Understanding Dynamic Range and Frequency-Specific Expansion

Before diving into the plugin’s controls, it is essential to grasp what a multiband expander is doing under the hood. At its core, downward expansion reduces the gain of a signal when it falls below a user-defined threshold. In practical terms, this means the quiet sections of your recording—which are often dominated by noise—are turned down even further. The louder, desired signal passes through unaffected, or is affected only minimally.

The Mechanics of Downward Expansion

A standard noise gate is a drastic form of downward expansion (an infinite ratio). A multiband expander gives you a variable ratio, allowing for gentle attenuation. A ratio of 1:1.5 means that for every 1 dB the input signal drops below the threshold, the output drops by 1.5 dB. This creates a smooth, graded reduction in noise rather than an abrupt on/off switch. The attack control determines how quickly the expander engages once the signal falls. The release control dictates how quickly the gain returns to normal once the signal rises above the threshold again. Getting these time constants right is the key to transparent processing.

Most modern expanders offer two detection modes: RMS (root mean square) and peak. RMS detection averages the signal level over a short window, making it ideal for sustained noises like tape hiss or room tone. Peak detection responds instantly to transient spikes and is better for percussive clicks or impulsive noise. Many multiband expanders, such as FabFilter Pro-MB, allow you to select the detection mode per band, giving you fine control over how each frequency region responds.

Why Single-Band Processing Fails for Archival Audio

Consider a vintage recording of a string quartet that has significant tape hiss in the high frequencies (above 8 kHz) and room rumble in the low frequencies (below 100 Hz). If you use a single-band expander, it listens to the overall level. When the cello plays a loud, sustained note, the total energy is high, so the expander stays open—letting the hiss and rumble through. During a quiet pizzicato passage, the overall energy drops, and the expander clamps down, simultaneously muting the hiss, the rumble, and the natural reverb tail of the room. The result is an unnatural, “breathing” effect where the background noise swells in and out. A multiband expander solves this by allowing you to treat each noise source independently.

The Essential Controls of a Multiband Expander

To use these plugins effectively, you need to understand the specific controls available to you. While layout varies between developers, the core parameters remain consistent.

Crossover Filters and Band Splitting

The crossover points determine how the frequency spectrum is divided. The slope of these filters (typically 12, 24, or 48 dB per octave) dictates the separation between bands. Gentler slopes create smoother transitions but allow more frequency overlap, which can lead to phase cancellation if you apply drastically different gain changes to adjacent bands. Steeper slopes provide cleaner separation but introduce more latency and potential phase shift at the crossover point. Linear-phase crossovers, available in advanced plugins like FabFilter Pro-MB, are often preferred for restoration work because they avoid the pre-ringing and phase distortion associated with minimum-phase filters, preserving the transient integrity of the original recording.

An often-overlooked parameter is the crossover filter’s shape. Minimum-phase filters introduce frequency-dependent delay, which can smear transients. Linear-phase filters use a constant delay across all frequencies, but they introduce a fixed latency (often 10–50 ms) that may be problematic for real-time monitoring or live broadcast. For offline restoration, linear-phase is almost always the better choice. A detailed comparison of these filter types can be found in the Audio Issues guide on phase vs. minimum-phase EQ—the same principles apply to expander crossovers.

Threshold, Ratio, Attack, Release, and Range

These five controls define the behavior of the expander within each band.

  • Threshold: Sets the level below which the expander begins to reduce gain. In restoration work, you typically set this just above the noise floor of the specific frequency band you are targeting.
  • Ratio: Controls the intensity of the gain reduction. A ratio of 1:1.5 to 1:3 is common for gentle cleanup. Higher ratios (4:1 and above) are used for aggressive noise removal.
  • Attack: Determines how quickly the gain reduction is applied. For sustaining noises like hiss or rumble, a medium attack (10–30 ms) works well. For percussive noise, a faster attack is needed.
  • Release: The trickiest control. A release that is too fast will cause the noise to “pump” up and down. A release that is too slow will make the expander sound sluggish and may inadvertently cut off the tails of musical notes. A good starting point is 100–300 ms.
  • Range: Caps the maximum amount of gain reduction applied. This is a safety net. Setting a range of –6 dB prevents the expander from fully muting the noise, which can sound more natural than complete silence. It leaves a consistent, low-level ambience rather than a “hole” in the sound.

Some plugins also offer a knee control, which smooths the transition around the threshold. A soft knee (2–6 dB wide) is almost always preferable for restoration, as it prevents the expander from “biting” into the signal harshly.

A Systematic Workflow for Audio Restoration

Applying a multiband expander effectively requires a methodical approach. Rushing into tweaking knobs without analyzing the source material will lead to subpar results. Follow this workflow to achieve consistent, professional-grade restorations.

Phase 1: Analysis and Pre-Processing

Load your audio into your DAW and insert a spectrum analyzer on the track. Solo the recording and listen to the quietest sections—the gaps between songs, the pauses in dialogue, or the tail end of a reverb. Identify the frequency profile of the noise. Is the hiss concentrated above 10 kHz? Is there a pronounced 60 Hz hum? Note these frequencies. Before applying the multiband expander, perform basic pre-processing. Use a high-pass filter to remove infrasonic rumble below 30 Hz that can confuse the expander’s threshold detection. If the recording has clicks or pops (common with vinyl transfers), use a dedicated de-clicker first. Clicks are high-energy, broadband events that can trick a multiband expander into triggering, causing a momentary “blip” in the gain reduction. Tools like iZotope RX’s Spectral De-click are excellent for this step.

Phase 2: Configuring Your Crossover Points

Set your crossover points based on the noise profile you identified. A standard four-band setup for restoration might look like this:

  • Band 1 (Low): 20 Hz – 150 Hz (Target: Rumble, HVAC noise, footfall)
  • Band 2 (Low-Mid): 150 Hz – 800 Hz (Target: Boxiness, muddiness, hum harmonics)
  • Band 3 (High-Mid): 800 Hz – 5 kHz (Target: Harshness, sibilance, hiss harmonics)
  • Band 4 (High): 5 kHz – 20 kHz (Target: Tape hiss, pre-echo)

Adjust these frequencies to match your recording. If the hiss is lower in pitch (common with older, slower tape speeds), you might drop Band 4 down to 4 kHz. If you are working with a 78 RPM shellac record, you may need an additional band around 2–4 kHz to handle the heavy surface noise known as “sandpaper” or “frying” noise. The goal is to isolate the noise as much as possible within a single band so that the other bands remain untouched.

Phase 3: Treating Low-Frequency Rumble (20–150 Hz)

Start with the lowest band. Solo this band so you can hear exactly what the expander is doing to the low end. Set the threshold low enough that it only reacts to the noise floor, not the fundamental frequencies of the bass instruments or kick drum. Use a ratio of 2:1 to 4:1. Apply a medium attack (20–40 ms) to allow the initial thump of a bass note to pass through. Use a moderate release (200–400 ms) to avoid a pumping sensation in the low end. Set a range of –6 dB to –10 dB. The goal here is to tighten the low end and remove the “woolly” feel of room rumble without draining the warmth. For recordings with a persistent 50/60 Hz hum, you may want to use a notch filter before the expander, as the expander alone cannot remove a steady-state hum that sits above the threshold.

Phase 4: Clearing Midrange Haze and Mud (150 Hz–2 kHz)

This region is the most sensitive because it contains the fundamental frequencies of most instruments and the human voice. Over-processing here will make the recording sound thin and hollow. Solo Band 2 and listen for a constant, muddy “wash” of sound. Set the threshold very carefully, just a few dB above the noise floor. Use a gentle ratio (1:1.2 to 1:2). Use a slow attack (30–60 ms) to preserve the transient snap of a guitar or the plosive of a vocal. A slow release (300–500 ms) is crucial here to prevent the expander from clamping down on the sustain of notes. A small amount of gain reduction (2–3 dB) can clear up “boxiness” significantly. If you hear unnatural warble or pitch variation in the noise floor—often present in wire recordings—consider using a slow attack and a soft knee to make the expansion nearly imperceptible.

Phase 5: Taming Presence and Harshness (2 kHz–8 kHz)

This band often contains the “edge” of the noise. Overly bright analog tape or digitized vinyl can have a grating quality in this range. Use a ratio of 1:2 to 1:3. Set the threshold to catch the constant hiss. A fast attack (10–20 ms) helps prevent the hiss from masking the initial transients of sibilant sounds. Be careful not to overdo the range; limiting the attenuation to –4 dB or –6 dB prevents the vocals from sounding like they are “ducking” or moving away from the microphone during quiet passages. If the recording contains artifacts from analog tape compression (such as “saturation noise”), you may want to use a slightly wider knee to avoid turning the noise on and off abruptly.

Phase 6: Eliminating High-Frequency Hiss (8 kHz–20 kHz)

This is where the multiband expander truly shines. Solo Band 4. You should hear mostly air and hiss. Set the threshold just below the level of the steady-state hiss. Use a high ratio (3:1 to 5:1). A medium attack (15–30 ms) is often better than a very fast attack, as it prevents the expander from “pinching” the air and reverb tails. A fast release (50–100 ms) can work well here, as the human ear is less sensitive to pumping in the extreme high frequencies. Aggressive range settings (–10 dB to –15 dB) are acceptable in this band because the information content is generally low. However, listen carefully for the loss of high-frequency detail in hi-hats, cymbals, or string overtones. If the track sounds dull after expansion, back off the ratio or increase the threshold slightly.

Restoring a Historical Recording: A Case Study

To illustrate the workflow, consider a 1940s acetate disc recording of a jazz big band. The transfer revealed significant issues: a low-frequency rumble from the cutting lathe motor (30–60 Hz), constant groove hiss concentrated at 7–12 kHz, and occasional clicks from dust. The dynamic range of the music was wide—loud brass hits contrasted with quiet drum brush sections.

After de-clicking with a dedicated tool, I set up a three-band expander (since the low-mid region was clean aside from a slight hum that I EQ’d out). Band 1: 20–150 Hz with a ratio of 3:1, medium attack, 300 ms release, range –8 dB. This eliminated the rumble without affecting the double bass fundamental. Band 2: 2–8 kHz with ratio 1:2.5, fast attack 15 ms, release 150 ms, range –5 dB—this smoothed out the hiss without dulling the trumpet transients. Band 3: 8–20 kHz with ratio 4:1, medium attack, release 80 ms, range –12 dB, which removed the tape hiss and left the cymbal shimmer intact. The result was a recording that sounded clear and natural, with no pumping artifacts.

Advanced Restoration Strategies

Once you are comfortable with the basic workflow, you can employ advanced techniques to achieve even more transparent and musical results.

Dynamic EQ vs. Multiband Expansion

It is important to know when to use a multiband expander versus a dynamic EQ. A dynamic EQ cuts or boosts a specific frequency based on the level of that frequency. A multiband expander reduces the level of an entire band based on the level of that band. For example, if you have a resonant 120 Hz hum that only appears during quiet sections, a dynamic EQ is the better tool because it can surgically target only the 120 Hz peak. If you have a broadband rumble across the entire 50–150 Hz region, a multiband expander is more effective. Many modern plugins, such as TDR Nova or FabFilter Pro-Q 3, combine dynamic EQ functionality, while FabFilter Pro-MB strictly functions as a multiband dynamics processor.

Parallel Expansion for Maintaining Ambience

One of the biggest criticisms of aggressive expansion is that it “dries out” the recording by removing the natural reverb and room tone along with the noise. To combat this, try parallel processing. Duplicate your track, or use the mix knob on your expander. Apply high-ratio expansion to the duplicate track to isolate the noise. Then blend this heavily processed signal back with the dry signal. This allows you to aggressively clean up the main signal while using the wet signal to reintroduce some of the natural ambience. This technique works exceptionally well for dialogue restoration or for preserving the air of a concert hall recording.

Sidechain Filtering Inside a Multiband Expander

Some multiband expanders allow you to apply an additional sidechain filter to the detector circuitry. For instance, you could tell Band 2 (150–800 Hz) to listen to a narrower slice of the spectrum, say 300–600 Hz, so that the expansion only reacts to energy in that sub-band. This is useful when a noise source has a strong tonal component, like a 500 Hz hum. By narrowing the sidechain, you prevent the expander from being triggered by unrelated low-frequency or high-frequency content. FabFilter Pro-MB and Waves C6 both offer this feature, and it can dramatically improve the precision of your cleanup.

Automation for Non-Stationary Noise

The noise profile of a recording is rarely constant. A tape recorder might have more hiss at the beginning of a reel than at the end. A recording of a live performance will have different noise levels between the quiet ballad and the loud chorus. This is where automation becomes critical. Automate the threshold parameters of your expander bands throughout the track. During a loud, dense section, you can raise the thresholds so the expander does not engage at all, preserving the energy. During quiet, exposed sections, you can lower the thresholds to aggressively suppress noise. This dynamic approach yields far better results than static settings.

Common Pitfalls and How to Avoid Them

Even experienced engineers can fall into these traps. Knowing them beforehand will save you hours of troubleshooting.

The “Breathing” Artifact: This is the most common issue, caused by release times that are too short. When the expander quickly opens and closes, the noise floor audibly pumps. To fix this, lengthen the release time on the affected bands. If lengthening the release causes the expander to sound sluggish, check your attack time as well—a slower attack can often pair with a faster release.

Phase Distortion at Crossover Points: Using steep crossover filters on adjacent bands that are applying different amounts of gain reduction can introduce phase cancellation. This results in a hollow, phasey sound. To mitigate this, use the gentlest crossover slope possible that still provides adequate separation. If your plugin offers it, switch to linear-phase mode, which resolves phase issues at the cost of additional latency. If you are working in a DAW that supports multichannel routing, you can also split the signal into separate tracks with dedicated EQ high-pass/low-pass filters and then apply independent expanders—this avoids crossover phase issues entirely.

Over-Expansion and Loss of Ambience: Applying too much gain reduction (high ratio and high range) across all bands will result in a sterile, unnatural sound. The recording will feel “dead.” Leave some of the low-level noise in the mix. A completely silent background on an old recording sounds eerily modern and can be jarring for listeners expecting vintage warmth. Use the Range control to limit the maximum attenuation. Additionally, consider using the expander’s make-up gain function to restore the overall perceived level after reduction—but be careful not to bring the noise back up.

Spectral Leakage from Overlapping Bands: When using gentle crossover slopes, energy from one band can bleed into an adjacent band’s detector. This can cause the expander to react to frequencies it wasn’t intended to. For example, a loud kick drum at 100 Hz might trigger the low-mid band (150–800 Hz) if the crossover slope is shallow, causing the midrange to dip momentarily. To avoid this, either steepen the crossover or use a high-pass sidechain filter on the affected band.

While the techniques discussed apply to any multiband expander, some plugins offer features that are particularly useful for archival restoration.

FabFilter Pro-MB is widely considered the most transparent and versatile option. Its dynamic phase mode is excellent for maintaining clarity, and the intuitive interface allows for quick band isolation and soloing. It is an ideal primary tool for restoration engineers. (Learn more here)

iZotope RX is the industry standard for audio repair. Its “Mouth De-click” and “Spectral De-noise” are excellent, but its Multiband Compressor/Expander module is also very powerful, integrating well with the suite’s other tools like the Spectral Shaper. (iZotope RX product page)

Waves C6 is a more budget-friendly workhorse. While it lacks the surgical precision of Pro-MB, it is simple to set up and effective for general cleanup of tape hiss and rumble.

TDR Nova is a free dynamic EQ that can function brilliantly as a multiband expander. It offers highly flexible bands and is a fantastic starting point for anyone on a tight budget. (Download TDR Nova)

Melda Production MAutoDynamicEQ is another flexible option that can be used as a multiband expander with up to eight bands, and it includes a free version with limited features. It is particularly useful for its advanced sidechain options.

For an in-depth technical primer on the underlying dynamics processing, the archived articles from Sound on Sound remain an excellent resource.

Conclusion

Enhancing old audio recordings is a delicate balance between removing the artifacts of the past and preserving the musical integrity of the performance. The multiband expander is one of the most powerful tools in the modern restoration engineer’s arsenal precisely because it allows for this nuanced, frequency-dependent approach. By carefully setting your crossover points, dialing in appropriate threshold and ratio settings for each band, and respecting the natural dynamics of the source material, you can lift the veil of noise from vintage recordings and present them with a clarity that was previously unattainable. The goal is not to make them sound modern, but to make them sound true to the original performance, revealing the skill and artistry that has been hiding in plain hearing for decades. With patience and practice, multiband expansion becomes an invisible craft—the listener hears only the music, not the process.