Why Did Removing a Low-Frequency Noise Band in Audacity Also Erase Sounds Below It?

Understand the low-frequency exception

Spectral Delete normally removes the frequency region identified by a spectral selection during its selected time interval. The command has no parameter dialog; the selection defines the operation. For example, Audacity illustrates deletion of a 5–7 kHz band without deleting the entire waveform.

The manual separately warns that very low frequencies are unsuitable for this effect. At a sample rate of 44,100 Hz, a lower boundary below approximately 100 Hz can cause the deletion to extend from zero to the selected upper frequency. The actual threshold depends on sample rate, so 100 Hz must not be treated as a universal fixed cutoff.

This is particularly important for narration. Low-frequency energy can include unwanted rumble and legitimate components of a speaker's voice. Deleting a broader low-frequency region than intended may change vocal weight even if the high-frequency consonants remain audible.

Inspect the actual selection before applying deletion

Work on a duplicate of the original narration. Display the spectrogram and enable the Spectral Selection Toolbar if available. Record the sample rate, lower frequency boundary, upper frequency boundary and selected time interval.

Do not rely solely on the apparent height of a selection rectangle. A logarithmic or otherwise configured spectrogram scale can make visual distance a poor substitute for numeric boundary values. The toolbar provides a more explicit record of the intended interval.

If the lower boundary enters the documented unsuitable region, stop before applying Spectral Delete. Consider a more appropriate low-frequency treatment or a cleaner recording, and review the relevant effect's own documentation. Do not assume that moving the selection upward by a visually small amount is sufficient without checking the actual frequency values.

Hypothetical rumble removal from a product demonstration

Hypothetical illustration—not a processed recording: a merchant has a 44,100 Hz narration track with a low rumble during 00:04–00:06. The editor marks 60–180 Hz, intending to remove only that band.

Because the selected lower boundary is below approximately 100 Hz, Audacity's documented limitation means the operation may instead remove frequencies from zero through 180 Hz during the interval. That would affect a larger part of the voice than the editor intended. The worksheet records this as a documented risk, not a measured result for the fictional track.

The editor preserves the source and evaluates another supported approach rather than applying Spectral Delete blindly. If a revised treatment makes the narration thinner or changes the sound of the spoken product name, the result should not be accepted merely because the rumble is quieter.

Verify the output without confusing waveform appearance and frequency loss

Audacity notes that a spectral deletion may not produce an obvious change in the waveform display. The waveform represents amplitude over time, not a complete visual inventory of retained frequency content. Compare the spectrogram and listen to the affected phrase in context.

Check the processed interval and the transitions immediately before and after it. If the voice loses body, or a low-frequency sound remains because it overlaps wanted speech, reconsider the intervention. Avoid repeatedly deleting larger bands until the result sounds unnaturally thin.

The central decision is whether the chosen operation can isolate the unwanted frequency range under the documented sample-rate conditions. If it cannot, select a different method or preserve the original recording for further work.

Record your own observations

The filled rows are illustrative examples. Use the blank row for your own files and observations.

Download the editable CSV worksheet

What the references establish

• Spectral Delete — Audacity Manual: Directly documents the low-frequency limitation, its dependence on sample rate and the approximately 100 Hz example at 44,100 Hz.
• Spectral Selection and Editing — Audacity Manual: Explains how to define and inspect time-frequency selections. The fictional narration case is not a reported test.