The keyboard displays any playing notes, showing relative velocity for each note as shades of gray. The darker the gray, the lower the velocity.
You can also play and record from the keyboard. Simply click on the desired key and the corresponding note plays for a second or so, then turns itself off.
When step recording, the keyboard sends note pitch for any clicked key to the respective sequencer.
To the left of the keyboard is a slider labeled Bend . The last value from either this slider or incoming MIDI bend data is available throughout the synth as a mod source.
Bend is bipolar: center position on the slider is no bend; higher values increase bend value; lower values decrease bend value.
The vertical slider labeled Whl sets the mod wheel level. The frq knob next to the modwheel slider sets the LFO frequency, with resulting output as follows:
- If the LFO frq knob is set to 0, then LFO modulation is turned off. Instead, the last value from either this slider or incoming MIDI mod wheel data is directly available throughout the synth.
In this case, the mod wheel source is unipolar. Setting the slider to its bottom causes mod wheel value to be zero; at the top, mod wheel level is at its max level.
- If the frq knob is set higher than 0, then a bipolar LFO modulation is instead available in the synth as the modwheel source.
The Whl slider and incoming mod wheel data sets the base level of the modulation; the frq knob sets the LFO frequency.
MIDI pitch and gate data is routed to the synth and sequencers. When step recording, the note plays while holding down the key, and the pitch and gate information is sent to the respective sequencer.
When activated, Hold causes all incoming MIDI notes to be sustained until all notes are released and the first new note is received.
The arpeggiator provides polyphonic arpeggiation of all active notes. Arpeggiation strictly follows the order of notes received. ClockDiv sets the note frequency by defining a division of the global clock rate. Gate length is half of clockdiv.
With MIDI Hold enabled, you can keep one note playing and add notes to the arpeggio without holding down all the rest of the keys. Then when you release all the keys, the pitch pattern continues to play and you can modify its tonality. When you want a new note, after releasing all notes, play one or more new ones. Immediately the new arpeggio starts.
The power control adjusts the CPU load, as follows:
A four-phase clock spreads overall processing over four consecutive control-rate cycles, reducing CPU peaks. Output values are updated at one quarter the control rate. CPU usage is <30% of normal; however, with slow attack envelopes or fast note runs, new notes may briefly begin with the pitch of a previous note. The resulting audio artefacts (effecting filter as well as pitch) are sometimes pleasing, but often undesirable.
Same as Tiny , but in addition, a note-on event causes all envelopes and gain levels to be completely recalculated in the one cycle. This provides low CPU without audio artefacts, and is most often the preferable setting. However, fast note runs can significantly increase CPU usage.
Same as Small , but with two-phase rather than four-phase clocking. So, all values are updated at half the control rate, rather than one quarter the control rate. Note-on events still force a complete update of output values.
All values are calculated every control-rate cycle, regardless when new notes are received.
Internally, the instrument's switchless modulation architecture only scales a modulation if the modulation amount is not zero (so setting modulation amounts to zero somewhat reduces CPU load). Also, it uses latched accumulators, instead of adders, to sum up multiple modulations. These techniques eliminate unnecessary recalculations for unused modulation sources, monophonic modulations, and also for circuitry after the modulation summing. Otherwise, as each modulation changes, all downstream values would be entirely (and unnecessarily) recalculated.
Actual CPU usage is higher than reported by Reaktor, and can be much lower when displaying the B panel than the A panel (depending on the graphics card). Low-end performance tests were performed on a 1.6 GHz Pentium 4 (with a Matrox G450 graphics card) running Windows XP, on 8 voices, 44.1kHz audio, and 400Hz control rate. With two oscillators, both filters enabled, no saturation, tempo delay, and all other effects disabled, the Windows Task Manager reported Normal CPU usage of ~50%; but Small CPU utilization was only ~21% for the B panel (increasing to ~31% for the A panel).
Without the switchless modulation architecture, similar designs yielded CPU usage in excess of 80%, even on the B panel. Thus, the instrument's switchless modulation architecture easily increases performance four-fold.
Higher performance CPUs can support much more sophisticated oscillator and effect configurations. With a 2.8GHz Pentium 4 on a Hercules 3D Prophet II MX, a similar Small configuration uses only ~11% of available CPU.
The pitch mapper can force all incoming pitch values to a particular category of key scale (major, minor, etc.). Also, pitch mapping may be enabled for the pitch offset of oscillators 2 and 3.
Selecting the "chromatic" option disables pitch mapping.
Notes passed into the pitch mapper are quantized to the nearest whole pitch value. Then, any difference from the whole pitch value is added back to the output value, after pitch mapping. So, the mapper correctly preserves pitches for slightly detuned voices. The design is similar to Bubu's in the Reaktor user library, but pitch is quantized, rather than using modulo values, for properly detuned voices. Also, the internal pitch table is expanded to cover the entire pitch range, eliminating a modulo division by 12. This especially increases performance for rapid note runs on Celeron machines.
A scope can display any of the oscillators, either of the morph mixer outputs, or either of the filter outputs for the last voice. Also it can display the output for the left or right channel; or it can be turned off to save CPU. The Y slider scales the displayed waveform vertically; horizontal resolution is scaled to match the pitch of the source voice.