Transfer printing for RF applications
Monolithic integration of GaN and SOI technologies via micro-transfer printing addresses the challenges of MMIC circuits by combining their advantages, reducing complexity and cost while enhancing performance and power management.
Patent Information
- Application Number
- FR2022000952
- Authority / Receiving Office
- FR · FR
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2019-05-13
- Filing Date
- 2022-02-03
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2040-05-13
AI Technical Summary
Existing MMIC circuits face challenges in simultaneously achieving high linearity, high power, low noise figure, and low energy consumption, which are often contradictory, leading to the use of system-in-package techniques that increase complexity and cost.
Monolithic integration of GaN and SOI technologies through micro-transfer printing, allowing for direct interconnections between stacked GaN and SOI transistors, enhancing performance by combining the advantages of both technologies.
This approach reduces parasitic elements, improves voltage breakdown, and achieves better power management, noise factor, and insulation, resulting in a smaller, less complex, and cost-effective RF circuit with improved linearity and power handling capabilities.
Smart Images

Figure 00000016_0000 
Figure 00000016_0001 
Figure 00000017_0000
Abstract
Description
Title of the invention: Transfer printing for RF applications
[0001] The present invention relates to transfer printing for RF applications and in particular to micro-transfer printing of GaN components on an SOI wafer.
[0002] GaN technology is relatively new compared to other semiconductors, such as Si and GaAs, but it has become the technology of choice for high-frequency and power-intensive applications, such as those required to transmit signals over long distances or at high power levels. GaN transistors offer high power density, high operating temperatures, improved efficiency, low on-state resistance, and can be used in various frequency bands from 1 GHz to 110 GHz. Figure 1 shows the schematic diagram of a cross-section of a GaN transistor 2. The GaN layer 4 is formed on a high-resistivity (HR) silicon substrate 6 and an AlGaN buffer layer 8. The heavily doped n1-l1 regions 10 of the GaN are connected to metallic contacts 12 (drain and source).The grid 14 is isolated from the GaN 4 layer by a thin layer of AIN 16.
[0003] GaN on SiC combines the high power density capabilities of GaN with the superior thermal conductivity and low RF losses of SiC (silicon carbide). GaN on SiC is the combination of choice for achieving good RF performance with high power density.
[0004] The GaN-on-Si combination is associated with significantly lower thermal performance and higher RF losses, but is less expensive. GaN-on-Si is the combination of choice for price-sensitive power electronics applications.
[0005] Silicon-on-insulator (SOI) technology is used in semiconductor manufacturing, particularly in microelectronics, to reduce parasitic capacitance through a silicon-insulator-silicon layered substrate structure (rather than bulk silicon). CMOS SOI technology reduces power consumption in mixed analog and digital RF circuits when designing low-power RF transceivers due to its low leakage capacitance.
[0006] To obtain some of the advantages of GaN and SOI in high-power applications, attempts have been made to grow GaN on SOI using trench etching to isolate the GaN transistors. Careful deformation machining is necessary to control the stress accumulating in the wafer during growth.
[0007] Alternatively, microtransfer printing [1] can be used to transfer a GaN device from a native / source wafer to a target SOI wafer. During this process, the GaN device is detached / removed from the native wafer on which it was grown and is reattached to the SOI wafer.
[0008] 1] R. Lener, et al., “Heterogeneous Integration of Microscale Gallium Nitride Transistors by Micro-Transfer-Printing". 2016 IEEE 66th Electronic Components and Technology Conference, pp 1186-1189.
[0009] Certain aspects of the present invention relate to semiconductor structures for RF applications and manufacturing processes as presented in the attached claims.
[0010] Preferred embodiments of the invention will be described below by way of non-limiting example and with reference to the attached drawings.
[0011] [Fig. 1] The [Fig. 1] represents a schematic cross-section of a GaN transistor on a silicon substrate;
[0012] [Fig.2a] The [Fig.2a] represents a schematic diagram of a monolithic hybrid switch branch according to one embodiment;
[0013] [Fig.2b] Fig.2b represents a schematic cross-section of a semiconductor structure according to one embodiment, the structure comprising a GaN transistor micro-printed by transfer onto an SOI wafer;
[0014] [Fig.3a] The [Fig.3a] represents the configuration of an SPDT (Single Pole Double Throw, or Monopolar Bidirectional) switch using hybrid branches according to one embodiment;
[0015] [Fig.3b] The [Fig.3b] represents the configuration of an SPDT using pure GaN branches according to one embodiment;
[0016] [Fig.4] The [Fig.4] represents a wiring diagram of a semiconductor structure according to one embodiment;
[0017] [Fig.5] The [Fig.5] represents a wiring diagram of a semiconductor structure according to another embodiment;
[0018] [Fig.6] The [Fig.6] represents a wiring diagram of a semiconductor structure according to another embodiment;
[0019] [Fig.7] The [Fig.7] represents a wiring diagram of an SPDT switch according to one embodiment;
[0020] [Fig.8] The [Fig.8] represents a wiring diagram of an SPDT switch according to another embodiment;
[0021] [Fig.9] The [Fig.9] represents a wiring diagram of an SPDT switch according to an embodiment comprising hybrid switching branches;
[0022] [Fig. 10] The [Fig. 10] represents a wiring diagram of an SPDT switch according to another embodiment comprising hybrid switching branches;
[0023] [Fig. 11] Figure 11 represents a schematic diagram of an amplification cell according to an embodiment;
[0024] [Fig. 12] Fig. 12 shows a wiring diagram including an amplifier low noise according to one embodiment;
[0025] [Fig. 13] [Fig. 13] shows graphs of S parameters on 2 ports simulated from a low-noise amplifier according to an embodiment, plotted as a function of frequency;
[0026] [Fig. 14] Figure 14 represents, on the left, the simulated noise factor (NF) and, on the right, the K factor of a low-noise amplifier according to the embodiment as a function of frequency;
[0027] [Fig. 15] Fig. 15 represents the upper and lower OIP3 as a function of the center frequency of a low-noise amplifier according to the embodiment;
[0028] [Fig. 16] Fig. 16 represents a wiring diagram of a semi-structure conductive according to an embodiment;
[0029] [Fig. 17] Fig. 17 represents a wiring diagram of a semi-structure conductive according to another embodiment;
[0030] [Fig. 18] Figure 18 shows a wiring diagram of a low-noise amplifier presenting a cascode structure according to an embodiment mode;
[0031] [Fig. 19] Fig. 19 represents a schematic diagram of a two-phase amplifier floors according to a method of implementation;
[0032] [Fig.20a] Fig.20a represents the electrical diagram of a two-phase amplifier floors;
[0033] [Fig.20b] Fig.20b represents the electrical diagram of a two-phase amplifier stages according to an embodiment using GaN transistors printed by microtransfer for the second stage;
[0034] [Fig. 21] Fig. 21 shows a wiring diagram of a low-noise amplifier according to an embodiment using microtransfer printed GaN transistors;
[0035] [Fig. 22] Fig. 22 shows a wiring diagram of a low-noise amplifier two stages according to an embodiment using GaN transistors printed by micro-transfer for the second stage;
[0036] [Fig.23] Fig.23 shows a wiring diagram of a low-noise amplifier two stages according to another embodiment using micro-transfer printed GaN transistors for the first and second stages;
[0037] [Fig.24] Fig.24 shows a wiring diagram of a low-noise amplifier N stages according to an embodiment using microtransfer printed GaN transistors; and
[0038] [Fig. 25] Fig. 25 shows a wiring diagram of a low-noise amplifier N stages according to another embodiment using an SOI cascode structure and microtransfer printed GaN transistors.
[0039] A growing number of applications today require high linearity, high power, low noise, and low power consumption. Since these requirements are often contradictory, they are very difficult to meet in Monolithic Microwave Integrated Circuits (MMICs). Therefore, it is common to use system-in-package (SiP) techniques to house several integrated circuits from different technologies in a single module, thus benefiting from the best of each. The chips are then connected to each other using connecting wires or bosses. This SiP technique enables system-level integration of multiple chips / technologies.
[0040] The embodiments described herein allow for the monolithic integration of GaN and SOI technologies by microtransfer printing. Both the CMOS devices and the high electron mobility transistors (HEMTs) are fabricated separately in their processing environments and can have different wafer diameters. The aim is to design most of the wiring using SOI technology and to print only a small number of HEMTs on the upper part of the SOI wafer in order to improve the overall performance of the RF circuit.
[0041] One advantage of this technique is that it uses a monolithic circuit, resulting in a smaller, less expensive, and less complex circuit than a packaged system. In addition to reducing complexity and cost, this heterogeneous design process allows for better control of parasitic elements and performance. Another advantage relates to the performance of the GaN transistor. The vertical isolation of existing GaN-on-silicon technology is poor. Microtransfer printing significantly reduces this vertical leakage and thus improves the voltage breakdown of HEMTs.
[0042] This heterogeneous design process makes it possible to combine the best of GaN and SOL II, thus achieving a good noise figure (NF), good isolation and very good power management capabilities in association with a high level of integration in a single chip.
[0043] Base station applications require switches capable of handling high power, ranging from a few watts (W) to over 100 W. Some existing switching solutions are achieved using PIN diodes, which are capable of handling high power. Their main drawback is their static power consumption, which reduces the overall energy efficiency of the system. GaN transistors could be promising candidates because they can handle high power and exhibit low static power consumption. (mainly due to gate leakage) as well as good linearity compared to MOS transistors.
[0044] Furthermore, SOI is the state-of-the-art technology for switching applications. SOI MOS transistors offer very good low-power performance with a low Ron x Coff (Ron x Coff, factor of merit, FOM, crucial for switches). SOI is a cost-effective technology for RF systems-on-chip (SoCs) intended for applications up to 10 GHz without significant performance loss, as well as for crosstalk reduction. To achieve such performance, it is necessary to use advanced SOI technology capable of maintaining a low voltage, the only drawback of this technology.
[0045] The power handling capacity can be improved by stacking several transistors, but it is limited due to the saturation effect of the stacked transistors: Zhu, Y., Klimashov, O., & Bartle, D. (November 2014). "Analytical model of voltage division inside stacked-FET switch". In 2014 Asia-Pacific Microwave Conference (pp. 750-752). IEEE.
[0046] According to one embodiment, microtransfer printing (pTP) of GaN onto a SOI wafer allows for monolithic designs with shorter interconnects, thus avoiding losses and transitions between chips and substrates. Interconnects are directly made between the two stacked chips using a metallic layer (e.g., thick copper), also called a redistribution layer (RDL). During microtransfer printing, one or more chips, each containing one or more GaN devices, are removed from the native silicon substrate and placed on the SOL wafer.
[0047] Some embodiments described herein can overcome the bottlenecks of SOI and GaN technologies by combining them using microprinting technology. In one embodiment, at least some of the existing problems are solved by a hybrid branch. This branch consists of one or more stacked GaN p-Hemts (high electron mobility transistors) and stacked SOI MOS devices.
[0048] Figure 2a shows a schematic diagram of a branch 18 of a monolithic hybrid switch. The branch 18 comprises two or more stacked SOI MOS transistors 20 connected in series with two or more stacked GaN p-Hemt transistors 22. The stacked SOI transistors 20 are driven by an SOI gate drive network 24, and the stacked GaN transistors are driven by a GaN gate drive network 26. In this configuration, the Coff is controlled by the lower value (SOI transistor), and the Ron is determined by the higher value (GaN transistor). It is possible to choose a compromise when sizing the two transistors to achieve a specified isolation, return loss, and insertion loss. The second parameter to consider is the voltage distribution between the transistors. The impedance of GaN is much lower than that of a MOS transistor in the OFF state. Large MOS transistors are therefore necessary to match the impedance of GaN in the OFF state. This combination of GaN and SOI technologies improves low-power performance compared to GaN alone and provides better high-power handling compared to SOI alone.
[0049] Figure 3a shows a top view of the configuration of a hybrid GaN and SOI concept 30, and Figure 3b shows, for comparison, a similar top view of a concept 32 using only GaN p-Hemt (also microprinted on an SOI wafer). Both configurations represent bidirectional single-pole switches (SPDTs). The operating frequency band of these concepts ranges from DC to 6 GHz. The power requirement is 47 dBm, and the dimensions are optimized to achieve the lowest possible insertion loss at 6 GHz. The upper thick copper layer 34 connects the GaN transistors to SOL components.
[0050] The simulated performance of the two SPDT concepts is shown in Table 1. The performance is established at the chip level and no destacking has been applied. The interconnects are simulated using Keysight's Momentum electromagnetic simulator, while the transistors are considered either as an electrical concept for GaN transistors or as an extracted view for SOI transistors. With the exception of return losses, the hybrid combination exhibits better performance across all considered operating bandwidths. Isolations are improved by an average of 5 dB.
[0051] Table 1 - Simulated performance of certain SPDT concepts Freq. (GHz) RL (dB) il (dB) Iso AàRF (dB) Iso RFàRF (dB) GaN Pure Hybrid GaN Pure Hybrid GaN Pure Hybrid GaN Pure Hybrid 1 43.1 32.8 0.47 0.49 24.3 29.6 24.8 30.1 3 35.5 25.9 0.58 0.58 16.2 21.2 16.7 21.7 6 27.6 19.9 0.94 0.76 10.5 15.7 11.1 16
[0052] Figures 4 to 9 show wiring diagrams of embodiments described herein. Similar or corresponding characteristic features of the embodiments shown in the various figures have been designated by the same reference numerals for ease of understanding. Figure 4 shows a GaN micro-printed transfer (pTP) transistor 40 (GaN transistor micro-printed on a SOI wafer) with a resistor 42 connected to the gate of the transistor 40. The resistor 42 is fabricated on the SOI and / or on the GaN micro-printed chip. The resistor produced on the SOI wafer can be connected to the GaN 40 transistor, after microprinting, by a thick redistribution layer (RDL) of copper.
[0053] Figure 5 shows two GaN transistors 40 connected in series using transfer microprinting. The two transistors 40 are connected to their respective resistors 42 on the SOI wafer (where "chip" refers to the chip cut after the wafer has been cut). Each resistor 42 may have the same or different resistances. Preferably, the two transistors 40 are transferred to the SOI wafer in a single processing step. Having two or more GaN transistors in series in a switching branch on an SOI wafer can improve switching performance by increasing the maximum voltage that can be handled. Each GaN transistor 40 should be able to handle a peak voltage of approximately 50 V. Ideally, two transistors in series should accept an RF voltage of 100 V, corresponding to 50 dBm / 100 W (with a 50 Q load).Having some margin on power handling can be advantageous, and one or more additional transistors can be added in series to reduce the drain-source voltage across the first two transistors. It is possible to increase the power handling capacity by stacking more GaN transistors.
[0054] Figure 6 shows a GaN transistor 40 micro-printed by transfer microprinting in series with one or more SOI transistors 44. The gate of the GaN transistor 40 is connected to a resistor 42, and the gates and bodies of the SOI transistors are connected to their respective resistors 42. The resistors 42 connected to the SOI transistors 44 are made of SOL. The resistor 42 connected to the GaN transistor 40 can be made of either GaN or SOL. The GaN transistor 40 can be connected to the SOI resistor 42 and to the SOI transistors 44 by a thick copper RDL layer.
[0055] Figure 7 shows a wiring diagram of an SPDT switch 46 according to one embodiment. The two switching branches 48 and 50 are connected to RF port 1 52 and RF port 2 54, respectively, and to an antenna 56. Each branch comprises three series-connected transfer microprinted GaN transistors 40.
[0056] All the gates are connected to a resistor 42 in parallel with a diode 58. The purpose of resistor 42 is to prevent any RF leakage from the drain or source to the gate, and also to provide a floating RF node at the gate so that the gate power level moves in sync with the drain or source. The purpose of diode 58 is to establish a path for the micro-printed GaN gate DC current. Without this diode 58, the gate DC current would lead to a voltage drop across the gate resistor Rg, and the voltage actually applied to the gate would be the DC bias modified by this voltage drop. This could cause a problem in the OFF state, where the gate DC current is not negligible. The gate voltage can be significantly altered. due to this voltage drop, so that the 40 transistors may be less open than expected.
[0057] All gates are biased using a bias block 60, which delivers a negative voltage to turn the switch to the OFF state. Micro-printed GaN transfer switches are normally in the ON state, so applying ground to the gate is sufficient to turn it ON. A bias block 62 is used to deliver the correct DC voltage to the correct gate. Other architectures may have 1, 3, or 4 arms / branches, etc., connected to respective RF ports.
[0058] Fig. 8 represents a wiring diagram similar to that of Fig. 7, but with only two transfer microprinted GaN transistors 40 in each switching branch 48 and 50. Each gate of the transistors 40 is connected to a resistor 42 in parallel with a diode 58.
[0059] Figure 9 shows a wiring diagram of an SPDT 46 switch comprising hybrid branches 64 and 66. Each branch comprises a transfer microprinted GaN transistor 40 and stacked SOI transistors 44. A second biasing block 68 and a second switching driver circuit 70 are used to drive the SOI transistors 44, although in other embodiments, a single biasing block and a single switching driver circuit may be used to drive all of the GaN transistors 40 and the SOI transistors 44. Figure 10 shows a wiring diagram of a similar SPDT switch 46, but with two transfer microprinted GaN transistors 40 in series in each branch 64 and 66.
[0060] Although each of Figures 7 to 10 illustrates a two-switching branch architecture, these concepts can be extended to include additional switching branches. By way of example, one embodiment may include four RF ports, each connected to a switching branch comprising one or more transfer-microprinted GaN transistors. The architecture can be generalized to an NPMT switch (with N antennas and M RF ports, i.e., N poles and M directions).
[0061] One embodiment includes a low-noise amplifier using heterogeneous integration of GaN technology with SOI technology to improve circuit linearity.
[0062] Figure 11 shows a cell 74 which can be used in a low-noise amplifier according to one embodiment for amplifying an RF signal. The cell 74 comprises a common-source nMOS SOI transistor 76 in series with a common-gate, transfer-microprinted GaN HEMT 78 forming a cascode structure (having a common-source transistor directly in series with a gate transistor) (municipality).
[0063] One advantage of this topology is the improved overall linearity of the SOI technology through the use of a discrete GaN HEMT. These GaN micro-printed transistors are capable of accepting a high supply voltage, the main weakness of silicon-based transistors. This allows the structure to develop a higher output power and therefore a higher IP3 (the FOM is used to characterize the linearity of the low-noise amplifier). The GaN HEMT is also used as a cascode because the IP3 of a system composed of n cascaded stages is imposed by the last stage, as shown in Equation 1.
[0064] This is why the use of GaN as a cascode transistor makes it possible to improve the linearity characteristics of the low-noise amplifier (OIP3, Third Order Output Intercept Point) compared to the use of an all-SOI low-noise amplifier.
[0065] Another reason for using this heterogeneous cell 74 is that it ensures low continuous power consumption compared to pure GaN designs. SOI technology allows for low leakage and a low-current MOS transistor, which can be particularly advantageous for this application.
[0066] Figure 12 represents a low-noise amplifier concept 80. This uses a A low-noise, degenerate amplifier architecture comprising cell 74 with the SOI transistor 76 and the micro-transfer-printed GaN transistor 78 shown in [Fig. 11] and used as an amplification cell. The various passive and active parts are mounted on a printed circuit board (PCB) 84, formed from SOI 82 (native SOI components), or micro-printed (for the micro-transfer-printed GaN transistor 78). The SOI chip can be flipped onto a PCB for characterization purposes.
[0067] The low-noise amplifier can meet the 5G NR standard (below 6 GHz). The operating band is from 5.9 GHz to 7.1 GHz and corresponds to a band not authorized in the United States. The supply voltage is set at 5.5 V and corresponds to the typical voltage of a base station. The gate voltage of the common-source transistor is set to comply with the IP3 specification of 28 dBm across the entire band. The common-gate bias voltage is then adjusted so that the drain voltage of the nMOS is set at 1.2 V, i.e., its typical value.
[0068] The circuit was simulated, and the simulation results for parameter S are shown in [Fig. 13], and the factors NF 86 and K 88 in [Fig. 14]. The continuous power consumption was 165 mW. The input and output are well matched at 50 Q within the band. The amplifier has a maximum gain of 16.6 dB within the band and a minimum isolation of -26 dB. The amplifier is unconditionally stable up to 10 GHz. NF is less than 1.7 dB within the band, with a minimum of 1.24 dB at 6.8 GHz.
[0069] The simulated results for OIP3 are shown in [Fig. 15]. The power of the two fundamental input tones was set at -30 dBm with a tone separation of 1 MHz. The upper OIP3 90 and the lower OIP3 92 are greater than 28 dBm from 6.35 GHz to 7.1 GHz. The maximum OIP3 is 30.6 dBm at 6.8 GHz.
[0070] The overall performance and specifications are summarized in Table 2 and given at 6.5 GHz, i.e. in the middle of the band.
[0071] Table 2 - Summary of DC performance, for small signals (SP) and noise (NF) Heterogeneous LNA Specifications Unit Power Supply, DC 110 165 mW Frequency 5.9-7.1 5.9-7.1 GHz Gain 16.2 >15 dB NF 1.28 1.7 dB SH -14.5 -10 dB S22 -41 -10 dB S12 -26 -20 dB OIP3 30.6 28 dBm
[0072] Table 3 compares the low-noise amplifier concept with existing products. Performance is equivalent, except for power consumption, which is reduced by a factor of 3 compared to analog devices and by a factor of 5 compared to the Qorvo [3]. The chip size is not representative as this is a demonstrator.
[0073] Table 3 - Comparison with existing devices Ref. product Technology Freq (GHz) Power (V) ldc| (mA) Gain [dB) Analog devices HMC392A [2] GaAs 3.5-7 5 59 17.2 Qorvo TGA2611 [3] GaN on SiC TQGaM25 2-6 10 110 25 Classic LNA GaN onXR013 Ojlprn 5.9-7.1 5.5 20 16.2
[0074] Table 3 continued Product Ref. Technology OIP3 (dBm) MF (dB) SH [dB] S22 (dB) Area (mm2) Analog devices HMC392A [2] GaAs 32.5 1.7 12 18 1.3 Qorvo TGA2511 [3] GaN on SIC TQGaN25 32 1.5 >10 >10 3.15 Classic LNA GaNsurXR013 0.1 pm 30.5 1.28 14.5 41.3 2.48
[0075] HMC982A Chip: Low-noise MMIC GaAs amplifier, 3.5-7.0 GHz Datasheet, https: / / www.analog.eom / en / products / hmc392a-die.html#product-documentation
[0076] Preliminary Product Data Sheet TGA2611 2-6GHz GaN LNA (LNA: low-noise amplifier), https: / / w
[0077] www.qorvo.com / products / p / TGA2611
[0078] Microprinting technology allows the co-integration of different technologies at the chip level. Circuit performance can be improved while maintaining a small footprint.
[0079] Figure 16 shows a cell 94 for a low-noise amplifier or power amplifier comprising two GaN p-HEMTs 96 in series on a single-chip SOI wafer / chip. The GaN transfer-microprinted transistors may be located on the same microchip or on separate microchips. Figure 17 shows a cell 94 for a low-noise amplifier comprising a GaN transfer-microprinted p-HEMT 96 with an inductor 98 connected to its source. The inductor 98 is fabricated on the SOI wafer and connected to the GaN HEMT by RDL.
[0080] Figure 18 shows a wiring diagram of a low-noise amplifier concept 80 comprising a cascode cell 74 with an SOI transistor 76 in series with a transfer-microprinted GaN p-Hemt 78 and a linearizer 100 between the two transistors to improve linearity. The diode in series with the capacitor in parallel with the resistor generates a nonlinear current at the third-order intermodulation frequency to cancel the MOS drain component at the same frequency. The linearizer 100 is implemented on the SOI wafer.
[0081] Another low-noise amplifier concept that complies with the 5G millimeter wave standard is described below. It is designed using two cascaded stages 102 and 104, as illustrated in [Fig. 19]. The IP3 of the overall system when cascading n stages is imposed by the last stage, as expressed in equation 1.
[0082] The use of GaN for the low-noise amplifier 104 of the second stage will thus improve the linearity performance of the low-noise amplifier (IIP3 and CPI) compared to the use of an all-SOI low-noise amplifier. For the first stage 102, an SOI transistor can be used to take advantage of its good per- noise-related standards.
[0083] To evaluate the advantage of using GaN in the second stage, two circuits were designed at 28 GHz:
[0084] • The first uses an all-self architecture for the first and second floors.
[0085] • For the second concept, a GaN transistor was used for the second stage.
[0086] The schematics of the two low-noise amplifiers are shown in Figures 20a and 20b, respectively. In [Fig. 20a], two cascoded stages 106 and 108 have been cascaded for the all-SOL circuit. Regarding the heterogeneous low-noise amplifier of [Fig. 20b], the first stage 110 is cascoded. Since the nominal drain-source voltage of the GaN transistor 112 must be higher than that of the SOI transistors, the second stage 114 is designed using a common-source architecture. The aim is to use the same power supply as the first stage 110.
[0087] The simulated DC performance for small signals (SP) and noise (NF) of the two versions of the low-noise amplifier are summarized in Table 4. The performance is similar between the two versions and both meet the specifications (i.e., the requirements of the 5G standard).
[0088] Table 4 - DC performance, for small signals (SP) and noise (NF), of two versions of the low-noise amplifier INA all SOI INA heterogeneous Specifications Unit Aim. CC 29.5 (2.4V * 12.3 mA) 29.5 (2.4V * 12.3 mA) 30 mW Frequency 27.5-30.5 27.5-30.5 27.5-30.5 GHz Gain at 29G 16.3 15.2 15 dB Max Gain 17.2 16.3 15 dB NF at 29G 3.7 3.4 4 dB SU -17.7 -15 -10 dB S22 -13.9 -23.7 -10 dB S12 -46.1 -20.9 -20 dB Kf >1.4 >1 >1
[0089] Two merit factors are used to quantify the linearity of the circuits:
[0090] • Input compression point CPI.
[0091] • Third IP3 input interception point.
[0092] The use of the GaN transistor 112 in the second stage 114 improves these parameters by nearly 10 dB. The simulated CPI of the heterogeneous solution is approximately -13 dBm compared to -21 dBm for the all-SOI low-noise amplifier. Regarding IP3, the simulation yielded 6 dBm for the heterogeneous low-noise amplifier and -10 dBm for the all-SOL version.
[0093] The low-noise amplifier designed with the micro-printed GaN transistor by transfer was compared to other low-noise amplifiers designed using different technologies different and exhibiting high integration capabilities (CMOS and SiGe). As can be seen in Table 5, the low-noise amplifier described here exhibits better linearity thanks to the use of the GaN transistor on SOI, with comparable performance for the other parameters.
[0094] Table 5 - Comparison of low-noise amplifiers Unit [4] [5] [6] [7] LNA Implementation Form Process 130 nm CMOS CMOS 40 nm ma ss 4 CMOS 90 nm SiGe 0.25 µm GaN / SOi 130 nm Topology 3-stage cascode 3-stage cascode 2-stage cascode 3-stage 2-stage Frequency GHz 27.2 27.1 28.5 29 29 Pdc mW 24 31.4 16.25 134 29.5 Gain dB 20 27.1 20 26.4 15.2 NF dB 5.2-7.5 3.3-4.3 2.9-4.2 24-3.5 3.4 CPI dBm - -21.6 ■17 ■14.9 ■13 IIP3 dBm - -12.6 -7.5 -5 6
[0095] J. Luo, et al. “A 28 GHz LNA using defected ground structure for 5G application”. Microw Opt Technol Lett. 2018;60:1067-1318.
[0096] M. Elkholy, et al. "A Wideband Variable Gain LNA With High OIP3 for 5G Using 40-nm Bulk CMOS". 2017 IEEE MICROWAVE AND WIRELESS COMPONENTS LETTERS.
[0097] E. Adabi, B. Heydari, M. Bohsali, and AM Niknejad, “30 GHz CMOS low noise amplifier,” in Proc. IEEE Radio Freq. Integrator Circuits (RFIC), June 2007, p. 625-628.
[0098] Q. Ma, DMW Leenaerts, and PGM Baltus, “Silicon-based truetime-delay phased-array front-ends at Ka-band,” IEEE Trans. Microw. Theory Techn. flight. 63, no. 9, pp. 2942-2952, September 2015.
[0099] Figure 21 shows a wiring diagram of an LNA 116 according to one embodiment. The low-noise amplifier 116 comprises a GaN transistor 112 micro-printed by transfer onto SOI connected to other SOI components. The source of the GaN transistor is connected to an inductor 118 fabricated on the SOI wafer.
[0100] Figure 22 shows a wiring diagram of a two-stage amplifier 120 comprising two connected low-noise amplifiers 116 and 122. The first low-noise amplifier 122 (first stage) comprises two SOI transistors 124 in a cascode structure, and the second low-noise amplifier 116 (second stage) comprises a common-source, transfer-microprinted GaN transistor 112. The second low-noise amplifier 116 is equivalent to that shown in Figure 21. A single 126 bias block made on the SOI wafer is used to power both stages 116 and 122. Both stages 116 and 122 preferably use the same power supply.
[0101] Fig. 23 represents a wiring diagram of another two-stage amplifier 120 comprising two low-noise amplifiers 116 connected on a wafer / chip SOL. The two low-noise amplifiers comprise a respective common-source microprinted transfer GaN transistor 112.
[0102] Figure 24 shows a wiring diagram of an N-stage amplifier 124 (with N > 2 amplification stages) according to one embodiment. Three low-noise amplifiers 116a, b, c are shown, and a dashed line between the output of the second low-noise amplifier 116b and the input of the third low-noise amplifier 116c indicates an arbitrary number of other low-noise amplifiers connected between them. Each amplification stage can be made from a low-noise amplifier such as the one shown in Figure 21. Each low-noise amplifier includes a transfer microprinted GaN transistor 112. Each low-noise amplifier 116a, b, c can be powered by the same power supply or by separate power supplies. The bias block 126 and the other circuit components are fabricated directly on the SOI board.
[0103] Fig. 25 represents a wiring diagram of an N-stage amplifier similar to that of Fig. 24, but in which the first stage comprises an SOI 122 cascode structure with SOI 124 transistors.
[0104] Although specific embodiments of the invention have been described above, it should be noted that the invention can be implemented in ways other than those described herein. The descriptions given above are provided by way of illustration only, and are not intended to be limiting. It will be evident to those skilled in the art that modifications can be made to the invention as described without departing from the scope of the appended claims.
[0105] Each characteristic element disclosed or illustrated in this document may be incorporated into the invention, alone or in appropriate combination with any other characteristic disclosed or illustrated herein.
Claims
Claims
1. A semiconductor structure for amplifying an RF signal, comprising: - a cascode structure comprising a GaN transistor micro-printed by transfer on an SOI wafer or chip; and a layer of glue confined between a GaN layer of said GaN transistor micro-printed by transfer and said SOI wafer or chip.
2. The semiconductor structure of claim 1, wherein said cascode structure further comprises an SOI transistor fabricated on the SOI wafer and connected to said microprinted GaN transistor by transfer.
3. The semiconductor structure of claim 1, wherein said cascode structure further comprises a second transfer-printed GaN transistor formed on the SOI wafer or chip and connected to said transfer-printed GaN transistor.
4. A semiconductor structure according to claim 1, 2 or 3, further comprising an inductor connected to the source of said transfer microprinted GaN transistor.
5. A semiconductor structure according to any preceding claim, further comprising a linearizer formed on the SOI wafer.
6. A method of manufacturing a structure according to any one of the preceding claims, using a micro-transfer printing process.
