- name
SKILL.md
技能檔案保留在發布者的儲存庫中。
在終端機中貼上複製的指令。它會將這個 SKILL.md 安裝到對應的代理中。
npx skills add matlab/simulink-agentic-toolkit --skill simulink-frequency-response -a cursor -y在終端機中貼上複製的指令。它會將這個 SKILL.md 安裝到對應的代理中。
npx skills add matlab/simulink-agentic-toolkit --skill simulink-frequency-response -a claude-code -y在終端機中貼上複製的指令。它會將這個 SKILL.md 安裝到對應的代理中。
npx skills add matlab/simulink-agentic-toolkit --skill simulink-frequency-response -a codex -y在終端機中貼上複製的指令。它會將這個 SKILL.md 安裝到對應的代理中。
npx skills add matlab/simulink-agentic-toolkit --skill simulink-frequency-response -a grok -y在 Grok Bot 聊天中傳送複製的訊息。Bot 會從 GitHub 取得這個 SKILL.md,並加入你所有 Bot 共用的技能中。
Install this skill: https://github.com/matlab/simulink-agentic-toolkit/tree/HEAD/skills-catalog/control-systems/simulink-frequency-response來自發布者
Obtain frequency response data from Simulink models using simulation-based estimation (frestimate) when analytical linearization is not viable or as a validation tool.
linearize — use simulink-linearize instead1. Define I/O → 2. Choose Signal → 3. Configure → 4. Estimate → 5. Fit (optional)
(linio) (PRBS/Sinestream) (constraints) (frestimate) (tfest)
Determine I/O points for the estimation using this decision sequence. Use the first case that applies:
Case A — IO points can be inferred from prompt or model context:
Use the first sub-case that matches:
User specifies explicit I/O signals or blocks (e.g., "from r to y") → define linio points. All linio points must reference a block's output port. If a candidate block has no output ports (Outport, Terminator, Scope) → trace upstream to find the source block and port with model_read.
io = [linio(sprintf("%s/InputBlock", mdl), 1, "input"); ...
linio(sprintf("%s/OutputBlock", mdl), 1, "output")];
Root-level Inport/Outport blocks exist → use model_read at root scope (depth "0") to identify root-level Inport/Outport blocks, then define linio at those blocks (trace Outport blocks upstream to their source for the output linio point).
Case B — Cannot determine IO points:
If none of the above apply → do not guess. Ask the user which signals to use as estimation input and output. Present the available blocks/signals from the model to help them decide.
Pre-flight checks (before choosing a signal):
Decision (follow in order):
frest.PRBS (DEFAULT)frest.PRBS with Ts matching the I/O sample time, or frest.createFixedTsSinestream if per-frequency data neededfrest.Sinestream (continuous) or frest.createFixedTsSinestream (discrete)Prefer frest.PRBS — it estimates the full frequency range in a single simulation. Sinestream simulates each frequency sequentially and is significantly slower for broadband estimation.
PRBS (default):
in = frest.PRBS(Ts=Ts, Amplitude=0.01, Order=10, NumPeriods=2);
If the I/O signal is discrete, set Ts to match the signal sample time. If continuous, set Ts to a value that provides sufficient temporal resolution. Choose Amplitude small enough to stay in the linear regime of saturations/nonlinearities.
Why PRBS first? A single PRBS simulation estimates the full frequency range at once. Sinestream simulates each frequency sequentially — for 30 frequencies with 8 periods each, this can take 10-100x longer. Use Sinestream only when you need precise per-frequency data (e.g., gain/phase margin at specific crossover frequencies).
Sinestream (continuous models):
in = frest.Sinestream(Frequency=logspace(-1, 2, 30), Amplitude=0.01);
in.NumPeriods = 8;
in.SettlingPeriods = 3;
The filtering constraint: NumPeriods - SettlingPeriods >= 3 when ApplyFilteringInFRESTIMATE = "on" (default). Violating this throws an error at estimation time.
Fixed-Ts Sinestream (discrete models):
Ts = 0.01;
in = frest.createFixedTsSinestream(Ts, {wmin, wmax});
in.Amplitude = 0.01;
in.NumPeriods = 8;
in.SettlingPeriods = 3;
Use the cell syntax {wmin, wmax} for the frequency range — this auto-selects frequencies that are valid integer divisors of the sampling frequency. Do NOT pass an explicit frequency vector:
% CORRECT — cell syntax auto-selects valid frequencies
in = frest.createFixedTsSinestream(Ts, {wmin, wmax});
% WRONG — explicit vector (most frequencies violate integer-multiple constraint)
in = frest.createFixedTsSinestream(Ts, logspace(-1, 2, 30)); % Error
When to use Sinestream instead of PRBS:
Check simulation time: Always verify that the signal duration is practical before launching the estimation:
tFinal = getSimulationTime(in);
fprintf("Estimated simulation time: %.1f seconds\n", tFinal);
If tFinal is too big compared to Ts, use larger lower frequency bounds for estimation.
Determine where to start the experiment. Choose one:
sysest = frestimate(mdl, op, io, in, opts);
op and opts are optional arguments. If op is not provided, the experiment will start at model initial conditions.
opts is a frestimateOptions object. Pass it when time-varying sources need to be disabled (see below).
The result is an frd (frequency response data) object.
Only perform this step if a parametric model (transfer function, state-space, zpk) is required. If the goal is frequency response data only (e.g., Bode plot, gain/phase margins from frd), stop after Stage 4.
Convert the non-parametric frd to a parametric model:
sysFit = tfest(sysest, np, nz);
fprintf("Fit: %.1f%%\n", sysFit.Report.Fit.FitPercent);
Time-varying source blocks (Step, Ramp, Signal Generator, etc.) can drive the model away from its steady-state operating point during estimation. When this happens, the system does not remain near the operating point and the estimated response is unreliable — gain estimates can be off by orders of magnitude while executing without error.
When to disable sources:
How to identify and disable them:
Use frest.findSources to identify time-varying source blocks in the estimation path, then set BlocksToHoldConstant so they are held at their initial value during estimation:
srcblks = frest.findSources(mdl, io);
opts = frestimateOptions;
opts.BlocksToHoldConstant = srcblks;
sysest = frestimate(mdl, io, in, opts);
Note: frest.findSources requires model compilation. The perturbation input is not affected by BlocksToHoldConstant.
When linearize returns zero, follow this sequence:
% 1. Try linearize
sys = linearize(mdl, io);
if dcgain(sys) == 0
% 2. Disable time-varying sources if present
srcblks = frest.findSources(mdl, io);
opts = frestimateOptions;
opts.BlocksToHoldConstant = srcblks;
% 3. Fall back to frestimate with PRBS
in = frest.PRBS(Ts=Ts, Amplitude=0.01, Order=10, NumPeriods=2);
sysest = frestimate(mdl, io, in, opts);
% 4. Fit parametric model
sysFit = tfest(sysest, 2);
end
Do NOT use manual block substitution (replace_block) as a workaround for zero linearization. The frestimate approach is generalizable to any discontinuous model without requiring domain knowledge of each block's averaged equivalent.
frest.findSources and BlocksToHoldConstant to disable themfrest.PRBS for broadband estimation — faster than Sinestream for most workflows{wmin, wmax} with frest.createFixedTsSinestreamNumPeriods - SettlingPeriods >= 3 when filtering is enabledgetSimulationTime — Validate signal duration before running. Long simulations relative to max solver step size should prompt redesign.frestimate over manual block substitution for discontinuous modelstfest or ssest for fitting parametric models to frd resultsreplace_block as a general linearization workaroundCopyright 2026 The MathWorks, Inc.
>
---name: simulink-frequency-responsedescription: >---# skills-catalog/control-systems/simulink-frequency-response/SKILL.md常見問題
從發布者下載 skills-catalog/control-systems/simulink-frequency-response/SKILL.md,或複製路徑後開啟儲存庫。PlugMCP 只收錄這項技能,不代管檔案。
發布者將它存放在 https://github.com/matlab/simulink-agentic-toolkit。
Claude、Cursor、Codex 和 Grok Build 可以讀取你從發布者儲存庫加入的 SKILL.md。在聊天中傳送技能的 GitHub 資料夾連結,Grok Bot 就會安裝它。