LAP TSL-A95
A 12-bit digital delay built as a component-level circuit model: the analogue board - input stage, pre-emphasis, limiter, anti-alias and reconstruction filters, de-emphasis and the output stage - is solved as a circuit in real time around a bit-exact model of the converter, the memory and the clock.
How it works - and how it differs from a plugin delay. TSL-A95 is a 12-bit digital delay modelled part by part: pre-emphasis, a limiter and an anti-alias filter in front of a 12-bit converter, a 32k-word memory, reconstruction filter and de-emphasis after it. The delay time is RANGE (2 … 512 ms in octaves) × FINE (×0.5 … ×2). FINE changes the clock of the converter, so turning it while sound is in the memory shifts the pitch like tape; with HOLD the memory loops and FINE transposes it by up to an octave. The delayed signal is band-limited to about 10 kHz and 12 bits; the direct signal is clean. The limiter sits in the delay path after the pre-emphasis, so bright material lights LIMIT first and the repeats come back compressed and darker - FEEDBACK returns them through the limiter and filter again, PHASE inverts the feedback. MODULATION modulates the clock (a vibrato of the delayed sound); its depth in cents grows with RANGE. BYPASS mutes the delayed path only; POWER switches the unit off. HI drives the input 6 dB harder.
The whole unit, solved live
Not a delay algorithm: the analogue board around the line is solved as a circuit at sample rate - op-amps with real bandwidth, the JFET limiter with its detector, every filter from its components. The digital line is bit-exact: a 12-bit converter, a 32k-word memory and a clock from 32 to 128 kHz. The sound of the repeats falls out of the parts.
RANGE × FINE - a clock, not a buffer
RANGE sets the length of the memory in octaves (2 … 512 ms), FINE the clock of the converter (×0.5 … ×2), together 1 … 1024 ms. Because FINE changes the clock, turning it while sound is in the memory bends the pitch like tape - and the bandwidth of the repeats stays at about 10 kHz whatever the setting.
Limiter in the delay path
Pre-emphasis lifts the treble by about 14.5 dB in front of the converter and a limiter keeps the line from clipping, so bright material lights LIMIT first and the repeats come back compressed and darker. FEEDBACK returns them through the filter and the limiter again; at 100 % the line is just short of self-oscillation. PHASE inverts the feedback.
HOLD · MODULATION
HOLD stops writing into the memory: its contents loop endlessly, FINE transposes the loop by up to an octave and DEPTH / SPEED wobble it. MODULATION modulates the clock - a vibrato of the delayed sound whose depth in cents grows with RANGE, from 10 Hz down to 0.1 Hz.
MIX · BYPASS · SENSITIVITY
One knob from the direct signal to the delayed one; INV. MIX R in the menu puts the repeats on the right with inverted polarity for width. BYPASS mutes the delayed path only, POWER switches the unit off. SENSITIVITY HI drives the input 6 dB harder into the limiter.
Output transformer · SATURATION
After the unit can sit a model of a real audio output transformer - the core with its magnetisation and hysteresis solved sample by sample - for weight in the low end and low-order harmonics that grow with the bass and the level. SATURATION drives it from -12 to +18 dB without changing the level. Off by default - switch TRAFO ON in the settings menu.
Settings menu
Right-click the panel: window SCALE 80 / 100 %, STEREO MIX / INV. MIX R, OVERSAMPLING 1x / 2x / 4x, TRAFO OFF / ON and SATURATION. Everything can be changed while playing - a new unit is built in the background and cross-faded in, without a drop-out.
Presets
A preset bar above the panel: categories, Save / Save as / Delete and ◀ ▶ to step through the settings. A preset holds the knobs, switches and the menu settings; the selected preset is saved with the project. One plain file per preset - copy them between machines and systems.
A mastering plugin
The analogue board is solved as a circuit at every sample: about 35 - 40 % of a CPU core for a stereo instance at the default 2x oversampling (about 20 % at 1x, 70 % at 4x). Made for the master bus, stems and a few key tracks - not for every track of a session.
First release. RANGE 2 … 512 ms × FINE ×0.5 … ×2 (1 … 1024 ms), FEEDBACK with PHASE, HOLD, clock MODULATION (DEPTH, SPEED), MIX ORIG … DELAY, BYPASS and SENSITIVITY HI / LO; oversampling 1x / 2x / 4x (default 2x); a model of a real audio output transformer (off by default) with SATURATION -12 … +18 dB; presets that include the right-click menu settings and are saved with the project.