared with the full-wave simulation tool. In chapter 3, a

TSV-based CMF in 3-D ICs is proposed to solve the CM-excited noise problem inside the chip package. By employing the parasitics of TSVs into design, the TSV-CMF has a high DM operating bandwidth and a relatively compact size. The CM stopband can be easily designed with the derived formula and can be enhanced by applying stacking technique of chips. In Chapter 4, a CMF with three designable TZs is proposed for different kinds of application. Once the center frequency of the stopband is decided, one of the TZ can be designed by the proposed method. Then the other two TZs can be used in the compromise between the bandwidth and suppressing level. In Chapter 5, a design method of CMF with absorption functionality is constructed. The proposed A-CMF can be easily implemented in cost-efficient PCB process with excellent performance. Not only absorption of CM noise, but also high cutoff frequency of DM to CM and high FBW of CM stopband can be all achieved.

6.2 Suggestions for Future Works

Strictly speaking, there are still something undone and to be improved in this dissertation. The MTL analysis method in Chapter 2 assumes that the propagating modes are TEM modes, but in the inhomogeneous system like microstrip coupled lines, the modes will be quasi TEM modes. Then the effect of resistance and conductance cannot be ignored. Although a possible solution is also discussed, but how the loss factor affects the modal decomposition has not been studied yet. In addition, theoretically this method can model the system with multiple differential pairs, but the validation is not done. More studies on these will make the mechanism of noise interference clearer. The TSV-CMF in Chapter 3 is designed with a lumped circuit model, which is known with a DM cutoff even under lossless environment. It is

suggested that higher order models or transmission-line-based models can be studied for more advanced design. Since the electrical behavior of transmission line is a transcendental function, there must be more than 3 TZs in the CMF design proposed in Chapter 4. Studying how to generate or involve more TZs in the stopband and how to control the TZs can make the CMF with better suppressing effect. Although step-impedance lines have been applied in the A-CMF in Chapter 5, there is still a DM TZ at the higher frequency, which may limit the application. In addition, to be implemented in PCB indeed reduces the cost, but the size is relatively large. It is suggested to study the duality circuit of the proposed one and to realize the CMF without slotlines.

REFERENCE

[1] S. H. Hall and H. L. Heck, Advanced Signal Integrity for High-Speed Digital System Design, Hoboken, NJ: Wiley, 2009.

[2] F. Jun, Y. Xiaoning, J. Kim, B. Archambeault, and A. Orlandi, “Signal integrity design for high-speed digital circuits: Progress and directions,” IEEE Trans.

Electromagn. Compat., vol. 52, no. 2, pp. 392–400, May 2010.

[3] T.-L. Wu, F. Buesink, and F. Canavero, “Overview of signal integrity and EMC design technologies on PCB: fundamentals and latest progress,” IEEE Trans.

Electromagn. Compat., vol. 55, no. 4, pp. 624–638, Aug. 2013.

[4] Y. Kayano, Y. Tsuda, and H. Inoue, “Identifying EM radiation from asymmetrical differential-paired lines with equi-distance routing,” in Proc. IEEE Int. Symp.

Electromagn. Compat., Pittsburgh, PA, Aug. 2012, pp. 311–316.

[5] L. O. Hoefi, J. L. Knighten, J. T. DiBene II, and M. W. Fogg, “Spectral analysis of common mode currents on fibre channel cable shields due to skew imbalance of differential signals operating at 1.0625 Gb/s,” in Proc. IEEE Int. Symp.

Electromagn. Compat., Denver, CO, Aug. 1998, pp. 527–531.

[6] J. L. Knighten, N. W. Smith, J. T. DiBene II, and L. O. Hoeft, “EMI common-mode current dependence on delay skew imbalance in high speed differential transmission lines operating at 1 gigabit/second data rates,” in Proc. IEEE Int.

Symp. Quality Electron. Design, San Jose, CA, Mar. 2000, pp. 309–313.

[7] D. M. Hockanson, J. L. Drewniak, T. H. Hubing, T. P. Van Doren, F. Sha, and M.

Wilhelm, “Investigation of fundamental EMI source mechanisms driving common-mode radiation from printed circuit boards with attached cables,” IEEE Trans. Electromagn. Compat., vol. 38, no. 4, pp. 557–566, Nov. 1996.

[8] D. M. Hockanson, J. L. Drewniak, T. H. Hubing, T. P. Van Doren, F. Sha, C.-W.

Lam, and L. Rubin, “Quantifying EMI resulting from finite-impedance reference planes,” IEEE Trans. Electromagn. Compat., vol. 39, no. 4, pp. 286–297, Nov.

1997.

[9] D. M. Hockanson, X. Ye, J. L. Drewniak, T. H. Hubing, T. P. Van Doren, and R. E.

DuBroff, “FDTD and experimental investigation of EMI from stacked-card PCB configurations,” IEEE Trans. Electromagn. Compat., vol. 43, no. 1, pp. 1–10, Feb.

2001.

[10] Y. Kayano, M. Tanaka and H. Inoue, “An equivalent circuit model for predicting EM radiation from a PCB driven by a connected feed cable,” in Proc. IEEE Int.

Symp. Electromagn. Compat., Portland, OR, Aug. 2006, pp.166–171.

[11] M. Leone, and V. Navratil, “On the electromagnetic radiation of printed-circuit-board interconnections,” IEEE Trans. Electromagn. Compat., vol.

47, no. 2, pp. 219–226, May 2005.

[12] P. Fornberg, M. Kanda, C. Lasek, M. Piket-May, and S. H. Stephen, “The impact of a nonideal return path on differential signal integrity,” IEEE Trans. Electromagn.

Compat., vol. 44, no. 1, pp. 11–15, Feb. 2002.

[13] B. Archambeault, S. Connor, and J. Diepenbrock, “EMI emissions from mismatch in high-speed differential signal trace and cables,” in Proc. IEEE Int. Electromagn.

Compat. Symp., Honolulu, HI, Jul. 2007, pp. 1–6.

[14] X. Duan, B. Archambeault, H.-D. Bruens, and C. Schuster, “EM emission of differential signals across connected printed circuit boards in the GHz range,” in Proc. IEEE Int. Electromagn. Compat. Symp., Austin, TX, Aug. 2009, pp. 50–55.

[15] C. R. Paul, Introduction to Electromagnetic Compatibility, 2nd ed., New York:

Wiely, 2006.

common-mode-current prediction method based on imbalance difference model for single-channel differential signaling system,” in Proc. Asia-Pacific Symp.

Electromagn. Compat., Singapore, Singapore, May 2012, pp. 21-24.

[17] C. Su, and T. H. Hubing, “Calculating radiated emissions due to I/O line coupling on printed circuit boards using the imbalance difference method,” IEEE Trans.

Electromagn. Compat., vol. 54, no. 1, pp. 212–217, Feb. 2012.

[18] K. Sejima, Y. Toyata, K. Iokibe, L. R. Koga, T. Watanabe, “Experimental model validation of mode-conversion sources introduced to modal equivalent circuit,” in Proc. IEEE Int. Electromagn. Compat. Symp., Pittsburgh, PA, Aug. 2012, pp. 492–

497.

[19] L. Niu, and T. H. Hubing, “Rigorous derivation of imbalance difference theory for modeling radiated emission problems,” IEEE Trans. Electromagn. Compat., vol.

57, no. 5, pp. 1027–1026, Oct. 2015.

[20] G.-H. Shiue, W.-D. Guo, C.-M. Lin, and R.-B. Wu, “Noise reduction using compensation capacitance for bend discontinuities of differential transmission lines,” IEEE Trans. Adv. Packag., vol. 29, no. 3, pp. 560–569, Aug. 2006.

[21] T. Matsushima, and O. Wada, “A method of common-mode reduction based on imbalance difference model for differential transmission line bend,” in Proc. IEEE Int. Symp. Eur. Electromagn. Compat., Brugge, Belgium, Sep. 2013, pp. 338–341.

[22] C.-H. Chang, R.-Y. Fang, and C.-L. Wang, “Bended differential transmission line using compensation inductance for common-mode noise suppression,” IEEE Trans.

Compon. Packag. Manuf. Technol., vol. 2, no. 9, pp. 1518–1525, Sep. 2012.

[23] C. Gazda, D. Vande Ginste, H. Rogier, R.-B. Wu, and D. De Zutter, “A wideband common-mode suppression filter for bend discontinuities in differential signaling

using tightly coupled microstrips,” IEEE Trans. Adv. Packag., vol. 33, no. 4, pp.

969–978, Nov. 2010.

[24] G.-H. Shiue, Y.-H. Kao, C.-M. Hsu, Y.-C. Tsai, and D. V. Ginste, “Common-mode noise reduction schemes for weakly coupled differential serpentine delay microstrip lines,” IEEE Trans. Compon. Packag. Manuf. Technol., vol. 3, no. 6, pp.

1016–1027, Jun. 2013.

[25] K. Yanagisawa, F. Zhang, T. Sato, K. Yanagisawa, and Y. Miura, “A new wideband common-mode noise filter consisting of Mn-Zn ferrite core and copper/polyimide tape wound coil,” IEEE Trans. Magn., vol. 41, no. 10, pp. 3571–3573, Oct. 2005.

[26] R. Lai, Y. Maillet, F. Wang, S. Wang, R. Burgos, and D. Boroyevich, “An integrated EMI chokes for differential-mode and common-mode noise suppression,”

IEEE Trans. Power Electron. Lett., vol. 25, no. 3, pp. 539–544, Mar. 2010.

[27] N. Oka, K. Misu, S. Yoneda, S. Saito, and S. Nitta, “A common mode noise filter for high speed and wide band differential mode signal transmission,” in Proc.

IEEE Int. Symp. Eur. Electromagn. Compat., York, UK, Sep. 2011, pp. 732–736.

[28] W.-T. Liu, C.-H. Tsai, T.-W. Han, and T.-L. Wu, “An embedded common mode suppression filter for GHz differential signals using periodic defected ground plane,” IEEE Microw. Compon. Lett., vol. 18, no. 4, pp. 248–250, Apr. 2008.

[29] S.-J. Wu, C.-H, Tsai, and T.-L. Wu “A novel wideband common-mode suppression filter for GHz differential signals using coupled patterned ground structure,” IEEE Trans. Microw. Theory Techn., vol. 57, no.4, pp. 848–855, Apr. 2009.

[30] J. Naqui, A. Fernández-Prieto, M. Durán-Sindreu, F. Mesa, J. Martel, F. Medina, and F. Martín, “Common-mode suppression in microstrip differential lines by means of complementary split ring resonators: Theory and applications,” IEEE

[30] J. Naqui, A. Fernández-Prieto, M. Durán-Sindreu, F. Mesa, J. Martel, F. Medina, and F. Martín, “Common-mode suppression in microstrip differential lines by means of complementary split ring resonators: Theory and applications,” IEEE

在文檔中 高速差動傳輸介面中共模衍生電磁暨射頻干擾之預測與消除 (頁 142-0)