A low power consumption parallel optoelectronic communication system

Abstract

The application discloses a low-power-consumption parallel photoelectric communication system, which comprises a transmitting subsystem, a receiving subsystem and a multi-channel imaging optical fiber link; the transmitting subsystem comprises a blue light Micro-LED transmitting array chip and a driving chip thereof; the receiving subsystem comprises a long-wavelength Micro-LED detecting array chip and a reading chip thereof; and the imaging optical fiber link transmits parallel optical signals between the transmitting and receiving ends. The long-wavelength Micro-LED detector adopts an InGaN / GaN multi-quantum well active region with adjustable indium components, so that the band gap is smaller than the energy of incident blue light photons, thereby realizing efficient and strong absorption of blue light. Meanwhile, the junction capacitance of the detector is compressed to the level of femto through a deep depletion effect, and the photo-generated carriers are driven to drift at a saturation speed. The system combines a 'wide and slow' architecture with high-density integration, and realizes high bandwidth, high responsivity and ultra-low power consumption.

Description

Technical Field

[0001] This invention relates to the technical field of optical communication and integrated circuits, and specifically to a low-power parallel optoelectronic communication system. Background Technology

[0002] The exponential growth in the number of training parameters for large artificial intelligence (AI) models, the expansion of high-performance computing (HPC) clusters, and the explosive growth in traffic of hyperscale data centers have shifted the bottleneck of computing power from the computing units themselves to data interconnect bandwidth. Between chips, between boards, and within server racks, data transmission bandwidth urgently needs to break through the "interconnect wall" of Tbps (terabits per second) and even Pbps (petbits per second).

[0003] Currently, short-distance high-speed interconnect technologies mainly face a competition between two mainstream technological paths: traditional electrical interconnects and laser-based optical interconnects. However, both exhibit insurmountable physical or engineering limitations when facing the demands of next-generation green, high-density computing. For a long time, copper wire electrical interconnects have been the mainstream solution for short-distance transmission. However, as single-channel signal rates approach 112Gbps PAM4 and 224Gbps PAM4, electrical interconnects face insurmountable physical obstacles, including severe frequency conversion losses and deterioration in energy efficiency. To address the distance limitations of electrical interconnects, optical interconnect technologies based on vertical-cavity surface-emitting lasers (VCSELs) or silicon photonics have emerged. While they solve the transmission distance problem, they generally adopt the "narrow-and-fast" architecture of the telecommunications field, relying on a small number of physical channels to carry extremely high single-channel rates (such as 4×100Gbps). This architecture has significant drawbacks: InP-based lasers are extremely sensitive to temperature, requiring sophisticated packaging and thermoelectric cooling (TEC) temperature control systems, resulting in high module costs; to support ultra-high single-channel speeds, high-power SerDes (serialization / deserialization) circuits are necessary, with end-to-end energy efficiency typically at 5~10 pJ / bit, which is difficult to reduce further; and it is difficult to achieve large-scale two-dimensional array integration of thousands of channels at low cost, limiting the improvement of spatial parallelism.

[0004] Existing optical interconnect technologies suffer from problems such as weak detector response, low energy efficiency, and high integration costs, which limit the further development of optical communication. Therefore, there is an urgent need to build a high-response parallel optical interconnect system that can overcome the physical limitations of single devices at the system architecture level, thereby completely breaking through the existing bottleneck of short-distance high-density interconnects. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the existing technology and provide a low-power parallel optoelectronic communication system to achieve excellent detection performance, and to achieve extremely high energy efficiency by combining a "wide and slow" architecture.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A low-power parallel optoelectronic communication system includes a transmitting subsystem, a receiving subsystem, and a multi-channel imaging fiber optic link; the transmitting subsystem includes a blue-light Micro-LED emitting array chip and a driver chip, and the receiving subsystem includes a long-wavelength Micro-LED detector array chip and a readout chip; the multi-channel imaging fiber optic link is used to transmit a parallel optical signal matrix between the transmitting subsystem and the receiving subsystem.

[0008] Among them, long-wavelength Micro-LEDs have In x Ga 1-x In an N / GaN multi-quantum-well active region, x is modulated such that the band gap Eg of the active region is less than the incident photon energy E of the blue micro-LED. ph The readout chip integrates a reverse bias circuit, which is configured to apply a high reverse bias voltage of -10V to -30V to the long-wavelength Micro-LED detector array chip.

[0009] The high-voltage bias generates a strong electric field, which expands the depletion region width of the PN junction to the physical limit, compressing the junction capacitance of the device to the fF (f-farad) level; at the same time, it causes photogenerated carriers to be collected at a saturation drift velocity, thus instantly transforming the "slow" LED structure into a high-speed photodiode that supports Gbps-level transmission.

[0010] Optionally, the center emission wavelength of the blue Micro-LED is 400 nm to 480 nm, and the electroluminescence peak wavelength of the long-wavelength Micro-LED is 500 nm to 650 nm.

[0011] Optionally, in the array of blue Micro-LED emitting array chip and long-wavelength Micro-LED detector array chip, the size (mesa diameter or side length) of each Micro-LED unit is 3 μm to 50 μm. At this tiny size, the device exhibits an extremely short carrier recombination lifetime under high current density driving, possessing an intrinsic GHz-level modulation bandwidth.

[0012] Optionally, the epitaxial structure of the long-wavelength Micro-LED detector array chip includes a substrate, an N-type layer, and an In layer arranged in sequence. x Ga 1-xThe active region of the N / GaN multi-quantum well and the P-type layer are formed by etching to form several arrayed Micro-LED units; the N-type layer and the P-type layer are electrically connected to the readout chip through N electrodes and P electrodes, respectively.

[0013] Optionally, the etching includes forming trenches separating each Micro-LED unit, with the trench depth extending to a portion of the substrate depth. The inner walls of the trenches are covered with a passivation layer and filled with a light-blocking material, forming a deep trench isolation structure that physically cuts off the lateral optical waveguide path between units (preventing optical crosstalk) and the leakage current path (preventing electrical crosstalk), significantly improving the signal-to-noise ratio of the high-density array.

[0014] Optionally, the deep trench isolation structure is also used between the Micro-LED light-emitting units of the blue Micro-LED emitting array chip to suppress optical coupling and electrical crosstalk between the light-emitting units.

[0015] Alternatively, ΔE = E ph - E g The range of ΔE is 0.5 eV to 0.8 eV. This optimal energy difference avoids insufficient absorption due to an excessively small energy difference or heat loss due to an excessively large energy difference, thus maximizing the responsiveness.

[0016] Optionally, the multi-channel imaging fiber optic link includes a transmitting microlens array, an imaging fiber bundle, and a receiving microlens array. The transmitting microlens array is integrated into the light-emitting surface of the blue Micro-LED emitting array chip, and the receiving microlens array is integrated into the light-incident surface of the long-wavelength Micro-LED detector array chip. They form a 1:1 conjugate imaging optical path through the imaging fiber bundle. Image transmission is achieved using spatial division multiplexing (SDM) technology.

[0017] Optionally, the readout chip includes a readout circuit without clock data recovery, the readout circuit including a transimpedance amplifier and a limiting amplifier; the transimpedance amplifier is configured based on an inverter structure to directly convert the photocurrent generated by the long-wavelength Micro-LED detector array chip into a voltage signal; the limiting amplifier is configured to shape the voltage signal output by the transimpedance amplifier into a digital level.

[0018] Optionally, the input front end of the readout circuit is provided with a virtual ground bias, which is used to clamp the potential of the long-wavelength Micro-LED.

[0019] Optionally, in the blue Micro-LED emitting array chip, a deep trench isolation structure is provided between adjacent Micro-LED units; the blue Micro-LED emitting array chip and the driving chip are flip-chip bonded together and driven by the driving chip.

[0020] The beneficial effects of this invention include:

[0021] Using band-engineered tunable long-wavelength InGaN Micro-LEDs as detectors ensures that blue photons can undergo strong interband absorption in the active region of the detector. Its absorption coefficient is far greater than that of traditional silicon-based detectors, which can achieve high responsivity and provide a higher signal-to-noise ratio (SNR) basis for the entire communication link.

[0022] By integrating a high-voltage reverse bias circuit on the receiver readout chip, the large capacitance of Micro-LED is compressed to the limit through the deep depletion effect, and combined with the saturation drift speed mechanism, single-channel Gbps-level transmission is supported.

[0023] Adopting a "wide and slow" architecture, it improves system energy efficiency and reduces the cost of large-scale manufacturing and packaging; it can be applied to data exchange in artificial intelligence (AI) computing clusters, interconnection of data center server boards, and high-density input / output (I / O) interfaces between high-performance computing (HPC) chips. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the overall architecture of a low-power parallel optoelectronic communication system as an example.

[0025] Figure 2 This is a schematic diagram of the structure of a single Micro-LED unit in a long-wavelength Micro-LED detector array chip, as shown in the embodiment.

[0026] Figure 3 This is a schematic diagram illustrating the principle of band transition, light absorption, and optimal energy difference matching for blue light signals using long-wavelength Micro-LED detection as an example.

[0027] Figure 4 This is a comparison of the responsivity of a green Micro-LED detector with an electroluminescence wavelength of 520 nm and a traditional silicon photodiode in the blue light band in the embodiment.

[0028] Figure 5 This is a schematic diagram illustrating the precision optical coupling principle of a single communication channel in an embodiment. Detailed Implementation

[0029] The present invention will be further explained below with reference to the accompanying drawings and specific embodiments. The accompanying drawings are merely illustrative to facilitate a better understanding of the invention, and their specific proportions can be adjusted according to design requirements. The "upper" and "lower" relationships of relative elements and the definitions of "front" and "back" in the graphics described herein should be understood by those skilled in the art to refer to the relative positions of the components; therefore, they can all be flipped to present the same component, and all of this should fall within the scope disclosed in this specification.

[0030] refer to Figure 1 The low-power parallel optoelectronic communication system of this embodiment employs a "wide-and-slow" architecture for an ultra-high-density parallel optical interconnect system. It aims to solve the problems of short distances in traditional copper interconnects and high power consumption in laser interconnects, achieving ultra-high bandwidth through a massive number of parallel channels. Specifically, it includes a transmitting subsystem, a receiving subsystem, and a multi-channel imaging fiber optic link. The transmitting subsystem includes a blue Micro-LED emitting array chip and a driver chip, while the receiving subsystem includes a long-wavelength Micro-LED detector array chip and a readout chip. The multi-channel imaging fiber optic link is used to transmit a parallel optical signal matrix between the transmitting and receiving subsystems.

[0031] The blue-light Micro-LED emitting array chip can be based on InGaN / GaN material. An array of Micro-LED light-emitting units is formed on its epitaxial structure through etching, and then bonded to a CMOS driver chip via flip-chip bonding, which drives the LEDs. The long-wavelength Micro-LED detector array chip can also be based on InGaN / GaN material. An array of Micro-LED detector units (each unit acting as a detector) is formed on its epitaxial structure through etching, and then bonded to a CMOS readout chip via flip-chip bonding.

[0032] The following describes the specific device structure and fabrication process of a long-wavelength Micro-LED detector array chip according to one embodiment. (Reference) Figure 2 The epitaxial layer of the long-wavelength Micro-LED detector array chip is grown on substrate 1 and fabricated using metal-organic chemical vapor deposition (MOCVD) technology. Substrate 1 is a patterned sapphire substrate (PSS) or a silicon substrate. The epitaxial layer, from bottom to top (from top to bottom after flip-chip fabrication), includes a nucleation layer 21, a buffer layer 22, an N-type contact layer 23, a stress relief layer 24, an active region 25, an electron blocking layer 26, and a P-type contact layer 27.

[0033] A 20-30 nm low-temperature GaN nucleation layer 21 and a 2-4 μm undoped u-GaN buffer layer 22 are used to reduce the dislocation density caused by lattice mismatch and improve crystal quality. Lower defect density is crucial for reducing the dark current of the detector. A 2-3 μm Si-doped n-GaN layer serves as an N-type contact layer 23, providing an electron channel. The stress relief layer 24 is an InGaN / GaN superlattice structure (SLS) grown for several periods before the active region 25. This layer alleviates the enormous compressive stress caused by the high indium composition quantum well, prevents excessive quantum confinement Stark effect (QCSE), and simultaneously improves the absorption cross-section for blue light. The active region 25 consists of 5-10 periods of In... x Ga 1-xThe system consists of N / GaN multiple quantum wells. All quantum well layers have a uniform indium (In) composition x and barrier height, and do not contain heterogeneous wavelength quantum wells. Optimal energy difference matching for blue light emission is achieved by precisely controlling the In composition x. The electron blocking layer 26 (EBL) is a p-type AlGaN layer that helps block leakage current under reverse bias and reduce noise in probe mode. The p-type contact layer 27 is a 100-200 nm Mg-doped p-GaN layer that provides hole channels. Furthermore, an ITO transparent conductive layer 3 can be fabricated on the p-type contact layer 27.

[0034] After the epitaxial wafer is grown, mesa etching is performed to a depth down to the surface of the N-type contact layer 23. Furthermore, to avoid optical crosstalk between adjacent units (pixels) in the array, a deep trench isolation structure is employed. Using deep trench physical isolation (DTI etching) technology, deep dry etching is performed with inductively coupled plasma (ICP) to a depth reaching a portion of the substrate 1, forming trenches a that separate the Micro-LED detection units. A dense SiO2 or Al2O3 passivation layer 4 is deposited using atomic layer deposition (ALD) technology. Passivation layer 4 covers the surface of trench a, effectively repairing etching damage and cutting off sidewall leakage channels, thereby significantly reducing the detector's dark current and improving detection sensitivity. Further, a light-blocking material 5 (such as a Ti / Au metal layer or black photoresist) is filled into trench a or its sidewalls are covered. This "light wall" effectively blocks photon propagation in the lateral waveguides between pixels, suppressing optical crosstalk in the array to below -40 dB. Metal electrodes, P-electrode 6 and N-electrode 7, are fabricated on p-GaN and n-GaN mesa, respectively. Micro-bumps are then made, and the long-wavelength Micro-LED detector array chip and CMOS readout chip are interconnected at the pixel level through thermo-press bonding or reflow soldering processes.

[0035] Blue Micro-LED emitting array chips are fabricated using a similar method, and also prevent optical crosstalk through deep trench isolation structures between Micro-LED light-emitting units. Their active region, for example, uses In... y Ga 1-yN / GaN multiple quantum wells allow for emission wavelengths in the blue light range by adjusting the γ-component content. The physical dimensions of the Micro-LED cells in the blue Micro-LED emitter array chip and the long-wavelength Micro-LED detector array chip are 3 μm to 50 μm, preferably matched. When the size is less than 50 μm, the device operates at high current density, current spreading is reduced, and self-heating is relatively controllable. More importantly, the smaller area means a smaller RC time constant, which is fundamental for achieving high-speed modulation. Sizes less than 3 μm lead to difficulties in fabrication and alignment accuracy, such as severe sidewall etching damage, increased non-radiative recombination, and decreased luminous efficiency (EQE) and detection efficiency. Considering bandwidth, efficiency, and fabrication difficulty, a size of 20 μm to 50 μm is preferred. At this size, the device can maintain high quantum efficiency while easily achieving GHz-level intrinsic bandwidth.

[0036] Blue Micro-LEDs, as the emitting light source, have a center wavelength λ. TX The wavelength can be flexibly selected within the range of 400 nm to 480 nm (e.g., 405 nm, 450 nm, or 470 nm). The corresponding incident photon energy E ph = hc / λ TX The range is approximately 2.58 eV to 3.10 eV. The receiver detector is a long-wavelength Micro-LED, and the bandgap E of the active region of the detector is adjusted by changing the indium flow rate and temperature during MOCVD growth. g "On-demand customization" is used to adapt the detector to the wavelength of the emitting light source. The peak wavelength of the detector's electroluminescence is between 500nm and 650nm (corresponding to bandgap E). g (≈1.9~2.5 eV). And always maintain an optimal energy difference ΔE = E between the incident photon energy and the detector bandgap. ph - E g (Controlled between 0.5 eV and 0.8 eV). For example... Figure 3 As shown, thanks to the aforementioned bandgap design, the energy of the incident blue photons is consistently significantly greater than the bandgap width of the detector material. When blue photons are incident on the active region of a long-wavelength Micro-LED, the photon energy far exceeds the bandgap width, allowing the photons to penetrate the valence band (E0). v Electrons deep within the electron matrix undergo direct excitation transitions to the conduction band (E). cThe higher energy levels of these deep-band transitions are direct transitions, not limited by phonon assistance, and the process is extremely fast (femtosecond level) with a very high absorption coefficient (α), typically greater than 10. 5 cm -1 This means that blue photons are completely absorbed and generate electron-hole pairs within an extremely short distance (<1μm) after entering the InGaN material, ensuring extremely high internal quantum efficiency.

[0037] like Figure 4 As shown, the responsivity of the long-wavelength Micro-LED detector (taking a green InGaN detector with an electroluminescence wavelength of 520 nm as an example) in this embodiment is compared with that of a conventional silicon (Si)-based photodiode in the blue light band. The silicon detector (prior art - curve B) is shown as the dashed line in the figure. Silicon generally has a low responsivity in the 400-480 nm band, typically below 0.2 A / W. This is because silicon is an indirect bandgap material, and blue light penetrates very shallowly in silicon (only about 0.2 μm). A large number of photogenerated carriers recombine at dangling bonds and defects on the silicon surface and cannot be collected by the electric field to form an effective photocurrent. In this embodiment, when the wavelength of the blue Micro-LED serving as the emitter is 450 nm, the long-wavelength Micro-LED serving as the detector is adjusted to the green light band (about 520 nm). In other embodiments, when the wavelength of the emitter is 470 nm, the detector is adjusted to the red light band (about 620 nm) to satisfy the optimal energy difference ΔE. By adjusting the indium composition, long-wavelength Micro-LED detectors can maintain a responsivity of >0.3 A / W in the 400 nm-480 nm range. This matching mechanism produces a signal amplitude that is more than 50% higher than that of silicon detectors, significantly improving the signal-to-noise ratio (SNR) of the communication link.

[0038] The CMOS readout chip integrates a reverse bias circuit and a readout circuit. The reverse bias circuit generates a deep reverse bias voltage of -10V to -30V, directly driving the upper long-wavelength Micro-LED detector to operate in full depletion mode. To accommodate bandgap-tunable long-wavelength detectors with different indium compositions, the reverse bias circuit employs a programmable high-voltage reverse bias setting. Specifically, the readout chip integrates a programmable high-voltage charge pump module. This module utilizes on-chip capacitors and switching logic to double a low-voltage power supply (such as 1.8V or 3.3V) to generate an adjustable DC negative high voltage of -10V to -30V. This voltage is applied to the common N-electrode of the long-wavelength Micro-LED detector array chip. The programmable high-voltage charge pump module adopts a multi-stage series charge transfer unit structure, utilizing a metal-insulator-metal (MIM) capacitor under standard CMOS technology as the energy storage element. This module drives the power transistors via an internal high-frequency non-overlapping clock, achieving a stepped voltage multiplication conversion from a standard 1.8V or 3.3V supply voltage to -10V to -30V. Furthermore, the reverse bias circuit includes a voltage feedback loop and a reference source, dynamically adjusting the charge pump gain via a digital control interface (such as I2C or SPI). This allows for precise setting of the detector's reverse bias intensity based on the selected long-wavelength Micro-LED adapted to the incident blue light wavelength, improving responsivity while minimizing dark current.

[0039] A potential drawback of Micro-LEDs as detectors is their potentially large junction capacitance (Cj), which limits bandwidth. Under high reverse bias, the device exhibits two physical effects:

[0040] 1) Capacitor compression effect: The approximate formula for the PN junction capacitance of a Micro-LED is C. j = εA / W d Where A is the area, W d The depletion layer width W increases with increasing reverse voltage. d It expands rapidly until it exhausts the entire active region and even extends into the buffer layer. For a tiny pixel with a diameter of 10 μm, the measured capacitance can be reduced to below 10 fF (flfar) at a bias of -30 V. The extremely small capacitance means an extremely small RC time constant, thus laying the foundation for high bandwidth.

[0041] 2) Carrier acceleration effect: The strong electric field causes the photogenerated electrons and holes to move at saturation drift velocity (approximately 10⁻⁶) within the depletion region. 7This significantly shortens the carrier transit time, eliminates the diffusion tail phenomenon common in low-field applications, and makes the detector's response to light pulses steeper and faster. Experiments show that, under a -30 V bias, the -3 dB bandwidth of the red Micro-LED detector can be increased from tens of MHz to over 1 GHz, meeting the requirements of high-speed communication.

[0042] This deep depletion effect not only reduces the junction capacitance, but also creates an extremely high built-in electric field (reaching 10) in the active region. ^5 (On the order of V / cm), causing the drift velocity of photogenerated carriers to reach saturation. Experimental measurements show that under deep reverse bias of -30 V, due to the depletion region W d The significant increase in width compresses the RC time constant of the device to the picosecond (ps) level, which enables long-wavelength Micro-LED detectors to effectively suppress carrier diffusion tailing effects, thereby obtaining steeper pulse rise and fall edges, supporting single-channel non-return-to-zero (NRZ) signal transmission at 2.5 Gbps or even higher baud rates.

[0043] The readout circuit employs a compact transimpedance amplifier (TIA) based on an inverter structure. This TIA directly converts the photocurrent into a voltage signal. Because it eliminates the need to handle large capacitors, the TIA achieves high gain and high bandwidth with low power consumption. The signal is then shaped into a digital level and directly output after a simple limiting amplifier. The entire receiver eliminates the need for complex clock data recovery (CDR) or equalization (EQ) circuitry, significantly reducing power consumption (single-channel power consumption < 1 mW) and chip area.

[0044] To further reduce electrical crosstalk in the detector array, a "virtual ground" bias design is adopted at the input front end of the readout circuit. Through the negative feedback mechanism of the transimpedance amplifier (TIA), the P-electrode potential of each long-wavelength Micro-LED unit is clamped to a stable level. Combined with the deep trench isolation (DTI) structure used in the long-wavelength Micro-LED detector array chip array, the lateral conductive path of the n-GaN layer is cut off, avoiding parasitic currents generated between adjacent units through the n-GaN layer. This "dual isolation" mechanism of circuit and device ensures that electrical crosstalk remains below -60 dB even with extremely small channel spacing. This means that even with strong light signal input to adjacent channels, the dark current noise floor of the tested channel remains at the pA level, ensuring the detection sensitivity of weak signals.

[0045] Reference Figure 5To achieve efficient energy transfer between the transmitter, transmission medium, and receiver, a "1:1 pixel-level" precision optical coupling design is employed. The multi-channel imaging fiber optic link includes a transmitting microlens array, an imaging fiber bundle, and a receiving microlens array. The transmitting microlens array is integrated into the light-emitting surface of the blue Micro-LED transmitter array chip, while the receiving microlens array is integrated into the light-incident surface of the long-wavelength Micro-LED detector array chip. These components form a 1:1 conjugate imaging optical path through the imaging fiber bundle. The microlens array is integrated with the chip using photolithography with thermal reflow or nanoimprint lithography. The pitch of the microlenses is strictly consistent with the pitch of the Micro-LED unit.

[0046] At the emitting end, Micro-LEDs typically exhibit Lambertian light emission characteristics with a large divergence angle (approximately 120 degrees). The function of the emitting microlens array is to collimate this divergent beam into a parallel beam with a smaller divergence angle (approximately 10-15 degrees) to match the numerical aperture (NA) of the imaging fiber, thereby improving fiber coupling efficiency and preventing light leakage.

[0047] When the light signal exits the imaging fiber, it diverges again. The role of the receiving microlens array is to refocus these beams, shrinking their spot diameter to less than 10 μm. The microlenses precisely focus the light energy onto the center of the active region of the long-wavelength Micro-LED unit, which is surrounded by a deep trench isolation (DTI) structure. This dual protection mechanism of "optical focusing + physical isolation" completely eliminates the possibility of edge light spreading to adjacent pixels, ensuring extremely low crosstalk.

[0048] The imaging fiber bundle serves as the intermediate transmission medium, with a total cross-sectional diameter of 3.5 mm, slightly larger than the diagonal size of the Micro-LED array, ensuring coverage of all pixel units. The imaging fiber bundle contains approximately 50,000 individual fiber cores. Due to the extremely high core density, the light spot emitted by each Micro-LED light-emitting unit is transmitted by 9-12 fiber cores on average (oversampling). This design allows the optical signal to be transmitted in the form of an "image" while maintaining the spatial topology within the fiber, enabling the parallel and independent transmission of thousands of signals.

[0049] In one embodiment, the system integrates a 64×64 physical channel array, with the blue Micro-LED emitting array chip and the long-wavelength Micro-LED emitting array chip each employing a 64×64 unit array, resulting in a total of 4096 independent parallel communication channels. These channels are spatially closely arranged, each operating independently without interference, collectively forming a massive data transmission pipeline. Under the "slow" condition of a single-channel data transmission rate set at 2.5 Gbps (compared to 50 Gbps for VCSELs), the total system bandwidth reaches 10 Tbps (4096 × 2.5 Gbps). The large number of channels (width) compensates for the insufficient single-channel rate (slowness), thereby achieving extremely high total throughput while avoiding high-frequency losses and eliminating high-power equalization circuitry.

[0050] The above-mentioned low-power parallel optoelectronic communication system achieves:

[0051] 1. Breakthrough Improvement in Detection Performance. Blue light detectors constructed using InGaN material achieve strong interband absorption by utilizing the principle of narrow bandgap absorption of high-energy photons, compared to the low responsivity (<0.2 A / W) of silicon-based detectors in the blue light band. Furthermore, by adjusting the indium composition, the system can maintain the optimal photon energy difference (ΔE) for any blue light emission wavelength in the 400nm-480nm range, ensuring a high responsivity exceeding 0.3 A / W under different wavelength configurations, significantly improving system flexibility.

[0052] 2. Effective suppression of array crosstalk. For high-density parallel transmission, a deep trench isolation (DTI) structure is introduced at both the transmitter and receiver. Compared with the traditional mesa structure, the DTI structure physically blocks the lateral diffusion of photons and electrons between pixels, effectively solving the crosstalk problem caused by "arraying" and ensuring the signal purity of 4096 channels working independently.

[0053] 3. Excellent high-frequency response and integration. A -30V high-voltage charge pump is integrated into the CMOS readout chip at the receiver. This on-chip integrated design not only solves the problem of external high-voltage power supply, but also compresses the large capacitance of Micro-LED to its limit through the deep depletion effect. Combined with the saturation drift velocity mechanism, this enables the InGaN detector to have bandwidth characteristics comparable to high-speed photodiodes, supporting single-channel Gbps-level transmission.

[0054] 4. Exceptional Energy Efficiency and Cost Advantages. Utilizing a "wide and slow" architecture, the high-power SerDes and DSP circuits are eliminated, resulting in a system energy efficiency better than 0.5 pJ / bit. Simultaneously, the receiver detector directly reuses mature LED epitaxy and chip manufacturing processes, eliminating the need for expensive InP substrates or customized CMOS image sensor processes. The highly standardized material systems at both the transceiver and receiver simplify the packaging process, resulting in significant mass production cost advantages.

[0055] The above embodiments are only used to further illustrate a low-power parallel optoelectronic communication system of the present invention. However, the present invention is not limited to the embodiments. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention shall fall within the protection scope of the technical solution of the present invention.

Claims

1. A low-power parallel optoelectronic communication system, characterized in that: It includes a transmitting subsystem, a receiving subsystem, and a multi-channel imaging fiber optic link; the transmitting subsystem includes a blue light Micro-LED emitting array chip and a driver chip, the receiving subsystem includes a long-wavelength Micro-LED detector array chip and a readout chip; the multi-channel imaging fiber optic link is used to transmit a parallel optical signal matrix between the transmitting subsystem and the receiving subsystem; Among them, long-wavelength Micro-LEDs have In x Ga 1-x In an N / GaN multi-quantum-well active region, x is modulated such that the band gap Eg of the active region is less than the incident photon energy E of the blue micro-LED. ph The readout chip integrates a reverse bias circuit, which is configured to apply a high reverse bias voltage of -10V to -30V to the long-wavelength Micro-LED detector array chip.

2. The low-power parallel optoelectronic communication system according to claim 1, characterized in that: The center emission wavelength of the blue Micro-LED is 400 nm to 480 nm, and the electroluminescence peak wavelength of the long-wavelength Micro-LED is 500 nm to 650 nm.

3. The low-power parallel optoelectronic communication system according to claim 1, characterized in that: In the array of blue light Micro-LED emitting array chip and long wavelength Micro-LED detector array chip, the size of each Micro-LED unit is 3 μm to 50 μm.

4. The low-power parallel optoelectronic communication system according to claim 1, characterized in that: The epitaxial structure of the long-wavelength Micro-LED detector array chip includes a substrate, an N-type layer, and an In layer arranged in sequence. x Ga 1-x The active region of the N / GaN multi-quantum well and the P-type layer are formed by etching to form several arrayed Micro-LED units; the N-type layer and the P-type layer are electrically connected to the readout chip through N electrodes and P electrodes, respectively.

5. The low-power parallel optoelectronic communication system according to claim 4, characterized in that: The etching includes forming trenches that separate each Micro-LED unit, the trenches extending to a portion of the depth of the substrate, the inner walls of the trenches being covered with a passivation layer and filled with a light-blocking material.

6. The low-power parallel optoelectronic communication system according to claim 1, characterized in that: ΔE = E ph - E g The range of ΔE is 0.5 eV to 0.8 eV.

7. The low-power parallel optoelectronic communication system according to claim 1, characterized in that: The multi-channel imaging fiber optic link includes an emitting microlens array, an imaging fiber bundle, and a receiving microlens array. The emitting microlens array is integrated on the light-emitting surface of the blue Micro-LED emitting array chip, and the receiving microlens array is integrated on the light-incident surface of the long-wavelength Micro-LED detector array chip. They form a 1:1 conjugate imaging optical path through the imaging fiber bundle.

8. The low-power parallel optoelectronic communication system according to claim 1, characterized in that: The readout chip includes a readout circuit without clock data recovery, the readout circuit including a transimpedance amplifier and a limiting amplifier; the transimpedance amplifier is configured based on an inverter structure to directly convert the photocurrent generated by the long-wavelength Micro-LED detector array chip into a voltage signal; the limiting amplifier is configured to shape the voltage signal output by the transimpedance amplifier into a digital level.

9. The low-power parallel optoelectronic communication system according to claim 8, characterized in that: The input front end of the readout circuit is provided with a virtual ground bias, which is used to clamp the potential of the long-wavelength Micro-LED.

10. The low-power parallel optoelectronic communication system according to claim 1, characterized in that: In the blue Micro-LED emitting array chip, a deep trench isolation structure is provided between adjacent Micro-LED units; the blue Micro-LED emitting array chip and the driver chip are flip-chip bonded together and driven by the driver chip.

Images