How to Use Spectral Frequency Display to Target and Remove Crackles Precisely

Crackles and unwanted noise can ruin the quality of audio recordings, from spoken-word tracks to music mixes. Using spectral frequency display tools allows audio engineers, podcasters, and enthusiasts to target and remove these imperfections with surgical precision. Instead of applying broad filters that damage desired content, spectral editing isolates each click and pop so only the problem area is treated. This article explains how to effectively utilize spectral frequency display to identify and eliminate crackles, and provides advanced techniques to refine your workflow.

Understanding Spectral Frequency Display

The spectral frequency display — also called a spectrogram — visualizes audio signals across different frequencies over time. It uses a color gradient to show intensity: quieter sounds appear as darker or cooler colors, while louder sounds are brighter or warmer. Time runs horizontally (left to right), frequency runs vertically (bottom to top). This two-dimensional representation gives engineers an instant overview of the entire audio file.

Because the human ear often struggles to localize very short, high-amplitude transients, crackles can go unnoticed until they are exaggerated by compression or limiting. In a spectrogram, crackles appear as sharp, isolated spikes — narrow vertical lines or bright dots that cut across the natural texture of the recording. These spikes are typically broadband, meaning they contain energy across many frequencies, which is why they sound harsh and unnatural. By analyzing this visual, you can pinpoint exact moments and frequency ranges where crackles occur, and apply corrections that your equalizers and compressors cannot achieve.

How Spectral Display Works Under the Hood

Most audio editors convert the time-domain waveform into a time-frequency representation using a Fast Fourier Transform (FFT). The FFT breaks the audio into small windows (usually 512 to 8192 samples per frame), and for each window calculates the amplitude of every frequency bin. The result is a set of frequency bins at each time step. The spectrogram you see is essentially a heatmap of those bins. The size of the FFT window determines the trade-off between time resolution and frequency resolution: a larger window gives better frequency resolution but worse time resolution, and vice versa. For removing short crackles, you typically want high time resolution, so you might choose a smaller window size (e.g., 1024) and a greater overlap between windows.

Common Sources of Crackles

Before you remove crackles, it helps to understand where they come from. Each source may leave a slightly different signature in the spectrogram, informing your repair strategy.

Analog tape or vinyl artifacts: Physical imperfections in the medium cause short, random bursts of noise. These often appear as thin, irregular vertical streaks that may have a slight pre-echo or post-echo.
Digital clock errors or buffer underruns: Glitches in the ADC/DAC or during processing create zero-sample dropouts or corrupted samples. In a spectrogram, these look like perfectly sharp, rectangular spikes at a single sample point.
Microphone handling noise or cable interference: Friction with cables or transient contact noise from the microphone body generates low-frequency thumps with high-frequency clicks. You will see both a dark vertical band in the low end and a bright spike higher up.
Plosives or close-mic breaths: While not always destructive, these can cause crackle-like transients that need selective attenuation rather than total removal.
Environmental noise (rain, static, electrical interference): Random sparks from nearby electronics or rain hitting a windscreen produce irregular, scattered spikes across the spectrogram.

Essential Tools for Spectral Editing

Several professional audio editors offer built-in spectral frequency displays and dedicated repair tools. Choosing the right one depends on your budget and workflow.

iZotope RX

iZotope RX is industry-standard for audio restoration. Its Spectral Repair module includes three algorithms — Attenuate, Replace, and Blend — that allow you to target specific time-frequency regions with great control. The Decrackle module (within Mouth De-click or standalone) automatically detects and removes short clicks based on a learnable threshold. (Learn more about iZotope RX).

Adobe Audition

Adobe Audition provides a robust spectral display (called the Spectral Frequency Display) that you can switch to from the waveform view. Its Spot Healing Brush tool works similarly to clone-stamp in image editing: you paint over a crackle and it uses neighboring clean samples to reconstruct the missing data. The DeClicker effect also offers a frequency-specific mode. (Adobe Audition spectral editing documentation).

Audacity

For free, open-source editing, Audacity has a Spectrogram view (set the waveform to Spectrogram in the track dropdown menu). While it lacks dedicated repair tools, you can use the Noise Gate or manually draw silence over crackles, though this is less precise. Advanced users can install plugins like Click Removal. (Audacity Spectrogram Guide).

Step-by-Step Guide: Removing Crackles with Spectral Editing

Follow this systematic workflow in any spectrogram-equipped editor. The exact key commands will vary, but the principles remain the same.

Step 1: Import and Visualize

Open your audio file and switch to the spectral frequency view. In most software, you can cycle between waveform and spectrogram with a single button. Adjust the viewer settings to maximize crackle visibility: set the frequency axis from 0 Hz to 8 kHz (most crackles live above 1 kHz), and set the color scale to highlight short peaks — often a “hot” color map (blue-to-red) works well. Set the FFT window size to around 1024–2048 for a good balance.

Step 2: Identify Crackles in the Spectrogram

Playback the audio at normal speed, then slow down to half speed or use a loop over suspect sections. Look for thin, bright vertical streaks that stand out from the surrounding noise floor. They often appear in rapid succession — a staccato pattern. If you see a cluster of spikes, zoom into the region. Use a high zoom level to see each individual crackle as a distinct dot or very short line.

Step 3: Select the Problematic Region

Most editors provide a rectangle selection tool (or lasso in iZotope RX). Draw a tight selection around the crackle — include a few milliseconds before and after to give the repair algorithm enough context. For a single crackle, the selection might be only 10–20 ms wide in time and span the frequency range that shows the spike. Do not overshoot: selecting too much clean audio may cause the repair to blend across the transient, softening the sound.

Step 4: Apply Spectral Repair / Decrackle

With the region selected, apply the appropriate tool:

Attenuate – reduces the amplitude of the selected frequency bin(s). Best for low-level crackles embedded in noise.
Replace – synthesizes new audio for the selected region based on surrounding clean material. Works well for short, isolated clicks.
Blend – a hybrid that mixes attenuated and replacement signals. Good for preserving transients while reducing harshness.

For batch removal of many similar crackles, use a dedicated Decrackle plug-in first, then manually fix any survivors.

Step 5: Adjust Parameters and Preview

After applying the repair, always preview the result. Toggle the original and processed audio (most editors have a “Bypass” or “Preview” button). Listen in solo mode to the region; if you hear an unnatural artifact, undo and try a different algorithm. Common adjustments include:

Threshold – lower values catch softer crackles but may damage desired content.
Frequency Skirt – controls how far above and below the selection the repair spreads.
FFT Window Overlap – more overlap improves accuracy but increases processing time.

Step 6: Refine and Iterate

Zoom out to listen to a longer section. Often crackles are not isolated — after removing the most obvious ones, quieter ones become audible. Work through the file in small chunks (5–15 seconds at a time). Use your spectrogram cursor to mark and fix each crackle before moving on. Save incremental versions (e.g., audio_cleaned_step1.wav) so you can revert if a repair introduces an unwanted change.

Advanced Techniques and Tips

Once you are comfortable with the basics, these refinements will help you achieve transparent results even on dense crackle patterns.

Working with Multiple Simultaneous Crackles

When crackles overlap in time, you may see a wide vertical band formed by many close spikes. Instead of selecting each one individually, you can use a frequency-selective expand tool (like Adobe Audition’s Marquee tool with shift-click) to select a frequency range across a time region, then apply a gentle Attenuate or Decrackle with a low threshold. This is faster and often preserves the underlying transient structure better than aggressive per-spike removal.

Preserving Transient Detail

Some crackles share frequency content with wanted transients (e.g., a click similar to a hi-hat strike). In this case, use the Blend algorithm with a ratio of 60–80 % original, 20–40 % cleaned. This attenuates the crackle component while keeping the natural attack of the instrument. Also consider applying a high-pass filter to the repair region if the crackle is above the frequencies of the instrument.

Batch Processing for Large Projects

For long recordings like podcasts or vinyl rips, manual editing of every crackle is impractical. Use a two-step approach:

Run a global Decrackle plug-in with conservative settings (low reduction, high threshold).
Then visually scan the spectrogram for remaining spikes and fix them manually with Spectral Repair.

Most restoration suites (iZotope RX batch processor, Adobe Audition’s Effects Rack) allow you to apply the same profile to multiple files. Always test on a representative sample first.

Using EQ Before and After

Crackles tend to contain high-frequency energy. Before applying spectral repair, you can add a gentle shelf cut above 6 kHz to reduce the overall harshness of the crackles, making the algorithm work less hard. After repair, restore the high end with a complementary shelf boost — but only if the repair has not removed wanted high-frequency detail.

Common Mistakes and How to Avoid Them

Even experienced engineers can make these errors. Being aware of them will save you time and frustration.

Overselecting: Selecting too much audio around a crackle forces the repair algorithm to synthesize a large area, causing audible smearing. Keep selections tight.
Aggressive settings: Turning the reduction knob to maximum removes crackles but also removes the natural body of the sound. Always start with 50 % reduction and go up gradually.
Not saving backups: Spectral repair is destructive if applied to the original file. Always work on a copy or use non-destructive processing (if your editor supports it).
Ignoring context: A crackle that lies directly over a sibilant or consonant may need a different algorithm than one in a vowel. Listen to the surrounding 2–3 seconds to understand the content.
Relying solely on automation: Batch processing often misses subtle crackles. Always do a manual pass through the spectrogram after the automatic step.

Conclusion

Using a spectral frequency display is the most precise method for targeting and removing crackles from audio recordings. By visualizing the time-frequency domain, you can identify each click as a discrete spike, apply tailored repairs, and preserve the natural character of your audio. With practice, you will develop an intuition for which algorithm to use and how to adjust parameters for different source materials — from vintage tape transfers to modern digital recordings. Combining the techniques described here with careful listening will allow you to produce clean, professional results that stand up to the highest editorial standards.

Continue learning: iZotope’s guide to Spectral Repair.