Preparing S-parameters/Touchstone® Model for High-Speed PCB Analysis

Menu

In past years, more design engineers started to use S-parameters/Touchstone® for Signal Integrity and Power Integrity analysis. The S-parameters/Touchstone® is nothing new. However, it is becoming more popular because it covers good information in a simplified behavioral format. Still, it does have some disadvantages if a lack of data accuracy or/and sufficiency is present.

In this post, I’ll introduce the history and content of S-parameters/Touchstone®. We also will discuss its advantages and disadvantages and how to use the tool to overcome the issues.

History of S-parameters

S-parameters refer to the scattering matrix of a microwave network; the “S” in S-parameters refers to scattering.

Before the 1950s and the extensive use of high-frequency systems, network parameters like Y- and Z-parameters were the primary method of characterizing circuit performance. At higher frequencies, the concept of voltages and currents became more difficult to relate to network performance – especially in networks using transmission lines like waveguides. S-parameters are related to power waves and eliminate many of these issues.

The 1948 Radiation Laboratory series, “Principles of Microwave Circuits,” introduced the emerging microwave community to practical approaches to solving microwave design problems using scattering parameters in the design of circuits.

It helped that during the 1960s, Hewlett Packard introduced the first microwave network analyzers. The scattering wave concept was further popularized around the time that Kaneyuki Kurokawa of Bell Labs wrote his 1965 IEEE article “Power Waves and the Scattering Matrix.” Also, Robert Collin’s textbook, Field Theory of Guided Waves, published in 1960, has a brief discussion on the Scattering matrix. Collin’s book is extensively annotated, including an author index, which reads like a Who’s Who of electromagnetic theory for the first half of the twentieth century.

The word “Touchstone” was issued by EEsof (a part of Agilent Technologies, now Keysight). In 2009, Touchstone 2.0 was ratified by IBIS. Touchstone 2.0 added resistance per port option for Power Delivery System and mixed-mode features for differential setups. Currently, Touchstone 1.0 is widely supported by most of the tools. Touchstone 2.0 is still not supported by a number of EDA vendors.

Contents of S-parameters

What do the S-parameters contain? An S-parameters model contains the information for the gain/loss and phase against reference resistance port(s) on the frequencies.

Original S-parameters and Touchstone 1.0 only allow the single reference port for one model, but Touchstone 2.0 enables multi-reference ports in an S-parameters model.

  • 2-port and n-port S-parameter/Touchstone® format and contents are shown below:
    • 2-port:
Figure 1a:2-port definition
Figure 1a:2-port definition
Figure 1b: 2-port graph
Figure 1b: 2-port graph
Figure 1c: 2-port format
Figure 1c: 2-port format
Figure 1d: 2-port DB&A & R&I
Figure 1d: 2-port DB&A & R&I
    • N-port:
Figure 1e: N-port definition
Figure 1e: N-port definition

Strengths and Weaknesses when Using S-parameters in Analysis

S-parameters are used more and more in high-speed digital circuit designs, so it’s not new to SI/PI engineers. But why are S-parameters used often in SI/PI analysis in multi-GHz channel analysis?

Consider these reasons:

  • S-parameters are always defined, though impedance or admittance may not be.
  • S-parameters are normalized. There are good numerical properties in the simulations.
  • S-parameters are easily measured, even at very high frequency. Better for simulation and measurement correlations.
  • Standardized format for S-parameters (Touchstone®). The measurement equipment uses Touchstone® for output. Most field solvers provide Touchstone® output as well.
  • Tabulated ASCII format frequency data does not disclose any design details but only electrical properties.
  • The best way to represent broadband EM/circuit interactions. It is the essence of Signal and Power Integrity.

With a lot of advantages from S-parameters, there are also some weaknesses in simulations when S-parameters data has:

  • Limited frequency range coverage
  • DC point absent
  • Coarse sampling
  • Unintentional non-passivity
  • Measurement noise

Effects of Passivity and Causality Issues in S-parameters/Touchstone®

  • Passivity

In passive interconnects, such as traces, vias, wires, etc., the signal from one end to another should not have gained. Instead, it should always have a loss. S-parameters/Touchstone® is designed for passive elements. The simulation will be “exploded” if the S-parameters/Touchstone® model is not passive. See the image below.

Figure 2a: Passivity explanation
Figure 2a: Passivity explanation
  • Causality
    • When a signal goes through a passive interconnect from one end to another, the input signal will always be earlier than the output signal. Given that the S-parameters/Touchstone® file is frequency-oriented, sometimes we find the output signal is earlier than the input signal in time-domain simulations. This is called a non-causal event. Non-causal events directly affect channel analysis accuracy, especially for timing analysis. The picture below shows what a non-causal event is.
Figure 2b: Causality explanation
Figure 2b: Causality explanation

Preparing S-parameters/Touchstone® for SI/PI Simulations

We discussed that S-parameters/Touchstone® has many advantages for SI/PI analysis. And we also know it is not a silver bullet. To optimize its advantages, we need to minimize its disadvantages by checking and fixing the possible issues that S-parameters/Touchstone® might have before using it in SI/PI simulations.

Many EDA tools support S-parameters/Touchstone® models for SI/PI simulations. Here, I use a 24-pin connector (S24P) S-parameter/Touchstone® 1.0 model (93.6 MB, 461060 lines) as an example to describe how Zuken’s CR-8000 tools support S-parameters/Touchstone® model for simulations.

Step 1: Load into the simulation library

At the beginning of the model file, the comments show P01 connects to P13 and P02 connects to P14 as circled in RED from the visual checker. We will focus on these four pins to find out how to check the data and import these pins only into the library to save the file size.

Figure 3a: Read the comments
Figure 3a: Read the comments

Step 2: Visualization check

The tool’s visualization checking feature provides a visual representation of insertion losses and data symmetrical arrangements. It also visually highlights data if it is out of range in the S-parameters/Touchstone® model files.

Figure 3b: Visual checker
Figure 3b: Visual checker

Step 3: Check the data for each frequency response

S(1, 1): Reflections

Figure 3c-1: S(1, 1)
Figure 3c-1: S(1, 1)

S(1, 13): Insertions

Figure 3c-2: S(1, 13)
Figure 3c-2: S(1, 13)

Step 4: Extracting and fitting

Most of the time, we only focus on specific ports inside a large S*P file to reduce the file size and simplify the fitting process. In this case, we have a huge S24P file (93.6MB). To start, we will extract only four ports (Port 1, 2, 13, and 14) for our differential pair simulations.

We take ports 1, 2, 13, and 14 as “only selected” ports to extract and convert to the SPICE netlist for future simulations.

In Zuken tools, “Direct pole-residue specification” is used for the conversions.

Figure 3d: Direct pole-residue specification
Figure 3d: Direct pole-residue specification

In the import assistant dialog, we can define the accuracy settings for fitting as well as passivity, reciprocity and causality enforcements before fitting and passivity enforcement after fitting. It also offers lowpass off band behavior enforcement and DC extrapolation in case it is needed.

Figure 3e-1: Selection
Figure 3e-1: Selection

We skip all the enforcement features for the first run of conversion.

S(1, 1):

Figure 3e-2: S(1, 1)
Figure 3e-2: S(1, 1)

S(1, 13):

Figure 3e-3: S(1, 13)
Figure 3e-3: S(1, 13)

From the results, we can see the fitted data are well-aligned with the original data except in the low-frequency area (< 100MHz)

Step 5: Tests for causality enforcements and DC extrapolation

  • The test with causality enforcement only

In the result shown below, the original data and the fitted data for S(1, 13), the part of insertion loss, pretty much agreed with each other. However, the original data and the fitted data for S(1, 1), the part of reflection, have some disagreements, especially for the low-frequency range. In other words, the reflection part of this model needs to be fixed for non-causal conversions.

Figure 3f-1: Selection
Figure 3f-1: Selection

S(1, 1):

Figure 3f-2: S(1, 1)
Figure 3f-2: S(1, 1)

S(1, 13):

Figure 3f-3: S(1, 13)
Figure 3f-3: S(1, 13)
  • The test with DC extrapolation only
    • As the results show below, both S(1, 1) and S(1, 13) mostly agree between the original and fitted data. We can tell that the data for DC and low-frequency area is sufficient in this S-parameters/Touchstone® file.
Figure 3g-1: Selection
Figure 3g-1: Selection

S(1, 1):

Figure 3g-2: S(1, 1)
Figure 3g-2: S(1, 1)

S(1, 13):

Figure 3g-3: S(1, 13)
Figure 3g-3: S(1, 13)

Step 6: Simulation with fitted models

After pole-residue conversions, the fitted model stored in the library is in broadband SPICE format.

Figure 4a: SPICE netlist
Figure 4a: SPICE netlist

It can be used as an N-port SPICE circuit for simulations. If the common reference is selected, it will contain only one reference node; if the floating reference is selected, each port will have a reference node separately. It is essential to connect the proper reference node(s) for accurate simulation results.

Figure 4b: Topology
Figure 4b: Topology

Conclusions

S-parameters/Touchstone® is very useful for multi-GHz high-speed channel analysis. Touchstone 2.0 is also an international standard ratified by IBIS Open Forum. It is easily used with measurement correlations and frequency responses. It uses a “black box” type ASCII format for greater interoperability and intellectual property protection.

The S-parameters/Touchstone® model is not a silver bullet. It has some weak points, especially if there is a mistake when the model was created. It is not easily found even with visual graphical curve checks. It is not possible to be fixed by hand either. Using a tool (such as Zuken’s CR-8000 Analysis Module) for checking and enforcing is essential to ensure accurate simulation results for SI/PI analysis.

References:

https://www.microwaves101.com/encyclopedias/s-parameters

What You Can Read, What You Have to Read, Shinichi Maeda, KEI Systems, IBIS Summit

Towards Real-time S-parameter qualification and macromodeling, Stefano Grivet Talocia, Politecnico di Torino, Italy, IBIS Summit

Optimum Frequency Sampling in S Parameter Extraction and Simulation, Jinhua Huang, Synopsys, China, IBIS Summit

Overcoming Signal Integrity Channel Modeling Issues | Signal Integrity Journal, John Baprawski, SerDes Systems, Signal Integrity Journal

Circuit Synthesis of Multiport Networks from Passive Poles and Residues, Chiu-Chih (George) Chou & José E. Schutt-Ainé, DesignCon IBIS Summit 2022

S-parameter Renormalization, The Art of Cheating | 2017-01-12 | Signal Integrity Journal, Gustavo Blando, Oracle Co. Signal Integrity Journal

Lance Wang
Lance Wang
Solutions Architect
Lance Wang is a solutions architect in Zuken SOZO Center. He supports CR-8000 product line, mainly focusing on high-speed PCB design and signal integrity features. When not behind the keyboard or in front of customers, he is a Tom Brady fan and enjoys playing ping pong in the spare time.
Industrial Automation: A focused engineer interacts with a futuristic holographic interface, showcasing the power of advanced technology in a modern industrial setting
  • Blog
Dezember 05, 2024
Die Zukunft gestalten: Herausforderungen und Innovationen im IoT der Industrie

Von der Automatisierung bis zur Echtzeitanalyse – das Industrial Internet of Things (IIoT) verändert die Industrie grundlegend. Doch hinter den Chancen lauern Herausforderungen: Können Sicherheit, Altgeräte und komplexe Systeme in Einklang gebracht werden? Erfahren Sie, wie intelligente Fabriken der Zukunft gestaltet werden und warum das IIoT der Schlüssel dazu ist.

Read now
wp-header-1920x844-pcb-510x310
  • Blog
November 28, 2024
Stack-Up Design und Impedanzkontrolle für High-Speed Leiterplatten

Der Layer-Stack-up und die präzise Berücksichtigung der Leitungsimpedanz sind entscheidend für robuste High-Speed-Designs. Erfahren Sie in unserem Blog, wie Sie Stack-ups optimal mit CR-8000 und Polar Speedstack™ gestalten können.

Read now
replacing obsolete compenents from a PCB
  • Blog
November 04, 2024
Effektive Strategien für das Management veralteter Bauteile

Die Aktualisierung bestehender Designs, um veraltete Komponenten zu ersetzen, ist eine zeit- und arbeitsintensive Aufgabe. Moderne PCB-Design-Tools wie CR-8000 können dabei helfen, diesen Prozess zu optimieren.

Read now
wp-header-1920x844-artificial-intelligence-510x310
  • Blog
Mai 06, 2024
Abhilfe für den Fachkräftemangel im PCB-Design durch KI

Angesichts des Fachkräftemangels in vielen Branchen finden sich immer mehr Ingenieure in Positionen wieder, die Fähigkeiten erfordern, für die sie nicht speziell ausgebildet wurden. Was wäre, wenn wir diese Lücke mit Hilfe der KI-Technologie schließen könnten? Erfahren Sie mehr.

Read now