The D-Lev is a (mostly) digital Theremin with many unique features.  The primary achievement is the linearity of the pitch and volume fields, which are much more linear than anything you've likely experienced.  Field sensitivity can be configured to best suit your playing style, and the volume field has an adjustable knee.  True pitch correction, full-featured pitch preview, and real-time visual pitch feedback via the LED "tuner" all provide much welcomed assists to performance.

The synth is sufficient to emulate human vocals, strings, brass, woodwinds, bells, percussion, crazy sounds, etc.  The variety of voices encourages the development of different playing techniques, and helps maintain an interest in playing itself.  MIDI out lets you control almost any outboard synth for further sonic explorations.

The D-Lev was designed from the ground up using FPGA technology, and incorporates my Hive soft processor core and phase locked peripherals into a highly responsive, tightly integrated whole.  Boot time and null calibration are almost instantaneous, so you won't lose any ideas or inspiration waiting to play.

Exactly like a traditional analog Theremin, the D-Lev employs oscillators that vary in frequency with the capacitance of the player's hands.  The exact same electrostatic interaction is at work, but in the D-Lev the oscillator frequencies are now represented by digital numbers, and the heterodyning process used to develop a difference frequency in the analog theremin is now replaced by a simple arithmetic subtraction.  Musicians who are accustomed to traditional analog Theremins won't have to alter their fundamental style or techniques in order to play the D-Lev.

How a digital Theremin works.

The variable oscillators in the D-Lev are almost entirely digital, which makes them quite stable.  The operating frequency and volume are therefore just a series of numbers that are fed to an internal DSP (digital signal processing) based music synthesizer in order to produce audio.

Separating the process of sensing hand capacitance (related to hand position) from the generation of sound opens up a new world of exciting possibilities for both the designer and the player.  The pitch and volume numbers are extensively digitally filtered to dramatically reduce environmental noise and hum.  The responses of the fields, which are key to making the D-Lev comfortable and predictable to play, are linearized mathematically and stored in any of four system profiles.  The octave and timbre of preset voices are entirely defined by the synth, with presets to easily switch among a variety of very different voices.  In terms of technology and features, the D-Lev stands out among Theremins, both past and present.

• •Nearly perfect pitch & volume field linearity, tailored by the user — right up to the antennas! 

• •Adjustable field sensitivities and offsets to accommodate any playing style. 

• •Volume field has two fully configurable / defeatable knees (to reduce "hand flapping"). 

• •Wide gestural bandwidth precisely tracks your every hand movement. 

• •High antenna voltage swings provide large fields immune to most forms of interference. 

• •Simultaneous auto-calibration of both pitch & volume fields via foot pedal or button. 

• •Zero lag LED visual tuner for real-time pitch & volume feedback during play. 

• •Full-featured chromatic pitch correction from none, to unobtrusive, to full hard quantization. 

• •Highly configurable pitch preview with extensive options. 

• •Inharmonic resonator produces drums, bells, gongs, human vocals, room ambiance, and pseudo stereo. 

• •Envelope generator with velocity sense for plucked and percussive effects. 

• •+/-3 octave register bank switch for instant access to a wide pitch range. 

• •MIDI out with CC, pitch bend range, trigger location, velocity, and octave controls. 

• •Sculpture mode for automated stand-alone operation over an extended period. 

• •250 voice presets and 6 system profiles. 

• •ALL parameters fully adjustable via the front panel interface. 

• •Easy to own, no internal adjustments (except for LCD contrast). 

Technology

• •Low-drift, high Q / high voltage, single layer, hand wound air core coils. 

• •Rock steady, high resolution, dithered phase locked digital oscillators. 

• •22nd order pitch & volume field digital filtering kills mains hum dead (50/60Hz). 

• •FPGA-based logic with custom 8 thread Hive soft processor core. 

• •Internal digital signal processor (DSP) 32 bits @ 48kHz; digital audio outs 24 bits @ 48kHz. 

• •LED visual tuner employs PWM to intuitively display high-res real-time pitch and volume. 

• •Subtractive / modal synthesis engine with pitch and volume hand modulations. 

• •User interface: 8 rotary encoders (knobs); 20x4 line backlit LCD. 

• •Outputs: Stereo monitor out (S/PDIF, TOSLINK); Stereo line out (S/PDIF); MIDI TX (DIN); Mute LED. 

• •Inputs: RS-232 over USB; ACAL & Mute momentary switches. 

• •Preset librarian software (Windows, Mac, Linux). 

The D-Lev is controlled by two columns of four knobs (rotary encoders) each.  These directly correspond to the two columns of text on the four line LCD display:

Each knob controls one associated value on the LCD display.

For example, on the main D-LEV screen above: turning knob 3 adjusts the Pcal parameter, and pressing knob 6 zeros out load.

How the knobs and LCD work:

• •Knob 7 selects one of 20 screens, or pages.  Knobs 0 through 6 adjust the associated display values on the current page. 

• •Turning a knob quickly changes the displayed value in larger steps, so you can make big edits with less work. 

• •Pressing a knob cycles through => 0 => maximum => minimum (with a few exceptions).  Use this to defeat or maximize a function, or to examine the range of the knob.  For example, press knob 7 to move quickly to the first page, and between the first and last pages. 

The 20 D-Lev screens / pages.

• •Pressing the Out knob on the D-LEV page page initiates an ACAL (auto-calibration of the fields) which blinks the center tuner LED during the period set by the Wait knob on the SYSTEM page and automatically mutes the audio.  To undo the auto-mute, move your hand to turn off all of the volume LEDs. 

• •Pressing the stor knob on the D-LEV page or Stor on the SYSTEM page initiates a two step preset / profile write.  After the first press the screen label will change to ?WR? and the center tuner LED will start blinking. Press it again to actually perform the write and return things back to normal.  If you want to abort the write, turn any knob, or press any knob other than stor / Stor. 

To tell them apart on the screen, system profile controls begin with a Capital Letter (Title Case) and voice preset controls are all lower case.

Most knob values have binary limits (such as 0 to 31) as opposed to decimal (such as 0 to 9) which may seem unusual at first.  These give finer grained ranges and are a better fit to the underlying hardware and software.  

In this manual these knob ranges are represented by bracketed values.  For example, the Vcal range of -127 to 127 is shown as Vcal[-127:127], and setting load to 45 is shown as load[45].

D-Lev main modules.

The D-Lev architecture consists of many interconnected functional modules as shown above.  

The player’s hands interact with the pitch and volume fields which are generated by the digital oscillators and resonant circuits.  After extensive filtering and mathematical manipulation, this pitch and volume information is fed to a digital audio synthesizer to generate sounds.  The LED visual tuner displays highly responsive pitch and volume information to visually assist setup and playing.  Pitch preview provides a means for the player to personally monitor pitch even when in full volume cutoff.  And the D-Lev sound palate is easily expanded by connecting it to an external MIDI synthesizer.

This section discusses the D-Lev modules in some detail, and is followed by a focus on the DSP music synthesizer and the sub-modules within it.

Pitch & volume field generation and processing.

Pitch corrector block diagram.

The pitch corrector is chromatic (all 12 notes in the octave are corrected identically) and employs multiple mechanisms to mask audible functioning even with fairly aggressive settings.  The correction rate can be sped up to produce hard pitch stepping quantization, slowed down to give unobtrusive pitch correction, or anything in-between.  A key innovative feature is modulation of the correction rate based on pitch hand position within the note, which accentuates note centers and blends note transitions.  A bi-modal filter enables smooth glissando.  Since it corrects to absolute pitch, it is most effective when used actively in conjunction with the tuner during play.

All controls are on the PITCH page.

PITCH page controls.

corr sets the pitch correction strength.  To enable pitch correction, both the span and corr knobs must be non-zero.  If you experience phasing effects with external audio sources due to overly perfect pitch correction, reduce corr to intentionally introduce some minor pitch tracking error.  corr[0] will disable pitch correction, and this provides a quick and simple means of disabling correction.

rate sets the correction rate.  Larger values will cause faster correction, and for maximum correction this should be set as high as possible for the piece you are playing.  Use rate[31] for hard quantization.  The basic corrective action is linear (to the ear) slewing of pitch to the note center, which can be dynamically altered by the other knobs.

cmod modulates the correction rate with the location of the pitch hand.  Higher settings speed up the correction rate when the pitch hand is near the center of a note, and slow down the correction rate near the note edges.  This enables slower note transitions with higher rate settings, imparts a “magnetic” feel to the note centers, and softens the otherwise constant correction rate pulling the pitch this way and that at the note boundaries.  Note that the rate knob sets the nominal correction rate here.

vmod modulates the correction rate with the location of the volume hand.  Positive values of vmod increase the correction rate with increasing volume, with the upper limit set by rate; negative values increase the correction rate with decreasing volume, with the lower limit set by rate.  When the volume is so low that the audience can’t hear anything, it might make sense to speed up correction so that the next note to be played is quickly centered.

span sets the distance from note center over which pitch correction is applied.  Typically span is maxed out to cover the entire note, but lower values will limit correction to a narrower span centered on each note.  span[0] will disable pitch correction.

Influence of the span control.

Post selects the LED tuner pitch and volume inputs (please see the table in the tuner section for what the settings do).  Post-correction is useful for visualizing and adjusting the often quite subtle sounding pitch correction controls.  Otherwise use pre-correction so that the pitch hand position is displayed on the tuner – this helps you better hit the note centers when playing, which enables the pitch corrector to operate optimally.

Envelope generator block diagram.

The envelope generator is used to alter the static and dynamic response of volume hand movements.  It consists of a knee unit and a velocity detector unit in parallel, followed by a peak hold / envelope generator.  The knee unit is used to introduce a secondary linear response to the volume hand location – which sounds bad but is actually highly desirable when done in a controlled manner.  The velocity detector enables more nuanced staccato / percussive play.  The peak hold and envelope generator together create rapid – or very slow and smooth – volume changes that would be difficult or impossible to accomplish with hand movements alone.

Most controls are on the VOLUME page.

VOLUME page controls.

knee causes the volume, when it is below a certain volume threshold – set by kloc – to react more strongly to the location of the volume hand.  This is known as expansion, and the volume response above kloc is unaffected.  Most natural sounds quickly ramp up to a certain loudness, with a more expressive region above that.  Moderate settings of knee facilitate this with smaller, slower, and more natural hand movements, while retaining good control over dynamics.  Very high settings of knee cause the volume to instantly “pop” to the kloc loudness, which can be used percussively.  knee[0] defeats the knee.

kloc sets the knee location in the volume field.  The range is -48dB to 0dB (kloc[0] to kloc[63], respectively).  An upper expressive region of 12dB or so (kloc[48]) works well for most voices.

velo controls the volume hand velocity gain, where the speed of the volume hand is sensed and added to the overall volume.  Velocity sensing is bi-directional and continuous over the full range of hand movement (sensing is not restricted to a threshold point).  Adding velocity to a high gain knee provides nuanced control to percussive presets.

rise and fall set the volume rise and fall time, with higher settings giving slower volume ramping rates.  Rise will continue until the output matches the peak hold value, after which it falls to the current hand location volume.

damp is a global modulator of fall and all filter reso settings.  If the volume hand is before the location given by dloc then damp is active, reducing fall and all reso values.  Step modulating the decay and resonance rates this way provides increased control over ringing and percussive presets.  Setting damp to a non-zero value causes the MIDI note-off location to be the position set by the dloc knob.

dloc sets the damp switching location in the volume field.  The range is -48dB to 0dB (dloc[0] to dloc[63], respectively) before the velocity and knee are applied.  Also sets the MIDI note-off location if damp is non-zero.

cvol (located on the PITCH page) modulates the volume down between notes.  It automatically reduces the volume at note transitions.

MIDI module block diagram.

The MIDI module takes the pitch and volume hand information and sends serial MIDI OUT data to the MIDI IN of an external music synthesizer.  This is strictly a MIDI transmit module, there are no MIDI IN or THRU connections nor processing.

Most controls are on the MIDI page.

MIDI page controls.

chan selects the MIDI channel.  MIDI channels are sort of sub-synthesizers with their own independent controls.  Valid channels are 1 through 16.  chan[0] disables the MIDI module, which can be handy when changing cc.  Negative values of chan insert 2ms gaps between MIDI data “clumps” which might help with MIDI “flooding” situations.

velo controls the note-on velocity gain.  The speed of the volume hand as it crosses the vloc threshold (to play a new note) is converted into MIDI note-on velocity.  Higher velo settings make it easier to play higher velocities.  Some special modes here are velo[0], which transmits a fixed velocity of 64 (the MIDI standard for non-velocity instruments) and velo[31], which transmits a fixed velocity of 127 (this may work better for some synths / patches).

vloc sets the note-on velocity sense point and note on / off trigger location in the volume field.  The range is -48dB to 0dB (vloc[0] to vloc[63], respectively).  Setting vloc below that of cloc can help hide note on/off trigger events, which can make the voice being played sound less percussive and more Theremin-y.  Setting damp on the VOLUME page to a non-zero value causes the MIDI note-off location to be the position set by the dloc knob on the VOLUME page, which can be useful for triggering drones.

bend sets the MIDI pitch bend span in terms of half steps.  bend[12] will produce a +/- 1 octave pitch bend range, and so on.  Pitch bend is re-centered whenever a note is played (when the vloc threshold is exceeded), and when playing exceeds the pitch bend range the note will be re-triggered with pitch bend re-centered on the new note.  In order to track correctly, the value of bend needs to match the pitch bend span setting in the external synthesizer.  bend[0] defeats pitch bend and note re-triggering, so notes just stick at the center of where they were played at until done.

oct sets the relative octave of the signal; this is an offset to the bank octave setting.

cc selects the MIDI control change manipulated by the volume hand.  Knob values [1:31] transmit 14 bit data to cc[1:31].  Knob values [-127:-1] transmit 7 bit data to cc[127:1].  Setting cc[0] disables control change transmission.  Here are some common MIDI CC’s:

Some MIDI cc values.

cloc sets the cc start location in the volume field.  The range is -48dB to 0dB (cloc[0] to cloc[63], respectively).  The full range of control is scaled to fit into the selected volume range.  For example: cloc[32] will set the cc start point to -24dB; when the volume is below this the CC data is zero; above this the CC data goes smoothly from 0 to maximum.

Out (located on the D-LEV page), when switched from 1 to 0, sends current MIDI note-off (note-on with velocity=0), as well as MIDI Channel Mode Messages 123 (All Notes Off) and 120 (All Sound Off) for the current MIDI channel (set by chan).

LED tuner display.

The LED visual tuner provides zero lag real-time pitch and volume feedback during play.  The pitch circle indicates the note within the octave being played, the octave is shown on a 7-segment display, and the volume level is indicated with a 4-LED thermometer type bargraph with 12dB graduations.  Pulse-width modulation (PWM) dimming is employed to dramatically increase the resolution by using peak brightness to indicate volume shading and exact note centers.

All controls are on the DISPLAY page.

DISPLAY page controls.

LED sets the brightness of all of the tuner LEDs.  If you find the tuner distracting, LED[0] turns all the LEDs off.

LCD controls the LCD backlight brightness.  LCD[0] turns off the backlight.

Qant controls the tuner pitch brightness quantization.  Higher quantization settings cause the brightness to be more centered on one LED, rather than smeared out among two or three LEDs, which makes it easier to see at a glance the note being played in the presence of pitch inaccuracies and/or vibrato.  Try Qant[1].

Post selects the tuner pitch and volume inputs:

Post knob functionality.

        Post pitch is useful for visualizing and adjusting the often quite subtle sounding pitch correction controls.  Otherwise use pre pitch so that the pitch hand position is displayed on the tuner – this helps you hit the note centers when playing, which helps the pitch corrector operate optimally.  Pre volume is with the knee and velocity applied, but not the envelope generator.  Post volume includes the envelope generator response.

Cent offsets the global frequency in cents (1/100 note).  This is a system-wide offset for the entire D-Lev including all oscillators and filters and MIDI, not just the tuner.

Note offsets the tuner pitch display and doesn’t affect the pitch.  It’s handy if you prefer to view one key pattern while actually playing in another key.  For instance, Note[-4] displays the key of C pattern on the tuner while the key of E is being played.

Oct also simply offsets the tuner octave display number and doesn’t affect the pitch.

Tips & Tricks The pitch circle is composed of staggered inner and outer rings.  Within a ring the notes are spaced a whole step apart, and the rings are offset from each other by a half step.  So sequential notes within a ring are whole steps, and switching rings produces a half step. The pitch circle LEDs form two distinct patterns when playing in major or natural minor keys, which are dubbed the “Crown” and “House” patterns: C / Am “Crown” pattern.                         C# / A#m “House” pattern. It’s easy to infer the key of the music you are playing by recognizing these two patterns and their rotation.  Once recognized, the patterns can be a powerful aid to improvisation. The location of C on the pitch circle was chosen mainly to give a vertically symmetric pattern on the pitch circle with C lying on the outer ring.  Conveniently, this also places C near the octave display, the value of which morphs over the B/C interval.  Clockwise rotation is used to indicate increasing pitch, which follows the convention of the controls. The point at which all volume bargraph LEDs just turn off is -48dB.  Similarly, the bottom LED fully on with the others fully off is -36dB; two bottom LEDs fully on is -24dB; three LEDs on is -12dB, and all four LEDs on is 0dB, or the loudest possible audio. -48dB (1/256) is right at the edge of audibility, so when all the volume bargraph LEDs are off the volume is almost certainly low enough that the audience won’t hear anything.  Actual volume control continues on to -96dB to ensure inaudibility, but this isn’t displayed. The octave is displayed in hexadecimal format, where the values 10 thru 15 are represented by the letters a through f – don’t confuse these letters with note names! Minimum and maximum pitch are represented by A8, so when you play the highest note possible, or when you step away from the D-Lev, it is normal to see this value “pegged” on the tuner. The D-Lev tuner is sort of the inverse of a guitar tuner (where the note is displayed as a character and the sharp/flat information is displayed graphically).  Guitar tuners are often quite sluggish, and their displays aren’t really designed for real-time intuitive feedback during play.

Pitch Preview block diagram.

Pitch preview is a separate oscillator that provides a way for the player to more effectively monitor the pitch during play, particularly during passages of low volume and silence.  This is a feature sometimes found on professional Theremins, but this implementation offers much more control to better tailor it to your needs.

All controls are on the PREVIEW page.

PREVIEW page controls.

prev sets the volume of the preview signal.

vmod modulates pitch preview volume with the volume hand.  Positive values of vmod increase pitch preview volume with increasing volume; vmod[32] makes the pitch preview volume perfectly track the volume of the preset voice.  Negative values increase pitch preview volume with decreasing volume and produce a “ducking” effect where the pitch preview signal fades out as the preset voice is fading in, and vice-versa.  Some find this useful, others may find it confusing.  vmod[0] is a safe starting point as it replicates the behavior of more basic pitch preview implementations on other Theremins (i.e. a constant tone).

pmod modulates pitch preview volume with the pitch hand.  Positive values of pmod increase pitch preview volume with increasing pitch.  Negative values increase pitch preview volume with decreasing pitch, which can help when higher notes seem much louder than bass notes.  Depending on the timbre of the preset, pmod[-15:-25] may give the perception of more balanced pitch preview loudness over the full pitch range.  The modulation pivot point is A440.

harm sets the fixed harmonic level; negative values here give odd rather than all harmonics and have a hollow sound, which may make pitch identification a little more difficult.  Use higher harmonic levels (positive or negative) to better hear the pitch, and consider using odd harmonics to make the pitch preview timbre more distinct from the preset voice.

oct sets the relative octave of the signal; this is an offset based on the oscillator octave setting.  oct[1] will shift the octave up one and may enhance pitch perception of low notes by making the pitch preview signal more distinct.  The bank knob offsets it.

tone boosts / cuts the lows and highs together.  Rather like the “loudness” knob on some stereos, increasing tone may enhance pitch perception and allow lower settings of prev.  Be aware that adding bass and treble may push a preset into digital distortion at high volumes; it maybe be necessary to decrease prev to compensate and stay safely below any distortion level. In mode[12] the tone knob controls treble boost / cut only.

mode controls the mixing of the pitch preview signal, as well as the pitch correction and quantization of it.  The pitch preview signal is injected into the L&R (left & right) monitor output only (not the L&R line output).  The 4th oscillator modes instead inject the preview oscillator into the synthesizer oscillator section for special effects and the like.  Any applied pitch correction is the same as that applied to the preset voice; any hard pitch quantization (note stepping) is entirely separate from pitch correction.  Special mode[12] injects a DC signal for stimulus for bells and such.

mode knob settings.

• Applying either no pitch correction (in the case where the preset voice uses pitch correction) or hard quantization to the pitch preview signal will produce phasing between the pitch preview and preset voices, providing further pitch cues.

DSP synthesizer modules.

The synthesizer section uses digital signal processing (DSP) to produce sounds via a combination of FM, subtractive, and modal synthesis.  A group of digital oscillators and a digital noise source feed a variety of digital filters.

The oscillators and noise source each have their own bass and treble controls, as well as dedicated multi-mode filters.  Following this are global bass and treble controls, a formant filter bank, and the inharmonic resonator.  

The formant bank is used to mimic human vocals, but it’s equally useful at providing the resonances found in wind and stringed instruments, as well as bells and other percussion.  

The inharmonic resonator provides resonance fill, pseudo stereo, and ambiance effects, but it can also do passable drums and gongs, and can even human vocals all by itself.

The following sections discuss these synth modules in more detail.

Triple oscillator block diagram.

The triple oscillator consists of a bank of three oscillators, an EQ section, and a multi-mode filter.  The oscillators generate reduced-alias harmonic content via phase modulation, with a blend of all and odd harmonics on tap.  The harmonic level can be modulated via the pitch and volume hands, and the oscillator phase modulation inputs and outputs can be interconnected for more complex timbres.  Oscillator 0 is the “center” and always on the pitch being played, 1 & 2 can be offset from it in multiple ways to give a thicker sound, harmonies, etc.

Most controls are on the 0_OSC, 1_OSC, and FLT_OSC pages.

0_OSC page controls (common to all oscillators).

harm sets the static harmonic level.  vmod and pmod can be used to modulate this level up or down.  Unmodulated harm[0] gives sine waves.  Negative harm makes oscillators 1 & 2 outputs sine waves.

vmod modulates the harmonic level with the volume hand.  Positive values of vmod increase harmonics with increasing volume; negative values increase harmonics with decreasing volume.  Harmonics for most natural sounds increase with increasing volume (i.e. positive vmod).  Setting harm[31] and vmod[31] will give sine to full harmonics over the -24dB to 0dB volume range; vmod[42] reduces this range to -12dB to 0dB.  harm[18] and vmod[16] are typical for human vocals.

pmod modulates the harmonic level with the pitch hand.  Positive values of pmod increase harmonics with increasing pitch; negative values increase harmonics with decreasing pitch.  The modulation pivot point is A440.

odd crossfades between all and odd harmonic content.  odd[0] gives all harmonics, odd[31] gives only odd harmonics.

oct sets the octave of the oscillators; this also affects pitch preview, but does not affect any filter cutoff frequencies, MIDI, nor the tuner.  It is offset by the bank knob.

bass and treb EQ the lows and highs.

1_OSC page controls (mostly oscillators 1 & 2).

offs are used to offset the pitches of oscillators 1 & 2 (+/- 1 octave).  These are multiplicative offsets, so pitch ratios are maintained.  They are cubic scaled to increase the resolution around the unison (offs[0]) and octave (offs[-127, 127]) regions:

Cubic plateau scaling of offs and hmul.

Some useful offs and hmul values for various musical intervals.

hmul are multiplicative pitch offsets of the phase modulation units within oscillators 1 & 2 (+/- 1 octave) and are relative to their associated offs values (offs offsets hmul).  They are scaled the same as offs, and are useful for introducing pitch-dependent beating and ring-modulation type harmonics.

sprd introduces an additive pitch offset to oscillator 1 and a subtractive pitch offset to oscillator 2; the pitch of oscillator 0 is unaffected.  This is useful for unison and beating type effects when you want the beating rate to be independent of pitch.

fm cross-fades the phase modulation input with the output of the phase modulation (PM) unit within each oscillator:

Single oscillator block diagram.

The phase modulation outputs of oscillator 1 & 2 are combined and sent to the phase modulation input of oscillator 0, and the phase modulation output of oscillator 0 is sent to the phase modulation inputs of oscillators 1 & 2.  With fm[0] all oscillators only use their internal phase modulation units to create harmonics; fm[31] makes them fully cross-coupled.

Triple oscillator phase modulation cross-coupling.

fm can radically transform the timbre and introduce some fairly wild harmonics, particularly when offs and/or hmul are non-zero.  Use it to produce scratchy records, trumpeting elephants, UFO invasions, and other sonic mayhem.

xmix cross-fades the audio output of oscillators 1 & 2 with that of oscillator 0.  xmix[0] gives only oscillator 0; xmix[31] only 1 & 2.  Similarly, it cross-fades phase modulation outputs of the oscillators, the mix of which is used to pulse width modulate the noise generator.  It is linearly scaled, so xmix[21] produces an equal mix of all three oscillators / noise modulation.  Negative settings of xmix cross-fade between oscillator 0, and the multiplication of it with the sum of oscillators 1 & 2, which produces a ring modulation effect.  In this mode, oscillators 1 & 2 are set to A440, which can be offset via the offs, sprd, and oct knobs, but not the bank knob on the main page.

FLT_OSC page controls (multi-mode filter).

osc sets the oscillator audio output level.  Reduce this if the filter is overloading.

freq sets the center or corner frequency of the multi-mode filter.  The frequency is displayed in Hz, and the increments are 50 cents.  It is offset by the bank knob.

vmod modulates the filter frequency with the volume hand.  Positive values of vmod increase the frequency with increasing volume, and negative values increase filter frequency with decreasing volume.

pmod modulates the multi-mode filter frequency with the pitch hand.  Positive values of pmod increase the filter frequency with increasing pitch, and negative values increase filter frequency with decreasing pitch.  The modulation pivot point is where the pitch hand normally plays an A440, so setting freq[440] and pmod[32] will make the filter frequency track the oscillator frequency (this is only true if oct[0];  if oct[1] then use freq[880], etc.).

reso sets the resonance of the multi-mode filter.  This is analogous to Q (or the inverse of damping) but is scaled to give useful range of resonance behavior, from highly damped to infinite sustain.  Higher reso increases peak gain, so you my have to reduce osc to keep the filter from overloading.

mode selects the filtering mode and order of the multi-mode filter:

Filter mode functionality.

Higher orders are more effective at filtering in general and tend to sound “synthy.”  When constructing a more natural sounding preset a second order filter is usually what you’re looking for.

The hollow “phasey” sounding notch filter modes benefit from lower reso settings, otherwise they can be hard to hear.

xmix cross-fades the filter input and output.  xmix[0] completely bypasses the filter (as does mode[0]); xmix[31] gives full filter output.  Mixing some input into the output can be useful when you need milder emphasis, but you usually want xmix[31] when using the filter.  Negative xmix settings invert the mixed signal, which might be useful for phasing effects.

Noise generator block diagram.

The noise generator consists of a Voss-McCartney pink noise source which can be amplitude modulated (turned on and off) by the phase modulation output of the triple oscillator.  This is followed by an EQ section and a multi-mode filter.  

Most controls are on the NOISE and FLT_NOISE pages.

NOISE page controls.

nois sets the noise audio level.  nois[0] turns off the noise section and also routes the oscillator through the noise EQ and multi-mode filter.  Reduce this if the noise multi-mode filter or bass and treble EQ are overloading.

vmod modulates the noise level above the knee point set by kloc – the noise level below kloc is unaffected.  Positive values of vmod make the noise level increase more than the oscillator level; negative values make the noise level increase less, or even reverse.  vmod[‑32] gives a constant noise level above the knee, and more negative values of vmod cause the noise level to dip back down.

pmod modulates the noise level with the pitch hand.  Positive values of pmod increase the noise level with increasing pitch; negative values increase the noise level with decreasing pitch.  The modulation pivot point is where the pitch hand normally plays an A440 (with oscillator oct[0]).

puls and duty modulate the noise level with the triple oscillator phase modulation output mix.  puls controls the depth of modulation: puls[0] turns modulation entirely off giving constant noise; below puls[8] or so the noise doesn’t turn fully off; above this the noise is fully modulated and higher puls values will tend to decrease the on time.  Whereas increasing duty will increase the on time.  These controls interact quite a bit, so you may have to play with them to find a setting you like.  The oscillator harm level will influence modulation depth too.

bass and treb EQ the lows and highs.

FLT_NOISE page controls (multi-mode filter).

nois is a copy of the same control on the NOISE page (for your convenience).

freq, vmod, pmod, reso, mode, xmix – same as the triple oscillator multi-mode filter (see that section for descriptions).

Formant bank block diagram.

The formant bank is a parallel combination of eight second order band-pass filters.  Each filter has its own center frequency and level control, but the filters are paired up in order to share common center frequency modulation and resonance controls.

All controls are on the 0_FORM, 1_FORM, 2_FORM, and 3_FORM pages.

0_FORM page controls (other pages are the same).

freq sets the center frequency of the filter.  The frequency is displayed in Hz, and the increments are 50 cents.  Each freq is associated with the levl to the right.  Offset by the bank knob.

levl sets the audio input level.  levl[0] turns off the filter.  Reduce this if the formant filter is overloading.  Setting all eight levl[0] will completely bypass the formant filter bank.  Negative values will invert the phase of the filter, which can be used to eliminate the anti-resonance notch between adjacent formants or to produce a steeper filter skirt.

vmod modulates the filter center frequency with the volume hand.  Positive values of vmod increase the frequency with increasing volume, and negative values increase filter frequency with decreasing volume.

pmod modulates the filter center frequency with the pitch hand.  Positive values of pmod increase the filter frequency with increasing pitch, and negative values increase filter frequency with decreasing pitch.  The modulation pivot point is where the pitch hand normally plays an A440, so setting freq[440] and pmod[32] will make the filter frequency track the oscillator frequency.

reso sets the resonance of the filter.  This is analogous to Q (or the inverse of damping) but is scaled to give useful range of resonance behavior, from highly damped to infinite sustain.  Higher reso increases peak gain, so you my have to reduce levl to keep the filter from overloading.

Inharmonic resonator block diagram.

The inharmonic resonator is a delay-based filter which consists of an absorptive all-pass reverberator in series with a simple delay.  The output is recirculated to form a nested comb filter, which increases the Q / decay time and accentuates the resonant peaks and anti-resonant valleys.  The all-pass reverberator is used to break up the strict harmonic resonance locations, which enables the resonator to produce realistic string body resonances, drums, bells, gongs, and even human vocals.  With longer delay times it can provide ersatz small room ambiance, and a pseudo stereo effect which can considerably pep up a preset.

Most controls are on the RESON page.

RESON page controls.

freq sets the inverse of the total delay time, where the lowest resonance roughly corresponds to the displayed frequency (Hz) if reso is positive.  If reso is negative the frequency of the lowest resonance drops by an octave.  Not offset by the bank knob.

tap sets the ratio of absorptive all-pass reverberator delay time to simple delay time.  tap[0] divides the total delay time equally between them; positive values make the all-pass delay longer than the simple delay; negative values make it shorter.  Non-zero values can help jumble the resonance peaks.

harm sets the all-pass internal feedback, and is used to jumble the otherwise evenly spaced resonances.  harm[0] disables all-pass feedback and converts the all-pass section into a simple delay.  Some settings of harm drive the lowest resonance even lower, which can dramatically boost the bass.

reso sets the overall feedback.  High positive and negative values will cause the resonator to “ring” like a high Q filter.  Negative values invert the feedback and provide a more hollow sound.  reso also controls a mild low-pass filter that tames “zinging” with low to moderate feedback levels.

hpf sets the cutoff frequency of the 4th order high pass filter at the input of the resonator.  Quite often you don’t want the resonator processing lower frequencies, such as when blending it with the formant bank or generating pseudo stereo.  Not offset by the bank knob.

xmix cross-fades the resonator input and output.  xmix[0] completely bypasses the resonator; xmix[-31,31] give full resonator output.  The input / output mix directly sets the height of the resonance peaks, and is also critical to getting a good pseudo stereo effect.  Negative values of xmix invert the output, which may help when blending with the formant bank.

mode selects a number of useful input/output mixing modes:

Resonator mode functionality.

Series modes place the resonator in series with the formant bank, which is useful for ambiance.  Parallel modes place the resonator in parallel with the formant bank, which combines all the resonances together and is generally used for formant fill.  

mode[-2,2] are pseudo stereo modes where the resonator output is added and subtracted from the input to produce L & R; the rest of the modes are mono (where the output is combined identically with L & R).  

mode[0] bypasses the resonator, but does not bypass the high pass filter at the input, so you can still use the hpf and xmix knobs if desired.

Many parameters in the DSP synthesizer can be modulated or controlled by the locations of the hands in the pitch and volume fields.  Understanding the sources of these inputs can help to remove some of the mystery behind their functionality and use.

Volume field processing – modulation sources / taps.

The processing of the volume field has four different tap points:

1. 1.The first tap is used to set the point in the volume field for damping the envelope generator and damping all filters.  It employs the most linear form of the volume hand response so the damping point can be located far away from more dynamic elements such as velocity and the knee, thus avoiding accidental damping during play. 

2. 2.The second tap is after the velocity has been applied, and this sets the location of the knee in the volume field.   

3. 3.The third tap point is after the knee has been introduced.  This is used to locate the MIDI thresholds for note on/off and control change in the volume field, as well as for pre-monitoring of the volume level on the LED tuner.   

4. 4.The final tap point is after the envelope generator and pitch center modulation.  This is used to modulate the pitch correction rate, as the vmod modulation source for all filters and such, and for post-monitoring of the volume level on the LED tuner. 

Pitch field processing – modulation sources / taps.

Pitch field processing has only two tap points:

1. 1.The first tap is used for pre-monitoring of the pitch on the LED tuner.   

2. 2.The second tap point is after pitch correction has been applied.  It is employed as the pmod modulation source for all filters and such, as well as for post-monitoring of the pitch on the LED tuner. 

Synth voice presets (left) – System profiles (right; note upper case Stor & Load).

In order get the most out of any Theremin, the fields must be properly aligned.  The D-Lev is uniquely capable of being calibrated to exhibit ideal playing characteristics for virtually any playing style, but this capability necessitates a fair number of parameters to adjust and knobs to twiddle.  But the good news is that once set, these parameters can be saved to a system profile, and then rarely if ever need to be touched again.  

The main goal is to linearize the fields, which is accomplished by adjusting the field controls until you get the same pitch / volume change with an open & closed hand gesture everywhere in the playing zone.  Aligning the fields doesn’t require any special equipment, just the built-in LED tuner and your hand.  And it isn’t particularly difficult, though it can be a bit confusing the first time through.

There are situations where it’s desirable for the D-Lev to operate unattended for long periods of time.  The D-Lev may be part of an interactive sculpture, or it could be sitting in a merchant’s showroom or on a trade show floor, or the musician may be investigating electrical noise.

Regardless of the Theremin technology, one will encounter the general problem of oscillator drift due to environmental factors that are beyond one’s control.  Over time the calibration will go out of whack, and take the response of the unit along with it.

To accommodate these scenarios, the D-Lev software has a so called "sculpture mode."  The mode is engaged and controlled via a single Auto[0:249] knob on the SYSTEM page.  Auto[0] disables the mode; Auto[x] cycles through x voice presets [0:x-1] with a pitch hysteresis range of [A0:A2].  Auto[1] is slightly special in that it only uses preset 0 (as expected) but never reloads it.

To use this mode, set Auto to however many voices you want to cycle through, where the sequence starts at voice preset 0 and proceeds in a positive direction through the presets.  Perform an ACAL and then walk away from the D-Lev.  The field null points will automatically float to a pitch of A0 and a volume of -48dB.  Moving closer will cause the null adaption rate to dramatically lower, so interaction with the unit won’t cause too noticeable of a droop in pitch and volume over a short period of time.  If one gets close enough for the pitch to reach A2 then preset cycling is armed.  If one then moves away until the pitch drops below A0 the voice preset will increment.

Use the provided librarian software and the following procedures to manage the software and presets / profiles in your D-Lev.  With the librarian you can:

• •Save and restore snapshots of the entire system state. 

• •Reset the D-Lev to the original “factory fresh” state. 

• •Update or roll back the software. 

• •Save / restore / arrange presets and profiles, in bulk or individually. 

CAUTION: These procedures can overwrite any and all editing and customization that you have performed!  It is strongly suggested that you take a system snapshot (see below) before making any changes to your system.