Switchable DBR Filters Using Semiconductor Distributed Doped Areas (ScDDAs)

: This paper presents a novel way to switch dual-behavior resonator (DBR) ﬁlters without any additional active surface-mount components. By using a semiconductor substrate, we were able to simultaneously co-design the ﬁlters and semiconductor distributed doped areas (ScDDAs) with integrated N + PP + junctions as active elements. These ScDDAs act as electrical vias in the substrate, which makes it possible to have an open-circuited resonator in the OFF state and a short-circuited resonator in the ON state, and, consequently, to control the transmission zeroes of the ﬁlters. This method o ﬀ ers degrees of freedom as the dimensions and positions of these doped areas can be chosen to obtain the best performances. In this study, four ﬁlters were simulated and fabricated to spotlight di ﬀ erent possibilities for the dimensions and positions of the ScDDA to control the low-or high-frequency transmission zero of the ﬁlters. The simulations were in very good agreement with the measured results. All the ﬁlters present insertion losses lower than 2 dB in the OFF and ON states, a great ﬂexibility in the frequency choice, and good agility compared with the state of the art.


Introduction
In the era of the Internet of Things and Internet of Everything, communication systems are omnipresent. It has never been as complicated as it is today, therefore, to obtain satisfactory tradeoffs between electrical and thermal performances, integrability, and cost. A system has to support increasing numbers of applications and its components such as antennas or filters consequently need reconfigurable functionalities. Filters have to select each desired frequency band related to the targeted applications. There exist many discrete reconfigurable filters making it possible to switch a filter from bandpass to bandstop [1][2][3][4], between different bandwidths [5][6][7] or between different frequency bands, sometimes with multi-states, i.e., with a combination of states of several active elements [8][9][10][11][12][13][14][15][16]. However, if active surface-mount components (SMCs) such as PIN diodes, RF MEMS or varactor diodes are used to effect such changes, the filters can have issues related to parasitic effects induced by these active components and their associated bias networks. Whole filter performances may thus be decreased by these parasitic effects, which are potentially exacerbated as the frequency increases.
In order to counter these issues, reconfigurable RF components can be co-designed using semiconductor distributed doped areas (ScDDAs), such as in [23][24][25][26][27]. By designing a tunable RF device with ScDDAs, we can obtain a tunability comparable to that of classical technology with SMCs while avoiding the need for other components. Indeed, ScDDAs are semiconductor junctions acting as electrical vias in the substrate thickness, i.e., they are integrated active components in the substrate. This method offers flexibility in terms of dimensions and positions of the active elements and avoids the drawbacks of adding SMCs.
In this context, the aim of the present paper is to propose narrow-band filters such as DBR ones using ScDDAs. This offers filter designers more possibilities for the control of the frequency zeroes and means to achieve good tradeoffs in terms of electrical performances, integrability, and cost.
In this paper, Section 2 deals with the switchable DBR concept. Then, Section 3 explains the cosimulation method and the measured results on a first demonstrator. Finally, in Section 4, the flexibility and degrees of freedom are discussed and compared with the state of the art and three other DBRs.

Switchable DBR Theory
DBRs are based on the parallel association of two bandstop structures that can be two openended stubs, a quarter wavelength long, at a specific frequency (Figure 1a). Each resonator has its own specific length to give a transmission zero. By appropriately selecting these transmission zeroes (TZ), a bandpass filter can be obtained.
Thus, the longer resonator in Figure 1a is for the low-frequency (LF1) transmission zero and the shorter one is for the high-frequency (HF1) transmission zero. Figure 1b shows an example of simplified simulated results, i.e., without end effects or dielectric losses, obtained using the Advanced Design Systems (ADS) electronics software from Keysight Technologies©, where the two TZ make a 5.7 GHz bandpass filter possible. Then, if one resonator is short-circuited, this modifies the transmission zero and, consequently, the central frequency of the bandpass filter. By using one active element, there are two alternative ways to create a switchable DBR. The first option is to short-circuit a stub as shown in Figure 2a; the LF transmission zero LF1 becomes LF2, thus modifying the central frequency from 5.7 GHz to 3.2 GHz (Figure 2b) in the OFF and ON states, Thus, the longer resonator in Figure 1a is for the low-frequency (LF 1 ) transmission zero and the shorter one is for the high-frequency (HF 1 ) transmission zero. Figure 1b shows an example of simplified simulated results, i.e., without end effects or dielectric losses, obtained using the Advanced Design Systems (ADS) electronics software from Keysight Technologies©, where the two TZ make a 5.7 GHz bandpass filter possible. Then, if one resonator is short-circuited, this modifies the transmission zero and, consequently, the central frequency of the bandpass filter. By using one active element, there are two alternative ways to create a switchable DBR. The first option is to short-circuit a stub as shown in Figure 2a; the LF transmission zero LF 1 becomes LF 2 , thus modifying the central frequency from 5.7 GHz to 3.2 GHz (Figure 2b) in the OFF and ON states, respectively. The second option is to short-circuit the high-frequency resonator ( Figure 3a); the high-frequency transmission zero HF 1 becomes HF 2 , which makes it possible to switch the central frequency from 5.7 GHz in the OFF state to 2.5 GHz in the ON state (Figure 3b).
Electronics 2020, 9, x FOR PEER REVIEW  3 of 11 respectively. The second option is to short-circuit the high-frequency resonator ( Figure 3a); the highfrequency transmission zero HF1 becomes HF2, which makes it possible to switch the central frequency from 5.7 GHz in the OFF state to 2.5 GHz in the ON state (Figure 3b).

Switchable DBR Demonstrator
Based on the switchable DBR concept, our objective here was to co-design a DBR and its active element on a high-resistivity silicon (HR-Si) substrate. Indeed, a high resistivity was chosen to minimize the losses of the propagating waves in the substrate. Furthermore, the advantage of using this kind of substrate, i.e., a semiconductor, is that it makes it possible to design transmission lines and ScDDAs at the same time and thus to co-design a switchable DBR. We can, therefore, optimize electrical performances of the two states and enhance the integrability by using a low-cost technology compatible with mass-production.

Design and Modeling
A P-type silicon substrate was chosen with a 675 µm thickness and doped with boron with a

Switchable DBR Demonstrator
Based on the switchable DBR concept, our objective here was to co-design a DBR and its active element on a high-resistivity silicon (HR-Si) substrate. Indeed, a high resistivity was chosen to minimize the losses of the propagating waves in the substrate. Furthermore, the advantage of using this kind of substrate, i.e., a semiconductor, is that it makes it possible to design transmission lines and ScDDAs at the same time and thus to co-design a switchable DBR. We can, therefore, optimize electrical performances of the two states and enhance the integrability by using a low-cost technology compatible with mass-production.

Design and Modeling
A P-type silicon substrate was chosen with a 675 µm thickness and doped with boron with a resistivity of 2500 Ω·cm. The active element was an N + PP + junction with a surface doping of around

Switchable DBR Demonstrator
Based on the switchable DBR concept, our objective here was to co-design a DBR and its active element on a high-resistivity silicon (HR-Si) substrate. Indeed, a high resistivity was chosen to minimize the losses of the propagating waves in the substrate. Furthermore, the advantage of using this kind of substrate, i.e., a semiconductor, is that it makes it possible to design transmission lines and ScDDAs at the same time and thus to co-design a switchable DBR. We can, therefore, optimize electrical performances of the two states and enhance the integrability by using a low-cost technology compatible with mass-production.

Design and Modeling
A P-type silicon substrate was chosen with a 675 µm thickness and doped with boron with a resistivity of 2500 Ω·cm. The active element was an N + PP + junction with a surface doping of around 3 × 10 19 atoms/cm 3 for the two N + and P + regions and doping depths of around 3 µm. Figure 4 shows the switchable DBR design with its integrated active element, located at the end of the HF resonator. The length × width dimensions of the low-and high-frequency stubs are noted Lstub HF × W HF and Lstub LF × W LF , respectively, and the dimensions of the doped area are noted L DOP × W DOP . The lengths and widths of the two resonators were calculated to obtain a low-frequency transmission zero at 4.8 GHz and a high-frequency transmission zero at 7.2 GHz, based on the synthesis in [17]. The access line widths were dimensioned to have a 50 Ω characteristic impedance and their lengths were chosen to be sufficiently long to be easily measured. The DBR dimensions are listed in Table 1. ctronics 2020, 9, x FOR PEER REVIEW 4 of tubLF × WLF, respectively, and the dimensions of the doped area are noted LDOP × WDOP. The length d widths of the two resonators were calculated to obtain a low-frequency transmission zero 8 GHz and a high-frequency transmission zero at 7.2 GHz, based on the synthesis in [17]. The acce e widths were dimensioned to have a 50 Ω characteristic impedance and their lengths were chose be sufficiently long to be easily measured. The DBR dimensions are listed in Table 1.  This demonstrator was simulated using an HFSS TM electromagnetic simulator to predict i havior. The semiconductor losses were taken into account using the substrate resistivity in th lculation loss tangent as follows: is the vacuum dielectric permittivity and e silicon dielectric permittivity, equal to 11.9.
The active element was simulated using two 3 µm deep layers of 7.1 × 10 5 S/m conductivit  This demonstrator was simulated using an HFSS TM electromagnetic simulator to predict its behavior. The semiconductor losses were taken into account using the substrate resistivity in the calculation loss tangent as follows: where ρ is the resistivity, ω is equal to 2π f req, ε 0 is the vacuum dielectric permittivity and ε r is the silicon dielectric permittivity, equal to 11.9. The active element was simulated using two 3 µm deep layers of 7.1 × 10 5 S/m conductivity, corresponding to the conductivity of heavily doped areas. Between these two layers, i.e., between the top and the bottom sides, within the substrate itself, a resistivity of 2500 Ω·cm in the OFF state and a resistivity of 0.1 Ω·cm in the ON state, which was estimated using Atlas TM from Silvaco © when the junction was forward biased. Figure 5 presents the electromagnetic simulated results of the switchable DBR 1 demonstrator in OFF and ON states. In the OFF state, the two transmission zeroes at 4.8 GHz and 7.2 GHz make it possible to obtain a bandpass filter at 5.52 GHz. When the N + PP + is simulated in forward bias, the high-frequency transmission zero is switched to DC allowing a transmission frequency band at 2.44 GHz. The simulated insertion losses are 0.9 dB at 5.52 GHz in the OFF state and 2.01 dB at 2.44 GHz in the ON state. Electronics 2020, 9, x FOR PEER REVIEW 5 of 11 Figure 5. HFSS TM -simulated results of switchable DBR1 in the OFF and ON states.

Fabrication and Measurements
This demonstrator was fabricated with only two masks: one for the doping steps and one for the metallization steps of the upper side. The manufacturing steps are detailed in [27]. Figure 6 shows a photograph of the switchable DBR1 demonstrator, placed between two SMA connectors for measurement. The two RF cables and DC source were connected to an R&S ® ZVA 67 Vector Network Analyzer (VNA). The DC bias voltage was applied with the RF signal. Because the DC ground was connected to the RF ground, a negative voltage was required to forward bias the N + PP + junction. A Short Open Load through (SOLT) calibration was performed to remove the losses of the cables but not the losses from the SMA connectors. The measured results are presented in Figure 7. In the OFF state, with a zero-bias voltage, the two transmission zeroes are measured at 4.83 GHz and 7.2 GHz, which implies a central frequency at 5.58 GHz with an insertion loss level of 1.97 dB. In the ON state, the lowest frequency transmission zero stays relatively constant, whereas the high-frequency transmission zero is moved to the DC frequency so the central frequency is switched to 2.55 GHz with a bias voltage of −1.5 V. The insertion loss level is then 1.9 dB.

Fabrication and Measurements
This demonstrator was fabricated with only two masks: one for the doping steps and one for the metallization steps of the upper side. The manufacturing steps are detailed in [27]. Figure 6 shows a photograph of the switchable DBR 1 demonstrator, placed between two SMA connectors for measurement. The two RF cables and DC source were connected to an R&S ® ZVA 67 Vector Network Analyzer (VNA). The DC bias voltage was applied with the RF signal. Because the DC ground was connected to the RF ground, a negative voltage was required to forward bias the N + PP + junction. A Short Open Load through (SOLT) calibration was performed to remove the losses of the cables but not the losses from the SMA connectors.

Fabrication and Measurements
This demonstrator was fabricated with only two masks: one for the doping steps and one for the metallization steps of the upper side. The manufacturing steps are detailed in [27]. Figure 6 shows a photograph of the switchable DBR1 demonstrator, placed between two SMA connectors for measurement. The two RF cables and DC source were connected to an R&S ® ZVA 67 Vector Network Analyzer (VNA). The DC bias voltage was applied with the RF signal. Because the DC ground was connected to the RF ground, a negative voltage was required to forward bias the N + PP + junction. A Short Open Load through (SOLT) calibration was performed to remove the losses of the cables but not the losses from the SMA connectors. The measured results are presented in Figure 7. In the OFF state, with a zero-bias voltage, the two transmission zeroes are measured at 4.83 GHz and 7.2 GHz, which implies a central frequency at 5.58 GHz with an insertion loss level of 1.97 dB. In the ON state, the lowest frequency transmission zero stays relatively constant, whereas the high-frequency transmission zero is moved to the DC frequency so the central frequency is switched to 2.55 GHz with a bias voltage of −1.5 V. The insertion loss level is then 1.9 dB. The measured results are presented in Figure 7. In the OFF state, with a zero-bias voltage, the two transmission zeroes are measured at 4.83 GHz and 7.2 GHz, which implies a central frequency at 5.58 GHz with an insertion loss level of 1.97 dB. In the ON state, the lowest frequency transmission zero stays relatively constant, whereas the high-frequency transmission zero is moved to the DC frequency so the central frequency is switched to 2.55 GHz with a bias voltage of −1.5 V. The insertion loss level is then 1.9 dB. two transmission zeroes are measured at 4.83 GHz and 7.2 GHz, which implies a central frequency at 5.58 GHz with an insertion loss level of 1.97 dB. In the ON state, the lowest frequency transmission zero stays relatively constant, whereas the high-frequency transmission zero is moved to the DC frequency so the central frequency is switched to 2.55 GHz with a bias voltage of −1.5 V. The insertion loss level is then 1.9 dB.   A good agreement was obtained overall, with slight differences that could be due to the substrate resistivity (given by the manufacturer as between 1 kΩ·cm and 10 kΩ·cm), the SMA connector losses themselves and the losses related to connection defects, i.e., the gap that exists between the connector and the substrate because these are separate elements.
Electronics 2020, 9, x FOR PEER REVIEW 6 of 11 Figure 8a,b show comparisons of the simulated and measured results in the OFF and ON states. A good agreement was obtained overall, with slight differences that could be due to the substrate resistivity (given by the manufacturer as between 1 kΩ·cm and 10 kΩ·cm), the SMA connector losses themselves and the losses related to connection defects, i.e., the gap that exists between the connector and the substrate because these are separate elements.

Discussion
The co-design approach used in the present study offers great flexibility and accuracy for the dimensioning and positioning of the doped areas, thanks to the semiconductor process. In order to show an overview of the possibilities, three other demonstrators (Figure 9) with different doped lengths and widths were designed and characterized. These switchable DBRs had the same metal layout as the first demonstrator DBR1, only the dimensions (listed in Table 2) of the doped areas (located at the end of the resonator) were modified.

Discussion
The co-design approach used in the present study offers great flexibility and accuracy for the dimensioning and positioning of the doped areas, thanks to the semiconductor process. In order to show an overview of the possibilities, three other demonstrators (Figure 9) with different doped lengths and widths were designed and characterized. These switchable DBRs had the same metal layout as the first demonstrator DBR 1 , only the dimensions (listed in Table 2) of the doped areas (located at the end of the resonator) were modified. dimensioning and positioning of the doped areas, thanks to the semiconductor process. In order to show an overview of the possibilities, three other demonstrators (Figure 9) with different doped lengths and widths were designed and characterized. These switchable DBRs had the same metal layout as the first demonstrator DBR1, only the dimensions (listed in Table 2) of the doped areas (located at the end of the resonator) were modified.     Figure 10a,b show comparisons of the simulated and measured results of switchable DBR 2 in the OFF and ON states. As for DBR 1 , a good agreement was obtained overall. It has a longer ScDDA than DBR 1 on the HF stub, with 1 mm length. Therefore, compared with DBR 1 , the central frequency is the same in the OFF state, i.e., equal to 5.53 GHz, with a 0 V bias voltage. The insertion loss level is 1.97 dB. In the ON state, with a −1.2 V bias voltage, the highest transmission zero is moved to DC and the transmission frequency band is at 2.6 GHz, with an insertion loss level of 1.95 dB. The bias voltage is lower than for DBR 1 because the doped area is longer.
Electronics 2020, 9, x FOR PEER REVIEW 7 of 11 Figure 10a,b show comparisons of the simulated and measured results of switchable DBR2 in the OFF and ON states. As for DBR1, a good agreement was obtained overall. It has a longer ScDDA than DBR1 on the HF stub, with 1 mm length. Therefore, compared with DBR1, the central frequency is the same in the OFF state, i.e., equal to 5.53 GHz, with a 0 V bias voltage. The insertion loss level is 1.97 dB. In the ON state, with a −1.2 V bias voltage, the highest transmission zero is moved to DC and the transmission frequency band is at 2.6 GHz, with an insertion loss level of 1.95 dB. The bias voltage is lower than for DBR1 because the doped area is longer. Switchable DBR3 has an ScDDA with a doped length of 0.2 mm, located at the end of the LF resonator. Figure 11a  Switchable DBR 3 has an ScDDA with a doped length of 0.2 mm, located at the end of the LF resonator. Figure 11a  Switchable DBR3 has an ScDDA with a doped length of 0.2 mm, located at the end of the LF resonator. Figure 11a The last demonstrator, DBR4, has a wider and longer ScDDA than DBR3. This implies a capacitive effect in the OFF state, which explains why the resonant frequency, at 5.2 GHz, is lower than for the other demonstrators (Figure 12a). It also has a shorter resonator in the ON state, which gives a resonant frequency of 3.8 GHz in the ON state (higher than with DBR3) (Figure 12b), with a bias voltage of −1 V. The insertion losses are 1.94 dB and 1.95 dB in the OFF and ON states, respectively. The last demonstrator, DBR 4 , has a wider and longer ScDDA than DBR 3 . This implies a capacitive effect in the OFF state, which explains why the resonant frequency, at 5.2 GHz, is lower than for the other demonstrators (Figure 12a). It also has a shorter resonator in the ON state, which gives a resonant frequency of 3.8 GHz in the ON state (higher than with DBR 3 ) (Figure 12b Moreover, with a 5.52 GHz filter in the OFF state, the central frequency in the ON state can be selected between 2.44 GHz and 5 GHz by choosing the length of the doped area and short-circuiting the LF-or HF-stub ( Figure 13). The ratio can be between 1:1.1 and 1:2.25, which offers many possibilities depending on the application. The greater the doped area surface, the lower the bias voltage. Thus, if the bias voltage is not an issue, the size can be minimized while maintaining the switched frequency. This can be a good solution in the case of multiple states, such as in [25]. Moreover, with a 5.52 GHz filter in the OFF state, the central frequency in the ON state can be selected between 2.44 GHz and 5 GHz by choosing the length of the doped area and short-circuiting the LF-or HF-stub ( Figure 13). The ratio can be between 1:1.1 and 1:2.25, which offers many possibilities depending on the application. The greater the doped area surface, the lower the bias voltage. Thus, if the bias voltage is not an issue, the size can be minimized while maintaining the switched frequency. This can be a good solution in the case of multiple states, such as in [25]. selected between 2.44 GHz and 5 GHz by choosing the length of the doped area and short-circuiting the LF-or HF-stub ( Figure 13). The ratio can be between 1:1.1 and 1:2.25, which offers many possibilities depending on the application. The greater the doped area surface, the lower the bias voltage. Thus, if the bias voltage is not an issue, the size can be minimized while maintaining the switched frequency. This can be a good solution in the case of multiple states, such as in [25].  Table 3 shows a comparison between the state-of-the-art and the results of the present study. Our work shows a great agility, with the best tradeoff between the highest frequency ratio and good performances, i.e., low losses, without any additional components.  Table 3 shows a comparison between the state-of-the-art and the results of the present study. Our work shows a great agility, with the best tradeoff between the highest frequency ratio and good performances, i.e., low losses, without any additional components. Although devices of this kind present measurement difficulties due to their fragility and size and, therefore, require measurement by SMA connectors without the possibility of soldering, the devices tested here show good performances in both states, with the same losses overall. Indeed, in the ON state, even though the demonstrators have different surface areas, their equivalent resistance values can be roughly the same with different bias voltages.

Conclusions
This paper shows a novel method for switching DBR filters without any additional components. Four demonstrators were characterized, offering a large range of reconfigurability without sacrificing performances, i.e., a low switching voltage, low losses, and a high level of integrability, all obtained with a well-known manufacturing process (the same as for semiconductor components), with a reduced number of masks and steps. Such a co-design offers flexibility in terms of positioning and dimensioning of the ScDDAs, which implies a good agility, with a large range of choice for the ratio: between 1:1.1 and 1:2.25.