Digital reverb splits into two broad families. Algorithmic reverbs build an ambience from networks of delays and filters; convolution reverbs start from a recording of the space itself. That difference isn’t just a matter of taste. It shapes how the sound settles into your mix and which situations each approach is best suited to.
What is convolution reverb?
A convolution reverb applies the measured acoustic fingerprint of a place or a device to your signal. The processor takes your track and blends it, sample by sample, with an impulse response — an IR for short. The result sounds as if your vocal, your snare or your synth pad had been played in the cathedral, studio or echo chamber that the IR came from.
The word “convolution” is simply the maths that performs this marriage. You don’t need the equation: just remember that the processor layers your sound with thousands of tiny delayed, attenuated copies, spread through time exactly as the IR dictates. Where an algorithmic reverb invents a believable acoustic, convolution reproduces one that genuinely existed.
The impulse response (IR): the heart of the process
An impulse response is the “sonic photograph” of a space. Picture a single hand clap in the middle of a church: the way that clap spreads, bounces off the walls and dies away describes the room’s acoustic signature completely. The IR captures that tail precisely — its length, its colour, the density of its reflections. Once you have the file, a convolution reverb can stamp that same signature onto any source.
How an IR is captured
In practice nobody uses an actual hand clap, which is too imprecise. Engineers play a controlled reference signal into the space — usually a slow sine sweep rising from lows to highs — and record it with one or more microphones. A deconvolution step then extracts the pure fingerprint of the room, stripped of the excitation signal. The choice and placement of the capture microphone weigh heavily on the IR’s character, exactly as in conventional recording.
This isn’t limited to rooms. The same method captures the fingerprint of spring reverbs, plates and echo chambers, but also of loudspeakers and guitar cabinets. That’s why IRs today serve as much to recreate a concert acoustic as to emulate a specific cab for a reamped guitar.
Convolution vs algorithmic: two philosophies
Neither approach is better in the abstract; they answer different needs. Convolution excels at realism: nothing recreates the scale of a real hall or the grain of a historic plate as faithfully. Algorithmic reverb offers a flexibility convolution long lacked: you shape size, diffusion or decay in real time, without swapping files.
| Criterion | Convolution | Algorithmic |
|---|---|---|
| Spatial realism | Very high, based on a real place | Believable but synthetic |
| Control flexibility | Limited (tied to the IR) | Wide (size, decay, diffusion) |
| CPU load | Often heavier | Generally lighter |
| Typical use | Post, credible acoustics, cabs | Creative mixing, textures, long pads |
The line is blurring, though: modern convolution reverbs add decay, envelope and pre-delay controls, while some high-end algorithmic units rival them for realism. So the choice is less about the technology than about the result you’re after.
Using it well in the mix
A few habits make all the difference when you drop a convolution reverb into a mix:
- Put it on an aux bus rather than as an insert on every track: you share one acoustic across several sources and keep control of the amount.
- Mind the pre-delay. A few dozen milliseconds between the dry sound and the start of the tail separate source from reverb and protect intelligibility, especially on vocals.
- EQ the reverb, not just the source. A high-pass filter on the return stops the low end clogging up; a gentle dip in the highs tames an over-bright IR.
- Beware over-long IRs on a fast track: the tail masks transients and drowns the mix. Shorten the decay if your plugin allows it.
- Check mono compatibility. A gorgeous stereo IR can expose phase issues once the mix is summed; a regular mono check (see Mid/Side processing) avoids nasty surprises.
Limitations worth knowing
Convolution has two trade-offs. First, an IR freezes a moment: it doesn’t “move” like a real room you walk through, and the subtle modulations that bring a plate reverb to life are sometimes missing from static versions. Second, the processing stays demanding, especially with long IRs at low latency; on a big session, group your convolution reverbs onto a few buses and, if the source is already compressed, watch how the tail reacts. Nothing fatal, but these points explain why many engineers keep both families on hand and pick by context.
My take
Convolution is still the ultimate tool when you want an acoustic that feels genuinely real: a vocal sitting in an actual chamber, drums stretching out in a big hall — nothing does it better. Its weakness isn’t really one, just the flip side of the coin: an IR is a frozen acoustic, and sometimes you’d rather have an algorithmic reverb you can stretch and sculpt at will. The smart move isn’t to pick a side but to know why you reach for each. Keep a handful of quality IRs you know inside out; they’ll serve you more often than the library of a thousand presets nobody ever auditions in full.
