Unexpected frequency halving and extra peaks in FFT output when changing sampling time (TS) in Simulink

Hi everyone,

I’m running a Simulink model using multiple sine waves generated by Sine Wave blocks, summed together, then passed through:

Zero-Order Hold → Buffer (size = 2048) → FFT → Gain → Abs → Vector Scope (or Spectrum Analyzer)

Each Sine Wave block has a specified frequency (e.g., 100 Hz, 300 Hz, 375 Hz, 500 Hz, 750 Hz) and amplitude.

The Zero-Order Hold has a sample time TS = 1/2048 s initially, and the FFT length and buffer size are both 2048.

When TS = 1/2048, the FFT result looks correct: all expected frequency peaks appear properly.

However, when I change TS (e.g., to 1/4096 s), some strange things happen:

• The 100 Hz component appears at 50 Hz instead.
• Extra peaks appear above my highest input frequency (700 Hz).
• These artifacts persist even when aliasing should not occur (FS ≫ 2× fmax).

I suspect the issue is related to the Spectrum Analyzer / Vector Scope sampling-rate setting or how Simulink scales the frequency axis, not aliasing.

If I manually set the Spectrum Analyzer Sampling Rate to 2/TS, the 100 Hz component displays correctly in one case, but not for other TS values.

I already tried:

• Matching buffer size and FFT length (2048 each).
• Confirming sample rates between blocks.
• Checking aliasing and Nyquist limits.

None of these fixed the inconsistent scaling or the extra peaks.

Question:

How should the Spectrum Analyzer (or Vector Scope) sampling rate and FFT configuration be set to correctly reflect frequency content when TS changes?

Is there a reliable way to synchronize the FFT and analyzer x-axis scaling with the actual sample rate in Simulink?

Any insights or example models would be greatly appreciated!

Answer 1

[sneha]

Hi,

The Spectrum Analyzer or FFT block is mismatched in its sample rate, it “scales” the x-axis incorrectly, creating fake peaks because the frequencies don’t align with the correct bin mapping.

Things you can try:

1. Always set the analyser’s Sample Rate = 1/TS — if you change TS, change the Spectrum Analyzer “Sample Rate (Hz)” so its x-axis matches your actual Fs. See Sample Rate (Hz) in https://www.mathworks.com/help/dsp/ug/configure-spectrum-analyzer.html
2. Make FFT length = samples-per-frame (N)
3. Remember sample vs frame rate — the analyser needs the underlying sample rate (Fs), not the slower frame-update rate. Confirm that ZOH sample time = TS and downstream blocks use that sample time. See Sample- and Frame-Based Concepts in https://www.mathworks.com/help/dsp/ug/sample-and-frame-based-concepts.html
4. Use a single-tone sanity test — feed one sine (100 Hz) with the TS you want and verify its bin; this quickly tells you whether the analyser/FFT scaling is wrong. See Spectrum Analyzer block examples / measurements in https://www.mathworks.com/help/dsp/ref/spectrumanalyzerblock.html
5. If you change TS frequently, automate the analyser Sample Rate — set the Spectrum Analyzer Sample Rate programmatically (or use a model variable) so it always equals 1/TS and you don’t have to update it manually. (See Spectrum Analyzer configuration properties / Sample Rate property in https://www.mathworks.com/help/dsp/ref/spectrumanalyzerconfiguration.html