A SPICE model generation method considering thermal noise of passive network

Abstract

The application discloses a SPICE model generation method considering passive network thermal noise, and comprises the following steps: a passive macro-modeling algorithm is used to model S parameters of a multi-port passive network, a passive state space model is obtained, a Riccati equation is constructed and solved, a state space model of a lossless network is obtained, a state space-based SPICE model generation method is used to obtain an equivalent SPICE circuit model, and joint simulation is carried out by using the SPICE circuit model and other modules in the circuit to obtain analysis results of field-circuit collaborative simulation. The state space model used for representing an original system is divided into two parts: a lossless network and a pure resistance network. The lossless network only considers transmission characteristics of the passive network, and can be synthesized by using a noiseless model device; the pure resistance network is used for equivalent thermal noise generated by the passive network, so that the accuracy of simulation results is ensured in simulation related to noise analysis.

Description

Technical Field

[0001] This invention relates to the field of radio frequency circuit design automation technology, and in particular to a method for generating SPICE models that takes into account the thermal noise of passive networks. Background Technology

[0002] Radio frequency (RF) integrated circuits typically operate at very high frequencies, reaching several gigahertz or even hundreds of gigahertz. Typical circuits include LNAs (Low Noise Amplifiers), mixers, and VCOs (Voltage-Controlled Oscillators). In the pre-simulation stage, passive devices are often simplified to ideal devices, ignoring their frequency-dependent characteristics and parasitic effects in the circuit. In schematic design, if the simulation results meet the design specifications, the next step is layout design, where electromagnetic simulation analysis of the physical layout is performed using an electromagnetic field solver based on Maxwell's equations. For passive networks in the layout, their multi-port S-parameters can be obtained through the electromagnetic field solver to characterize the transmission characteristics, noise behavior, and delay characteristics of the passive network. After the acquired S-parameters are converted into an equivalent SPICE (Simulation Program with Integrated Circuit Emphasis) model, they can participate in circuit simulation together with other modules in the circuit (including active devices), thereby realizing the function of field-circuit co-simulation, improving the accuracy of circuit performance analysis and the completeness of design verification.

[0003] S-parameters are essentially samples of the passive network response at a set of discrete frequency points, while the state-space model constructed using passive macro-modeling algorithms represents the network as a continuous mathematical function in the frequency domain. Some commercial simulators support directly importing state-space models and performing circuit simulations together with other circuit modules. The circuit simulator extracts thermal noise information from the passive network and, in simulation types that include noise analysis, treats the thermal noise as an equivalent noise voltage or current source. However, the applicability of state-space models is limited, and currently only a few commercial simulators support them. In contrast, SPICE circuit netlists, as a universal language in the field of circuit simulation, have greater compatibility. SPICE models can be directly integrated into any circuit simulation environment without modification, thus better achieving the goal of field-circuit co-simulation.

[0004] In radio frequency (RF) circuit design, passive networks are commonly used to transmit specific power signals. Due to the Brownian motion of electrons within passive networks, thermal noise is inevitably generated. If only the transmission characteristics of the passive network are considered during SPICE circuit modeling, while ignoring its thermal noise effect, the noise performance in simulations involving noise analysis—such as Transient Noise, Scattering-Parameters Noise, and Periodic Noise—will be underestimated, thus affecting the accuracy of the overall circuit performance analysis. Commonly used components in passive network synthesis include resistors, capacitors, inductors, VCVS (Voltage-Controlled Voltage Source), VCCS (Voltage-Controlled Current Source), CCVS (Current-Controlled Voltage Source), and CCCS (Current-Controlled Current Source). In SPICE circuit simulators, only resistors have a built-in thermal noise model. If only transmission characteristics are considered during synthesis, the resistor may be replaced by other components (such as VCCS). However, the noise voltage that was originally equivalent to the thermal noise of the resistor will be missing, affecting the accuracy of circuit simulation.

[0005] Therefore, in the process of generating SPICE circuit models, how to fully consider the thermal noise effect of passive networks in order to construct a high-precision equivalent circuit model is a technical problem that urgently needs to be solved. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a SPICE model generation method that considers the thermal noise of passive networks. The state-space model representing the original system is decomposed into two parts: a lossless network and a purely resistive network. The lossless network only considers the transmission characteristics of the passive network and can be synthesized using noise-free model devices. The purely resistive network is used to account for the thermal noise generated by the equivalent passive network, thus ensuring the accuracy of results in simulations involving noise analysis. The high-precision broadband SPICE equivalent model constructed based on this invention can be imported into any circuit simulator for joint simulation analysis with other devices in the circuit. Furthermore, when problems such as simulation non-convergence occur in field-circuit co-simulation, the SPICE circuit model is more intuitive and easier to debug and optimize compared to the abstract passive macro model, which helps improve simulation stability and design efficiency.

[0007] The embodiments of the present invention provide the following solutions: This invention provides a method for generating SPICE models considering passive network thermal noise, the method comprising: S1. Obtain the multi-port S-parameters of the passive network as the original input parameters, and establish a real state space model that satisfies the passive property condition through the passive macro-modeling algorithm. S2. Based on the real state-space model that satisfies the passivity condition, construct and solve the Riccati equation to obtain the state-space model of the lossless network. S3. Based on the state-space model of lossless network, the equivalent SPICE circuit model is constructed using the SPICE model generation method of state space. S4. Use the SPICE circuit model to perform joint simulation with other modules in the circuit to obtain the analysis results of field-circuit co-simulation.

[0008] In an optional embodiment, the real state-space model satisfying the passivity condition described in step S1 is represented as: , in, H (s) represents a real state-space model that satisfies the passivity condition. A Represents the state matrix, B Represents the input matrix, C Indicates the output matrix. D Represents a constant matrix. s Represents the Laplace variable. Represents the identity matrix. , , N The order of the fit is . p This represents the number of ports in the system.

[0009] In an optional embodiment, step S2 includes the following process: S2.1 For a real state-space model that satisfies the passivity condition, construct the following Riccati equations: , in, ; S2.2 Solve the Riccati equation to obtain the symmetric positive definite matrix. X ; S2.3, Let the matrix P = X Constructing a matrix W sum matrix L , so that: , in, yes The upper triangular matrix obtained by Cholesky decomposition ; S2.4. Thus, the state-space model of the lossless network is obtained: , in, H L (s) represents the state-space model of a lossless network, and the matrix is... A L , B L , C L and D L The construction methods are as follows: , in D L satisfy .

[0010] In one optional embodiment, during the process of generating the SPICE circuit model using the state-space-based SPICE model generation method, the resistors added to the lossless network are all noise-free resistors, and the values ​​of the added resistors and capacitors are positive.

[0011] In one alternative embodiment, during the generation of the SPICE circuit model, the resistance values ​​of the resistors added to the lossless network are greater than 1 ohm, and the capacitance values ​​of the added capacitors are greater than 1 fF.

[0012] The beneficial effects of this invention based on its technical solution are as follows: (1) The present invention decomposes the state space model used to characterize the original system into two parts: a lossless network and a pure resistive network. The lossless network only considers the transmission characteristics of the passive network and can be synthesized using noiseless model devices. The pure resistive network is used to represent the thermal noise generated by the equivalent passive network, thereby ensuring the accuracy of the results in simulations involving noise analysis.

[0013] (2) No special matrix structure requirements are required for the passive state space model, and it has a wide range of applications; the generated SPICE circuit model has good versatility and can be directly read and simulated by mainstream circuit simulators; the SPICE model can replace the original S-parameter data and supports various types of circuit simulation, including noise analysis; without any modification or adaptation to the circuit simulator, it can significantly improve the simulation speed and convergence, effectively shorten the chip design cycle, and has good engineering practical value and economic benefits.

[0014] (3) The high-precision broadband SPICE equivalent model constructed based on the present invention can be imported into any circuit simulator and used for simulation analysis together with other active devices in the circuit. In addition, when problems such as simulation non-convergence occur in field-circuit co-simulation, the SPICE circuit model is more intuitive and easier to debug and optimize than the abstract passive macro model, which helps to improve the stability of simulation and design efficiency. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this specification or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating a method for generating a SPICE model that considers the thermal noise of a passive network, provided as an example.

[0017] Figure 2 For primitive passive networks H ( s Decompose into a lossless network H L (s) and schematic diagram of the purely resistive network Rnet.

[0018] Figure 3 for H L (s) j Normalized incident power of each port u L ( j The corresponding circuit diagram.

[0019] Figure 4 for H L (s) for the first k A schematic diagram of the equivalent circuit of an internal state.

[0020] Figure 5 For the first j A schematic diagram of the equivalent circuit model of each output port.

[0021] Figure 6 This is a schematic diagram of the SPICE circuit model (Spectre netlist) obtained by S-parameter modeling of the 2-port inductor in the embodiment. Detailed Implementation

[0022] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention are within the protection scope of the embodiments of the present invention.

[0023] This embodiment provides a method for generating SPICE models that considers the thermal noise of passive networks, referring to... Figure 1 ,include: S1. Read the multiport S-parameters of the p-port passive network. If the original data is in Y-parameter or Z-parameter form, it can be converted to S-parameter form using standard transformation methods. (Frequency...) f k At that location, the form of the S-parameter matrix for port n is: , The complete S-parameters consist of NF elements. S k The matrix is ​​constructed, where NF is the number of frequency sampling points.

[0024] The real-state-space model established by the passive macro-modeling algorithm is as follows: , in, H (s) represents the real state space model obtained through the passive macro-modeling algorithm. A Represents the state matrix, B Represents the input matrix, C Indicates the output matrix. D Represents a constant matrix. s Represents the Laplace variable. Represents the identity matrix. , , N The order of the fit is . p This represents the number of ports in the system.

[0025] S2. Based on the real state-space model that satisfies the passivity condition, construct and solve the Riccati equation to obtain the state-space model of the lossless network.

[0026] The principle behind this step is: based on the bounded real lemma of the S-parameters of passive networks, the state-space model... The conditions for satisfying passivity are: (1), In particular, if passive networks H If (s) is also a lossless network, then there exists a symmetric positive definite matrix.P ,satisfy: (2), Therefore, according to equations (1) and (2), it is possible to obtain information from passive networks. H Extracting the pure resistive network Rnet and the lossless network from (s) H L (s). It should be noted that lossless networks H L (s) needs to satisfy the lossless condition of formula (2), and at the same time H L ( s The cascading of Rnet satisfies the original passive network. H The transmission characteristics of (s). For example... Figure 2 The diagram shows a lossless network. H L (s) is cascaded with the purely resistive network Rnet to form the original network. H A schematic diagram of (s), in which H L ( s The number of ports between Rnet and Rnet is ρ . Figure 2 Middle left p indivual H (s) The original port connection to the external circuit remains unchanged, and the connection added on the right side is... ρ Each port is connected to the purely resistive network Rnet. According to... H L (s) Port distribution, H L (s) can be divided into: , in , In Rnet, all resistors are S-parameter reference impedances. Figure 2 middle H L (s) The incident power input to Rnet does not have reflection, which makes it easy to obtain Therefore, lossless networks H L ( s The lossless condition of formula (2) must be satisfied simultaneously in the process. .

[0027] Based on the above principles, this step can first solve for matrices P, W, and L, including the following process: S2.1 For real state-space models that satisfy the passivity condition Construct the following Riccati equation: , in, .

[0028] S2.2 Solve the Riccati equation to obtain the symmetric positive definite matrix. X .

[0029] S2.3, Let the matrix P = X Constructing a matrix W sum matrix L , so that: , in, yes The upper triangular matrix obtained by Cholesky decomposition The matrices P, W, and L obtained here are actually the solutions to equation (1).

[0030] S2.4. Thus, the state-space model of the lossless network is obtained: , in, H L (s) represents the state-space model of a lossless network, and the matrix is... A L , B L , C L and D L The construction methods are as follows: (3), in D L satisfy .

[0031] To obtain the matrix D 12 , D 22 Constructing a matrix ,calculate Null space and front ρ The column is obtained V null : , Combining formulas (1) and (3), we get: , That is, constructed from formula (3) H L (s) is lossless and also satisfies .

[0032] If you want to maintain good sparsity in the generated SPICE circuit model, you can... A L Perform eigenvalue decomposition: , Where the diagonal matrix yes A L eigenvalue matrix, T It is a matrix composed of the corresponding eigenvectors.

[0033] right H L (s) Perform similarity transformation , This similarity transformation can effectively reduce the number of components required to generate a SPICE model.

[0034] S3. Based on the lossless network state-space model, the SPICE model generation method for state space is used to... H L (s) Perform circuit synthesis, and its port input / output expressions are as follows: , Where input u L With output y L Let S be the normalized incident power and normalized reflected power of the S-parameters.

[0035] S-parameter normalized incident power a i and normalized reflected power b i The definition of is: , in V i and I i These are the ports. i voltage and current, Z 0 is the port i reference impedance of the S-parameters. Scattering parameters. Defined as In fact, it is only necessary to satisfy the definition of scattering parameters. The SPICE circuit model does not necessarily have to strictly satisfy the definitions of normalized incident power and normalized reflected power.

[0036] H L (s) j Normalized incident power of each portu L ( j ) adopt as Figure 3 The circuit shown is used to describe this, where the value of the resistor element Resistor is... r Current I j The controlled current source CCCS value is f ,Voltage V j The controlled current source VCCS value is g .

[0037] H L (s) for the first k One internal state expression: , The corresponding equivalent circuit description is as follows: Figure 4 The value of capacitor Capacitor c is not necessarily equal to 1 pF. It can be set arbitrarily as long as it is within a reasonable range. Other set capacitor values ​​should also be within the protection scope of this patent.

[0038] Figure 4 middle( N -1) A voltage-controlled current source (VCCS) is used to describe the internal state. Relationship with other states, the final The voltage-controlled current source VCCS represents the internal state. Relationship with normalized incident power.

[0039] H L (s) j The expressions for the normalized reflected power of each port in relation to its internal state and normalized incident power are as follows: , Figure 5 The first j The equivalent circuit model of each output port, where N The voltage-controlled current source VCCS describes the first j The relationship between the normalized reflected power of each port and its internal state. p The voltage-controlled current source VCCS describes the first j The relationship between the normalized reflected power and the normalized incident power of each port. Figure 5 DC voltage source introduced in V j The resistor is used to convert the voltage and current at the port into normalized reflected power.

[0040] S4. After converting the above circuit structure into a standard circuit netlist description, it can be imported into any circuit simulator for simulation analysis. H L (s) All resistors added during synthesis are zero-noise resistors. In the Spectre netlist, isnoisy=no, and in the HSPICE netlist, noise=0, to explicitly instruct the simulator not to introduce resistor noise. Resistors and capacitors added during SPICE model generation should also meet the following conditions: the device values ​​should be positive, and avoid being too large or too small. A reasonable value range can usually be set as the range. Resistors added to passive networks should have a resistance value greater than 1 ohm, and capacitors added should have a capacitance value greater than 1 fF. The parameters of the controlled source should be adjusted according to the actual simulation requirements to ensure the accuracy of the circuit behavior. The resistor network Rnet, used to characterize the thermal noise characteristics of passive networks, is a noisy resistor and is considered an equivalent noise source in circuit simulation, thus realistically reflecting the thermal noise effect of passive networks in high-frequency environments.

[0041] The equivalent resistance model generated by the above method can be exported as a SPICE netlist in any standard format, including but not limited to Spectre format, HSPICE format, and the general SPICE netlist format.

[0042] For a simple 2-port inductor layout, its 2-port S-parameters are obtained through simulation using an electromagnetic field solver. Following the steps outlined in this embodiment, the following can be obtained: Figure 6 The circuit netlist is shown. To facilitate the integration of other modules in the circuit, the lossless network is... H L (s) is merged with the purely resistive network Rnet into a sub-circuit named ebsckt_ind1. The following circuit netlist is in Spectre format, but it can also be converted to other formats, which should also be within the scope of protection of this invention.

[0043] For a typical 57-port voltage-controlled oscillator (VCO) circuit layout, the S-parameter description of the passive network is first obtained using an electromagnetic field solver. Following the steps outlined in this embodiment, the equivalent SPICE circuit netlist description of this S-parameter description is then obtained. The obtained passive macromodel and equivalent SPICE circuit are then simulated together with the active circuit modules of the VCO, yielding the results shown in the table below: PN@10KHz -91.475 dBc / Hz -91.541 dBc / Hz PN@100KHz -113.473 dBc / Hz -113.506 dBc / Hz PN@500KHz -119.458 dBc / Hz -119.482 dBc / Hz PN@1MHz -124.788 dBc / Hz -124.806 dBc / Hz PN@10MHz -131.319 dBc / Hz -131.332 dBc / Hz PN@50MHz -151.737 dBc / Hz -151.750 dBc / Hz Idc 0.028 A 0.028 A It can be observed that the simulation results of the passive macro model and the active circuit are very close to those of the equivalent SPICE circuit model and the active circuit, proving that the SPICE equivalent model is completely accurate.

[0044] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0045] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (modules, systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0046] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0047] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0048] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0049] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for generating SPICE models considering passive network thermal noise, characterized in that, The method includes: S1. Obtain the multi-port S-parameters of the passive network as the original input parameters, and establish a real state space model that satisfies the passive property condition through the passive macro-modeling algorithm. S2. Based on the real state-space model satisfying the passivity condition, construct and solve the Riccati equations to obtain the state-space model of the lossless network; including the following process: S2.1 For a real state-space model that satisfies the passivity condition, construct the following Riccati equations: ， in, ; A Represents the state matrix, B Represents the input matrix, C Indicates the output matrix. D Represents a constant matrix. I Represents the identity matrix. , , N The order of the fit is . p This refers to the number of ports in the system. S2.2 Solve the Riccati equation to obtain the symmetric positive definite matrix. X ; S2.3, Let the matrix P = X Constructing a matrix W sum matrix L , so that: ， in, yes The upper triangular matrix obtained by Cholesky decomposition ; S2.

4. Thus, the state-space model of the lossless network is obtained: ， in, H L (s) represents the state-space model of a lossless network, and the matrix is... A L , B L , C L and D L The construction methods are as follows: ， in D L satisfy ; S3. Based on the lossless network state-space model, the SPICE model generation method for state space is adopted, including the following process: right H L (s) Perform circuit synthesis, and the expressions for its port inputs and outputs are as follows: ， Where input u L With output y L The normalized incident power and normalized reflected power of the S-parameters are used to construct an equivalent SPICE circuit model; In the process of generating the SPICE circuit model using the state-space-based SPICE model generation method, the resistors added in the lossless network are all noise-free resistors, and the values ​​of the added resistors and capacitors are positive. S4. Use the SPICE circuit model to perform joint simulation with other modules in the circuit to obtain the analysis results of field-circuit co-simulation.

2. The SPICE model generation method considering passive network thermal noise according to claim 1, characterized in that: The real state-space model that satisfies the passivity condition described in step S1 is expressed as follows: ， in, H (s) represents a real state-space model that satisfies the passivity condition. A Represents the state matrix, B Represents the input matrix, C Indicates the output matrix. D Represents a constant matrix. s Represents the Laplace variable. I Represents the identity matrix. , , N The order of the fit is . p This represents the number of ports in the system.

3. The SPICE model generation method considering passive network thermal noise according to claim 1, characterized in that: During the generation of the SPICE circuit model, the resistance values ​​of the resistors added to the lossless network are greater than 1 ohm, and the capacitance values ​​of the added capacitors are greater than 1 fF.

Images