Mantissa manual
Mantissa is a digital reverb modelled on a studio rack unit from 1978. It reconstructs the original's eight reverb programs and its Chorus program with their defaults and ranges, and runs them at 20.48 kHz, the rate the original most probably used, through models of its 12-bit converters and its 16-bit word limits. A switch moves it to the 1984 revision's faster rate and wider bandwidth. On top of that are the features later hardware added (Size, Definition, Stop Decay), a Mix control, presets, and predelay in note values.
Formats: VST3, AU, CLAP and Standalone on macOS; VST3, CLAP and Standalone on Windows. Free.
Installing
On macOS, open the installer (.pkg) and choose the formats you want. It is universal (Apple silicon and Intel), notarised, and needs macOS 11 or later. On Windows, run the installer (.exe) on Windows 10 or later, 64-bit. Rescan plugins in your DAW afterwards. Logic and GarageBand pick up the Audio Unit the next time they start.
Signal path
The input passes through Drive, then an input transformer model and a pre-emphasis shelf. It is then band-limited and resampled to the engine's rate: 20,480 Hz in the 1978 era, with about 8 kHz of bandwidth, or 34,133 Hz in the 1984 era, with about 15 kHz.
The 12-bit converter chooses one of four gains (0, 6, 12 or 18 dB) before every sample, as the original's "instantaneous floating point" converters did. Each sample becomes a 12-bit mantissa with a 2-bit exponent, which is where Mantissa gets its name. Its resolution follows the level of the signal: loud passages are converted coarsely, quiet ones finely. Every write to the reverb's delay memory clips at the limits of a 16-bit word, and the Ovfl lamp lights when it does. With Stepped on, the gains Diffusion and Definition set inside the network move in steps of 1/32. The output converter ranges the same way as the input. After it come a reconstruction filter back to your session's rate, de-emphasis, Output and Mix.
The hardware's outputs were 100 % reverb. Mix is our addition. The dry signal is delayed to line up with the reverb. Mantissa reports about 3.7 ms of latency to the host at any sample rate, and it reports its tail length, so a host that stops processing silent tracks doesn't cut the reverb off.
Mono and stereo inputs both work, and the output is always stereo. Chamber and Dense Plate sum their input to mono, as on the original.
The panel
From top to bottom:
- The display shows the program and the last control you moved. A few seconds later it goes back to the reverb time (the predelay on Chorus). Preset, register and tape messages appear here too. In, at the right, switches the reverb in and out; unlit, it is bypassed.
- The nine program keys.
- The mode row: Mode Enhancement and Decay Optimisation with their steppers, the Outputs pair, the Era pair, and the Clean, Stepped and Spill options.
- The register row: eight registers and Store, the tape Save key, the two Page keys, and the preset window with its arrows.
- The fader bay: seven faders that change with the page, Mix, and the Input meter.
Click the name in the top left to see the mainframe and the version. Click anywhere to go back to the panel. Each time the window opens, the panel lights every lamp for a moment and then reads Self Test OK.
Hover over any control for a tooltip. The text changes with the program where the control behaves differently.
Programs
The nine programs keep the original's final firmware order and defaults. Small Hall, Large Hall and Bright Hall have the same echo timings in the reference renders, so they share one hall network with different defaults and damping. Room and Chamber are that network scaled down, because the original's own room and chamber networks were never published.
| # | Program | What it is | Loads with | Predelay range |
|---|---|---|---|---|
| 1 | Small Hall | A small concert hall with a moderately dense start. | 2.0 s mid, 3.0 s bass, 24 ms | 24 to 152 ms |
| 2 | Vocal Plate | A plate for vocals, with a sparse, clear start and the treble damping fully open. | 2.2 s, no predelay | 0 to 107 ms |
| 3 | Large Hall | The hall network set up for long decays. Its Diffusion of 1 lets early echoes come through separately; raise it for drums. | 2.6 s mid, 3.4 s bass, 24 ms | 24 to 152 ms |
| 4 | Chamber | A chamber with a sparse start. Its input is summed to mono and its diffusion is fixed, so the Diffusion fader does nothing here. | 2.2 s mid, 2.8 s bass, 25 ms | 25 to 255 ms |
| 5 | Drum Plate | Denser from the start than Vocal Plate, with more diffusion inside its loop. | 2.2 s, no predelay | 0 to 107 ms |
| 6 | Bright Hall | Small Hall's network with gentler treble damping, starting at 6.6 kHz. | 2.0 s mid, 3.0 s bass, 24 ms | 24 to 152 ms |
| 7 | Room | The hall network scaled down, so the echoes are close together and the build-up is quick. | 1.8 s, 24 ms | 24 to 255 ms |
| 8 | Dense Plate | Dense from the start, and it keeps the same density as it decays. Its input is summed to mono. Decay Optimisation and Definition don't apply to it. | 1.8 s, 5 ms | 5 to 185 ms |
| 9 | Chorus | Eight moving delays, four per side, with no reverb tail. | Every voice at 50 %, 12 ms | 0 to 253 ms |
Pressing a program key sets the page 1 faders and the Mode Enhancement and Decay Optimisation amounts to that program's values, with Predelay in milliseconds. Page 2, Mix and the other keys stay as they are. Store your changes in a register or save them as a preset first if you want to keep them.
On the hall programs, Room and Chamber, the predelay reading includes part of the network's own delay, as it did on the original. On Small Hall the reverb starts about 20 ms before the value shown, and about 23 ms before it at the top of the range, within a millisecond of the reference. The plates start within about 2 ms of the value shown.
Page 1 faders
| Fader | Range | What it does |
|---|---|---|
| Bass Decay | 0.6 to 70 s | Reverb time below the Crossover frequency: how long the tail takes to fall by 60 dB. Set it longer than Mid Decay for a longer low end. |
| Mid Decay | 0.6 to 70 s | Reverb time above the Crossover frequency. This is the main length control. |
| Crossover | 100 Hz to 10.9 kHz | The frequency that splits Bass Decay from Mid Decay. It does nothing when the two are equal. |
| Treble Decay | 100 Hz to 10.9 kHz | Damping inside the reverb loop. Above this frequency the decay is shorter, so the tail darkens as it fades. The damping is lightest at the top of the fader. |
| Depth | 0 to 71 | How far away the reverb sounds. Near 0 the early part of the network reaches the outputs directly, so the sound is close, with a strong onset. At 71 only the late reverb is heard. The halls, Room and Chamber respond most, the plates less. Chorus doesn't use it. |
| Diffusion | 0 to 63 | How dense the echoes are. Turn it down to hear separate echoes. Chamber's diffusion is fixed. On Chorus it blurs voices 3 and 4. |
| Predelay | Per program, within 0 to 255 ms | Delay before the reverb starts. A setting outside the program's range stays at the nearest limit. See "Predelay in note values" below. |
On Chorus, faders 1 to 4 set the levels of the four voices. Each voice is a delay on each side that moves within its range while Mode Enhancement is on, after the predelay:
| Voice | Left | Right |
|---|---|---|
| 1 | 7 to 12 ms | 7.9 to 12.9 ms |
| 2 | 9.8 to 14.8 ms | 9.1 to 14.1 ms |
| 3 | 19 to 27 ms | 21.5 to 29.5 ms |
| 4 | 31 to 39 ms | 27.5 to 35.5 ms |
Voices 1 and 2 overlap, so at equal levels they phase against each other. Diffusion blurs voices 3 and 4 into a cluster of close echoes.
Page 2 faders
| Fader | Range | What it does |
|---|---|---|
| Size | 0.5x to 4x, in seven steps | Scales every delay in the network. Larger sizes spread the echoes further apart and slow the build-up, while Bass and Mid Decay still set the length. A change fades the reverb out and starts it again. |
| Definition | 0 to 100 % | Lowers the diffusion inside the loop, so the tail keeps more distinct echoes and turns grainier as it decays. At 0 % it has no effect. Dense Plate's loop diffusion is fixed, so it ignores Definition. |
| Stop Decay | Off, then 0.2 to 20 s | A second decay time for when the input stops. Once the input, after Drive, has stayed below about −54 dBFS for the Hold time, the decay glides to this time. When sound returns it glides back. Bass keeps its ratio to Mid. The display reads Stopped while it is in effect. |
| Hold | 20 ms to 2 s | How long the input must stay quiet before Stop Decay takes over. With a short hold it acts in the gaps between notes. |
| Drive | −12 to +15 dB | Level into the input transformer and converters. The same amount comes off after the converters, so the level stays put. Higher settings saturate the transformer and can clip the converters; lower ones bring out the converters' grain on quiet passages. With Clean on, Drive does nothing. |
| Spread | 0 to 100 % | Mixes the other output pair into the one picked under Outputs, with the level compensated. On the plates and Chorus the other pair is the same signal with the sides swapped, so raising Spread narrows the image, to mono at 100 %. |
| Output | −12 to +12 dB | Level of the reverb before Mix. The dry signal isn't affected. |
Mix (0 to 100 %, default 30 %) stays at the right on both pages. It is an equal-power crossfade between the dry signal and the reverb. Use 100 % on a send.
The Input meter shows the level after Drive in 6 dB steps from −24 to 0 dB, left and right. Ovfl lights when the converters clip or the 16-bit arithmetic overflows, which can happen below 0 dB on bright material or long decays. Turn down Drive or the source.
Moving the faders
Drag a fader, or use the mouse wheel over it. Double-click a fader, or Alt-click it, to return it to the current program's value (page 1) or its default (page 2). The readout under each fader shows the value you hear. With Stepped on, that is the snapped value.
Mode Enhancement and Decay Optimisation
Mode Enhancement moves some of the delays inside the network in small random steps, at a constant speed, changing direction at random. This breaks up ringing resonances and gives the tail a slight pitch drift that changes direction at random. Switching it on or off starts the reverb again, as on the original.
- Amount (1 to 16) is the time between steps. Lower values move the delays more often.
- Pitch (1 to 16) is the size of each step. Higher values give a wider drift.
On Chorus, Mode Enhancement moves the voices. With it off they are fixed delays.
Decay Optimisation follows the input level. About a second after the input falls well below its recent peak, the diffusion inside the loop is lowered in steps, so the late tail rings less and fades a little sooner. It steps back when the input returns, or once the held peak has faded. Amount (1 to 16): lower values hold on to the input's peak for longer, so the effect stays in longer after the input drops. Dense Plate and Chorus don't use it.
Outputs
The original had four outputs. Front takes the reverb from the main pair. Rear takes it from the other pair, which on the halls, Room and Chamber comes from other points in the same network, so the early pattern and the stereo image change while the decay stays the same. On the plates and Chorus the rear pair is the front pair with its sides swapped, as on the original. Spread on page 2 mixes the two.
Era
1978 runs at 20.48 kHz with about 8 kHz of bandwidth, for a darker top. 1984 runs at 34.13 kHz with about 15 kHz of bandwidth and the later revision's emphasis curve, for a brighter top. The networks and controls are the same in both. Switching fades the reverb out and back in, and both eras report the same latency, so switching doesn't move anything in time.
Options
Clean takes out the converters, the input transformer, the 16-bit clipping and the noise floor. The sample rate, bandwidth and networks stay as they are.
Stepped (on by default) snaps Bass Decay, Mid Decay, Crossover and Treble Decay to the 31 values each that the original offered, and moves the gains Diffusion and Definition set inside the reverb in the original's 1/32 steps. Depth and the Chorus voice levels still move smoothly. Off, the controls move smoothly; Decay Optimisation still works in whole steps. The 31 decay times are 0.6, 0.8, 0.9, 1.1, 1.2, 1.3, 1.4, 1.5, 1.7, 1.8, 2.0, 2.2, 2.4, 2.6, 2.8, 3.0, 3.4, 3.8, 4.2, 4.6, 5.2, 5.7, 6.5, 7.5, 8.5, 10, 12, 16, 22, 35 and 70 s.
Spill: when you change program, the old one keeps ringing and fades over 5 seconds while the new one starts. Off, the old program fades out in about 10 ms. Size, era and Mode Enhancement changes still restart the reverb.
Predelay in note values
Hold Shift while you move the Predelay fader and it moves in note values at the host's tempo: 1/64 triplet, 1/64, 1/32 triplet, 1/32, 1/16 triplet, 1/16, 1/8 triplet, 1/8, 1/4 triplet and 1/4. Only the values that fit the current program's predelay range are offered. Shift with the mouse wheel steps one note at a time. On a Mac, a Shift-wheel scrolls sideways, and Mantissa treats that as up and down.
Click the readout under the fader to switch between milliseconds and note values. Moving the fader by hand without Shift, or double-clicking it, goes back to milliseconds. The display shows the note and the time you hear, for example "Predelay 1/16" and the milliseconds.
The note is held as a tempo-synced setting, so the predelay follows tempo changes in the session. If the host doesn't send a tempo, Mantissa uses 120 BPM. Predelay Sync and Predelay Note are automatable. Loading a program switches sync off.
At 120 BPM a 1/64 note is 31 ms, a 1/32 is 63 ms and a 1/16 is 125 ms. A program whose range ends at 107 ms (Vocal Plate and Drum Plate) can't reach a 1/16 at that tempo, so the fader stops at the longest note that fits.
Presets
Mantissa comes with 56 factory presets. Click the preset window in the register row for the menu:
| Group | Presets |
|---|---|
| Basics | 1 (Init) |
| Halls | 8 |
| Plates | 8 |
| Rooms and Chambers | 7 |
| Vocals | 6 |
| Drums | 6 |
| Instruments | 6 |
| Synced | 6 |
| Special | 8 |
The arrows beside the window step through the menu's order: the factory presets, then your own. A preset sets every control on the panel except In and Spill, which stay as they are. The window shows an asterisk once you change anything. The preset's name is saved with your project, so the asterisk is still right when you open it again.
Save writes your changes back to one of your own presets. Save As asks for a name in the window; press Return to save or Escape to cancel. Factory names can't be reused, and saving over one of your own presets asks first. Delete removes one of your own presets after asking. Show Preset Folder opens the folder in Finder or Explorer.
User presets are XML files you can copy between computers or share:
- macOS:
~/Library/Application Support/Sliding Cheesecake/Mantissa/Presets - Windows:
%APPDATA%\Sliding Cheesecake\Mantissa\Presets
Recalling a register or loading a tape leaves no preset loaded, since the panel no longer matches one.
Registers
The eight registers work like the hardware's memories. Press Store, then a register key, to keep the current settings there; a register that already holds settings is overwritten. Press Store again to cancel. Press a register key to recall it: the program and every control. A brighter number means the register holds settings.
The registers are shared by every copy of Mantissa on the computer and kept between sessions, in Registers.xml beside the Presets folder. They are a quick way to move a sound from one track or project to another.
Tape
Save under Tape writes the current settings as a WAV file of data tones (Kansas City Standard, 300 baud), as the 1984 hardware's remote could save to audio tape. Keep it like any audio file, or record it to cassette. To load a tape, drag the WAV file onto the panel. The display reads Tape Loaded, or Not A Tape if the file doesn't decode. A tape resampled to another sample rate still loads.
Firmware quirks
The original's final firmware had known faults. Click the screw at the bottom right of the panel to bring them back; click it again to switch them off. The display confirms which. For now this brings back the pops in the right channel of the Chorus program. It is off by default.
Automation
Every control is automatable. The parameters, as your host lists them:
Program, Bass Decay, Mid Decay, Crossover, Treble Decay, Depth, Diffusion, Predelay, Mix, Mode Enhancement, Enhance Amount, Enhance Pitch, Decay Optimisation, Decay Opt Amount, Outputs, Clean Converters, Bypass, Era, Size, Definition, Stop Decay, Stop Hold, Drive, Output, Spread, Stepped Controls, Spill Tails, Firmware Quirks, Chorus Voice 1 to 4, Predelay Sync and Predelay Note.
Automating Size, Era or Mode Enhancement restarts the reverb each time. Switching Mode Enhancement did the same on the original.
Tips
- On a send, set Mix to 100 % and use the send level. On an insert, the presets' Mix values are a starting point.
- The 1978 era has about 8 kHz of bandwidth, so it sounds dark. If a hall sounds dull on a bright mix, try the 1984 era before opening Treble Decay all the way.
- Plates with Spread above 0 get narrower, not wider. Leave Spread at 0 on the plates unless you want a narrower plate.
- Low Diffusion with Definition up gives sparse, separate echoes and a grainy tail. High Diffusion with Mode Enhancement on gives the smoothest tails.
- If Ovfl lights on long decays, lower the input or Drive. The original's overflow lamp also lit on long decays and loud inputs.
- Stop Decay keeps a reverb short under a busy part and lets it ring on when the part stops. Set Hold a little longer than the gaps you want ignored.
Specifications
| Formats | VST3, AU, CLAP and Standalone on macOS. VST3, CLAP and Standalone on Windows. |
| macOS | 11 or later, Apple silicon and Intel (universal) |
| Windows | 10 or later, 64-bit |
| Programs | 9: Small Hall, Vocal Plate, Large Hall, Chamber, Drum Plate, Bright Hall, Room, Dense Plate, Chorus |
| Engine | 20,480 Hz (1978) or 34,133 Hz (1984); 16-bit word limits; diffusion gains in 1/32 steps with Stepped on |
| Converters | 12-bit with four-step gain ranging per sample (0, 6, 12, 18 dB), or Clean |
| Bandwidth | About 8 kHz (1978) or 15 kHz (1984) |
| Decay | 0.6 to 70 s, separate bass and mid, 31 steps or continuous |
| Predelay | 0 to 255 ms depending on the program, or note values from 1/64 triplet to 1/4 |
| Inputs and outputs | Mono or stereo in, stereo out |
| Latency | About 3.7 ms, reported to the host |
| Presets | 56 factory presets, plus your own; eight registers; tape save and load |
| Price | Free |