rus
TS_pub
technospheramag
technospheramag
ТЕХНОСФЕРА_РИЦ
Issue #3/2024
A. A. Kokolov, F. I. Sheyerman, L. I. Babak, D. A. Konkin, A. V. Ubaichin, A. S. Koryakovtsev, E. A. Shutov
Experimental Study and Modeling of High-Frequency Performances of Ge-Photodiode for Microwave Optical Receiver Integrated Circuits
Experimental Study and Modeling of High-Frequency Performances of Ge-Photodiode for Microwave Optical Receiver Integrated Circuits
DOI: 10.22184/1993-7296.FRos.2024.18.3.230.244
Experimental Study and Modeling of High-Frequency Performances of Ge-Photodiode for Microwave Optical Receiver Integrated Circuits
A. A. Kokolov, F. I. Sheyerman, L. I. Babak, D. A. Konkin, A. V. Ubaichin, A. S. Koryakovtsev, E. A. Shutov
Tomsk State University of Control Systems and Radioelectronics, Tomsk, Russia
A technique and setup for the probe measurement of high-frequency performances of integrated Ge-photodiode are considered accounting for the diode’s actual environment in a photonic (PIC) or electronic-photonic (EPIC) integrated circuit. The key feature of the technique is the application of two coherent laser optical sources with different wavelengths. The measured data are presented for the optoelectronic conversion coefficient of a Ge photodiode in a specially designed measuring (test) PIC based on the electronic-photonic SiGe BiCMOS technology. At the wavelength of 1 550 nm, the frequency band of the Ge photodiode reaches ~30 GHz that makes it possible to use it as a part of integrated optical receivers with a data transmission rate of at least 25 Gbit / s. Using the electromagnetic simulation, a low-signal equivalent circuit model of a Ge-photodiode placed in the PIC (EPIC) is developed allowing the calculation of characteristics of a monolithically integrated optical receiver.
Keywords: Ge photodiode, probe measurements, model, optical receiver, microwave, photonic integrated circuit.
Article received: 27.12.2023
Article accepted: 15.02.2024
INTRODUCTION
The important parts of fiber-optic transmission systems (FOTS) are the optical receivers (OR), the main components of which include a photodiode (PD) and a transimpedance amplifier (TIA) [1, 2]. The PD destination is to convert an optical signal into an electrical one, and the TIA is used to amplify the demodulated signal taken from the PD. Moreover, in the high-speed FOTSs the bandwidths reach tens of gigahertz, i. e. the modulation frequencies of optical signal are located in the microwave range.
As a rule, the up-to-date microwave ORs are fabricated on the basis of semiconductor technologies. In this case, there exists a distinction between the hybrid and monolithically integrated optical receivers [1]. In the hybrid-integrated ORs, the photodiode and TIA are implemented as the separate integrated circuits (ICs). In such receivers intended for the telecommunication wavelength bands (1 310 nm and 1 550 nm), the photodiodes are often produced using InP technology.
In order to manufacture the TIA ICs for the hybrid microwave ORs, the RF CMOS or silicon-on-insulator (SOI) technologies are now most often selected that are cheaper, provide a greater integration level and broader IC functionality compared to other technical processes (InP, GaAs) [3]. The PD and TIA ICs are connected to each other by the wire bonding or flip-chip technique. In both cases, the influence of parasitic capacitances and inductances of such interconnections reduces the OR bandwidth.
In the monolithically integrated receivers, the PD and TIA are fabricated in a single technological cycle and are placed on a single chip that in this case is an electronic-photonic IC (EPIC) [1]. The combined placement of a photodiode in the EPIC with a feeding integrated optical waveguide and a subsequent TIA makes it possible to reduce the distance between these components and reduce parasitic effects at the interconnections. This design solution helps to achieve a significant expansion of the OR bandwidth that mainly determines the optical system performance.
One of the most suitable semiconductor technologies for the fabricating monolithic optical receivers is BiCMOS technology based on silicon-germanium (SiGe) that is also belonging to the group of silicon technological processes [4]. The usual “electronic” SiGe BiCMOS technology makes it possible to produce on a single silicon substrate the microwave CMOS field-effect transistors with the operating frequencies up to 60–80 GHz, heterobipolar (HBT) transistors with even higher operating frequencies (up to 150–200 GHz) and passive electronic components (such as resistors, capacitors, inductors and microwave transmission lines) [5, 6]. For the manufacture of EPICs used in the optoelectronic systems, a special combined electronic-photonic SiGe BiCMOS technology [7–9] has been developed. It allows to additionally fabricate Ge-photodiodes within the same technological process, as well as the SOI-based passive optical and optoelectronic components (silicon optical waveguides, dividers/adders, ring resonators, optical input/output devices, Mach-Zehnder modulators, etc.) [3–5, 10].
This paper is devoted to the issues of measuring and modeling the high-frequency characteristics of integrated photodiodes used in the optical receivers’ EPIC for data transmission systems [8, 11]. The main characteristics of such PDs include the dependence of the optic-to-electronic transducer coefficient on the modulating microwave signal frequency, as well as the microwave signal bandwidth. A knowledge of the PD parameters and its small-signal model with accounting for the interconnections of optical, optoelectronic and electronic components in the EPIC over a wide frequency range is necessary for predicting and simulating the performances of both the OR and the entire FOTS. It should be noted that the experimental study of an integrated PD is not a trivial problem: due to the small component size, it is necessary to solve the tasks of introducing light beam into the EPIC and reducing the influence of optical and microwave connections on the measurement results.
The paper describes the technique and probe station setup for measuring the high-frequency characteristics of integrated photodiodes in the photonic integrated circuits (PIC) and EPIC. The technique and setup can also be applied to other fabrication technologies of the PDs, PICs and EPICs. The experimental data for the frequency response of optic-to-electronic conversion and bandwidth of the Ge-photodiode are given. A small-signal PD model is developed on the basis of the electromagnetic (EM) simulation in CAD with account for the influence of interconnections of various components in the PIC (EPIC). A comparison of the PD characteristics obtained from the model and experiment is presented.
It should be noted that the Ge photodiode under study in this paper was a part of electronic-photonic IC of 25 Gbit/s monolithically integrated single-chip optical receiver for digital communication, that is developed at the Research Institute of Microelectronic Systems (RIMS) of the Tomsk University of Control Systems and Radioelectronics (TUSUR) and fabricated with the combined SiGe BiCMOS technology [12, 13].
Photonic integrated circuit for measuring the Ge photodiode characteristics
In the electronic-photonic SiGe technology [7–9] the optical components are formed on a SOI substrate with a recessed oxide layer thickness of 2 μm and a Si layer with the thickness of 220 nm, optimized for photonic applications. The top thin silicon layer and the deep oxide layer are removed from the areas where the IC electronic part is located by the reactive ion and liquid etching. Next, an area of bulk silicon is formed in the etched areas using the selective epitaxy, after which the chemical-mechanical surface planarization is performed. After this, the standard SiGe BiCMOS process with the CMOS- and HBT transistors is implemented. A germanium pin photodiode [8, 10] is fabricated using the selective Ge epitaxy on a substrate and subsequent p+ and n+ ion implantation; its construction is shown in Fig. 1. When the light beam hits an area of undoped germanium located between p+ (base) and n+ (collector), the free charge carriers occur that are additionally accelerated by the electric field of the reverse biased p-n junction enhancing the PD speed and sensitivity.
Here, we investigate performances of a Ge photodiode that is fabricated with using the 0.25 μm electronic-photonic SiGe technology [14]. According to the published experimental data [7], the Ge photodiode parameters in this technology at an optical wavelength of 1550 nm are as follows: sensitivity SPD of at least 0.9 A/W with a reverse bias voltage of up to −2 V; dark current Idark less than 400 nA; the PD capacitance CPD is about 50 fF that potentially provides a photodiode frequency bandwidth Δf of up to 60 GHz. However, the objective of this study is not to determine the potential parameters of an isolated PD, but its actual performances achieved in the environment (optical waveguides, microwave feeding lines, specific constructions and layouts of elements) that is existed in the integrated circuit. Therefore, the measurement conditions should be close to the operating conditions of the PD in the PIC or EPIC.
For investigating the integrated PD characteristics, we have developed a special measuring (test) PIC; a sketch of its layout is shown in Fig. 2a. The layout contains a photodetector, two optical inputs to provide light input to the IC, feeding ridge optical waveguides, 50 Ohm microstrip lines (MSL) for receiving the detected microwave signal and applying a bias voltage to the diode, as well as five contact pads with a pitch of 150 μm for probe connection or the IC wire bonding. The contacts of the PD cathode and anode are made in the upper thick metallized layer of the semiconductor SiGe structure. Two available optical inputs provide a more convenient use of the PIC during the measurements.
To reduce the influence of optical and microwave inhomogeneities caused by the interconnections, it is worthwhile to measure the PD high-frequency parameters and bandwidth with a probe station using a test PIC on a semiconductor wafer. This also makes it possible to conveniently measure the characteristics of several Ge photodiodes within their test PICs that are simultaneously located on the wafer. When using a probe station, the light emission can be supplied to the PIC optical inputs through the optical probes to which a fiber-optic cable is connected. The contact pads connected to the PD allow to apply a reverse bias voltage Ucm to the diode and receive the detected microwave signal through the electrical probes.
The optical input in PIC is implemented using a diffraction Bragg grating [15, 16]. Its operation is based on the use of periodic irregularities {\displaystyle \Lambda} to leading to the spectral selectivity of reflection from the grating. However, this principle makes the diffraction- grating-based input very sensitive to the incidence angle of light beam as well as to the direction of polarization vector of the light. When performing the high-frequency probe measurements of PD, we have used a vertical injection of the light into the PIC directly using the optical probes [17]. Our experiments have shown that at a wavelength of 1550 nm, the minimal optical power losses (5–6 dB) are achieved if the light beam falls on the PIC diffraction grating through a cleaved optical fiber at an angle of 15.5 degrees to the normal direction.
A photograph of the fabricated PIC with an optical probe connected to one of the inputs as well as a connected 5‑pin microwave probe is shown in Fig. 2b, the PIC dimensions are 2.26 × 1.26 mm2.
A technique, setup and measurement results of PD high-frequency characteristics
The main high-frequency characteristics of PDs include the optic-to-electronic transducer coefficient and the microwave signal bandwidth. For a fixed optical wavelength, the PD optic-to-electronic transducer coefficient КOE-RF is defined as the ratio of the detected microwave signal power delivered in a matched 50 Ohm load to the power of modulated optical signal supplied to the PD at a given modulation frequency f. The PD bandwidth Δf for the microwave signal is determined from the amplitude-frequency response (AFR) КOE-RF(f) at a level of −3 dB relative to the maximum transducer coefficient value. Below, we investigate the mentioned high-frequency characteristics of the Ge photodiode at the central wavelength of the common telecommunication band, i. e., 15 50 nm; the measurements at other optical wavelengths are performed in a similar way.
The block diagram and general view of the probe station setup for measurements are shown in Fig. 3a and 3b, respectively. The setup block diagram (Fig. 3a) has two key features. The first of them is the application of a negative bias voltage Ucm that is supplied to the PD from the power supply unit through the bias tee BT. This increases the speed and bandwidth of the PD operating in the photoelectric mode by reducing its own capacitance when selecting an operating point in the reverse bias region. The second feature is the application of the input resistance of the spectrum analyzer SA as a PD load for an alternating current; in this case, it becomes possible to measure the integrated photodiode AFR when connecting to only one of its terminals (anode). The photodiode cathode is connected to a matched load ML (50 Ohm) through a 5‑pin microwave probe and a coaxial connector.
During the measurements, a chip or wafer with a test PIC (Fig. 3b) is arranged at the probe station PS. At the outputs of channels L1 and L2 of the optical generator OG, the coherent light beams with the frequencies fL1 and fL2 close to each other are generated, and then the signals L1 and L2 are added in the optical adder A. The resulting signal from the adder is fed to the polarizer P, connected to the optical probe OP. The polarizer is used to adjust the light polarization, to which the diffraction grating is sensitive. The OP is optimally positioned relative to the optical input of the PIC chip by the way that will be described below.
When the optical signals from two OG channels are applied to the PD input, an alternating current is generated in the photodiode with a frequency equal to the difference of frequencies fL1–fL2 and an amplitude proportional to the light intensity. Passing through the SA input resistance, this current leads to the appearance of a microwave signal at the SA input, also with a difference frequency. Thus, by determining the microwave signal level at the SA input when changing the difference frequency of optical signals in the channels L 1 and L2, it is possible to measure the frequency response of the integrated PD.
The setting of the optical probe is implemented with a micropositioner installed on the PS as follows. The PD current is controlled by using a precision ammeter (not shown in the diagram in Fig. 3a) connected to a 5‑pin probe. By changing the angle of the incident light beam as well as using the three-coordinate positioning of the optical fiber in OP relative to the diffraction grating, the position at which the PD photocurrent is maximum is found. After this, using manual adjustment of the polarizer, the maximum photocurrent is determined again, this corresponds to the coincidence of the polarization direction of the optical signal with the optimum for the diffraction grating.
The conditions for the PD measurements were as follows: power of the OG optical signals in the channels L1 and L2 are PL1 = PL2 = 10 mW; frequency fL1 of the optical signal in the channel L1 is 193.4 THz, frequency fL2 in the channel L2 varies from 193.401 THz up to 193.450 THz, with this the frequency of the microwave signal varies in the range from 1 GHz to 50 GHz; the photodiode bias voltage is Ucm = –0.5 V.
Figure 4 shows the normalized measured AFR of the optic-to-electronic transducer coefficient of the integrated Ge photodiode, obtained by the technique presented. The normalization is performed relative to the transducer coefficient value КOE-RF at the lowest frequency of the microwave signal (1 GHz), which is equal to –12 dB. When processing the AFR measurement results, the microwave signal losses in the bias tee and the cable connecting the probe station to the spectrum analyzer were accounted. The measured PD bandwidth at –3 dB is about 30 GHz.
Development of a small-signal model of an integrated Ge-photodiode with accounting for influence of parasitic parameters of interconnections in the PIC or EPIC
In order to compute the joint high-frequency characteristics of a Ge-photodiode and a TIA interconnected in an optical receiver, it is necessary to develop a small-signal PD model, for example, in the form of an equivalent circuit [1, 2]. For monolithically integrated optical receivers, it is worthwhile to prepare a model that will characterize not the isolated photodiode, but the PD with its actual environment in the PIC or EPIC. Such the model should describe the parasitic parameters of the connection between the PD and the TIA in a single-chip microwave OR.
As an example, let us consider the PD model development in the measuring PIC presented in Fig. 2. In this case, the model must consider as the PD parameters as the parasitic parameters of the diode’s connection with the microwave path elements; these parameters are influenced by the layout of the feeding microstrip transmission lines jointly with the microwave contacts (pins). The photodiode equivalent circuit in the measuring PIC taken by us is shown in Fig. 5a. It includes a photocurrent source IPD, parallel capacitance CPD and series resistance Rs that characterize the PD alone [1, 2] as well as two microstrip transmission lines TL1 and TL2 describing the connection elements. It should be noted that in the lines TL1 and TL2 not only the signal conductors are curved, but also the grounded metalized areas have a complicated shape (Fig. 2a). Therefore, the computation of the twoport scattering parameters of these connecting microwave lines on the PIC’s silicon substrate should be based on the electromagnetic (EM) simulation.
Such the computation was carried out using the Momentum simulator in the ADS CAD system. For this, we have used the geometric and electrophysical parameters of the metal-dielectric PIC stack available from the technology data [14]. The stack consists of five layers of aluminum metallization (three layers of thin metallization – 0.7 μm and two layers of upper thick metallization – 2 and 3 μm) and dielectric layers between them. The layouts of the signal and grounding conductors of the simulated microwave lines in CAD tool are shown in Fig. 6.
The small-signal model of the Ge-photodiode accounting for its environment in the measuring PIC is obtained as follows. The schematic with the PD equivalent circuit similar to the structure of Fig. 5a (elements IPD, C'PD and R's) is inserted and simulated in the ADS CAD tool (Fig. 5b). This PD equivalent circuit is connected by its terminals to the inputs of two two-port circuits S1 and S2 that describe the transmission lines TL and TL2, respectively, and are characterized by the S-parameters calculated through EM simulation. The load RL = 50 Ohm is connected to the outputs of these two-port circuits. After simulating the schematic in Fig. 5b, it is possible to determine its transimpedance ZMT as the ratio of the voltage UL on the load RL (between the output terminals A and B of the two-port circuits) to the source current IPD caused by the detected microwave signal:
UL
ZMT =−. (1)
IPD
While knowing the PD optical sensitivity SPD, it is easy to determine the optic-to-electronic transducer coefficient KMOE-RF for the circuit (model) in Fig. 5b from the transimpedance ZMT using the following formula:
PRF IPD· RL RL
ΚMΟΕ-RF =− =−=−−. (2)
Popt SPD · U2L SPD · ZMT · UL
Using the CAD tool, in the frequency band from DC to 50 GHz we have computed with the formulas (1) and (2) the AFR of the model optic-to-electronic transducer coefficient KMOE-RF(f) normalized to the value of KMOE-RF(0) at zero frequency. Next, by varying the elements C'PD and R's in Fig. 5b, the simulated AFR of the model transducer coefficient KMOE-RF(f) was made close to the measured normalized frequency response of the conversion coefficient KMOE-RF(f) of PD in the measuring PIC (Fig. 4). In general, such an approximation can be performed using the circuit parameter optimization available in the ADS CAD tool. However, as in this case there are only two variable model parameters, it is simpler and more obvious to apply the “visual optimization” CAD procedure by displaying both normalized frequency responses of the conversion coefficients (the measured and simulated ones) and changing the values of elements C'PD, R's using a special CAD instrument (a tuner). In this case, it is possible to observe in real time any changes in the AFR shape of the model conversion coefficient KMOE-RF(f) and move it closer to the measured AFR KOE-RF(f) by varying C'PD and R's.
The normalized AFR of the model optic-to-electronic transducer coefficient obtained by this technique that approximates the measured AFR, is shown in Fig. 4. It corresponds to the element values C'PD = 20 fF and R's = 5 Ohm. As it can be seen, the model in Fig. 5b describes the high-frequency response of the PD integrated in the measuring PIC with rather good accuracy. It was found in the study that the configuration of connecting microstrip lines greatly influences the AFR of conversion coefficient of the PD in the measuring PIC.
It is impermissible to believe that the values of C'PD and R's found with such the optimization procedure are exactly equal to the elements CPD and Rs in the equivalent circuit of the isolated PD used. This is due to the fact that the elements C'PD and R's in the model in Fig. 5b describe not only the parameters of the isolated PD, but also integrally reflect various effects occurred in the measuring PIC and considered in the EM modeling (such as the EM field irregularities arising in the connections of the PD and microstrip lines, in the contact pads for connecting the microwave probes, etc.). However, these effects influence the measurement results and are accordingly accounted in the model in Fig. 5b.
It should also be noted that the model presented can describe with sufficient accuracy not only the normalized value, but also the absolute level of the optic-to-electronic transducer coefficient КOE-RF of the PD in the PIC or EPIC. However, in the case under consideration, the measurements did not imply any calibration of the measuring setup for the КOE-RF leve. Therefore, when constructing the model, the normalized AFR of the PD was used.
The layout of a part of single-chip monolithically integrated optical receiver EPIC [12, 13], containing the Ge-photodiode and connecting microstrip lines, is almost identical to one used in the measuring PIC (Fig. 2a). Therefore, the proposed equivalent circuit model of a Ge-photodiode in PIC can be used to compute performances of a SiGe monolithically integrated optical receiver, taking into account the interconnection between the PD and the TIA.
It can be concluded that as the Ge-photodiode under study as its small-signal equivalent circuit model can be applied in the development of the optical receiver EPICs for data transmission systems with a bandwidth of up to 30 GHz (transfer rate of at least 25 Gbit/s).
A technique and setup for the probe measurement of high-frequency performances of integrated Ge-photodiode are considered accounting for the diode’s actual environment in a photonic (PIC) or electronic-photonic (EPIC) integrated circuit. The key feature of the technique is the application of two coherent laser optical sources with different wavelengths. Using the electromagnetic simulation, a low-signal equivalent circuit model of a Ge-photodiode placed in PIC (EPIC) is developed allowing the calculation of characteristics of a monolithically integrated optical receiver.
CONCLUSION
The paper describes a technique and setup for the probe measurement of high-frequency performances of an integrated Ge-photodiode accounting for the diode’s actual environment in a PIC or EPIC. According to the experimental results obtained, the Ge-photodiode placed in the specially designed measuring (test) PIC provides a bandwidth of about 30 GHz at an optimal bias voltage. Thus, the study has showed that characteristics of integrated Ge-photodiodes are rather high and close to the InP-based PDs. Their advantages compared to the InP-based photodiodes include the ability of the integration with electronic devices in single-chip optical receivers based on the electronic-photonic SiGe BiCMOS technologies, as well as their lower cost. By using the EM simulation, a small-signal equivalent circuit model of a Ge-photodiode placed in the PIC (EPIC) is developed allowing the calculation of characteristics of a monolithically integrated optical receiver.
ACKNOWLEDGMENTS
The research and modeling were performed with the financial support from the Ministry of Science and Higher Education of the Russian Federation (unique identifier No. FEWM‑2022-0006).
ABOUT AUTHORS
Kokolov Andrey Alexandrovich, Candidate of Technical Sciences, Head of Laboratory IC and SoC, TUSUR, Tomsk, Russia.
ORCID: 0000-0002-8910-4329
Scientific interests: microwave ICs, photonic integrated circuits, A3B5, CMOS, nonlinear models of transistors.
Sheyerman Feodor Ivanovich, Candidate of Technical Sciences, R&D Director, Research Institute of Microelectronic Systems, TUSUR, Tomsk, Russia.
ORCID: 0000-0001-6482-2108
Scientific interests: microwave ICs, photonic integrated circuits, A3B5, CMOS, microwave measurements.
Babak Leonid Ivanovich, Doctor of Technical Sciences, Professor, Head of the Research Institute of Microelectronic Systems, TUSUR, Tomsk, Russia.
ORCID: 0000-0002-2333-0518
Scientific interests: visual design methods, microwave ICs, photonic integrated circuits, CAD, A3B5, CMOS, models of microwave elements.
Konkin Dmitry Anatolievich, Senior lecturer, RSS, TUSUR, Tomsk, Russia.
ORCID: 0000-0002-5024-0825
Scientific interests: photonic integrated circuits.
Ubaychin Anton Viktorovich, Associate Professor, RSS, TUSUR, Tomsk, Russia.
ORCID: 0000-0001-6284-4645
Scientific interests: new measurement methods, scientific instrumentation, microwave radiometry.
Koryakovtsev Artem Sergeevich, Junior researcher of Laboratory IC and SoC, TUSUR, Tomsk, Russia
ORCID: 0000-0002-5024-0825
Scientific interests: microwave ICs, optical receivers.
Shutov Evgeniy Alexandrovich, Junior researcher of Laboratory IC and SoC, TUSUR, Tomsk, Russia
ORCID: 0000-0002-6199-7022
Scientific interests: microwave ICs, microwave measurements.
CONTRIBUTION BY THE MEMBERS
OF THE TEAM OF AUTHORS
Babak L. I.: scientific supervisor, literature review, approved the final version of the article before submitting it for publication; Konkin D. A.: development of the technique for measuring high-frequency characteristics of Ge-photodiode; Ubaichin A. V.: data processing and analysis; Shutov E. A.: conducting experimental research; Koryakovtsev A. S.: EM simulation, small-signal photodiode model extraction; Kokolov A. A.: description of results, formed the conclusions of the study; Sheyerman F. I.: statement of the research problem, contributed to the extraction of a small-signal photodiode model.
A. A. Kokolov, F. I. Sheyerman, L. I. Babak, D. A. Konkin, A. V. Ubaichin, A. S. Koryakovtsev, E. A. Shutov
Tomsk State University of Control Systems and Radioelectronics, Tomsk, Russia
A technique and setup for the probe measurement of high-frequency performances of integrated Ge-photodiode are considered accounting for the diode’s actual environment in a photonic (PIC) or electronic-photonic (EPIC) integrated circuit. The key feature of the technique is the application of two coherent laser optical sources with different wavelengths. The measured data are presented for the optoelectronic conversion coefficient of a Ge photodiode in a specially designed measuring (test) PIC based on the electronic-photonic SiGe BiCMOS technology. At the wavelength of 1 550 nm, the frequency band of the Ge photodiode reaches ~30 GHz that makes it possible to use it as a part of integrated optical receivers with a data transmission rate of at least 25 Gbit / s. Using the electromagnetic simulation, a low-signal equivalent circuit model of a Ge-photodiode placed in the PIC (EPIC) is developed allowing the calculation of characteristics of a monolithically integrated optical receiver.
Keywords: Ge photodiode, probe measurements, model, optical receiver, microwave, photonic integrated circuit.
Article received: 27.12.2023
Article accepted: 15.02.2024
INTRODUCTION
The important parts of fiber-optic transmission systems (FOTS) are the optical receivers (OR), the main components of which include a photodiode (PD) and a transimpedance amplifier (TIA) [1, 2]. The PD destination is to convert an optical signal into an electrical one, and the TIA is used to amplify the demodulated signal taken from the PD. Moreover, in the high-speed FOTSs the bandwidths reach tens of gigahertz, i. e. the modulation frequencies of optical signal are located in the microwave range.
As a rule, the up-to-date microwave ORs are fabricated on the basis of semiconductor technologies. In this case, there exists a distinction between the hybrid and monolithically integrated optical receivers [1]. In the hybrid-integrated ORs, the photodiode and TIA are implemented as the separate integrated circuits (ICs). In such receivers intended for the telecommunication wavelength bands (1 310 nm and 1 550 nm), the photodiodes are often produced using InP technology.
In order to manufacture the TIA ICs for the hybrid microwave ORs, the RF CMOS or silicon-on-insulator (SOI) technologies are now most often selected that are cheaper, provide a greater integration level and broader IC functionality compared to other technical processes (InP, GaAs) [3]. The PD and TIA ICs are connected to each other by the wire bonding or flip-chip technique. In both cases, the influence of parasitic capacitances and inductances of such interconnections reduces the OR bandwidth.
In the monolithically integrated receivers, the PD and TIA are fabricated in a single technological cycle and are placed on a single chip that in this case is an electronic-photonic IC (EPIC) [1]. The combined placement of a photodiode in the EPIC with a feeding integrated optical waveguide and a subsequent TIA makes it possible to reduce the distance between these components and reduce parasitic effects at the interconnections. This design solution helps to achieve a significant expansion of the OR bandwidth that mainly determines the optical system performance.
One of the most suitable semiconductor technologies for the fabricating monolithic optical receivers is BiCMOS technology based on silicon-germanium (SiGe) that is also belonging to the group of silicon technological processes [4]. The usual “electronic” SiGe BiCMOS technology makes it possible to produce on a single silicon substrate the microwave CMOS field-effect transistors with the operating frequencies up to 60–80 GHz, heterobipolar (HBT) transistors with even higher operating frequencies (up to 150–200 GHz) and passive electronic components (such as resistors, capacitors, inductors and microwave transmission lines) [5, 6]. For the manufacture of EPICs used in the optoelectronic systems, a special combined electronic-photonic SiGe BiCMOS technology [7–9] has been developed. It allows to additionally fabricate Ge-photodiodes within the same technological process, as well as the SOI-based passive optical and optoelectronic components (silicon optical waveguides, dividers/adders, ring resonators, optical input/output devices, Mach-Zehnder modulators, etc.) [3–5, 10].
This paper is devoted to the issues of measuring and modeling the high-frequency characteristics of integrated photodiodes used in the optical receivers’ EPIC for data transmission systems [8, 11]. The main characteristics of such PDs include the dependence of the optic-to-electronic transducer coefficient on the modulating microwave signal frequency, as well as the microwave signal bandwidth. A knowledge of the PD parameters and its small-signal model with accounting for the interconnections of optical, optoelectronic and electronic components in the EPIC over a wide frequency range is necessary for predicting and simulating the performances of both the OR and the entire FOTS. It should be noted that the experimental study of an integrated PD is not a trivial problem: due to the small component size, it is necessary to solve the tasks of introducing light beam into the EPIC and reducing the influence of optical and microwave connections on the measurement results.
The paper describes the technique and probe station setup for measuring the high-frequency characteristics of integrated photodiodes in the photonic integrated circuits (PIC) and EPIC. The technique and setup can also be applied to other fabrication technologies of the PDs, PICs and EPICs. The experimental data for the frequency response of optic-to-electronic conversion and bandwidth of the Ge-photodiode are given. A small-signal PD model is developed on the basis of the electromagnetic (EM) simulation in CAD with account for the influence of interconnections of various components in the PIC (EPIC). A comparison of the PD characteristics obtained from the model and experiment is presented.
It should be noted that the Ge photodiode under study in this paper was a part of electronic-photonic IC of 25 Gbit/s monolithically integrated single-chip optical receiver for digital communication, that is developed at the Research Institute of Microelectronic Systems (RIMS) of the Tomsk University of Control Systems and Radioelectronics (TUSUR) and fabricated with the combined SiGe BiCMOS technology [12, 13].
Photonic integrated circuit for measuring the Ge photodiode characteristics
In the electronic-photonic SiGe technology [7–9] the optical components are formed on a SOI substrate with a recessed oxide layer thickness of 2 μm and a Si layer with the thickness of 220 nm, optimized for photonic applications. The top thin silicon layer and the deep oxide layer are removed from the areas where the IC electronic part is located by the reactive ion and liquid etching. Next, an area of bulk silicon is formed in the etched areas using the selective epitaxy, after which the chemical-mechanical surface planarization is performed. After this, the standard SiGe BiCMOS process with the CMOS- and HBT transistors is implemented. A germanium pin photodiode [8, 10] is fabricated using the selective Ge epitaxy on a substrate and subsequent p+ and n+ ion implantation; its construction is shown in Fig. 1. When the light beam hits an area of undoped germanium located between p+ (base) and n+ (collector), the free charge carriers occur that are additionally accelerated by the electric field of the reverse biased p-n junction enhancing the PD speed and sensitivity.
Here, we investigate performances of a Ge photodiode that is fabricated with using the 0.25 μm electronic-photonic SiGe technology [14]. According to the published experimental data [7], the Ge photodiode parameters in this technology at an optical wavelength of 1550 nm are as follows: sensitivity SPD of at least 0.9 A/W with a reverse bias voltage of up to −2 V; dark current Idark less than 400 nA; the PD capacitance CPD is about 50 fF that potentially provides a photodiode frequency bandwidth Δf of up to 60 GHz. However, the objective of this study is not to determine the potential parameters of an isolated PD, but its actual performances achieved in the environment (optical waveguides, microwave feeding lines, specific constructions and layouts of elements) that is existed in the integrated circuit. Therefore, the measurement conditions should be close to the operating conditions of the PD in the PIC or EPIC.
For investigating the integrated PD characteristics, we have developed a special measuring (test) PIC; a sketch of its layout is shown in Fig. 2a. The layout contains a photodetector, two optical inputs to provide light input to the IC, feeding ridge optical waveguides, 50 Ohm microstrip lines (MSL) for receiving the detected microwave signal and applying a bias voltage to the diode, as well as five contact pads with a pitch of 150 μm for probe connection or the IC wire bonding. The contacts of the PD cathode and anode are made in the upper thick metallized layer of the semiconductor SiGe structure. Two available optical inputs provide a more convenient use of the PIC during the measurements.
To reduce the influence of optical and microwave inhomogeneities caused by the interconnections, it is worthwhile to measure the PD high-frequency parameters and bandwidth with a probe station using a test PIC on a semiconductor wafer. This also makes it possible to conveniently measure the characteristics of several Ge photodiodes within their test PICs that are simultaneously located on the wafer. When using a probe station, the light emission can be supplied to the PIC optical inputs through the optical probes to which a fiber-optic cable is connected. The contact pads connected to the PD allow to apply a reverse bias voltage Ucm to the diode and receive the detected microwave signal through the electrical probes.
The optical input in PIC is implemented using a diffraction Bragg grating [15, 16]. Its operation is based on the use of periodic irregularities {\displaystyle \Lambda} to leading to the spectral selectivity of reflection from the grating. However, this principle makes the diffraction- grating-based input very sensitive to the incidence angle of light beam as well as to the direction of polarization vector of the light. When performing the high-frequency probe measurements of PD, we have used a vertical injection of the light into the PIC directly using the optical probes [17]. Our experiments have shown that at a wavelength of 1550 nm, the minimal optical power losses (5–6 dB) are achieved if the light beam falls on the PIC diffraction grating through a cleaved optical fiber at an angle of 15.5 degrees to the normal direction.
A photograph of the fabricated PIC with an optical probe connected to one of the inputs as well as a connected 5‑pin microwave probe is shown in Fig. 2b, the PIC dimensions are 2.26 × 1.26 mm2.
A technique, setup and measurement results of PD high-frequency characteristics
The main high-frequency characteristics of PDs include the optic-to-electronic transducer coefficient and the microwave signal bandwidth. For a fixed optical wavelength, the PD optic-to-electronic transducer coefficient КOE-RF is defined as the ratio of the detected microwave signal power delivered in a matched 50 Ohm load to the power of modulated optical signal supplied to the PD at a given modulation frequency f. The PD bandwidth Δf for the microwave signal is determined from the amplitude-frequency response (AFR) КOE-RF(f) at a level of −3 dB relative to the maximum transducer coefficient value. Below, we investigate the mentioned high-frequency characteristics of the Ge photodiode at the central wavelength of the common telecommunication band, i. e., 15 50 nm; the measurements at other optical wavelengths are performed in a similar way.
The block diagram and general view of the probe station setup for measurements are shown in Fig. 3a and 3b, respectively. The setup block diagram (Fig. 3a) has two key features. The first of them is the application of a negative bias voltage Ucm that is supplied to the PD from the power supply unit through the bias tee BT. This increases the speed and bandwidth of the PD operating in the photoelectric mode by reducing its own capacitance when selecting an operating point in the reverse bias region. The second feature is the application of the input resistance of the spectrum analyzer SA as a PD load for an alternating current; in this case, it becomes possible to measure the integrated photodiode AFR when connecting to only one of its terminals (anode). The photodiode cathode is connected to a matched load ML (50 Ohm) through a 5‑pin microwave probe and a coaxial connector.
During the measurements, a chip or wafer with a test PIC (Fig. 3b) is arranged at the probe station PS. At the outputs of channels L1 and L2 of the optical generator OG, the coherent light beams with the frequencies fL1 and fL2 close to each other are generated, and then the signals L1 and L2 are added in the optical adder A. The resulting signal from the adder is fed to the polarizer P, connected to the optical probe OP. The polarizer is used to adjust the light polarization, to which the diffraction grating is sensitive. The OP is optimally positioned relative to the optical input of the PIC chip by the way that will be described below.
When the optical signals from two OG channels are applied to the PD input, an alternating current is generated in the photodiode with a frequency equal to the difference of frequencies fL1–fL2 and an amplitude proportional to the light intensity. Passing through the SA input resistance, this current leads to the appearance of a microwave signal at the SA input, also with a difference frequency. Thus, by determining the microwave signal level at the SA input when changing the difference frequency of optical signals in the channels L 1 and L2, it is possible to measure the frequency response of the integrated PD.
The setting of the optical probe is implemented with a micropositioner installed on the PS as follows. The PD current is controlled by using a precision ammeter (not shown in the diagram in Fig. 3a) connected to a 5‑pin probe. By changing the angle of the incident light beam as well as using the three-coordinate positioning of the optical fiber in OP relative to the diffraction grating, the position at which the PD photocurrent is maximum is found. After this, using manual adjustment of the polarizer, the maximum photocurrent is determined again, this corresponds to the coincidence of the polarization direction of the optical signal with the optimum for the diffraction grating.
The conditions for the PD measurements were as follows: power of the OG optical signals in the channels L1 and L2 are PL1 = PL2 = 10 mW; frequency fL1 of the optical signal in the channel L1 is 193.4 THz, frequency fL2 in the channel L2 varies from 193.401 THz up to 193.450 THz, with this the frequency of the microwave signal varies in the range from 1 GHz to 50 GHz; the photodiode bias voltage is Ucm = –0.5 V.
Figure 4 shows the normalized measured AFR of the optic-to-electronic transducer coefficient of the integrated Ge photodiode, obtained by the technique presented. The normalization is performed relative to the transducer coefficient value КOE-RF at the lowest frequency of the microwave signal (1 GHz), which is equal to –12 dB. When processing the AFR measurement results, the microwave signal losses in the bias tee and the cable connecting the probe station to the spectrum analyzer were accounted. The measured PD bandwidth at –3 dB is about 30 GHz.
Development of a small-signal model of an integrated Ge-photodiode with accounting for influence of parasitic parameters of interconnections in the PIC or EPIC
In order to compute the joint high-frequency characteristics of a Ge-photodiode and a TIA interconnected in an optical receiver, it is necessary to develop a small-signal PD model, for example, in the form of an equivalent circuit [1, 2]. For monolithically integrated optical receivers, it is worthwhile to prepare a model that will characterize not the isolated photodiode, but the PD with its actual environment in the PIC or EPIC. Such the model should describe the parasitic parameters of the connection between the PD and the TIA in a single-chip microwave OR.
As an example, let us consider the PD model development in the measuring PIC presented in Fig. 2. In this case, the model must consider as the PD parameters as the parasitic parameters of the diode’s connection with the microwave path elements; these parameters are influenced by the layout of the feeding microstrip transmission lines jointly with the microwave contacts (pins). The photodiode equivalent circuit in the measuring PIC taken by us is shown in Fig. 5a. It includes a photocurrent source IPD, parallel capacitance CPD and series resistance Rs that characterize the PD alone [1, 2] as well as two microstrip transmission lines TL1 and TL2 describing the connection elements. It should be noted that in the lines TL1 and TL2 not only the signal conductors are curved, but also the grounded metalized areas have a complicated shape (Fig. 2a). Therefore, the computation of the twoport scattering parameters of these connecting microwave lines on the PIC’s silicon substrate should be based on the electromagnetic (EM) simulation.
Such the computation was carried out using the Momentum simulator in the ADS CAD system. For this, we have used the geometric and electrophysical parameters of the metal-dielectric PIC stack available from the technology data [14]. The stack consists of five layers of aluminum metallization (three layers of thin metallization – 0.7 μm and two layers of upper thick metallization – 2 and 3 μm) and dielectric layers between them. The layouts of the signal and grounding conductors of the simulated microwave lines in CAD tool are shown in Fig. 6.
The small-signal model of the Ge-photodiode accounting for its environment in the measuring PIC is obtained as follows. The schematic with the PD equivalent circuit similar to the structure of Fig. 5a (elements IPD, C'PD and R's) is inserted and simulated in the ADS CAD tool (Fig. 5b). This PD equivalent circuit is connected by its terminals to the inputs of two two-port circuits S1 and S2 that describe the transmission lines TL and TL2, respectively, and are characterized by the S-parameters calculated through EM simulation. The load RL = 50 Ohm is connected to the outputs of these two-port circuits. After simulating the schematic in Fig. 5b, it is possible to determine its transimpedance ZMT as the ratio of the voltage UL on the load RL (between the output terminals A and B of the two-port circuits) to the source current IPD caused by the detected microwave signal:
UL
ZMT =−. (1)
IPD
While knowing the PD optical sensitivity SPD, it is easy to determine the optic-to-electronic transducer coefficient KMOE-RF for the circuit (model) in Fig. 5b from the transimpedance ZMT using the following formula:
PRF IPD· RL RL
ΚMΟΕ-RF =− =−=−−. (2)
Popt SPD · U2L SPD · ZMT · UL
Using the CAD tool, in the frequency band from DC to 50 GHz we have computed with the formulas (1) and (2) the AFR of the model optic-to-electronic transducer coefficient KMOE-RF(f) normalized to the value of KMOE-RF(0) at zero frequency. Next, by varying the elements C'PD and R's in Fig. 5b, the simulated AFR of the model transducer coefficient KMOE-RF(f) was made close to the measured normalized frequency response of the conversion coefficient KMOE-RF(f) of PD in the measuring PIC (Fig. 4). In general, such an approximation can be performed using the circuit parameter optimization available in the ADS CAD tool. However, as in this case there are only two variable model parameters, it is simpler and more obvious to apply the “visual optimization” CAD procedure by displaying both normalized frequency responses of the conversion coefficients (the measured and simulated ones) and changing the values of elements C'PD, R's using a special CAD instrument (a tuner). In this case, it is possible to observe in real time any changes in the AFR shape of the model conversion coefficient KMOE-RF(f) and move it closer to the measured AFR KOE-RF(f) by varying C'PD and R's.
The normalized AFR of the model optic-to-electronic transducer coefficient obtained by this technique that approximates the measured AFR, is shown in Fig. 4. It corresponds to the element values C'PD = 20 fF and R's = 5 Ohm. As it can be seen, the model in Fig. 5b describes the high-frequency response of the PD integrated in the measuring PIC with rather good accuracy. It was found in the study that the configuration of connecting microstrip lines greatly influences the AFR of conversion coefficient of the PD in the measuring PIC.
It is impermissible to believe that the values of C'PD and R's found with such the optimization procedure are exactly equal to the elements CPD and Rs in the equivalent circuit of the isolated PD used. This is due to the fact that the elements C'PD and R's in the model in Fig. 5b describe not only the parameters of the isolated PD, but also integrally reflect various effects occurred in the measuring PIC and considered in the EM modeling (such as the EM field irregularities arising in the connections of the PD and microstrip lines, in the contact pads for connecting the microwave probes, etc.). However, these effects influence the measurement results and are accordingly accounted in the model in Fig. 5b.
It should also be noted that the model presented can describe with sufficient accuracy not only the normalized value, but also the absolute level of the optic-to-electronic transducer coefficient КOE-RF of the PD in the PIC or EPIC. However, in the case under consideration, the measurements did not imply any calibration of the measuring setup for the КOE-RF leve. Therefore, when constructing the model, the normalized AFR of the PD was used.
The layout of a part of single-chip monolithically integrated optical receiver EPIC [12, 13], containing the Ge-photodiode and connecting microstrip lines, is almost identical to one used in the measuring PIC (Fig. 2a). Therefore, the proposed equivalent circuit model of a Ge-photodiode in PIC can be used to compute performances of a SiGe monolithically integrated optical receiver, taking into account the interconnection between the PD and the TIA.
It can be concluded that as the Ge-photodiode under study as its small-signal equivalent circuit model can be applied in the development of the optical receiver EPICs for data transmission systems with a bandwidth of up to 30 GHz (transfer rate of at least 25 Gbit/s).
A technique and setup for the probe measurement of high-frequency performances of integrated Ge-photodiode are considered accounting for the diode’s actual environment in a photonic (PIC) or electronic-photonic (EPIC) integrated circuit. The key feature of the technique is the application of two coherent laser optical sources with different wavelengths. Using the electromagnetic simulation, a low-signal equivalent circuit model of a Ge-photodiode placed in PIC (EPIC) is developed allowing the calculation of characteristics of a monolithically integrated optical receiver.
CONCLUSION
The paper describes a technique and setup for the probe measurement of high-frequency performances of an integrated Ge-photodiode accounting for the diode’s actual environment in a PIC or EPIC. According to the experimental results obtained, the Ge-photodiode placed in the specially designed measuring (test) PIC provides a bandwidth of about 30 GHz at an optimal bias voltage. Thus, the study has showed that characteristics of integrated Ge-photodiodes are rather high and close to the InP-based PDs. Their advantages compared to the InP-based photodiodes include the ability of the integration with electronic devices in single-chip optical receivers based on the electronic-photonic SiGe BiCMOS technologies, as well as their lower cost. By using the EM simulation, a small-signal equivalent circuit model of a Ge-photodiode placed in the PIC (EPIC) is developed allowing the calculation of characteristics of a monolithically integrated optical receiver.
ACKNOWLEDGMENTS
The research and modeling were performed with the financial support from the Ministry of Science and Higher Education of the Russian Federation (unique identifier No. FEWM‑2022-0006).
ABOUT AUTHORS
Kokolov Andrey Alexandrovich, Candidate of Technical Sciences, Head of Laboratory IC and SoC, TUSUR, Tomsk, Russia.
ORCID: 0000-0002-8910-4329
Scientific interests: microwave ICs, photonic integrated circuits, A3B5, CMOS, nonlinear models of transistors.
Sheyerman Feodor Ivanovich, Candidate of Technical Sciences, R&D Director, Research Institute of Microelectronic Systems, TUSUR, Tomsk, Russia.
ORCID: 0000-0001-6482-2108
Scientific interests: microwave ICs, photonic integrated circuits, A3B5, CMOS, microwave measurements.
Babak Leonid Ivanovich, Doctor of Technical Sciences, Professor, Head of the Research Institute of Microelectronic Systems, TUSUR, Tomsk, Russia.
ORCID: 0000-0002-2333-0518
Scientific interests: visual design methods, microwave ICs, photonic integrated circuits, CAD, A3B5, CMOS, models of microwave elements.
Konkin Dmitry Anatolievich, Senior lecturer, RSS, TUSUR, Tomsk, Russia.
ORCID: 0000-0002-5024-0825
Scientific interests: photonic integrated circuits.
Ubaychin Anton Viktorovich, Associate Professor, RSS, TUSUR, Tomsk, Russia.
ORCID: 0000-0001-6284-4645
Scientific interests: new measurement methods, scientific instrumentation, microwave radiometry.
Koryakovtsev Artem Sergeevich, Junior researcher of Laboratory IC and SoC, TUSUR, Tomsk, Russia
ORCID: 0000-0002-5024-0825
Scientific interests: microwave ICs, optical receivers.
Shutov Evgeniy Alexandrovich, Junior researcher of Laboratory IC and SoC, TUSUR, Tomsk, Russia
ORCID: 0000-0002-6199-7022
Scientific interests: microwave ICs, microwave measurements.
CONTRIBUTION BY THE MEMBERS
OF THE TEAM OF AUTHORS
Babak L. I.: scientific supervisor, literature review, approved the final version of the article before submitting it for publication; Konkin D. A.: development of the technique for measuring high-frequency characteristics of Ge-photodiode; Ubaichin A. V.: data processing and analysis; Shutov E. A.: conducting experimental research; Koryakovtsev A. S.: EM simulation, small-signal photodiode model extraction; Kokolov A. A.: description of results, formed the conclusions of the study; Sheyerman F. I.: statement of the research problem, contributed to the extraction of a small-signal photodiode model.
Readers feedback
