Every compressor needs a threshold. It's one of the first things you learn about dynamics processing and one of the last things you truly master. Setting it wrong means the compressor either ignores the peaks you want to control or clamps down on everything indiscriminately. The problem is inherent to the design: a threshold-based detector measures how loud a signal is, and "how loud" is relative to the input gain, the arrangement, the session.
In 1998, Ruben Tilgner at SPL Audio in Germany built something different. His Transient Designer needed no threshold. It processed a quiet violin and a loud kick drum with identical proportional accuracy, without asking the engineer to touch a threshold knob at all. The underlying technology — Differential Envelope Technology, or DET — is mathematically elegant and practically underappreciated, even by engineers who use transient shapers daily.
The Problem with Threshold Detection
A compressor detects transients by watching for level to exceed a user-defined threshold. The problem is that this measurement conflates two different things: the magnitude of a sound and the shape of its onset.
Consider a snare hit recorded at -20 dBFS versus the same hit recorded at -6 dBFS. The transient shape — the sharp rise from silence to peak — is acoustically identical. But a threshold-based processor treats them completely differently. Set the threshold for the louder recording and the compressor ignores the quieter one. Set it for the quieter and the louder gets crushed. Every gain-staging decision upstream changes how the processor behaves.
Tilgner's insight was to measure something different: not level, but rate of change.
Differential Envelope Technology
The original SPL Transient Designer (first shown at the Frankfurt ProLight + Sound fair in spring 1998) was a fully analog device, built around THAT 2181 VCA chips as the gain-control elements. The key was the envelope detection circuit Tilgner developed, which he called Differential Envelope Technology.
The principle: generate two envelope signals from the same input, one with a fast attack time constant and one with a slow attack time constant. Subtract the slow from the fast. The result is a control signal that is close to zero when the input is steady-state — both envelopes have had time to converge — and strongly positive when a transient arrives, because the fast envelope has jumped ahead of the slow one.
This difference signal represents the rate at which the signal's amplitude is rising, not the amplitude itself. It is, mathematically, an approximation of the first derivative of the amplitude envelope.
The consequence: the response is level-independent. A soft transient and a loud transient produce the same shaped difference signal, scaled proportionally. You can process a session without touching any threshold, and the processor still finds every attack.
The DSP Implementation
In a software plugin, the circuit maps directly to two first-order peak envelope followers with different time constants:
float fast_env = 0.0f;
float slow_env = 0.0f;
// Attack coefficients: fast ~1ms, slow ~20ms
const float fast_coeff = std::exp(-1.0f / (0.001f * sampleRate));
const float slow_coeff = std::exp(-1.0f / (0.020f * sampleRate));
for (int i = 0; i < numSamples; ++i)
{
float rectified = std::abs(input[i]);
fast_env = fast_env * fast_coeff + rectified * (1.0f - fast_coeff);
slow_env = slow_env * slow_coeff + rectified * (1.0f - slow_coeff);
// Positive during attacks, near-zero at steady state
float transient_signal = fast_env - slow_env;
// Use transient_signal to modulate gain of attack portion
output[i] = input[i] * applyAttackGain(transient_signal, attackParam);
}The fast envelope (1ms attack) closely tracks instantaneous peaks. The slow envelope (20ms attack) represents the running average of signal level. Their difference is a narrow pulse aligned to each transient onset — exactly the control signal needed to boost or cut attack without touching the sustain.
The ratio between the two time constants determines the sensitivity to fast versus slow onsets. Narrow the ratio and you only respond to the sharpest drum hits. Widen it and slower-rising material like guitar plucks also gets caught.
Attack and Sustain as Separate Measurements
What distinguishes the SPL design from a simple envelope subtractor is that it processes attack and sustain independently, using two separate difference signals.
The attack circuit fires during the rising edge — when the fast envelope pulls ahead of the slow one. The control signal is positive and quickly decays as both envelopes converge on the peak.
The sustain circuit operates on the falling edge — after the peak, when the signal is decaying and the slow envelope is now above the fast one (which is chasing the falling signal downward faster). The resulting difference signal during decay is what the sustain parameter acts on.
In practice: you can simultaneously boost the transient onset and shorten the tail of a snare, or reduce the attack of a guitar strum and extend its sustain. These are independent operations running in parallel, not sequential stages. A compressor cannot do this — its attack and release parameters control the timing of the gain reduction, not separate aspects of the sound's envelope shape.
Look-Ahead and Phase Coherence
Modern plugin implementations often add a look-ahead buffer of 5 to 15 milliseconds. The signal being processed is delayed by this amount; the envelope detection and gain calculation run on the un-delayed signal. This lets the processor calculate the required gain change before the transient actually reaches the output stage, eliminating the slight lag inherent in causal envelope following.
The tradeoff is latency — the plugin reports a fixed plugin delay to the DAW for compensation. In practice this is invisible during mixing but rules out use on live monitoring chains without dedicated hardware compensation.
What This Changes in Practice
Knowing how the measurement works clarifies something counterintuitive: a transient shaper is not a specialized compressor. It is a gain processor that operates on the shape of the amplitude envelope rather than its magnitude. This makes it complementary to compression rather than redundant with it.
A compressor on a drum bus controls the integrated loudness of the performance — ensuring the quieter hits match the louder ones, managing overall level. A transient shaper on the same bus controls the feel — making hits punchier or softer without changing their relative levels to each other, because it responds proportionally to each hit's own attack shape.
Ruben Tilgner left SPL in 2005 and founded elysia GmbH, continuing DET-based designs. The SPL Transient Designer 4 Mk2 (the four-channel version of the original hardware) is still in production. The number of software transient shapers that have followed in the 28 years since — from Waves Smack Attack to Sonnox Oxford TransMod to the native transient tools in every major DAW — suggests Tilgner identified something real about how dynamics processing can work when you stop measuring level and start measuring change.