Devices and methods for intensity modulation in photonic computing systems based on interference coupling

By designing the coupling method of ring waveguides and control elements in photonic devices and using the interference coupling modulation method, the optical crosstalk problem introduced by multiple micro-ring resonators was solved, realizing efficient modulation of photonic devices and integration of large-scale neural networks.

CN116783528BActive Publication Date: 2026-11-17HUAWEI TECH CANADA CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202180091506.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-01-25
Publication Date
2026-11-17
Estimated Expiration
2041-01-25

AI Technical Summary

Technical Problem

In the prior art, when multiple microring resonators operate in the same photonic circuit, they are prone to optical crosstalk, which affects the scalability of the circuit. Furthermore, existing photonic modulators have wavelength channel limitations when integrating large neural networks.

Method used

Design a photonic device including a ring waveguide and control elements. By coupling the waveguide in a symmetrical or asymmetrical manner in structure, the propagation characteristics of the optical signal are adjusted by the control elements to achieve controllable modulation of the optical signal. An interference-based modulation method is adopted to maintain a constant resonant wavelength while adjusting the coupling strength to reduce crosstalk.

Benefits of technology

It effectively reduces optical crosstalk, improves the modulation efficiency and scalability of photonic devices, allows for the integration of more wavelength channels, and is suitable for the implementation of large-scale neural networks.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116783528B_ABST
    Figure CN116783528B_ABST
Patent Text Reader

Abstract

The waveguide is coupled at both ends to a microring (MRR) modulator to form a Mach-Zehnder interferometer from the resulting unit cell. The coupled waveguide can be modulated independently of the MRR modulator itself using a resonant photoconductive heater or a geometrically suitable PN junction. Independent modulation of the MRR can be applied during modulation of the voltage applied to the coupler to compensate for wavelength shift. An ideal length for the coupler, and a helical configuration, were determined to minimize the space occupied by the coupler on the photonic integrated circuit. The resulting device reduces crosstalk between channels in the cascaded MRR system. A second coupler can be symmetrically coupled to each MRR to reduce insertion loss, thereby allowing for increased weighting range at the downlink and through-ports.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of optical computing technology, and more particularly to a photonic device. Background Technology

[0002] The von Neumann computing architecture has been widely used in personal computing because its central processing unit and memory are well-suited for running sequential, procedural programs. However, the von Neumann architecture can be inefficient for distributed, massively parallel, and adaptive computing models. These models typically include neural networks in machine learning (ML) for artificial intelligence (AI) and neuromorphic computing applications. To address the shortcomings of current computer architectures in implementing neural networks, increasing efforts from academia and industry are focused on developing specially customized architectures to improve computational speed and reduce power consumption.

[0003] Photonic solutions are considered a mature industrial reality in high-speed, high-throughput data communication and switching infrastructure. For many years, photonic solutions for optical communication and processing have evolved along the same path, aiming to improve transmission speed and energy efficiency. Therefore, optical implementations of neural networks are being investigated to leverage high parallelism through degrees of freedom such as wavelength, polarization, and mode, as well as the high connectivity achievable optically.

[0004] Large-scale analog optical processors have not yet been widely adopted, partly because they cannot be integrated into conventional architectures. Several issues still limit the realization of integrated silicon photonic modulators for large-scale artificial neural network (ANN) systems, including optical crosstalk between different wavelength channels and the limited total number of wavelength channels achievable within a free spectral range (FSR).

[0005] For example, tunable microring resonators (MRRs) have been proposed for use as photonic modulators. Modulation is typically achieved by modulating the refractive index of the MRR, which causes a shift in the resonant wavelength. However, when multiple MRRs are used in the same photonic circuit and operate simultaneously, this wavelength shift introduces optical crosstalk between adjacent resonant peaks, thus affecting the scalability of the circuit.

[0006] Therefore, there is a need for photonic modulation devices and associated methods that eliminate or mitigate one or more limitations of the prior art. Summary of the Invention

[0007] The object of this invention is to provide a photonic device and associated method. The photonic device can be used to controllably modulate optical signals. The photonic device includes a ring waveguide and operates substantially as a controllable MRR (Mean Resonance Regulator). According to some embodiments, the photonic device is balanced, for example, to limit insertion loss and provide a device with structural and functional symmetry. In these embodiments, a first waveguide is operatively coupled to the ring waveguide at the first two coupling locations, and a second waveguide is operatively coupled to the same ring waveguide at the latter two coupling locations. The photonic device may be structurally, functionally, or structurally and functionally symmetrical with respect to the first and second waveguides. According to some embodiments, the photonic device is unbalanced but includes other properties or is coupled to a controller that operates in a particular manner. An unbalanced photonic device includes only a first waveguide operatively coupled to the ring waveguide at two coupling locations, while omitting the second waveguide. In each case, the photonic device includes two, three, or more control elements operatively coupled to two or more waveguides, such as a first waveguide, a second waveguide, and a ring waveguide. The control element can be thermally based or carrier-injection based, used to adjust the optical signal propagation characteristics of a photonic device with respect to frequency. The control element can be referred to as a controllable tuner or an exponential modulator. The control element responds to electronic control signals (e.g., adjustable input voltage or current levels) to adjust the optical properties of the waveguide to which the control element is coupled, for example, through temperature adjustment, carrier density adjustment, or adjustment of another physical property affecting the optical properties of the waveguide, such as the refractive index.

[0008] According to some embodiments, a balanced or unbalanced photonic device includes a first waveguide (and a second waveguide, in a balanced device), characterized as follows. The waveguide has two coupling locations as described above, providing optical coupling to a ring waveguide. The waveguide also has an intermediate portion extending between the two coupling locations into the ring waveguide, such that the intermediate portion is further away from the ring waveguide than the two coupling locations. Furthermore, the intermediate portion is longer than in prior art implementations. This increased length allows the longer portion of the waveguide to be modulated by control elements. This, in turn, allows for increased modulation efficiency of the photonic device to alter the coupling strength. The length can be increased while maintaining a limited footprint by appropriately shaping the intermediate portion. For example, the intermediate portion can have a spiral shape, a meandering shape, etc. The intermediate portion can be shaped as a nearly planar filled curve, or a curve with a length-to-area ratio of at least a given minimum, depending on device and space requirements. The limited footprint of the intermediate portion can be achieved by having at least two opposing bends.

[0009] According to some embodiments, balanced or unbalanced photonic devices include a controller for operating the photonic device by adjusting its coupling power, extinction ratio, or related characteristics, while maintaining a substantially constant resonant wavelength. The resonant wavelength can be maintained, for example, within one percent, a fraction of a hundredth, or a few percent. This can be accomplished by operating two control elements in a complementary manner, one control element coupled to a first (or second) waveguide and the other to a ring waveguide. More specifically, when adjusted, each control element tends to change the resonant wavelength of the photonic device, and the control elements are operated such that a decrease in the resonant wavelength caused by controlling one control element is partially or completely offset by an increase in the resonant wavelength caused by controlling the other control element. Each control element has a different effect on the photonic device, making the combined control influential on the operation of the photonic device. For example, the extinction ratio of the photonic device can be adjusted in this way while maintaining the resonant wavelength. This control method can be advantageously implemented in systems comprising multiple such photonic devices, for example, connected in series.

[0010] According to an embodiment of the present invention, a photonic device, also known as an MRR-based modulator, is provided. The device includes a ring waveguide, also known as the MRR portion of the device. The device includes a first waveguide coupled to the ring waveguide and a second waveguide coupled to the ring waveguide. The first waveguide is coupled to the ring waveguide at a first location and a second location. The first waveguide has a first portion extending from the first location; a second portion extending from the second location; and an intermediate portion located between the first portion and the second portion. The first and second portions may terminate at ports of the photonic device. The intermediate portion extends between the first location and the second location. The intermediate portion is further away from the ring waveguide than the first and second locations (e.g., bent outwards, and may be helical or otherwise bent). Similarly, the second waveguide is coupled to the ring waveguide at a third location and a fourth location. The second waveguide has a third portion extending from the third location; a fourth portion extending from the fourth location; and a second intermediate portion located between the third portion and the fourth portion and extending between the third and fourth locations. The second intermediate portion is further away from the ring waveguide than the third and fourth locations. The first and second waveguides may be substantially symmetrical. The device also includes a first control element operably coupled to the intermediate portion, a second control element operably coupled to the second intermediate portion, and another control element operably coupled to the ring waveguide. The control elements may be PN junction-based or IRPH-based exponential modulation elements. The control elements may operate in a complementary manner to adjust the extinction ratio of the photonic device (or otherwise adjust the intensity variation in the photonic device's transfer function) while maintaining the resonant wavelength of the photonic device at a substantially constant value.

[0011] In various embodiments, the intermediate portion and the first control element interface along a lengthened extension of the intermediate portion. This length provides a coupling efficiency for the first control element sufficient to control the amount of optical power coupled from the first waveguide to the ring waveguide between a minimum value substantially zero and a maximum value substantially equal to all optical power input to the photonic device. In some cases, the length can be several times the circumference of the ring waveguide (e.g., 20 times). The intermediate portion can be at least partially formed into a compact shape by incorporating at least two opposing bends.

[0012] According to an embodiment of the present invention, a photonic device is provided having a ring waveguide and a first waveguide coupled to the ring waveguide at a first location and a second location, such as as described above. A second waveguide may be present, but is not required. If present, the second waveguide may be configured symmetrically with respect to the first waveguide. The photonic device includes a first portion extending from the first location. The photonic device includes a second portion extending from the second location. The photonic device includes an intermediate portion located between the first portion and the second portion and extending between the first location and the second location, the intermediate portion being further away from the ring waveguide than the first location and the second location. The photonic device includes a first control element operatively coupled to the intermediate portion and another control element operatively coupled to the ring waveguide. The intermediate portion and the first control element interface along an extension of the intermediate portion having a length. The length is used to provide a coupling efficiency of the first control element sufficient to control the amount of optical power coupled from the first waveguide to the ring waveguide between a minimum value and a maximum value, the minimum value being substantially zero and the maximum value being substantially equal to all optical power input to the photonic device.

[0013] According to an embodiment of the present invention, a photonic device is provided having a ring waveguide and a first waveguide coupled to the ring waveguide at a first location and a second location, as described above. Similarly, a second waveguide may be present, but is not required. If present, the second waveguide may be configured symmetrically with respect to the first waveguide. As described above, the photonic device includes a first portion extending from the first location, a second portion extending from the second location, and an intermediate portion located between the first portion and the second portion and extending between the first location and the second location. As described above, the intermediate portion is further away from the ring waveguide than the first location and the second location. Also as described above, the device includes a first control element operatively coupled to the intermediate portion and another control element operatively coupled to the ring waveguide. In this embodiment, the device includes a controller for complementaryly operating the first control element and the other control element to adjust an extinction ratio applied by the photonic device at the resonant wavelength of the photonic device. The extinction ratio is between the input to the first waveguide and the output of the first waveguide, while maintaining the resonant wavelength. In some embodiments, the first control element is operated to adjust the coupling strength between the first waveguide and the ring waveguide. The adjustment of the coupling strength tends to change the resonant wavelength, and the other control element is operated to counteract the change in the resonant wavelength.

[0014] In some embodiments, the device includes a second waveguide and a second control element operatively coupled to the second intermediate portion. In these embodiments, the controller is also configured to operate the second control element and the other control element or another control element in a complementary manner to adjust a second extinction ratio applied by the photonic device at the resonant wavelength of the photonic device. The second extinction ratio lies between the input to the second waveguide and the output of the second waveguide, while maintaining the resonant wavelength.

[0015] Embodiments of the present invention provide a method for controlling a photonic device. The photonic device is as described above, and the method includes operations corresponding to the operation of a controller as described above.

[0016] Embodiments of the present invention provide a system comprising, for example, multiple photonic devices connected in series or cascade as described above. Such a system can be integrated into a photonic neural network or other photonic computing device.

[0017] Embodiments have been described above in conjunction with various aspects of the present invention, and these embodiments can be implemented based on these aspects. Those skilled in the art will understand that embodiments can be implemented in conjunction with the aspects describing these embodiments, but may also be implemented together with other embodiments of that aspect. It will be apparent to those skilled in the art that embodiments are mutually exclusive or incompatible with each other. Some embodiments may be described in conjunction with one aspect, but may also be applicable to other aspects, as will be apparent to those skilled in the art. Attached Figure Description

[0018] Other features and advantages of the invention will become apparent from the following detailed description, taken in conjunction with the accompanying drawings, in which:

[0019] Figure 1a This paper illustrates a prior art microring resonator (MRR) modulator with carrier density tuning and thermal tuning.

[0020] Figure 1b It is a set of transmission spectra of existing MRR modulators tuned with voltages from 0V to 4V.

[0021] Figure 2a The prior art “broadcast and weight” configuration is shown, in which an MRR modulator is used as a weighted reassembly, a balanced photodetector is used for output summation, and another MRR modulator operates as part of a nonlinear activation function.

[0022] Figure 2b The power at the down-port and through-port of a prior art MRR modulator is shown, as well as the weight values ​​of the MRR modulator as a function of the bias voltage.

[0023] Figure 3 The prior art in-resonator photoconductive heater (IRPH) MRR modulator (IRPH-MRR) with doped waveguide cross section is shown.

[0024] Figure 4a The resonant spectrum of four cascaded MRRs in the prior art is shown, and how modulation of one MRR can cause crosstalk in the outputs of adjacent MRRs.

[0025] Figure 4b A diagram illustrating how the bias voltage applied to an MRR modulator can affect the optical modulation amplitude (OMA) of adjacent MRRs is presented for the purposes of the prior art.

[0026] Figure 5a The diagram illustrates how vector multiplication can be performed using an MRR modulator in an electro-optic circuit, according to existing technology.

[0027] Figure 5b The diagram illustrates the transmission outputs of the downlink and through-port of an MRR modulator with different weights, as shown in the prior art.

[0028] Figure 5c The diagram illustrates how an MRR modulator can be integrated into a feedforward neural network, using existing techniques.

[0029] Figure 5d A diagram illustrating how an MRR modulator can be integrated into a recurrent neural network, based on existing techniques.

[0030] Figure 5e The diagram illustrates how an MRR modulator can be integrated into a convolutional neural network, using existing techniques.

[0031] Figure 6a An embodiment of an interference-coupled MRR modulator is shown.

[0032] Figure 6b Five (5) transmission spectra of an interference-coupled MRR modulator based on an embodiment are shown, wherein the interference coupler and MRR operate in a complementary manner to allow for a wavelength-locked scheme.

[0033] Figure 7a The prior art interference-coupled MRR modulator is shown, along with mathematically related component parameters, including angle and coupling coefficient.

[0034] Figure 7b An example of an interference-coupled MRR modulator is shown, along with mathematically related component parameters, including angles and coupling coefficients.

[0035] Figure 8a This is a schematic diagram of an existing MRR modulator.

[0036] Figure 8b This is a schematic diagram of an MRR modulator based on interference coupling, as shown in the embodiment.

[0037] Figure 8c Multiple transmission spectra of an MRR modulator with an interference coupler arm are shown in the embodiment, each spectrum corresponding to the output at a through port or down port at a different modulation level.

[0038] Figure 8d The graph shows how the total power output of the MRR modulator with an interference coupler arm varies with the bias voltage, as illustrated in the example.

[0039] Figure 8e The graph shows the power output difference between the down-port and through-port of an MRR modulator with an interference coupler arm as a function of bias voltage.

[0040] Figure 8f The transmission spectrum of an analog MRR modulator with one coupler arm is shown in the embodiment for applying different modulation levels at the interference coupler arm.

[0041] Figure 8g The transmission spectrum of an analog MRR modulator for applying different modulation levels at the MRR is shown in the embodiment.

[0042] Figure 8h The transmission spectra at the downport and through port of an interference-based analog MRR modulator of an embodiment are shown, wherein the wavelength is kept locked by interdependent modulation at the coupler and the MRR.

[0043] Figure 8i The transmission power at the through port and down port of the analog MRR modulator of the embodiment is shown.

[0044] Figure 8j The normalized power output and weight range at the down-port and through-port of the simulated MRR shown in the embodiment vary with the bias voltage to the modulation control element.

[0045] Figure 9a This is a schematic diagram of an interference-coupled MRR modulator with two coupler arms, as an example.

[0046] Figure 9bThe transmission spectra at the through port and down port of an analog MRR modulator with different modulation voltages are shown in an embodiment with two heatable symmetrical interference couplers.

[0047] Figure 9c The graph shows the power output difference between the down-port and through-port of an MRR modulator with two interference coupler arms as a function of bias voltage.

[0048] Figure 10a The diagram illustrates an interference-coupled MRR modulator with two coupler arms, as described in the embodiment. The arms are elongated (e.g., to facilitate higher modulation efficiency) and optionally spiral to reduce their footprint.

[0049] Figure 10aa Another view of an interference-coupled MRR modulator with two coupler arms, as an example.

[0050] Figure 10b This is a schematic diagram of an MRR modulator based on interference coupling, which has two helical coupler arms with PN junctions.

[0051] Figure 10c The simulation of the transmission spectrum of a PN junction-based, interference-coupled MRR modulator based on an embodiment is shown, varying with the bias voltage, where the upper coupler arm πR is longer than the MRR perimeter, and R is the MRR radius.

[0052] Figure 10d The simulation of the transmission spectrum of a PN junction-based, interference-coupled MRR modulator based on an embodiment is shown, varying with the bias voltage, where the upper coupler arm is 20 times the MRR perimeter.

[0053] Figure 11 This is to illustrate how the interference-coupled MRR modulator of the embodiment can be used to develop other embodiments.

[0054] Figure 12a The output spectrum of a cascaded analog reference system consisting of three connected MRRs is shown, where the modulation of Channel_1 affects the optical modulation amplitude (OMA) of Channel_2.

[0055] Figure 12b The output spectrum of a cascaded analog reference system consisting of three connected MRRs is shown, where the modulation of Channel_2 affects the optical modulation amplitude (OMA) of Channel_3.

[0056] Figure 12c The output spectrum of a cascaded analog reference system containing three connected MRRs is shown, where the modulation of Channel_3 affects the optical modulation amplitude (OMA) of the unseen adjacent channel on the right.

[0057] Figure 13a The output spectrum of an analog embodiment containing three cascaded connected MRRs, wherein the modulation of Channel_1 does not affect the OMA of Channel_2.

[0058] Figure 13b The output spectrum of an analog embodiment containing three cascaded connected MRRs, wherein the modulation of Channel_2 does not affect the OMA of Channel_3.

[0059] Figure 13c The output spectrum is a simulated embodiment containing three cascaded connected MRRs, where the modulation of Channel_3 does not affect the OMA of adjacent channels.

[0060] Figure 14a This is a series of overlapping transmission spectra of two cascaded MRR modulators in the prior art, which have two different diameters and no coupler arms, wherein each spectrum corresponds to a different tuning voltage.

[0061] Figure 14b Two diagrams of prior art are shown illustrating the output weights (difference between the down-port transmission output and the through-port transmission output) of two cascaded MRR modulators with two different diameters and no coupler arms.

[0062] Figure 14c As an example, there is a series of overlapping transmission spectra of two cascaded MRR modulators with two different diameters, each MRR modulator having a pair of symmetrical coupler arms, wherein each spectrum corresponds to a different tuning voltage.

[0063] Figure 14d Two graphs illustrating the output weights (difference between the downstream port transmission output and the through port transmission output) of two cascaded MRR modulators are shown in the embodiment. These MRR modulators have two different diameters, and each MRR modulator has a pair of symmetrical coupler arms.

[0064] Figure 15a A graph illustrating the accuracy of dot product calculations performed using the examples and reference simulations as a function of the bit resolution of the modulation driver.

[0065] Figure 15bA graph is shown illustrating how the accuracy of matrix product calculations performed using the examples and reference simulations varies with the range of the input matrix.

[0066] Figure 16a An example of a convolutional neural network that can be constructed using an MRR modulator is shown.

[0067] Figure 16b A diagram illustrating the common weights and normalized weights of nine cascaded MRR modulators is shown as an example.

[0068] Figure 16c The transmission diagrams, in decibels and milliwatts, are shown for the downlink and through-link ports of the nine cascaded MRR modulators in the embodiment.

[0069] It should be noted that the same features are identified by the same reference numerals throughout the accompanying drawings. Detailed Implementation

[0070] Silicon photonics is an emerging technology that can be used to build large-scale photonic integrated circuits (PICs) into a single package. These silicon PICs can be used in a variety of applications, most of which can be manufactured using established facilities, including complementary metal oxide semiconductor (CMOS) foundries, which are cost-effective for mass production. Optical signal transmission can be achieved through tunable waveguide elements such as microring resonators (MRRs), modulated by carrier density tuning through the addition of a PN-doped region (a junction between positively and negatively doped materials) in the waveguide core, or by thermal tuning, which is achieved by adding a metal heater near the waveguide (e.g., at the top of the waveguide). The transmitted output signal of the waveguide element can be used for data communication by biasing the intensity modulator with an AC signal.

[0071] Figure 1a An MRR modulator 110 is shown, in which the refractive index can be changed with carrier density tuning and thermal tuning. Cross-sectional image 120 shows fabrication details of the modulator; the PN junction doped in the waveguide and the metal heater can both be used for optical signal modulation. In both the carrier density control element and the thermal control element, a single tuning technique may be sufficient.

[0072] Figure 1b This is a set of transmission spectra of an MRR modulator tuned from 0V to 4V. It is worth noting that in typical (e.g., prior art) systems, the resonant wavelength of the MRR shifts as the voltage applied to the control element is adjusted.

[0073] Neuromorphic computing can be implemented using a “broadcast and weights” architecture. This architecture can be used to implement neuromorphic processors for PICs using MRR modulators, which in turn can be used to implement generalized fully programmable network models. In this approach, the output of each neuron is assigned a unique wavelength carrier, which is wavelength division multiplexed (WDM) and broadcast. The input WDM signal is first weighted by a set of continuous-value filters used as configurable photon weight groups, and then the output of the weighted reassembly is summed by total power monitoring. This architecture can be implemented using a set of tunable add-drop MRR modulators that recreate on-chip synaptic weights and simulate biological neurons. By modulating the refractive index (RI) of the waveguide core in each modulator, the transmission intensity at their downlink and through-ports will change accordingly. This individually changes the weights in the weighted reassembly array, i.e., the intensity difference (IR) between the input and output of the MRR intensity. Drop -I Thru ).

[0074] Figure 2a The “broadcast and weighted” configuration is shown, in which the MRR modulator acts as a weighted reassembly unit. The MRR has a downlink port 205 and a through port 210, each of which is connected to the input 215 of a balanced-photodetector (BPD) 220, acting as a subtractor between the two ports (each individual PD adds the total optical power in the operating optical band).

[0075] Figure 2b The transmitted power at the down-port and through-port of the MRR modulator is shown, along with the weighted values ​​as a function of the bias voltage.

[0076] Microring resonators (MRRs) and Mach-Zehnder interferometers (MZIs), along with their derivatives and variants, are important optical modulation architectures that can be studied for the high-speed, high-throughput data communication required by modern data centers. They can also be used in neural network systems as key components of neuron simulators.

[0077] Many modulator architectures can be investigated for high-speed optical interconnects in data centers and high-performance computing systems. Among them, MRR modulators using in-resonator photoconductive heaters (IRPH) show advantages in applications requiring control of large-scale optical systems. The control element in an IRPH-MRR modulator can be based on a doped silicon waveguide. By combining the photoconductive effect of the doped waveguide with its thermo-optical tuning capability, the IRPH-MRR modulator according to an embodiment can be used to monitor and track the resonant conditions of the MRR separately. This can reduce or eliminate fabrication steps such as Ge deposition or Si+ implantation for on-chip PDs, and such an architecture can allow for a reduction in the number of electrical inputs / outputs of the system. The IRPH-MRR modulator can be studied as a weighted reassembly in a large-scale artificial neural network (ANN) and achieves 4.2-bit accuracy without any direct transmission measurements.

[0078] Figure 3 This is a schematic diagram of an in-resonator photoconductive heater (IRPH) MRR modulator 305 with an in-resonator photoconductive heater 310 (IRPH; combined IRPH-MRR), which has an n-doped waveguide cross section 315.

[0079] However, since the refractive index (RI) tuning scheme of the IRPH-MRR modulator is mainly based on thermo-optic modulation, the tuning bandwidth is only about 175 kHz, which is far below the requirements of high-speed, high-throughput communication in neural networks. Therefore, carrier density modulation inside the waveguide core using a PN junction can be used for modulators implemented in weighted arrays. This can provide electro-optic (EO) modulation bandwidths up to 40 GHz. Furthermore, by optimizing the optical mode overlapping with the depletion region in the waveguide, for example, using L-shaped or U-shaped vertical PN junctions, tuning efficiency can be effectively improved, thereby increasing the optical modulation amplitude (OMA). For IRPH-MRR or PN-MRR modulators, the output intensity modulation depends on the peak shift of the resonant wavelength when a bias voltage is applied; unfortunately, this affects the OMA of adjacent resonant peaks, leading to optical crosstalk.

[0080] Figure 4aFour transmission spectra are shown for three different stages, specifically for the signal at the operating wavelength λ2. When multiple MRR modulators are cascaded in a row, each modulator's transmission spectrum has a peak centered at a different wavelength (corresponding to maximum attenuation). As shown, the wavelength shift (typically wavelength reduction, also known as redshift) of the peak at λ1405 affects the OMA of the peak at λ2410. This is called optical crosstalk. The more MRR modulators cascaded in a row, the greater the crosstalk effect.

[0081] Figure 4a The top figure 415 shows the individual transmission spectrum of each of the four MRRs overlaid on the same graph. The middle figure 420 shows the total transmission spectrum of the four MRRs when they are coupled together, for example, in series. Crosstalk exists because the bandwidth of the peaks is non-zero and the peaks are relatively close, preventing the transmission spectrum from reaching its maximum value between the peaks. This affects characteristics such as extinction ratio and can negatively impact the operation of the corresponding optical circuitry. The bottom figure 425 shows a further degradation of these characteristics due to optical crosstalk when the resonant wavelength of one of the MRRs is adjusted. This illustrates a problem in the general operation of circuits involving multiple MRRs, where a wavelength shift in one MRR can affect the operation of the others when different MRRs have similar resonant wavelengths.

[0082] As shown in Figure 425 at the bottom, when the resonant peak from λ1 moves to position 412 between λ1 and λ2, the corresponding MRR will affect adjacent wavelengths, as indicated by the dashed curve extending from the peak at position 430. This will interfere with the operation of the next MRR by reducing the height of the curve leading to the peak at position 410 of λ2 (the other dashed curves). Therefore, if the first MRR operates at the resonant peak at position 412, and the second MRR operates at the second resonant peak modulated from λ2, then regardless of how much the second resonant peak drifts, the maximum power will only be the point where the dashed curve leading from the resonant peak 412 overlaps with the dashed line leading to the resonant peak at λ2 410.

[0083] Figure 4b The diagram shows the optical modulation amplitude (OMA) of the MRR resonant peaks as the resonant peaks of adjacent MRRs shift. As the resonant peak λ1 shifts, the power transmitted at the downport of the λ2 MRR decreases by 430, and the power transmitted at the throughport of the λ2 MRR increases by 440, but the detectable power difference of λ2 cannot exceed the reduced OMA of 450 and the available weight of 460.

[0084] To address crosstalk, absorption-based modulation schemes can be used. For example, MRR modulators fabricated on silicon-on-insulator (SOI) platforms can utilize graphene-based electroabsorption mechanisms to tune the intensity at a fixed wavelength. However, this concept can lead to other problems, such as the low electroabsorption efficiency of graphene, which may result in small OMA (Original Motion Amplification). Furthermore, the additional system heat caused by absorption can cause the resonant peak to drift to longer wavelengths. Additionally, the post-fabrication process for graphene layer growth is more complex compared to PN or IRPH doping.

[0085] Reducing or eliminating crosstalk can allow MRR modulators to be cascaded and used as weight reassemblies in neural networks (e.g., large-scale deep neural networks). Examples of neural networks include feedforward neural networks, recurrent neural networks, and convolutional neural networks.

[0086] Figure 5a This diagram illustrates how a vector multiplication can be performed using an MRR modulator in an electro-optic circuit. Wavelength 510 is used to represent the vector dimension, and the first vector component is encoded as an MRR modulation intensity 520 at that wavelength. The second vector component is encoded as another intensity modulation by an MRR modulator 530, which effectively performs the product of the first and second vector components. If necessary, the component products are collected at a balanced photodiode 540 and added to other products.

[0087] Figure 5b This is a diagram showing the transmission output power and output weight (downstream port power - through port power) 560 at the downstream port 550 and through port 555 of the MRR modulator 530.

[0088] Figure 5c This is a schematic diagram of a feedforward neural network. A set of MRR modulators 530 can play a role in weight reconfiguration during the computation of the feedforward neural network. Embodiments include feedforward neural networks in which the MRR modulators are configured according to embodiments described further.

[0089] Figure 5d This is a schematic diagram of a recurrent neural network. A set of MRR modulators 530 can play a role in weight reconfiguration during the computation of the recurrent neural network. Embodiments include recurrent neural networks in which the MRR modulators are configured according to embodiments described further.

[0090] Figure 5e This is a schematic diagram of a convolutional neural network. A set of MRR modulators 530 can play a role in weight reconfiguration during the computation of the convolutional neural network. Embodiments include convolutional neural networks in which the MRR modulators are configured according to embodiments described further.

[0091] Silicon photonics integration offers fundamental new opportunities for large-scale analog information processing systems, potentially at a low cost. The embodiments aim to address the aforementioned problems, particularly crosstalk, to facilitate the development of large-scale optical processors on SOI platforms. The embodiments employ an interference-coupled modulation method, applying intensity tuning at a substantially fixed wavelength instead of tuning the wavelength. This approach can potentially eliminate crosstalk. According to intensity tuning, the modulator's transmission spectrum is adjusted relative to the height (depth) of the resonant peak while maintaining the resonant wavelength at a relatively constant value. That is, the extinction ratio applied by the modulator at the resonant wavelength is adjusted, where the extinction ratio measures the change in optical power between a specific input and a specific output of the photonic device (e.g., between an uplink port and a through port). This can also be described as intensity tuning or coupling power adjustment. The refractive index of the coupler can be changed based on a control signal, which alters the proportion of optical input coupled to the ring waveguide of the MRR.

[0092] Because the intensity variations in this embodiment occur at a fixed spectral location, i.e., at the resonant wavelength, the impact on adjacent wavelength channels is minimal (or nonexistent). Furthermore, if there were an impact on adjacent channels, adjusting the extinction ratio of the photonic device rather than the wavelength might have a smaller effect on other channels. Additionally, compared to a standard modulator with the same radius, this embodiment provides twice the free spectral range (FSR), allowing for a significantly larger number of wavelength channels in large-scale integrated systems.

[0093] In one embodiment, a waveguide length, acting as a two-point interference coupler, is introduced as an adjacent portion of the MRR. The length of the additional waveguide can be configured to provide the modulation efficiency required by the internal modulation elements. This allows embodiments to adjust the coupling strength to regulate the power coupled at the resonator and externally, thereby providing a substantially complete range of intensity variation at the output. The intensity variation can range from substantially zero to substantially 100% of the input power. A device that includes the MRR, the additional waveguide, and any components used to modulate the properties of the MRR can be collectively referred to as a modulator, or more generally as a photonic device. The additional waveguide can be referred to as a coupler, coupler arm, and interference coupler.

[0094] Figure 6aAn embodiment of an MRR modulator based on interference coupling is shown. A microring resonator (MRR) ring waveguide 605 is coupled to a separate waveguide 610 at two locations. The coupled waveguides serve as arms of a Mach-Zehnder interferometer and can be referred to as interference couplers. The input signal 615 can be split at a first coupling location 620 and recombined at a second coupling location 625, such that the throughput signal 630 has an intensity depending on the path length difference between the two waveguide portions (i.e., the middle portion of waveguide 610 and the ring waveguide 605) from which the two parts of the split signal propagate. A portion of the signal in the ring can also be released from the lower port 635 of the MRR modulator. PN junction A 640 can be used to modulate the refractive index of the coupler, and PN junction B 645 can be used to modulate the refractive index of the MRR modulator separately. PN junctions A and B are part of a control element. An in-resonance photoconductive heater (IRPH) 650 can also be used to adjust the MRR modulator. In one embodiment, modulation can be performed using electronic circuitry including a first controllable voltage source 652 operably coupled to PN junction A 640 and a second controllable voltage source 653 operably coupled to PN junction B 645. The first controllable voltage source and PN junction A may be referred to as a first control element, and the second controllable voltage source and PN junction B may be referred to as another control element. The controller can operate the first and second controllable voltage sources 652 and 653 in a complementary manner to adjust the extinction ratio of the MRR modulator, as described herein.

[0095] To achieve intensity modulation at a fixed wavelength, embodiments may include electrical modulation components in the coupler arm (PN junction A) and the MRR (PN junction B). By tuning the voltage of the PN junction (PN junction A), with N-doped and P-doped materials located on either side of the coupler, the power coupling coefficient (κ2) from each of the two coupling interfaces can be adjusted, resulting in a modulated transmission output. However, this may cause a slight drift in the resonant wavelength (i.e., peak wavelength) of the MRR due to a change in the effective refractive index of its waveguide. To lock the resonant wavelength at a fixed value, compensatory modulation can be applied by simultaneously and independently tuning the PN junction (PN junction B) of the MRR. This compensatory modulation can re-adjust the resonant wavelength to its original wavelength. In other words, PN junction A and PN junction B are controlled and operated in a complementary manner to adjust the extinction ratio of the photonic device at the resonant wavelength while maintaining the resonant wavelength. Maintaining the resonant wavelength can involve keeping the resonant wavelength constant, substantially constant, or significantly limiting changes to the resonant wavelength. Some changes to the resonant wavelength are tolerable; however, since it is not the primary means of achieving modulation, such changes are considered secondary and can be reduced. Adjusting the extinction ratio is also known as adjusting the amount of attenuation applied by the photonic device at the resonant wavelength, or intensity tuning of the photonic device. For further clarity, operating multiple control elements in a complementary manner (also known as complementary operation) involves controlling multiple control elements in a complementary manner such that the resonant wavelength remains substantially constant.

[0096] In some embodiments, complementary operation involves the following. The coupling strength of light from the first or second waveguide to the ring waveguide can also be adjusted when the refractive index of the middle portion of the first or second waveguide is adjusted (tuned). Adjusting the coupling strength of light from the first or second waveguide to the ring waveguide (e.g., using a first control element) can be used to control the extinction ratio at the output of the photonic device. However, an exponential change in the coupler (e.g., as part of the first control element) will also tend to cause the resonant peak position to drift (in the shorter wavelength direction). To fix the resonant peak at a substantially constant wavelength while also achieving extinction ratio modification, embodiments of the invention use another control element (exponential modulation element) in the ring waveguide. This other control element is operated to shift the resonant peak back (towards the higher wavelength direction). This other control element can be another exponential modulation element in the ring and can be referred to as a wavelength complementary element. Thus, complementary operation is based on wavelength complementarity to lock the peak, achieved by using two exponential modulation elements (the first and the other control elements) in the coupler and the ring, respectively.

[0097] Figure 6b The transmission spectrum at the output of the MRR modulator of the embodiment is shown, for example, as Figure 6aAs shown, the MRR modulator has an interference coupler. When a reference voltage Bias1 PN A is applied to PN junction A 640, a first transmission spectrum 675 is obtained. When different voltages Bias2 PN A are applied to junction PN A 640, the transmission spectrum 680 exhibits different intensity levels (extinction ratios) and wavelength shifts. However, since PN junction B 645 of the MRR can be tuned individually, by applying a voltage Compensate2 PN B to PN junction B 645 in addition to Bias2 PN A, the transmission spectrum 680 can be shifted to coincide with the original spectrum position 675 685. Similarly, when a voltage Bias3 PN A is applied to PN junction A, a transmission spectrum 690 is obtained. When a voltage Compensate3 PN B is applied to PN junction B in addition to Bias3 PN A at PN junction A, a transmission spectrum 695 is obtained.

[0098] Therefore, by applying different voltage combinations at PN junctions A and B, i.e., by operating the two control elements in a complementary manner, the intensity of the transmission spectrum can be adjusted while keeping the resonant wavelength substantially constant. Different spectra 675, 685, and 695 illustrate this effect, where the extinction ratio is adjusted. The extinction ratio refers to the amount by which the input signal at the resonant wavelength can be adjusted. The extinction ratio can be between the input to the photonic device (e.g., the input to the first waveguide) and the output of the photonic device (e.g., the output of the first waveguide, such as a through port). A continuum of different spectra can be realized, such as an intermediate between spectra 675 and 695.

[0099] In one embodiment, coupling strength modulation between the interference coupler arm and the MRR can be used to implement an output intensity interrogation scheme at a fixed wavelength. This effectively eliminates optical crosstalk problems encountered when using wavelength-shifting modulation schemes. The interrogation scheme may involve a feedback control scheme designed to maintain the desired input / output power level behavior of the photonic device. Feedback can be obtained during a separate offline calibration operation. Depending on the interrogation scheme, tests can be performed to obtain lookup tables or other relational information. This information indicates a combination of control signals to be sent to multiple control elements of the photonic device, resulting in different extinction ratios while maintaining a substantially fixed resonant wavelength. The calibration operation can be performed at a constant (e.g., minimum) power level of the optical signal.

[0100] In one embodiment, the coupler arm can be coupled to the MRR ring waveguide at two locations, and this can be modulated to adjust the extinction of the resonant peak. This dual-location coupling is characterized in that the middle portion of the coupler arm waveguide is typically further away from the MRR ring waveguide than the two coupling locations. For example, in… Figure 6aIn the process, the middle portion of waveguide 610 is further away from the ring waveguide 605 than either coupling position 620 or 625. The power coupling strength between the MRR and the dual-position coupler arm can be adjusted by applying a bias voltage to the control element in the coupler, where both the MRR and the dual-position coupler arm act as arms of a Mach-Zehnder interferometer (MZI). For example, if a thermally based modulation element is added, the middle portion of the dual-position coupler can be further away from the coupling portion of the ring (L... MRR ) πR. However, for PN junction-based modulation elements, the middle section can be 20 times longer than the coupled section of the ring in order to achieve similar modulation efficiency for full-coupling-strength modulation (between 0% and 100% of the input power).

[0101] Figure 7a The geometry of an MRR modulator based on interference coupling is shown. It is an exemplary topology that can be adapted for use in embodiments of the invention. Adaptation may include making the topology symmetrical by adding a second waveguide coupled to the bottom of the ring waveguide, adding and manipulating control elements in a complementary manner, adjusting the shape and length of the middle portion of the waveguide, or combinations thereof. The second waveguide is shown. Regarding... Figure 7a An input signal 705 is applied at waveguide 710, which extends as a two-point interference coupler 715 to a length πR, where R is the radius of the microring resonator ring (MRR) waveguide 720. In other embodiments, this length may be further extended. A first coupling occurs at the MRR input interface 725 and has a certain power coupling coefficient. A second coupling occurs at the MRR output interface 730 and also has a power coupling coefficient. In one embodiment, the couplings have a symmetrical geometry, and the power coupling coefficient at each interface is κ. 2 With the center 735 of the MRR ring waveguide located on the first horizontal axis 740 and the curvature center 745 of the input waveguide 710 located on the second horizontal axis 750, the angle θ755 can be defined as the angle between the first horizontal axis and the line 760 connecting the two curvature centers.

[0102] Further in Figure 7a In the case where the curvature center of the two-point interference coupler is set on the third horizontal axis 765°, the angle It can be defined as the angle between the third horizontal axis 765 and the line 775 connecting the curvature center 780 of the two-point interference coupler and the curvature center 745 of the input waveguide.

[0103] Further in Figure 7aIn embodiments where the MRR ring waveguide and coupler waveguide have the same width, the common width can be referred to as w 792. In embodiments where the first coupling interface 725 and the second coupling interface 730 each have similar gaps and exhibit symmetry, the gaps can have the same width, which can be referred to as s1 794.

[0104] Further in Figure 7a In the middle, when the MRR ring waveguide and coupler are heated to 785, the power coupling coefficient κ 2 Modulated 790. Using the parameters defined in Figure 7, the power coupling coefficient κ in the embodiment can be calculated. 2 By choosing the path difference ΔL between the coupler arm and the MRR to be equal to half the length of the ring, as with the MZI, the second resonant peak from the MRR can be suppressed by the double FSR of the MZI, where the coupling strength is completely suppressed. Geometric considerations indicate that the correct path length is derived from the following constraints:

[0105]

[0106]

[0107] Similarly, θ and φ are as follows: Figure 7a The angles depicted in the equations are: R is the MRR radius, w is the waveguide width, s1 is the spacing between adjacent waveguides, and s3 is the bend (i.e., fault line) introduced between waveguide arcs to improve mode matching. Combining the equations above, we obtain the solution for the following angle φ.

[0108]

[0109] In addition, κ 2 It can be obtained using the following equation:

[0110]

[0111] Figure 7b This illustration shows a configuration where two coupler arms are symmetrically positioned to the MRR ring. Equation parameters can be labeled in a manner similar to the single-arm case. This embodiment is discussed in other paragraphs.

[0112] By modulating the refractive index (RI) of the waveguide used as the coupler arm, the power coupling coefficient (κ2) can be tuned from critical coupling to overcoupling or undercoupling, thus producing an intensity variation at the resonant wavelength. Intensity modulation is achieved because a portion of the input power propagates to the through port, while another portion is trapped within the resonator. However, this tuning not only alters the output intensity but also changes the wavelength of the resonant peak.

[0113] To achieve modulation intensity without altering the resonant wavelength, one embodiment utilizes a compensation tuning scheme comprising two refractive index (RI) control elements, one in the MZI coupler arm (also referred to as the first waveguide) and one in the MRR (ring waveguide) section. When the control element in the MZI coupler arm (e.g., the first or second waveguide coupled to the ring waveguide, which may be a PN junction or a heater) is adjusted, the transmission intensity changes due to the altered coupling strength. In various embodiments, the operation of this control element adjusts the coupling strength between the first waveguide and the ring waveguide. This adjustment ends to change the resonant wavelength of the photonic device. The other control element coupled to the ring waveguide is operated to counteract this change in resonant wavelength. The control elements are typically electrically controllable, for example, controlled by digital control circuitry via a digital-to-analog converter, or by analog control circuitry. For example, the control elements can be controlled by voltage or current, as readily understood by those skilled in the art.

[0114] According to an embodiment, an MRR modulator based on an interference coupler can have a single interference coupler heated by an IRPH. This can be referred to as an asymmetric IRPH MRR modulator design.

[0115] When the refractive index (RI) of the MZI coupler arm waveguide is tuned, the resonance will also be shifted. In one embodiment, this shift can be compensated for by adding an independently controllable RI control element to the MRR section. By applying an appropriate bias voltage to the second control element, the resonant wavelength (also known as the resonant peak) can be shifted in the opposite direction. Intensity tuning at a fixed resonant wavelength can be achieved by fine-tuning the two control elements. In one embodiment, sufficiently fine adjustment can be achieved by real-time detection of the pass power. Both control elements are independently controllable. However, the control elements can cooperate in the sense that the control signal provided to one control element compensates for the resonant wavelength drift caused by the control signal provided to the other control element. Two, three, or more control elements can be controlled cooperatively in this manner.

[0116] In one embodiment, a coupler arm may be added to each MRR in the weighted reassembly of the “broadcast and weight” protocol architecture, and in other embodiments, a downlink port for monitoring modulation may also be added to each MRR.

[0117] Figure 8a An MRR-based modulator 805 without an interference coupler 805 is shown for use with subsequent... Figure 8bA comparison is made. A substantially ring-shaped waveguide 806 is provided, tightly coupled to a pair of straight waveguides 808a and 808b, which provide an add-in port, a drop-out port, and a through port. The device is a add-drop or dual-bus MRR-based modulator. The upper waveguides of the two waveguides 808a and 808b terminate at two ports, referred to as the add-in port and the drop-out port. The lower waveguides of the two waveguides 808a and 808b also terminate at two ports, referred to as the input port 803 and the through port 804. Light received at the input port, if not resonantly coupled to the ring waveguide, will exit at the through port. Light received at the input port, to the extent that it is resonantly coupled to the ring waveguide, will exit at the drop-out port 801. An add-in port 802 is also shown, which can operate similarly to the input port, or can be used to add light to the drop-out port 801.

[0118] Figure 8b An embodiment of an asymmetric IRPH MRR modulator design is illustrated, showing an MRR-based modulator photonic device 810 with an interference coupler. A ring waveguide 811 is provided and coupled to a first waveguide 812 at a pair of coupling locations 813. An intermediate portion of the first waveguide 814 extending between the coupling locations bends away from the ring waveguide. Control elements are operatively coupled to the intermediate portion of the ring waveguide 811 and the first waveguide 814. The first waveguide may terminate at an add-on port and a through port.

[0119] Due to the imbalance in coupling loss between the two waveguides, the insertion loss (IL) of the interference coupler can be very high, which may limit the weight range at the downlink and through ports.

[0120] Figure 8c Multiple transmission spectra of an MRR modulator (unbalanced photonic device) with a single interference coupler arm of an embodiment are shown. A first plurality of spectra 825 corresponds to the output at a through port. A second plurality of spectra 825 corresponds to the output at a downlink port 830. Each of the plurality of spectra includes a different spectrum at a different modulation level. Both sets of spectra show an adjustable extinction ratio while maintaining the same resonant wavelength. That is, the peaks or valleys in the figure appear at the same wavelength. However, due to intensity control (extinction ratio control), the levels at the peaks or valleys can be adjusted in a complementary manner by appropriate operation of the control elements of the photonic device, as described above. Figure 8c Can be with Figure 9b Comparison, Figure 9b This is for an embodiment with two symmetrical interference coupler arms, not... Figure 9b The illustrated embodiment features reduced insertion loss.

[0121] When a signal propagates from one device to another, power loss due to design flaws can be expected. This power loss can be referred to as insertion loss. In one embodiment, the insertion loss can be expected between the optical power injected into the device and the total output optical power of the MRR modulator (the combined power at the through port and the dropout port).

[0122] Figure 8d This is a graph showing the total power output 835 of an MRR modulator with an interference coupler arm as a function of bias voltage in an embodiment. In such an embodiment, the insertion loss 840 can be higher than desired. Insertion loss can be defined as the difference between the total input power and the total output power of a device. Due to inherent loss mechanisms, the total output power is always lower than the input. Figure 8d The diagram shows the total output power of the add-drop photonic device, which includes power from the through port (T) and the drop-down port (D). For example, when the bias voltage is between 0 and 1.5V, the input power can be 1mW, and the total output power can be 0.25mW(D+T). Therefore, when using an asymmetric design, the insertion loss can be significant. A symmetric design can improve the insertion loss.

[0123] Figure 8e The weight W 845 for the signal is shown as a graph of the bias voltage applied to the MRR modulator based on a single-arm interference coupler, where the weight is the difference between the down-port power D and the through-port power T (W = D – T). The resulting weight ranges from -0.6 to 0.25 (mW). This can also be referred to as an unbalanced weighted 850 photonic device.

[0124] Figure 8f A series of transmission spectra of an MRR modulator based on a single-arm interference coupler, according to an embodiment, are shown, with a voltage 855 applied to the coupler arm of the MRR modulator. Each square 860 is a peak, and they typically exhibit a blue shift (increase in resonant wavelength) as the heater voltage 865 increases.

[0125] Figure 8g A series of transmission spectra of an MRR modulator based on a single-arm interference coupler, as illustrated in this embodiment, are shown, with a voltage applied to the MRR modulator itself 870. Each square 875 is a peak, and they typically exhibit a redshift as the heater voltage increases 880.

[0126] Figure 8hA series of transmission spectra of an MRR modulator based on a single-arm interference coupler according to an embodiment are shown. The output power spectrum at the through port is shown from the top 885, and the output power spectrum at the lower port is shown from the bottom 890. The peak values ​​remain near the same wavelength because the independent tuning between the coupler arm and the MRR modulator according to the embodiment allows a user or automatic control scheme to readjust the peak position to a constant wavelength. By simultaneously tuning two independent modulating elements, one for each arm, the intensity of the resonant wavelength (resonant peak) is changed. This can be referred to as a locked wavelength spectrum. The figure shows that the wavelength can remain constant for any selected power.

[0127] Figure 8i The output power 891 at the downstream port, the output power 892 at the through port, the total output power 893, and the resulting weight 894 are shown as functions of the voltage applied to the MRR modulator based on the interference coupler according to the embodiment.

[0128] By subtracting the power between the downlink port and the through port, a weighted range that balances to 0 at large swings can be obtained as a function of the bias voltage applied to the control element of the coupler. To redistribute the weighted range to ±1, a normalization step can be added. Examples include a modulator based on symmetrical interferometric coupling utilizing an intensity-tuned interrogation scheme at a fixed wavelength.

[0129] Figure 8j The following is illustrated as a function of the voltage applied to the interference coupler arm of the MRR modulator according to an embodiment: Figure 8j The original weights 894 are obtained, but they are normalized in the range [-1,+1] using a series of common weights 896.

[0130] To limit or minimize the insertion loss (IL) of the modulator, embodiments of a modulator based on add-drop interference coupling may include two coupler arms located on opposite sides of the MRR, and modulation may be performed using photoconductive heaters (IRPHs) within the first resonators of the two coupler arms and a second IRPH of the MRR. By controlling the first IRPH control elements of the coupler arms and the second IRPH control elements of the MRR respectively, the resonant peak can be shifted in the desired direction or maintained at a constant value. In some embodiments, when tuned using the IRPHs in the coupler, the transmission strength at the through port can vary significantly from 0 to -60 dB.

[0131] By introducing two symmetrical add-drop interference couplers into the modulator, the insertion loss of the embodiment can be limited or minimized due to the equal losses on both arms. This allows the total power from the drop port and the through port to be relatively constant. The weighting value can be obtained by subtracting the power at the drop port from the power at the through port, achieving zero-point balance with a larger swing compared to a single arm. This can be an important consideration in "broadcast and weighted" protocol applications.

[0132] Figure 9a This is a schematic diagram of an MRR modulator with two symmetrical interference couplers, as illustrated in the embodiment. The device may include a central MRR ring waveguide 910, an upper coupler arm first waveguide 920, a lower coupler arm second waveguide 930, an input port 940, a lower port 950, and a through port 960. The second waveguide 930 has a portion extending to the input port 940 and another portion extending to the through port 960. The first waveguide has a portion extending to the lower port 950. The first waveguide 920 is coupled to the ring waveguide 910 at a first coupling position 962 and a second coupling position 963. The second waveguide 930 is coupled to the ring waveguide 910 at a third coupling position 964 and a fourth coupling position 965.

[0133] The first waveguide 920 has an intermediate portion 924 located between the other two portions of the waveguide and extending between a first coupling position and a second coupling position, while bending further away from the ring waveguide than the coupling positions. The second waveguide 930 has another intermediate portion 934 located between the other two portions of the second waveguide and extending between a third coupling position and a fourth coupling position, while bending further away from the ring waveguide than these coupling positions. Control elements, such as PN junction-based control elements or IRPH-based control elements, are operatively coupled to the first waveguide, the second waveguide, and the ring waveguide.

[0134] In some embodiments, a single control element is coupled to a ring waveguide and operates in a complementary manner to a control element coupled to a first waveguide. This single control element may also operate in a complementary manner to a control element coupled to a second waveguide, for example, simultaneously or at different times. In some embodiments, multiple control elements may be coupled to a ring waveguide and operate individually or jointly in a complementary manner to control elements coupled to the first and second waveguides. For example, the ring waveguide may be coupled to a first control element for complementary operation with a control element of the first waveguide, and the ring waveguide may be coupled to a second control element for complementary operation with a control element of the second waveguide.

[0135] More in detail, Figure 9aAn example of a symmetrical photonic device (e.g., modulated by an IRPH control element) is shown. The first part of the device, comprising the upper half of a first waveguide 920 and a ring waveguide 910, is mirror-symmetrical to the second part of the device, comprising the lower half of a second waveguide 930 and a ring waveguide 910.

[0136] Figure 9b The transmission spectrum is shown when a bias voltage is simultaneously applied to both IRPH control elements of an interferometric MRR modulator to vary the output intensity while locking the resonant peak at a fixed wavelength. Due to their symmetrical geometry, both the dropout waveguide and the punch-through waveguide exhibit similar propagation and coupling losses, which contributes to low insertion loss at the bottom port. Figure 8c The comparison shows that using two symmetrical interference coupler arms reduces insertion loss.

[0137] Figure 9c The resulting weights in the range [-0.8, +0.8] are shown as a function of the voltage applied to the interference coupler arms of the MRR modulator according to an embodiment. The two-arm symmetrical interferometric device with first and second waveguides coupled to a ring waveguide can be considered balanced. Bias voltages are applied simultaneously to the first and second waveguides (both arms). Figure 8e In comparison, a device with two coupler arms is much heavier than a device with one coupler arm.

[0138] In some embodiments, a PN junction can be used to modulate the waveguide instead of an in-resonator photoconductive heater (IRPH). An IRPH can have higher tuning efficiency than a PN junction because the change in waveguide refractive index can be greater for the same applied voltage. However, a drawback of the IRPH is its low electro-optic bandwidth. The use of a PN junction can increase the electro-optic bandwidth to up to 40 GHz, which is suitable for high-speed data communication and processing.

[0139] To make the embodiments applicable to the field of RF speed communications, which is also crucial for artificial neural network (ANN) systems, the embodiments include designs of interferometric coupling-based modulators based on at least one PN junction. In some embodiments, the tuning efficiency of the PN junction can be approximately 20-50 pm / V, which is too low for some applications. To improve the modulation efficiency of the PN junction, some embodiments may include coupling arms that are much longer than simple πR arms. PN junction-based Mach-Zehnder modulators (MZMs) with longer modulation arms also have higher tuning efficiencies. To reduce footprint, in one embodiment, the longer arm coupler may have a helical geometry, and regions of P-doped and N-doped material may be positioned parallel to the silicon helical waveguide to form a PN junction with a helical geometry.

[0140] More generally, according to embodiments of the invention, the middle portion of one of the waveguides (the waveguide portion between the two coupling points of the waveguide coupling to the ring waveguide) is elongated. This elongation can be present in embodiments where the control element is a PN junction control element as described above, in order to improve coupling efficiency. However, this elongation can also be present in embodiments using other control elements, such as IRPH or other thermally based control elements. The longer waveguide portion allows for a longer control element, which correspondingly interfaces with the waveguide portion along the longer length. This, in turn, produces higher coupling efficiency. High coupling efficiency contributes to intensity modulation as described herein. The length of the waveguide portion and the corresponding control element (and the interface between the control element and the waveguide portion) can be used to provide a predetermined coupling efficiency required for a particular application. It should be noted that in some embodiments, a longer length of the middle portion can reduce FSR, and therefore the length may be required to be within a specific range suitable for the application.

[0141] The length of the intermediate portion and the corresponding control element length can be used to provide coupling efficiency high enough to facilitate the desired operability of the photonic device. For example, the coupling efficiency can be high enough that the control element and the entire photonic device can swing all their input optical power between two output ports (e.g., a dropout port and a through port). This, in turn, can be achieved with sufficient coupling efficiency such that substantially all optical power (maximum) is coupled into the ring waveguide at one end of the operation, and substantially no optical power (minimum) is coupled into the ring waveguide at the other end of the operation.

[0142] It is possible to achieve a slender middle section of the waveguide while maintaining a limited two-dimensional footprint. That is, the middle section itself can be bent or otherwise twisted and rotated to accommodate a limited area. This can be achieved by shaping part or all of the middle section into, for example, a helix or meander. More generally, the footprint can be limited by incorporating two, three, or more opposing bends into the middle section waveguide. In other words, a compact footprint for the middle section and the entire device can be achieved. When one bend is on the left and another on the right, the first and second bends can be considered opposing when traveling along the middle section waveguide in the same direction. It can be seen that the illustrated helical middle section waveguide has two such opposing bends. The helical shape is particularly suitable for cases where the control element is based on a PN junction.

[0143] Figure 10a An embodiment of a PN junction-based, interference-coupled modulator is shown, wherein the coupler arms are long and helical. In one embodiment, two coupler arms 1010 are symmetrically positioned on either side of the MRR, and modulation is applied using a PN junction with helical geometry as shown in illustration 1020. Figure 10a Illustration 1020 shows a cross-section of a helical PN junction, in which the P and N doped regions follow the helical waveguide in parallel.

[0144] Figure 10aa It shows something similar to Figure 10a An embodiment of the modulator (photonic device) based on interference coupling. A first coupler arm (waveguide) 1010a and a second coupler arm (waveguide) 1010b are coupled to an MRR (ring waveguide) 1012. The first waveguide 1010a is coupled to the ring waveguide 1012 at two coupling positions 1014a and 1014b. The second waveguide 1010b is coupled to the ring waveguide 1012 at two coupling positions 1014c and 1014d. The first waveguide 1010a has an intermediate portion 1013a extending between coupling positions 1014a and 1014b and bending away from coupling positions 1014a and 1014b. The second waveguide 1010b has an intermediate portion 1013b extending between coupling positions 1014c and 1014d and bending away from coupling positions 1014c and 1014d. A first control element 1016a is coupled to the middle portion 1010a of a first waveguide, a second control element 1016b is coupled to the middle portion 1010b of a second waveguide, and another control element 1016c is coupled to a ring waveguide. The first and second control elements operate in a complementary manner to perform extinction ratio adjustment at a fixed wavelength. Alternatively, the second and third control elements can also operate in a complementary manner to perform extinction ratio adjustment at a fixed wavelength.

[0145] Figure 10b This is a schematic diagram of a symmetrical MRR modulator based on interference coupling, which has a pair of symmetrical helical PN junctions 1030 as modulation components in the coupler. Insert image 1040 is a cross-section of the helical PN waveguide, where the P and N doped regions extend parallel to the helical waveguide.

[0146] Simulation results show that an embodiment using a coupler arm twenty (20) times longer than the MRR section (ring waveguide) can achieve greater intensity modulation than an embodiment with a coupler arm only πR longer. In the first case, a variation of approximately 30 percentage points can be achieved, while in the second case, the variation is approximately 12 percentage points. This corresponds to a range of -30 dB to -60 dB.

[0147] Therefore, in various embodiments, the middle portion of the waveguide has an extension along which the middle portion is coupled to a control element, including a middle portion of the extension that is at least 5 times, at least 10 times, or at least 20 times longer than the ring waveguide. The length of the middle portion can be used to provide at least a given coupling efficiency. The coupling efficiency can be given by the requirements of the application, the modulation scheme in use, and the type of control element in use.

[0148] Figure 10c The diagram shows the transmission spectrum of a simulated PN junction-based, interference-coupled MRR modulator as a function of bias voltage, where the upper coupler arm is longer than the MRR perimeter by 2RπR, as a subsequent... Figure 10d For comparison, it can be seen that the transmission modulation from 0V to -5V varies between approximately -39% and -53%, a difference of 14 percentage points.

[0149] Figure 10d The diagram shows the transmission spectrum as a function of bias voltage for a simulated PN junction-based, interference-coupled MRR modulator, where the upper coupler arm is 20 times the MRR perimeter 2R. It can be seen that the transmission modulation varies between approximately -7% and -35% from 0V to -5V, a difference of 28 percentage points, which is significantly greater than... Figure 10c The case of a shorter MZI arm.

[0150] Some embodiments include a single-bus, interference-coupled modulator. However, to meet the weight reassembly requirements of large-scale “broadcast and weight” protocol applications, they eliminate insertion loss within the resonator and provide a wider range of balanced weights. Other embodiments include two or more add-drop buses instead of a single bus.

[0151] Furthermore, to allow RF signal modulation, some embodiments include a helical upper coupler arm and a helical PN junction. These embodiments can increase the electro-optic speed (i.e., frequency) to 40 GHz.

[0152] Figure 11 This diagram illustrates a roadmap for developing an MRR modulator based on interference coupling. In the roadmap, once the MRR 1110 is acceptablely manufactured, an asymmetric design with MZI coupler arms 1120 can be developed. A symmetric IRPH design 1130 can then be developed. The PN junction MRR design can then be implemented using either an asymmetric configuration 1140 or a symmetric configuration 1150.

[0153] According to one embodiment, a modulator with a coupler arm can be used as a component in building a large system; therefore, this is also one embodiment. Due to crosstalk problems in wavelength-tuned modulators, building large systems using MRR modulators without any optical interactions can be difficult. This is especially true when hundreds or thousands of wavelength-tuned modulators are operating simultaneously. However, the optical crosstalk problem can be solved by using an interference-coupled modulator based on an intensity-tuning scheme according to an embodiment.

[0154] Multiple interferometric coupling-based modulators based on an intensity-tuning scheme, according to the embodiments, can be used to construct a computing system. Since the intensity-tuning mechanism has no wavelength drift, there are no crosstalk issues. This allows each modulator to have a large optical modulation amplitude (OMA), which can be crucial when the driver used in the system has a low resolution.

[0155] In one embodiment, a plurality of interferometric coupling-based modulators based on an intensity-tuning scheme according to the embodiment are used to construct a large-scale artificial neural network (ANN) system.

[0156] The crosstalk performance of the optical modulation system according to the embodiment can be evaluated by comparing it with another type.

[0157] In the reference system, three MRRs are connected in series, each with a different radius. Each MRR transmits an optical signal at a wavelength corresponding to its radius and resonant wavelength, due to its different radius. Therefore, each MRR corresponds to a wavelength channel. The output of each MRR is a spectrum with a Lorentzian profile, centered at the MRR's resonant wavelength. Thus, the outputs of the three MRRs connected in series are three consecutive Lorentzian curves, each representing one channel.

[0158] Figure 12aThe spectral curves 1220 at the outputs of the three connected MRR 1210s of the analog reference system are shown. When the intensity of channel 1 is modulated, the peak wavelength of channel 1 is also modulated, i.e., the curve shifts to the right 1230. Due to the Lorentz shape of the curve, the leading and trailing ends overlap with the curve of channel 2, which reduces the vertical distance 1240 between the minimum and maximum values ​​in channel 2, which is the optical modulation amplitude (OMA) of channel 2. This overlap is referred to as optical crosstalk 1250.

[0159] Figure 12b The spectral curves at the outputs of the three connected MRRs of the analog reference system are shown. Figure 12a Similarly, when the intensity of channel 2 is modulated, the peak wavelength of channel 2 is also modulated, and due to the Lorentz shape of the curve, the beginning and end of channel 2 overlap with the curve of channel 3.

[0160] Figure 12c The spectral curves at the outputs of the three connected MRRs of the analog reference system are shown. Similar to... Figure 12a and 12b When the intensity of channel 3 is modulated, the peak wavelength of channel 3 is also modulated.

[0161] In the system according to an embodiment, three MRRs are connected in series, each MRR having a different radius. According to one embodiment, each MRR interfaces with a pair of coupler arms, one coupler arm symmetrically positioned on either side of the MRR, and each coupler arm interfaces at two points on the MRR. The coupler arms are present to allow wavelength tuning to prevent wavelength shift, and the symmetrical pairing is to minimize and balance insertion loss at the interface between the coupler arms and the MRR. In such a system, crosstalk between channels can be reduced or even eliminated, with each channel corresponding to a different MRR radius.

[0162] Figure 13a The spectral curves at the outputs of three connected MRRs in a simulated symmetrical embodiment are shown. Figure 12a Unlike the reference system, when the intensity of channel 1 is modulated, the resonant wavelength does not shift and can remain fixed at 1310. The beginning and end of the Lorentz curve 1320 of channel 1 do not overlap with the Lorentz curve 1330 of channel 2, therefore the optical modulation amplitude (OMA) of channel 2 is not affected by the intensity modulation of channel 1.

[0163] Figure 13b This shows that intensity modulation of channel 2 does not affect channel 3, similar to... Figure 13a Modulation of the middle channel 1.

[0164] Figure 13c This demonstrates that intensity modulation of channel 3 does not affect other channels, similar to... Figure 13a Modulation of middle channel 1 and Figure 13b Modulation of the middle channel 2.

[0165] The absence of wavelength shift is a significant advantage because it prevents a reduction in OMA between adjacent channels, meaning that crosstalk is rarely or never observed.

[0166] For systems using typical MRR modulators, the peak tail of channel 1 can affect the OMA of the peaks in channels 2 and 3 due to the Lorentz shape of the resonant peak. The more modulators cascaded in a row, the greater the crosstalk effect. However, by using the interference-coupled modulator according to the embodiment, no crosstalk is observed because the intensity modulation is performed at a fixed wavelength. Therefore, the OMA of individual channels can remain unchanged.

[0167] The common weight range can also be improved by using a coupling-based modulator according to the embodiment. Since each MRR in the cascaded MRR has a different radius, not only is the resonant wavelength different for each MRR, but the transmission performance of each MRR is also different. Ideally, the modulation weight range of each MRR should be normalized to the common weight range. For normalization, the minimum weight swing among all modulators in the system can be selected as the common weight of the system.

[0168] Because the first wavelength-tuned MRR modulator causes crosstalk issues, adjacent MRR modulators will exhibit a degraded OMA compared to the first modulator, i.e., a shorter weight range. However, by using the modulator according to the embodiment, the minimum weight range is not shortened because the Lorentz spectrum has no wavelength shift, i.e., no crosstalk issues. The weight range remains greater than that of the wavelength-shifted modulator.

[0169] Crosstalk can be illustrated by comparing two existing cascaded MRR modulators with two cascaded MRR modulators according to an embodiment.

[0170] Figure 14a A series of overlapping transmission spectra 1425 of two cascaded MRR modulators (i.e., Device_1 1426 and Device_2 1427) with two different diameters and no coupler arms are shown, where each spectrum corresponds to a different tuning voltage. When Device_1 1426 is tuned, its peak wavelength is shifted toward the peak wavelength of Device_2 1427.

[0171] Figure 14bTwo figures are shown illustrating the output weights (difference between the downstream port transmission output and the through port transmission output) of two cascaded MRR modulators (Device_1 1426 and Device_2 1427), which have two different radii and no coupler arms. The wavelength tuning of Device_1 1426 reduces the weight range of Device_2 1427, with a common weight of ±0.37 1430.

[0172] Figure 14c A series of overlapping transmission spectra 1435 of two MRR modulators (i.e., MZI-Device_1 1432 and MZI-Device_2 1434) according to an embodiment are shown. In this case, the intensity in MZI-Device_1 1432 is tuned, but not its peak wavelength 1440, and since there is no wavelength shift, there is no crosstalk, and therefore no impact on the weighting range of the adjacent MZI-Device_2 1434.

[0173] The interferometric MRR modulator according to the embodiments can exhibit an enhanced weighting range compared to wavelength-tuned MRR, especially when multiple MRR modulators are cascaded. The enhanced weighting range allows for more channels within a free spectrum. The larger common weighting range resulting from the embodiments can also provide a better signal-to-noise ratio when the resolution of its voltage driver is low.

[0174] Figure 14d Two plots are shown of the output weights (the difference between the downstream port transmission output and the through port transmission output: W = DT) of two cascaded MZI-MRR modulators (MZI-Device_1 1432 and MZI-Device_2 1434), which have two different diameters and each has a pair of symmetrical coupler arms. Since there are no crosstalk issues, both modulators exhibit similar weight ranges, in this case ±0.81440. When two modulators in a row are cascaded, the use of modulators according to the embodiment allows for a common weight range that is twice that of a similar system using a reference MRR. This can result in a more robust system overall, as a larger common weight range can provide a better signal-to-noise ratio (SNR) when the MRR driver resolution is low.

[0175] about Figure 1a , Figure 1b , Figure 2a , Figure 3 , Figure 4a , Figure 5a , Figure 5b , Figure 5c , Figure 5d and Figure 5e If any MRR modulator is an MRR modulator according to an embodiment, then the electro-optic circuit in the corresponding figure is also an embodiment.

[0176] An interferometric modulator with a coupler arm, according to an embodiment, can be used as part of a system to perform matrix multiplication, and such a system is another embodiment.

[0177] An example could be a matrix multiplication system constructed by connecting two interferometric modulators according to an embodiment in series using an add-drop configuration and feeding the output into a balanced photodetector (BPD). The photocurrent generated by the BPD can be represented as the dot product of the input optical signal and the input matrix represented by the input modulation of the interferometric MRR according to the embodiment.

[0178] In the embodiment performing the dot product of two matrices, each matrix is ​​first vectorized, and the vectors are normalized. The components of the first matrix can be mapped to the weight range of a first MRR modulator, and the components of the second matrix can be mapped to the weight range of a second MRR modulator. Each vector component can then be converted into a bias voltage, which can be applied to an MRR modulator. Each weighted modulation of the input signal performs the multiplication of the two components of the dot product, and the BPD performs the summation of the dot product.

[0179] Return to Figure 5a It illustrates an electro-optic circuit capable of performing the dot product of matrices. If the MRR modulator for the first vector component 520, the second vector component 530, or both vector components is the MRR modulator according to the embodiment, then... Figure 5a The electro-optical circuit is also an example.

[0180] about Figure 5e This is a diagram illustrating how an MRR modulator can be integrated into a convolutional neural network that performs matrix multiplication. If the MRR modulator is the MRR modulator according to the embodiment, then... Figure 5e The electro-optical circuit is also an example.

[0181] To monitor the accuracy of the dot product calculated using the example, a lookup table representing all output dot products from the BPD can be used as a reference.

[0182] The optical processing performance of the matrix multiplication system according to the embodiment can be compared with that of the reference system. When compared to the dot product of the reference MRR modulator, the matrix multiplication (MM) according to the embodiment using 7-bit driver resolution shows improved computational accuracy and mean squared error (MSE), especially at larger input matrix sizes or input ranges. This indicates that the embodiment is more suitable for building large optical processors than typical wavelength-tuned modulators.

[0183] Figure 15a Simulation results of two-component multiplication for MRR system 1505, MZI-MRR system 1510 according to an embodiment with fixed weight step size, and MZI-MRR system 1515 according to an embodiment with fixed voltage step size are shown. The resulting accuracy 1520 is expressed as a function of driver resolution 1525, with mean squared error (MSE). It can be seen that the accuracy increases with driver resolution, and the best results are obtained using MZI-MRR system 1510 according to the embodiment with fixed weight step size.

[0184] Figure 15b The results of a matrix multiplication simulation with 7-bit resolution are shown compared to the results of a lookup table. The figure shows that the accuracy of calculating 1530 is improved when the matrix input range of 1535 is small, and it outperforms conventional MRR systems in all cases. For example, the results of the MZI-MRR 1×1 1540 example, the MZI-MRR 2×3 1545 example, and the MZI-MRR 3×3 1550 example are respectively superior to the results of the MRR 1×1 1555 example, the MRR 2×2 1560 example, and the MRR 3×3 1565 example.

[0185] The technical advantage of these embodiments is that they prevent crosstalk problems. Because of their ability to maintain a fixed wavelength when modulating signal strength, there is no optical crosstalk between channels caused by wavelength drift. The lack of optical crosstalk allows all modulators to operate at or near their full OMA without interfering with each other, or with limited interference, when multiple modulators are cascaded and operated simultaneously.

[0186] Another technical advantage of these embodiments is that they offer a larger range of common weights compared to typical modulators based on wavelength tuning schemes. This means they are more tolerant of driver resolution and system noise, making them more robust overall. This can be particularly crucial for large systems and artificial neural networks (ANNs).

[0187] Figure 16a It shows something similar to Figure 5e A convolutional neural network with nine cascaded MRR1605s instead of four. Each cascade can be connected to a tuning circuit system 1610. Weight reconfiguration of large neural networks can be achieved through similar assemblies. Reduced crosstalk and improved weight range allow the MRR modulator according to the embodiment to be preferred for such large systems.

[0188] Figure 16b This is a graph showing how the weights 1615 of an 8-channel 9×9 core system vary with the voltage 1620 applied to each coupler arm of the MRR modulator according to an embodiment. The graph is shown as normalized 1625, but also shows the range of the common weights 1630 of the original weights 1635.

[0189] Figure 16c The same transmission diagram is shown in two different power units, one in decibels (dB) 1630 and the other in milliwatts (mW) 1635. Each power unit has a transmission diagram of a cascaded downport port 1640 and a through port 1645 of a 9-unit MZI-MRR modulator according to an embodiment. It can be noted that when the power is modulated, the wavelength peak remains in its original position, thus there is no crosstalk.

[0190] The embodiment includes a signal modulator that can tune the signal strength without changing its wavelength. By introducing a dual-position interference coupler into the MRR modulator, the coupling strength at these positions can be adjusted by a control element in the bias coupler, thereby producing a modified MRR transmission at the resonant wavelength. However, to compensate for any unwanted wavelength shift, a second control element is added to the MRR section. This second control element can be used to lock the wavelength.

[0191] The embodiment includes an MRR modulator with two interference couplers, each with two interface locations, and each interference coupler is symmetrically positioned on the opposite side of the MRR. To use the embodiment as part of a weight reassembly in a "broadcast and weight" protocol, a balanced weight range is required, with each weight being a function of a bias voltage in the modulator with low insertion loss. The embodiment introduces a symmetrical interference coupler for constructing the add-drop modulator. Because the symmetrical interference coupler arms have equal propagation and coupling losses, the insertion loss at the downlink port is minimized, thus achieving a larger weight range at zero balance, since the weight is the difference between the output power at the downlink port and the output power at the through port (W = D – T).

[0192] The embodiment includes an MRR modulator with two interference couplers, each of which is a PN junction-based coupler with a helical shape. To enhance the electro-optic modulation efficiency of the PN junction in the embodiment, longer coupling arms are implemented, which not only improves tuning efficiency but also maintains electro-optic speeds on the order of tens of gigahertz (GHz). Furthermore, the long coupling arms can be formed in a helical shape, with the p- and n-doped regions parallel to the silicon waveguide core of the coupler. This geometry reduces their footprint and also allows for electrical control of the traveling wave. The traveling wave electrode can be used for the PN junction-based control element. The modulation efficiency can depend on the doping concentration of the PN junction of the control element, or, alternatively, on the distance between the modulated waveguide and the heater control element. The coupling can depend on the refractive index variation in the waveguide adjacent to the control element.

[0193] Embodiments of the present invention include a controller. The controller may be an electronic or electro-optical circuit. The controller may include integrated circuits, such as microprocessors, microcontrollers, FPGAs, ASICs, etc., or combinations thereof. The controller may be implemented as a purely hardware circuit or as a combination of hardware and software or firmware. The controller may include analog-to-digital converters, digital-to-analog converters, amplifiers, comparators, clock circuits, logic circuits, and various other digital circuit systems, analog circuit systems, or combinations thereof, as readily understood by those skilled in the art. The controller may be partially configured to respond to input signals by generating output signals for complementary control of two or more control elements, as described elsewhere herein. Control signals may be generated based on pre-configured operations (e.g., lookup tables or other rules). Control signals may be generated at least partially based on feedback. For example, wavelength offset may be monitored and compensated by the controller. Control signals may be generated based on a prior calibration procedure.

[0194] It should be understood that although specific embodiments of the technology have been described herein for illustrative purposes, various modifications can be made without departing from the scope of the technology. Therefore, the specification and drawings are to be considered merely as a description of the invention as defined by the appended claims, and are contemplated to cover any and all modifications, variations, combinations, or equivalents within the scope of the invention. Specifically, computer program products or program elements for storing machine-readable signals, or program storage or storage devices such as magnetic wires, magnetic tapes, disks or optical fibers, optical tapes or optical discs, are provided within the scope of this technology for controlling the operation of a computer according to the methods of this technology and / or constructing some or all of its components according to the systems of this technology.

[0195] The actions associated with the methods described herein can be implemented as coded instructions in a computer program product. In other words, a computer program product is a computer-readable medium on which software code is recorded to execute the methods when the computer program product is loaded into memory and executed on the microprocessor of a wireless communication device.

[0196] Furthermore, each operation of this method can be executed on any computing device (e.g., personal computer, server, PDA, etc.) and is performed based on one or more program elements, modules, or objects, or a portion thereof, generated from any programming language (e.g., C++, Java, etc.). Additionally, each operation, or the file or object implementing each operation, can be executed by dedicated hardware or a circuit module designed for this purpose.

[0197] Based on the description of the above embodiments, the present invention can be implemented solely in hardware, or it can be implemented using software and a necessary general-purpose hardware platform. Based on this understanding, the technical solution of the present invention can be embodied in the form of a software product. The software product can be stored in a non-volatile or non-transitory storage medium, such as a compact disc read-only memory (CD-ROM), a USB flash drive, or a removable hard drive. The software product includes a number of instructions that enable a computer device (personal computer, server, or network device) to perform the methods provided in the embodiments of the present invention. For example, such performance may correspond to the simulation of the logical operations described herein. According to embodiments of the present invention, the software product may additionally or alternatively include multiple instructions that enable a computer device to perform operations configuring or programming digital logic devices.

[0198] Although the invention has been described with reference to specific features and embodiments thereof, it will be apparent that various modifications and combinations thereof can be made without departing from the invention. Therefore, the specification and drawings are to be regarded only as a description of the invention as defined by the appended claims, and are contemplated to cover any and all modifications, variations, combinations, or equivalents within the scope of the invention.

Claims

1. A photonic device, comprising: include: Ring waveguide; A first waveguide, coupled to the ring waveguide at a first position and a second position, the first waveguide having: The first part extends from the first position; The second part extends from the second position; The middle portion is located between the first portion and the second portion, and extends between the first position and the second position, the middle portion being further away from the annular waveguide than the first position and the second position; A second waveguide, coupled to the ring waveguide at the third and fourth positions, has the following characteristics: The third part extends from the third position; The fourth part extends from the fourth position; The second intermediate portion is located between the third portion and the fourth portion, and extends between the third position and the fourth position, and the second intermediate portion is further away from the ring waveguide than the third position and the fourth position; A first control element is operatively coupled to the intermediate portion; A second control element is operatively coupled to the second intermediate portion; Another control element is operatively coupled to the ring waveguide; A controller is configured to operate the first control element and the other control element in a complementary manner to adjust the extinction ratio applied by the photonic device at the resonant wavelength of the photonic device, the extinction ratio being between the input to the first waveguide and the output of the first waveguide, while maintaining the resonant wavelength; The controller is also configured to operate the second control element and the other control element or another control element in a complementary manner to adjust a second extinction ratio applied by the photonic device at the resonant wavelength of the photonic device, the second extinction ratio being between the input to the second waveguide and the output of the second waveguide, while maintaining the resonant wavelength.

2. The photonic device of claim 1, wherein, One or more of the first control element, the second control element, and the other control element are electrically controllable in-resonator photoconductive heaters (IRPH).

3. The photonic device of claim 1, wherein, One or more of the first control element, the second control element, and the other control element are electrically controllable PN junction carrier density control elements.

4. The photonic device according to any one of claims 1 to 3, characterized in that, The intermediate portion and the first control element are interfaced along an extension of the intermediate portion having a length, the length being used to provide coupling efficiency for the first control element, the coupling efficiency being sufficient to control the amount of optical power coupled from the first waveguide to the ring waveguide between a minimum value and a maximum value, the minimum value being substantially zero and the maximum value being substantially equal to all optical power input to the photonic device.

5. The photonic device according to any one of claims 1 to 3, wherein, The middle section includes at least two opposing bends.

6. The photonic device of any one of claims 1-3, wherein, At least a portion of the intermediate section is spiral-shaped.

7. The photonic device of any one of claims 1-3, wherein, One or more of the first control element, the second control element, and the other control element include traveling wave electrodes.

8. The photonic device of any one of claims 1-3, wherein, The intermediate portion and the first control element are interposed along an extension of the intermediate portion, wherein the intermediate portion of the extension has a length at least 20 times that of the ring waveguide.

9. The photonic device of any one of claims 1-3, wherein, At least one additional control element includes a first additional control element for complementary operation with the first control element and a second additional control element for complementary operation with the second control element, the complementary operation being performed to adjust the extinction ratio applied by the photonic device at the resonant wavelength of the photonic device.

10. The photonic device of any one of claims 1-3, wherein, The first part of the photonic device, which includes the first waveguide and the front half of the ring waveguide, is symmetrical to the second part of the photonic device, which includes the second waveguide and the rear half of the ring waveguide.

Citation Information

Patent Citations

  • Upload and download filter and optical add-drop multiplexer

    EP3703287A1

  • Tunable Optical Element

    US20160356959A1