Let's Split!

A studio pro explains how split-band (aka multiband) compression works and how it can improve your mixes and solve everyday audio problems. Then he looks at several choices you have when selecting a split-band compressor

Illustration by Mike CruzA close friend of mine once told me he couldn't understand why recording engineers felt they needed a split-band compressor in their arsenal of studio gear. He held that a split-band compressor (also known as a multiband compressor) was a complex solution for a simple task — controlling audio dynamics — and therefore unnecessary. But he was dead wrong.

Image placeholder title

With the twist of a few knobs or virtual controls, a split-band compressor can pull off amazing feats that a single-band (aka broadband) compressor — or any other type of processor — couldn't accomplish in a million years. For example, a split-band compressor can easily prevent an acoustic guitar, cello, or upright bass from sounding boomy on low notes without thinning out warm formants when only high notes are present. Have a vocal track that's too shrill on several isolated phrases? Taming the sirens is child's play for a split-band compressor, and it won't eliminate intelligibility and detail in the process or force you to park one hand on your EQ knob for the entire mix. Split-band compressors can also make the bottom end of a mix sound huge while simultaneously increasing headroom. Try doing that with an equalizer!

Image placeholder title

FIG. 1: The Arboretum Systems Ionizer applies dynamics processing across 512 discrete frequency bands.

What enables a split-band compressor to perform such heroics is its multiband nature. Unlike a broadband compressor, which squeezes the dynamic range of its audio-input signal across its entire frequency spectrum, a split-band compressor uses a divide-and-conquer strategy to independently treat discrete frequency bands. It splits the audio signal into multiple bands, compresses each band independently of the others as needed, and then recombines the outputs of the bands into a single (mono or stereo) broadband signal. Because a different set of controls is dedicated to each band, you can, for example, slam the bass while leaving the highs untouched or, conversely, de-ess the highs without squashing the low end.

You can do a lot more with split-band compressors than that. This article will describe several applications and look at a representative sampling of the split-band compressor products — both hardware and software — that are currently available on the market.

But before I explore those subjects, an overview of key features and controls of split-band compressors is in order. I'll forgo introductory explanations of how threshold, ratio, attack, release, and makeup-gain controls work; you can read about the fundamentals of compression in my article “The Big Squeeze” in the February 2001 issue of EM, available online at www.emusician.com.

Because of their complexity and their ability to destroy pristine audio if used indiscriminately, most split-band compressors are poorly suited for beginning recordists who don't have a firm understanding of the basics. But if you're comfortable with using broadband compressors and want to take dynamics control to a much higher level, don't wait — let's split!


Most currently available split-band compressors offer from three to five independent bands per channel and handle both mono and stereo audio. I find that three or four bands are plenty for most applications in which individual tracks are treated. For mixing and mastering work — which often combines a broad palette of instruments, a wide range of spectra, and percussive and nonpercussive elements — I like to work with as many as five independent bands for each channel.

A good split-band compressor gives you broad control over the width of each of its bands. Typically, you'll adjust the bandwidth by specifying the crossover frequencies at the lower and upper ends of the band. For example, you might choose crossover frequencies of 250 Hz and 2.5 kHz when working with a 3-band compressor. In that case, the first band would range from 250 Hz down to the lower limit of the compressor's frequency response (typically 20 Hz in analog split-band compressors, and as low as DC in digital units). The second band would range from 250 Hz up to 2.5 kHz, and the third band would cover frequencies from 2.5 kHz to the upper frequency-response limit (generally 20 kHz or higher). Using that setup, you could separately compress the bass frequencies (below 250 Hz), midrange (from 250 Hz to 2.5 kHz), and highs.

Image placeholder title

FIG. 2: The BSS DPR-901 II can be configured for either single-channel, 4-band operation or 2-channel, 2-band processing.

For optimal flexibility, try to find a split-band compressor with crossover-frequency choices that overlap to a high degree. That might allow you to, for example, assign two bands to the bass frequencies, one to deal with the lowest frequencies and the other to deal with the upper bass.

A user-friendly and powerful split-band compressor should offer separate threshold, attack-time, release-time, makeup-gain, and ratio or range controls for each band. A range control is similar to a ratio control in that each adjusts the intensity of compression applied to a signal above threshold. However, a range control also limits the maximum amount of gain adjustment that can occur as a result of dynamics processing in its associated band. So although a high ratio can cause an ever-increasing amount of gain reduction as the input signal rises farther above the threshold, a range control specifies the maximum amount of gain reduction that can take place, even with a high ratio applied.


To illustrate why separate in-band (as opposed to global) controls are so useful, imagine a situation in which you're applying split-band compression to a problematic mix that has a generally weak bottom end, specific bass-guitar notes that are too loud, and occasionally severe vocal sibilance.

To make the bottom end of your hypothetical mix sound full without accentuating the louder bass-guitar notes, set a fairly low threshold and a high ratio or range for the split-band compressor's bass band (treating, say, all frequencies below 250 Hz). By setting the threshold slightly higher than the level of the generally weak bass energy, you allow the weak bottom end to pass through the compressor untreated. However, loud bass notes would trigger the bass band's compressor and be reduced dramatically in level, thanks to the high ratio or range control setting. You'd probably (but not necessarily) want fairly slow attack and release times for the bass band, in order to avoid potential pumping when the compressor kicks in and distortion when it releases. Now that you've tamed all the spikes in bass-energy level, you can boost the bass band's makeup gain to raise the overall level of the bottom end and achieve a beefy mix.

Image placeholder title

FIG. 3: MOTU's MasterWorks Compressor offers three independent bands of processing and a comprehensive feature set.

After you've dealt with the thin and erratic bottom end of a problematic mix, you turn your attention to the searing vocal sibilance that remains. Applying a static EQ cut to the sibilant frequencies is a poor solution because that would also dull the crack of the snare drum and decrease the intelligibility of the vocals during phrases that aren't sibilant. Split-band compression is the answer here as well. A suitably high threshold in the high-frequency band (with the highest crossover frequency set to roughly 5 kHz) allows loud vowel sounds to pass untreated but causes sibilant consonants to trigger the in-band compressor. You want a high ratio or range setting and very fast attack and release times for this application so the compressor will kick in quickly, extinguish the sibilance, and cease gain reduction before any nontransient material in the high-frequency band is audibly affected. Since you don't want to boost the highs at all in this example, keep the makeup-gain setting at 0 dB (unity gain).

Such a treatment of your hypothetical mix clearly illustrates the power of split-band compression. Aside from the high ratio or range settings used in both the low and the high band, the settings for the two treated bands are as different as night and day. That is, each band was processed independently and in dissimilar fashion, and the midrange band was left completely alone. That power and flexibility make a split-band compressor one of the ultimate tools for engineers.


In addition to the above-mentioned controls, a split-band compressor should ideally have independent bypass controls for each band. Bypassing and then activating each band in turn allows you to hear the effect that compression is having in each band, while preserving your painstakingly wrought settings. In-band bypass controls also allow you to prevent unintentional compression in any band (such as the midrange band in our example above) without having to assign a 1:1 ratio or a 0 dB range setting. Audio fidelity will also be best preserved if the bypass is the hardwire type (which routes the input directly to the output in hardware units).

Image placeholder title

FIG. 4: The creamy-sounding VintageWarmer from PSPaudioware provides both broadband and split-band compression modes.

Knee adjustment is another useful feature for split-band compressors. The knee adjustment (typically a global control that affects all bands simultaneously) might offer a choice of different fixed settings, or it might be continuously variable.

Some split-band compressors offer a toggle switch (whether real or, in software products, virtual) that alternates the response of the compressor's attack and release times between your manual settings and program-dependent or program-sensitive time constants. A program-dependent mode ignores your manual adjustments and adjusts attack and release times in response to the characteristics of the input signal. A program-sensitive function, on the other hand, modifies your manual settings as needed to better handle the current program material without pumping or causing distortion.

The most user-friendly split-band compressors offer a variety of meters to help you fine-tune your settings. At the very least, I like to see a main output-level meter that shows the summed signal from all bands, as well as separate gain-reduction meters for each band. If the compressor also features global clipping meters for each channel and metering to indicate how in-band energy compares to the threshold level, I'm in hog heaven.


Now that I've covered the key features and controls of split-band compressors, I'll take a brief look at a sampling of both hardware- and software-based split-band compressors currently available. This is neither a roundup of every multiband compressor in existence nor a shootout of the products mentioned. However, I will point out especially noteworthy attributes of each product.

Most of the split-band compressors I'll discuss actually fall into the broader category of multiband processors. That's because they perform other types of multiband processing — such as gating or expansion — in addition to compression. Most multiband compressors can also perform multiband limiting, simply by increasing the ratio, attack time, and knee hardness. Except where noted, all of the offerings listed accommodate both single- and dual-channel operation.

Image placeholder title

FIG. 5: Most parameter values for the TC Electronic Finalizer Express are determined by the unit's 25 presets.

One last point: split-band compressors are sometimes referred to as dynamic equalizers. There really is no practical difference between a split-band compressor and an equalizer that dynamically boosts or cuts the energy in a given band in response to varying levels in that band. A dynamic equalizer might use center-frequency and bandwidth controls to accomplish the same thing that split-band compressors achieve with crossovers, but what distinguishes the two product categories from each other is mostly a matter of semantics.


Much more than just a split-band compressor, the software-based Arboretum Systems Ionizer ($499) also performs multiband noise reduction (downward expansion), upward expansion, limiting, equalization, and frequency morphing (applying one sound's frequency characteristics to another sound, for creating vocoding and other effects). Ionizer is available in AudioSuite, MAS, Premiere, and standalone versions for the Mac, and as a DirectX plug-in for the PC.

Ionizer's split-band-compression function sounds great, but the graphical user interface lacks certain controls and metering, making compression applications quite unwieldy (see Fig. 1). The interface generates red, blue, and black graphical curves alongside an x-y plot of your audio's frequency response. Your subsequent placements of those colored curves determine the threshold, ratio, and range of processing across 512 discrete frequency bands. You determine the crossovers for each band by creating and dragging breakpoints along each curve. Because Ionizer provides so many bands to work with, in-band makeup gain and bypass controls are necessarily omitted. Also, attack and release controls are global, affecting all bands at once. The only metering provided is an imprecise color bar that dynamically alters its color and lightness to indicate the relative amount of processing across the frequency spectrum.


The BSS DPR-901 II ($1,549) is a 1U rackmountable processor that can operate as a single-channel, 4-band unit (for mono applications) or a 2-channel, 2-band device (see Fig. 2). It provides a per-band maximum 30 dB of compression or 16 dB of expansion.

The DPR-901 II approaches split-band compression in the guise of an equalizer, using center-frequency and bandwidth controls in lieu of crossover controls to determine each channel's bandwidth. Each band sports its own control for adjusting the intensity of processing, along with an independent threshold control and a bypass switch.

Image placeholder title

FIG. 6: The 3-band Tube-Tech SMC 2B is the world's only all-tube split-band compressor.

Comprehensive multisegment LED meters for each band show gain reduction or expansion on an arbitrary scale rather than in decibels. A Threshold LED meter shows signal level above and below the threshold point in decibels. Manual and program-dependent time constants are offered. The filters for bands 1 and 4 can be switched between bell-curve and shelving response. Bands 2 and 3 can be switched to provide either narrow- or wide-band operation, allowing — when used in conjunction with the other bands — simultaneous split-band and broadband compression and expansion in one unit.

I've worked with the original DPR 901 only, and found it to be an extremely transparent unit. So transparent, in fact, that I often found myself slamming the meters in order to achieve more than a subtle effect.


MOTU offers the MasterWorks Compressor as a standard MAS plug-in with Digital Performer (Mac, $795; reduced pricing is available for upgrades and crossgrades). This split-band compressor offers three discrete bands with adjustable crossovers (see Fig. 3). The crossovers can be set over an exceedingly wide range, with one notable exception: the lowest crossover can be adjusted only down to 125 Hz, which is a tad high for some bass-guitar, kick-drum, and mastering applications. The plug-in instantiates in mono-to-mono or stereo-to-stereo variants.

Each of the MasterWorks Compressor's bands provides controls for independently adjusting attack and release times, threshold, ratio, and makeup gain. Separate bypass and solo controls also serve each band. The solo function allows you to hear the compression effect in one band while the other bands are muted, which can be quite helpful. Each band also sports input, output, and gain-reduction meters and a compression graph (depicting the in-band compression slope and certain parameter values). Master input- and output-level controls, with associated meters and peak-clear function, round out the MasterWorks Compressor's comprehensive feature set.


PSPaudioware's VintageWarmer (Mac/Win, $149) delivers creamy, analog-sounding split-band and broadband compression modes at a bargain-basement price. This downloadable plug-in is available in VST and RTAS versions for the Mac (including OS X) and in VST and DirectX formats for Windows.

Now the bad news: VintageWarmer has a very quirky control interface, which is split between two alternate windows (see Fig. 4). Each of VintageWarmer's three bands has its own threshold and release-time controls, but only the low and high bands offer in-band boost/cut controls. (There is an unwieldy work-around for adjusting mid-band gain.) The gain controls are placed before dynamics, forcing you to move threshold controls higher whenever you boost in-band gain (assuming you want the degree of compression to remain the same). Master threshold and Speed (ganged attack and release) controls complicate operation by interacting with the above-mentioned in-band threshold and release controls. A global knee control is offered in lieu of in-band ratio knobs. You can't solo or bypass individual bands, nor can you adjust the attack time for each band independently. In-band meters (including threshold indicators) are also absent. The owner's manual is vague and confusing, screen redraws are slow, and save routines (for storing custom presets) are unfriendly. On the positive side, you get master input- and output-level controls, excellent global metering (with clipping indicators and switchable ballistics), and a low-band crossover that can be adjusted as low as 25 Hz.

Is it quirky? Yes. Is it awesome-sounding? Yes! Don't be put off by VintageWarmer's interface. You'll quickly become accustomed to it and get incredible results.


For recordists who want to use split-band compression on their projects but don't want to sweat the learning curve, TC Electronic's Finalizer Express Studio Mastering Processor ($1,595) is the answer (see Fig. 5). This stereo digital multiband processor provides 25 presets, each with different fixed attack, release, threshold, and ratio values for its 3-band compressor and its limiter. Its crossovers use linear-phase filters fixed at 315 Hz and 3.15 kHz. Makeup gain is applied automatically, but you can independently adjust the output level of each band downstream of the processing block. Defeatable in-band emphasis controls add additional compression and gain in their respective bands by lowering the thresholds and kicking in more automatic makeup gain.

Image placeholder title

FIG. 7: Waves C4 can apply compression, expansion, limiting, and EQ in any combination (one process per band) to its independent bands.

A global input-level normalizer, an output-stage soft clipper, and an output fader add to the 1U rackmountable unit's value, as does a fairly generous offering of analog and digital I/O. High-resolution LED-ladder metering keeps you apprised of gain-reduction levels for each band, master I/O levels, and soft-clipper action. The Finalizer Express can operate at 44.1 or 48 kHz nominal sampling frequency.

Although the Finalizer Express is a simple solution for quick demo production tasks, I'd be wary of using a preset split-band compressor for critical mastering applications. If you want to work at higher sampling rates or you require much more manual control, the TC Electronic Finalizer 96 kHz ($2,995) offers 88.2 kHz and 96 kHz operation, upsampling from and downsampling to lower rates, additional multiband processing options, and in-band manual controls for attack, release, threshold, ratio, and crossover frequencies.


The world's only all-tube split-band compressor, the dual-channel Tube-Tech SMC 2B ($4,999), incorporates six optical compressors (see Fig. 6). The unit's two crossovers are continuously variable, with the low crossover adjustable down to 60 Hz. In-band threshold, ratio, attack, release, and makeup gain controls (all continuously variable) are joined by 11-segment LED displays that show gain reduction for each band. Global controls include master gain, L/R balance, and bypass. Unfortunately, individual bands cannot be bypassed.

Yes, it's extremely pricey. Nonetheless, the SMC 2B is simply the sweetest-sounding multiband compressor I've heard.


The Waves C4 plug-in (Mac/Win; native version, $400; TDM version, $800) supports TDM, RTAS, AudioSuite, and VST formats for both the Mac and Windows (see Fig. 7). A MAS version for the Mac and a DirectX version for Windows are also available.

C4 offers four independent bands of compression, expansion, limiting, and equalization in any combination. You can use band-limited compression for bass frequencies, for example, coupled with expansion for the highs only. The slope of the C4's widely adjustable crossover filters can be varied globally. In-band controls include threshold, makeup gain, range, attack, and release. Each band can be soloed or bypassed. Master threshold, makeup gain, range, attack, and release controls each enable you to increase or decrease associated in-band parameter values as a group, maintaining interband offsets — a great feature.

A wide-ranging knee control, toggling release-characteristic (Opto or Electro) function, master output-level fader, and Waves' outstanding ARC program-sensitive release-time mode complete this plug-in's comprehensive set of master controls. Metering is excellent, providing a threshold indicator for each band, master L/R output meters and clipping indicators, and an ingenious graphical display that dynamically shows gain reduction (or potential boost, for processes other than compression) in each band.


Linear Phase Multiband is sold as part of the Waves Masters Bundle (Mac/Win; native versions, $900; TDM version, $1800). The Masters Bundle is available in TDM, RTAS, AudioSuite, and VST formats for both the Mac and Windows. Waves also offers a MAS version for the Mac and a DirectX version for Windows.

Essentially an improved and expanded version of the company's C4 Multiband Parametric Processor, the Linear Phase Multiband offers five independent bands with linear-phase crossovers (see Fig. 8). Enhancements include a trim control that optimizes headroom at the plug-in's output, dither, an automatic makeup-gain function, and Waves' pioneering Adaptive Threshold function, which automatically varies the threshold of a band to counteract frequency-masking effects. The only downside to the Linear Phase Multiband is that it's a voracious CPU hog.

The Waves Linear Phase Multiband sounds extremely transparent and offers more control than any other multiband plug-in I'm aware of. For DAW-based multiband mastering applications, it can't be beat.


Now that you've lusted over numerous split-band compressors, it's time to put away your drool bib and get to work. Here's a look at some real-world applications for split-band compressors, beginning with treating vocals.

Vocal tracks can present a number of challenges at mixdown. One hurdle is the varying proximity effect caused by a singer moving closer to and farther away from the mic during a performance. Simply rolling off bass frequencies with a static EQ setting will cause high vocal phrases to sound thin. A split-band compressor, however, will put a lid on the proximity effect without reducing fullness when the singer reaches for high notes.

Image placeholder title

FIG. 8: Waves Linear Phase Multiband offers five independent bands of processing, ultratransparent sound, and a feature-rich interface.

A typical approach might be to set your lowest crossover frequency to somewhere between 200 and 300 Hz, depending on the mic's response at close distances to the source and other factors. Set your low-band threshold so that when vocal phrases that aren't bass-heavy occur, they barely trigger low-band compression; then raise the threshold slightly higher so that those passages then pass through the compressor untreated. Any vocal phrases that are even slightly boomy will then exceed the low-band threshold and will be compressed according to your ratio (or range), attack, release, and other settings. By the way, you can use similar settings and adjustments to take the boom out of an acoustic-guitar track.

It's easy to de-ess a vocal track with a split-band compressor. Set your highest compressor crossover to 5 kHz for treating a male singer or to 6 or 7 kHz for a female vocalist. Then set the high-band threshold so that only sibilant phrases trigger in-band compression. De-essing usually requires a high ratio (10:1 to 50:1 is often called for), a high range (as much as 20 dB of gain reduction for extreme cases), or both, though much lighter treatment should be used for dealing with subtle problems. Set your high-band attack time as fast as possible (50 µs works great) and set the release time to between 50 and 60 ms.

Most of the time, you can also perform light de-essing on an entire mix without noticeably affecting the sound of most of the included instruments; cymbals, triangles, and other sources that produce prominent upper partials are notable exceptions. Whereas standard de-essing techniques using a broadband compressor will usually punch “holes” in a mix (causing noticeable dips and quick recoveries in level for the entire mix), carefully de-essing a mix with a split-band compressor will fully preserve the dynamics of bass and midrange content and avoid audible pumping.


A vocal track that sounds shrill on select phrases can be tamed by setting two of your split-band compressor's crossovers to roughly 3 kHz and 5 kHz. Set your threshold in this band so that only the shrill passages trigger in-band compression. A light to moderate ratio or range setting is usually all that's required to take the edge off everything but the most obnoxious banshees. On the other hand, heavy compression of low highs will dull the vocal track, so use moderation.

You can also lightly compress low highs on a guitar- or keyboard-heavy mix to prevent fatiguing edginess from rearing its ugly head during the loudest sections of your program material. But tread lightly, or your mix's detail and clarity will suffer.

Perhaps the most common use of split-band compression in modern recordings is to simultaneously fatten and tighten up the bottom end of a mix. Here, your ratio setting should be set according to how much the bottom end varies in intensity; widely varying bass energy requires a higher ratio than is called for when dealing with more constant levels. You'll have to use your ears to adjust the attack and release times for your particular mix, but bass frequencies are often best compressed using moderate or slow time constants. Lower your threshold until the bass frequencies sound tight, so that they don't bloom or dip too much at any point in the mix. Then increase the makeup gain in the bass band to the point where the bottom end sounds huge but does not overwhelm other elements and frequencies in the mix.

Adjusted properly, you'll often find that the mix's peak bass energy will now be lower than before you compressed it, which translates into overall greater headroom for the entire mix. That is the best of both worlds: the bottom end will sound tighter and you can make your mix louder if you want to.


Split-band compressors have many more applications, but they all begin with careful listening and evaluation of your program material. Split-band compressors should be used only when undesirable, excessive, and dynamic variances in spectral energy are evident (that is, when static EQ won't fix the problem). Routine application of split-band compression on material that doesn't need it — or heavy-handed use on material that needs only small amounts — will make a dynamic and compelling performance sound lifeless and impotent, or worse. It's all too easy to go overboard with split-band compression and ruin a good thing. However, deliberate and deft use of split-band compression will give your tracks and mixes a polished sound that you can't achieve any other way.

EMcontributing editorMichael Cooperis the owner of Michael Cooper Recording, located in beautiful Sisters, Oregon. Cooper's studio offers recording, mixing, and mastering services.


Arboretum Systems tel. (800) 700-7390; e-mail sales@arboretum.com; Web www.arboretum.com

BSS Audio tel. (615) 360-0277; e-mail bssaudiousa@harman.com; Web www.bss.co.uk

Mark of the Unicorn (MOTU) tel. (617) 576-2760; e-mail sales@motu.com; Web www.motu.com

PSPaudioware.com s.c. tel. 48-601-963-173; e-mail contact@pspaudioware.com; Web www.pspaudioware.com

TC Electronic tel. (805) 373-1828; e-mail infous@tcelectronic.com; Web www.tcelectronic.com

Tube-Tech/TC Electronic (distributor) tel. (805) 373-1828; e-mail infous@tcelectronic.com; Web www.tube-tech.com

Waves tel. (865) 546-6115; e-mail info@waves.com; Web www.waves.com