Ambika - User manual

Getting started


You have successfully built Ambika, Congratulations!

Ambika is a multi-voice hybrid synthesizer. You can play it as a 6-voice polysynth, an ensemble of 6 monosynths, or anything inbetween due to its easily configurable voicing architecture.

The sound generation is hybrid, combining the warmth and sonic character of a true 4-pole analog filter, with the large array of waveforms offered by digital wavetables, fm and phase modulation. The digital control of the analogue filter and VCA also means a very large palette of modulation possibilities.

Some of the key features of Ambika include:


The following connectors are available on the rear panel:

Some terminology…

A voice is a physical monophonic sound production device, consisting of digital oscillators, CV sources, an analog VCF and a VCA. A voice is only capable of producing a single-note sound. Ambika contains 6 voices, each of them being a physically different circuit board.

A part is one or many voices sharing the same synthesis settings. Ambika can manage up to 6 parts. Each part stores its own synthesis, arpeggiator and sequencer settings. Each part listens to a MIDI channel, and is assigned a range of keys on the keyboard.

Each of the 6 voices in Ambika needs to be linked to (assigned to) a part. This is a bit like showing each musician (voice) in an orchestra which staff they must play on a musical score! If you assign the 6 voices to the same part, Ambika will behave like a classic monotimbral polysynth. If you assign each voice to a different part, Ambika will behave like 6 independent monophonic synths. If you want to play a bassline on the lower part of the keyboard, and a brass riff on the upper part of the keyboard, you need to use two parts: one part with 1 voice for the bass, and a second part with 5 voices for the brass sound.

A patch is a specific combination of synthesis settings stored into a part.

A program consists of a patch, and additional sequencer/arpeggiator settings.

A multi stores 6 programs (one for each part of Ambika) along with the mappings between voices, parts, midi channels and keyboard range. This is a complete snapshot of the Ambika configuration!

Controls and navigation


Editing parameters

The Ambika parameters are organized in pages. To jump to a page, press one of the 7 switches S1S7. Some pages share the same switch ; and you will need to repeatedly press a switch to cycle between those pages. The active page is indicated by the LEDs L1L7 next to the navigation switches.

The following table lists which page is associated with each switch:

Switch Pages
S1 Oscillators, Mixer
S2 Filter
S3 Envelopes and LFOs, voice LFO
S4 Modulation matrix
S5 Keyboard & tuning, sequencer & arpeggiator, sequence editor
S6 Voice and parts mappings, Tempo/clock
S7 Performance, knob assignments

Each page displays up to eight related synthesis parameters. Each parameter can be edited by turning the knob sitting at its top (for the first row of the display) or at its bottom (for the second row of the display).

Here is an example:

After having powered up the unit, press S1 to bring the oscillators page. L1 lights up in green, and the LCD display shows the following parameters:

 Wave  saw|para   0|rang    0|tune   0
 wave squa|para  32|rang  -12|tune  12

You can use the first row of knobs to edit the shape, parameter, range and detune of the first oscillator ; and the second row of knobs to edit the shape, parameter, range and detune of the second oscillator. Observe that when you are turning a knob, the explicit name of the parameter is temporarily shown on the screen:

 oscillator 1       |waveform        saw
 wave squa|para   32|rang  -12|tune   12

After a short delay, the four names and values are shown again.

Press S1 again to bring the mixer page. L2 lights up in yellow, and the LCD display shows the following parameters:

 Mix    32|nois    0|sub     0|wave squ1
 xmod  off|amnt   31|fuzz    0|crsh    0

Press S1 again to get back to the oscillators page.

Using the encoder

When Ambika displays a parameters page, the rotary encoder can be used to scroll through the parameters. The name of the active parameter is capitalized. For example, oscillator 1 range is here the active parameter:

 wave  saw|para    0|Rang    0|tune    0
 wave squa|para   32|rang  -12|tune   12

Rotate the encoder clockwise to make tune the active parameter, rotate the encoder counter-clockwise to make para the active parameter. If you continue rotating the encoder clockwise, the next page will become active.

Once a parameter is selected (capitalized), click the encoder to edit it. The full name of the parameter is displayed on the screen. The encoder can now be used to increment/decrement the parameter value. Once the value has been set, click the encoder again.

Knob and encoder editing can be combined. Use a knob to rapidly adjust the value of a parameter, and then, while the parameter name is still displayed on the screen, use the encoder to fine-tune the value.

Shortcut. Hold the S8 switch while turning the encoder to increment/decrement values by 8 instead of 1.


This section describes in details each page and parameter of Ambika.

Voice allocation

This page, accessible by the S6 switch, serves two purposes:

 Part    1|chan    1|low    C-|high   G9
 voicecard| [1]  [2]  [3]   4    5    6

The first knob on the upper row is used to select a part. Notice how the LP1LP6 LEDs indicate by a green light which part is active. All settings on all the other pages apply to the part indicated by this green light.

chan (channel) sets the MIDI channel the active part responds to. Use omni if you want a part to respond to notes from all MIDI channels. Several parts can share the same MIDI channel. This is useful for controlling two patches from the same MIDI controller, in split or dual mode.

low and high set the range of MIDI notes the active part responds to. This can be used to create keyboard splits: set the range of part 1 to C- .. B3 and the range of part 2 to C4 .. G9 ; and set both part 1 and part 2 to listen to the same MIDI channel. The result is that part 1 is played on the lower half of the keyboard and part 2 on the upper half.

The lower part of the screen displays which voices are assigned to the active part. For example, in the display capture shown above, voices 1, 2 and 3 are assigned to part 1. Use the second knob on the lower row to assign/deassign voice 1 and 2 to the active part. Use the third and fourth knobs to assign/deassign voices 34 and voices 56.

Note that you can assign to a part only voices which are not currently in use by another part. For example, when Ambika boots, voices 1, 3 and 5 are assigned to part 1 ; and voices 2, 4, 6 are assigned to part 2. You will notice that it is not possible to assign voices 2, 4, 6 to part 1 before having de-assigned them from part 2. It might not be convenient, but a voice can only be used by one part, so this constraint has to be enforced!

Assigning/de-assigning a voice causes quite a bit of data shuffling between the processors running each voicecard, and this causes interruptions/reset notes. Don’t do that during a live performance!

Shortcut Hold the S1 switch while turning the encoder to change the active part. This works on any page!


Finally, let’s get to the real thing! Each voice of Ambika is built according to the diagram drawn below. Obviously, it would be tedious to edit the settings of each individual voice… Instead, you edit parts and all the voices assigned to a part automatically inherit its settings!

Here is a day (or rather a millisecond) in the life of a voice’s signal:

Each of these sound generation and modification modules have parameters which can be controlled by any of the modulation sources listed below. However, some connections are already “hardwired” (or rather “softwired” in the firmware):

Besides this, it is up to you to route modulations to parameters. By default, the following routings are wired:

Source Destination Amount
Env 1 Oscillator 1 parameter 0
Env 1 Oscillator 2 parameter 0
Lfo 1 Oscillator 1 pitch 0
Lfo 1 Oscillator 2 pitch 0
Lfo 2 Oscillator 1 parameter 0
Lfo 2 Oscillator 2 parameter 0
Lfo 3 Mixer balance 0
Lfo 4 Filter cutoff 0
Seq 1 Filter cutoff 0
Seq 2 Mixer balance 0
Envelope 3 VCA gain 100%
Velocity VCA gain 25%
Pitch-bend Oscillator 1+2 pitch 2 semitones
Lfo 4 Oscillator 1+2 fine pitch 2 semitones

Let us now review the different synthesis parameters.

Digital oscillators (S1)

 Wave  saw|para    0|rang    0|tune    0
 wave squa|para   32|rang  -12|tune   12

Each row displays the settings of an oscillator. The parameters are the following:

The following is a list of all the available waveform families, with some applications and a description of what adjusting the parameter setting actually does.

none: silence

This simply switches off the oscillator. Switching the oscillators off is useful if you want to use the sine-wave produced by the filter’s self-oscillation as the sole sound source.

saw: sawtooth

This waveform is perfect for basses and brass sounds. The parameter controls the waveshapping - when its value is increased, an increasingly large section of the waveform is shifted up. This waveform is band-limited. Thus, only a limited amount of aliasing artifacts will be heard when playing high-pitched notes.

square: square wave

The parameter controls the pulse-width. This waveform is perfect for simulating a clarinet, for basses, “hollow” sounds or Depeche Mode-like leads. This waveform is band-limited and only a limited amount of aliasing will be heard when playing high-pitched notes.

You will observe that there is a slight difference in sound when moving the parameter from 0 to 1. To offer the best sound quality, the pulse width = 50% flavor is read straight from a wavetable at full sample rate, while the pulse width > 50% flavor is obtained from two dephased sawtooth waves, evaluated at half the sample rate. For bass sounds, for which aliasing is not going to be a problem, it is recommended to use pwm instead of square to get a beefier sound.

triang: triangle wave

A pure waveform, which serves as a good basis for flute or soundchip-like leads. The parameter controls a kind of waveshapping, clipping the bottom of the waveform. This waveform is band-limited and will still sound fine above C5.

sine: sine wave

A pure and chaste sine wave lost her virginity. At some point she started listening to Nine Inch Nails.

zsaw: phase-distortion sawtooth with filter sweep

This waveform uses phase distortion to recreate a low-pass filtered sawtooth by progressively “pinching” the phase of a sine wave. The parameter controls the brightness of the sound: from a sine wave to a sawtooth, then from a sawtooth to a sawtooth gone through an ugly transistor amp. Good for dirty bass guitar sounds or clavinets.

lpzsaw, pkzsaw, bpzsaw, hpzsaw: resonant filter sweeps on a sawtooth wave.

This waveform family directly simulates the sound of a sawtooth wave processed by a low-pass, peaking, band-pass, or high-pass resonant filter. The parameter controls the cutoff frequency of the filter.

lpzpulse, pkzpulse, bpzpulse, hpzpulse: resonant filter sweeps on a trapezoidal wave.

This waveform family simulates the sound of a trapezoidal pulse wave processed by a low-pass, peaking, band-pass, or high-pass resonant filter. The parameter controls the cutoff frequency of the filter. pkzpulse is particularly good at recreating the dirty, saturated sound, of a sawtooth filtered by the least academic of the 2-pole analog filters.

ztriangle: mystery…

This waveform vaguely evokes two hardsync’ed oscillators – the parameter controlling their frequency ratio. It may or may not have been used in the Casio CZ-101.

pad: swarm of 4 buzzing detuned sawtooth waves

As the name implies, this waveform made of four stacked sawtooth waves is useful for pads (when a copious amount of filtering is applied) or for buzzing trance leads. The parameter controls the amount of detuning between the four waves. Note that no bandlimiting is happening here, so this thing doesn’t sound quite good above C5… but it’s doing a perfect job in the bass range!

fm: minimal 2-operators FM

The parameter controls the modulation strength. This oscillator provides the base material for metallic sounds, bells, metallophones, or the next 386 DX hit.

When the fm oscillator is selected, the range parameter plays a slightly different role than usual: instead of controlling the main pitch of the note, it controls the modulator frequency, and has a drastic impact on the timbre.

8bits: bitwise anarchy

A palette of 8-bits sounding waveforms obtained by applying bitwise operations to a basic sawtooth wave (something now known as “biscuiting”).

pwm: my first Arduino synth

This waveform is a shamelessly naive square wave. The parameter controls the pulse-width. Contrary to square, this waveform stinks aliasing - but for notes below C2 it is not a real problem: it becomes much more aggressive and “in your face” than square.

noise: filtered noise generator

The parameter controls the frequency of a simple 1-pole low-pass/high-pass filter in which is sent white noise. From 0 to 63, high-frequency content is progressively added. From 63 to 127, low-frequency content is progressively removed. Perfect as a raw material for percussions or sound effects.

vowel: low-tech formant synthesis

Changing the parameter will sweep between different vocal-like sounds (14 vowels and 2 consonants).


The remaining waveform families are wavetables – collection of single cycle waveforms, synthetic or sampled from real instruments. The parameter is used to smoothly “scan” the wavetable. For example, if a wavetable has 2 waveforms, parameter = 0 plays the first waveform ; parameter = 127 plays the second ; and parameter = 64 plays a mix of both. Wavetables can contain up to 16 waves.

Wavetable name Description
male Vocal formants, best played in the lowest octaves
female Vocal formants, best played in the medium octaves
choir Depeche mode-like vocal sound
tampura Single cycle waveforms from an indian tampura note
bowed Transwave extracted from a cello sample
cello Transwave extracted from a less dramatic cello sample
vibes Transwave extracted from a vibraphone sample
slap Transwave extracted from a slap bass sample
epiano Various single cycle waveforms extracted from electric piano samples
organ Various combinations of sine waves, as produced by drawbars organs
waves 16 basic waveforms, ready to be morphed into one another
digital Abstract, PPG-wave style wavetable
drone 1 Abstract, PPG-wave style wavetable
drone 2 Abstract, PPG-wave style wavetable
metallic Single cycle waveforms from classic D50 patches
bell A Depeche mode classic
wavquence Individually addresses each single cycle waveform

The last wavetable contains all the single cycle waveforms in memory. It is not recommended to scan the parameter with LFOs or envelopes ; but rather to control this from the step sequencer or a random modulation (for the infamous a new sound at each note effect).

Mixer (S1)

 Mix    32|nois    0|sub     0|wave squ1
 xmod  sum|amnt   31|fuzz    0|crsh    0

Here is a list of sub-oscillator/transient generator waveforms. The first 6 settings correspond to the sub-oscillator, which is an oscillator of its own, playing always one or two octave lower than the oscillator 1. The next settings disable the sub-oscillator and enable various flavors of transient generators, which will produce a short, clicky sound at the beginning of the note:

The crossmod parameter controls how oscillator 1 and oscillator 2 are blended together. The different modulation modes are:

Mode Description What does the xmod parameter controls?
off Mixing: Oscillators 1 and 2 are mixed together nothing
sync Mixing and sync: Oscillators 1 and 2 are mixed together, with oscillator 2’s waveform phase-synced to oscillator 1’s nothing
ring Ring modulation: Oscillators 1 and 2 are mixed together, and simultaneously sent to a ring-modulator balance between the original mixed signal and the ring-modulated signal
xor Xor modulation: Oscillators 1 and 2 are mixed together, and simultaneously sent to a digital XOR cross-modulator balance between the original mixed signal and the XOR signal
fold Foldback distortion: Oscillators 1 and 2 are mixed together, and sent to a foldback distortion Amount of foldback distortion
bits Bit reduction: Oscillators 1 and 2 are mixed together, and sent to a bit depth reducer Amount of quantization noise

Filter (S2)

 Freq   96|reso    0           mode   lp
 env2   24|lfo2    0

Envelopes and LFOs (S3)

Ambika has three ADSR envelopes, three part LFOs and one voice LFO.

A part LFO is shared among all voices assigned to a part and is used for creating global sweeps, tremolo effects, etc.

A voice LFO is specific to each voice, and there will actually be a small amount of detuning/dephasing between them, for a richer sound.

Consider the example of a guitarist playing a guitar through a tremolo pedal, and applying a vibrato on a particular note. If you want to program something similar, you will use the voice LFO for the vibrato (because it is slightly different and desynchronized for each individual note) ; and you will use a part LFO for the tremolo, since it uniformly affects all the notes. Another example: you have found the perfect combination of LFOs to create a trancey gated strings sound. The last thing you want is to have each note of the chord triggered at a slightly different instant due to difference in timings when the chord is played. You need a part LFO for the gating effect. A voice LFO is good for richness and randomness, a part LFO is good for sync’ing things up.

As a result, only the voice LFO can be controlled in the modulation matrix (because it can be controlled by velocity, aftertouch, etc. – all kind of things that might be different for each played note) ; and only the part LFOs can be synchronized to tempo.

This part LFO vs voice LFO distinction sounds confusing, and there is actually little documentation about it. Most classic analog synthesizers do not have one LFO circuit per voice, so their behaviour is accurately simulated by the use of part LFOs. Some VAs, like the access Virus, use voice LFOs. On the Waldorf Blofeld, each LFO is a voice LFO by default and can be turned into a part LFO through the use of the sync setting.

 Lfo/eg  1|rate   24|wave  tri|trig free
 attk    0|deca   40|sust   20|rele   60

Here is a list of LFO waveforms:

Voice LFO (S3)

          rate   24|wave  tri          

Modulation matrix and modifiers (S4)

 Modul.  1|srce env1|dest prm1|amnt    0
 modif.  1|in1  lfo1|in2  lfo2|oper  off

The upper part of this page allows the routing of up to 14 modulation sources to synthesis parameters. The first knob is used to select one of the 14 virtual “modulation patchcords”, while the other knobs on the first row are respectively used to edit its source (srce), destination (dest), and the attenuation amount (amnt).

For example, this line:

 Modul.  4|srce env3|dest  vca|amnt   63

Indicates that the third envelope is routed to the VCA gain.

The modulation sources are the following:

  1. env1, env2, env3: ADSR envelopes.
  2. lfo1, lfo2, lfo3: Part LFOs (bipolar).
  3. lfo4: Voice LFO.
  4. mod1, mod2, mod3, mod4: Modifiers (see next section).
  5. seq1, seq2: Step sequences 1 & 2.
  6. arp: Arpeggiator gate pattern.
  7. velo: Note velocity.
  8. aftr: Note aftertouch.
  9. bend: Pitch-pend wheel (bipolar).
  10. mwhl: Modulation wheel.
  11. whl2: Modulation wheel 2, Joystick negative Y axis or Breath controller.
  12. pdal: Expression pedal.
  13. note: MIDI note number (bipolar).
  14. gate: Keyboard gate (null when the key is released).
  15. nois: White noise sampled at 1kHz.
  16. random: Random value, updated every time a note is retriggered.
  17. [256]{style=“text-align:center;“} to [4]{style=“text-align:center;“}: Boring constant values.

The modulation destinations are the following:

  1. prm1, prm2: Oscillator 1&2 parameters.
  2. osc1, osc2: Oscillator 1&2 pitch.
  3. ~1+2: Oscillator 1&2 coarse pitch.
  4. vibr: Oscillator 1&2 fine pitch.
  5. mix: Oscillator balance.
  6. xmod: Oscillators cross-modulation amount. This has no effect when the off or sync mixing modes are selected.
  7. nois: Noise amount.
  8. sub: Sub-oscillator amount.
  9. fuzz: Fuzz amount.
  10. crsh: Sample-rate reduction amount.
  11. freq: Filter cutoff.
  12. reso: Filter resonance.
  13. attk: Envelopes attack time.
  14. deca: Envelopes decay time.
  15. rele: Envelopes release time.
  16. lfo4: Voice LFO rate.
  17. vca: VCA gain.

One thing that requires a bit of clarification is the notion of bipolar modulation source. Let’s take an example. We have a triangle LFO mapped to the filter cutoff, with a modulation amount of 30. If the cutoff is set to 80, the actual value of the cutoff will oscillate between 50 (80-30) and 110 (80+30). On the other hand, if we have an envelope mapped to the cutoff with a modulation amount of 20, the cutoff will go from 80 to 120 (80 + 2 * 20), then down to 80 after the release. This is something to remember if you want to do PWM for example. If you set the PWM modulation amount to 40, you also have to set the oscillator parameter value to 40, so it will oscillates between 40-40 = 0 and 40+40=80. Otherwise, it will spend half of the time stuck at 0. But this makes things nicer for vibrato, tremolo, wah-wah or growl effects!

LFOs, pitch-bend and MIDI note number modulations are bipolar. If you are used to electronics/Modular speak, it is equivalent to say that the LFOs are always “AC-coupled” instead of “DC-coupled”.

Another thing worth noting is that the modulation amount of the last patch in the modulation matrix is always attenuated by the modulation wheel. Let us say that you have connected patch 14 from LFO 4 to the oscillators fine pitch (actually this is done by default in the init patch!):

 Modul. 14|srce lfo4|dest vibr|amnt   16

This modulation will be applied with an amount proportional to the mod-wheel position - by default it will not be active until you start moving the modulation wheel, and to get it at full strength, you will have to push the modulation wheel to the max. In the modular synth world, you would have obtained the same effect by patching the LFO at the signal input of a VCA, the joystick at the control input of the VCA, and routed the resulting signal to the exponential FM input of the VCO. Pheww!

The lower page of the modulation/modifiers page allows the combination of up to 4 pairs of modulation sources to create new modulation signals. The first knob on the bottom row is used to select one of the 4 modulation modifiers, while the other knobs on the lower row are respectively used to edit the two inputs of the modulation modifier (in1 and in2), and the operation (oper) to apply.

 modif.  1|in1  lfo1|in2  lfo2|oper prod

In the example above, the modulation source mod 1 is defined as the product of lfo1 and lfo2. The following operations are available:

Keyboard and sequencing

Keyboard and tuning (S5)

 Volu  120|mode poly|lega  off|port    0
 octv    0|tune    0|sprd    0|raga equa

The different scales/keyboard mappings are:

Name Description
equal Equal temperament
just Just intonation (frequency ratios to C are rational numbers)
pythagorean Pythagorean scale
14 eb E and B are 14 tone lower
14 e E is 14 tone lower
14 ea E and A are 14 tone lower
bhairarasia A bunch of ragas of the Maihar gharana . Note that it is not possible to play notes outside of the raga - if you attempt to do so, the previously played note will be retriggered. The suggested mode of operation is to find out which keys are active in the raga, and improvise with them!

Sequencer and arpeggiator settings (S5)

Each part of Ambika has a sequencer / arpeggiator. The sequencer consists of:

All parts follow the same global clock, which can be set to a fixed BPM or synchronized to an external MIDI clock. However, since polyrythmy can be fun, each part can be synchronized to a different subdivision of the MIDI clock and use a different step-sequence / pattern length.

3 sequencing modes are available - they define how notes played on the keyboard are translated into the notes played by the voices:

 A/sq step|dire   up|rang    1|patt    1
 grid 1/16|len1   16|len2   16|lenp   16

Let us now discuss the most devilish, confusing thing about Ambika sequencer: sequencer memory sharing. Ambika only has memory for 16-steps long patterns. Yet, the len1, len2, lenp lengths can go up to 32. How is it possible? Simply by reusing steps from one pattern into the other!

For example, if you set the step sequence 1 length to 18, the last 2 steps of step sequence 1 and the first 2 steps of step sequence 2 will store the same value. If you set the step sequence 1 and 2 length to 32, the last 16 steps of step sequence 1 will be the first 2 steps of step sequence 2 and reciprocally. When you set the notes sequence length to 18, the first 4 steps of step sequence 1 will be scrambled with apparently random data from the note sequence.

So remember: as soon as one of sequence 1 or sequence 2 exceeds 16 steps they will scramble each other ; and as soon as the note sequence exceeds 16 steps it will scramble the step sequences. Whether you will prudently avoid using the scrambled sequence or take creative advantage of this scrambling is up to you!

Sequence editor (S5)

This page shows the note pattern and the step sequences 1 & 2.

   | note |velocity |stepseq 1|stepseq 2
 01|. C4  |.   o100 |.    255 |.      0

What about the dots shown in some columns? They indicate the start of the sequence. For example, if the note sequence is 16 steps long, but the step sequence 1 is only 5 steps long, you will see a dot on step 6, 11 and 15 to remind you that the step sequence 1 “cycles” over 5 steps only.

Trivia: the longest non-repeating pattern you can build with Ambika is 20677 steps long (set sequence 1 to 23 steps ; sequence 2 to 29 steps ; and the note sequence to 31 steps). With a clock speed of 116 and a BPM of 86, the sequence would last for 1 hour!

Multi settings

These settings are not related to a specific part. They are saved (along with each part’s individual settings) in a multi.

Clock settings (S6)

 Bpm   120|groo swin|amnt    0|ltch   0

Performance page (S7)

 Para    0|freq   96|para    0|freq   96
 env2   27|volu  120|env2   24|volu  120

This page allows 8 parameters from different pages and for different parts to be displayed together on the same page. Observe that when a parameter is edited, the part number part is shown on screen:

 pt 1 oscillator 1  |waveform        saw
 wave squa|para   32|rang  -12|tune   12

A particular quirk of this page is the specific function assigned to the encoder. Press it to simultaneously send a C4 note to all parts. Press it again to send a note off message to all parts. This allows Ambika to be used without a MIDI keyboard. While a note is being played through this mean, you can turn the encoder to transpose all parts up and down.

To select the parameters/parts shown on the performance page, press S7 again to bring the customization page:

 performance page customization
 knob 1|part 1|oscillator 1 parameter

Use the first knob to select the index of the knob you want to customize. Use the second knob to select a part number. Use the third and fourth knobs, or the encoder, to scroll through the list of parameters.

The library

Everything related to loading/saving patches, multi, etc. is done on this page. The data is stored on a SD card (SDHC cards are supported) formatted in FAT16 or FAT32 format, inserted at the back of Ambika. In case the SD card is incorrectly formatted, incorrectly inserted, or damaged, the following text will be displayed every time you try to display the library page:

 /!\ SD card I/O error

Press S8 to continue and try blowing on the SD card connectors.

If the card is correctly inserted and formatted, the main library page will be displayed:

 pt 1 program  .A000 junon
  |   init|send|save|versions  more|exit

The first row displays the following information:

The second row is a list of commands. Each command is associated with a switch. For example, S2 performs the init command, S3 the send command, and S8 the exit command. To quit the library page, you thus have to press S8 (exit command).

Selecting an object

Press S1 to repeatedly cycle through the different types of items that can be loaded/written from/to the memory card:

Remember that you can hold S1 while turning the encoder to change the active part.

Ambika can store 3328 (26 banks of 128) patches/sequences/programs/multis.

Loading programs/multis/sequences/patches

Turn the encoder to move to the previous/next program. Remember that you can hold S8 while turning the encoder to scroll faster in the list. To change the current bank:

Reinitializing programs/multis/sequences/patches

Press S2 (init command) to reinitialize or randomize the select object. This brings the following confirmation page:

 pt 1 program             [  randomize ]
                              yes | no

Turn the encoder to select an action (randomize will replace all settings by random values; while initialize will replace all settings by the values they had at startup). Press S7 to confirm, or S8 to cancel the initialization.

Dumping programs/multis/sequences/patches as SysEx

Press S3 (send command) on the library page to send the selected object on the MIDI output, encoded in SysEx messages. Note that programs are sent as 2 messages (1 block with synthesis settings and 1 block with part settings) ; and multi are sent as 13 messages ; so you need to set your SysEx Librarian program to receive many messages.

Saving programs/multis/sequences/patches

After having selected with S1 which object type you want to save (multi, program, etc.), press S4 (save command).

 pt 1 program  .A000 junon

Before saving a program/multi/sequence/patch, its bank letter, number and name must be set. Use the encoder to increment/decrement the value or letter at the cursor. To move the cursor, click the encoder, turn it to move the cursor, and click it again.

When you are done, press S7 (save command). If you want to cancel the operation, press S8 (cancel command).

A warning will be displayed in case the memory card is damaged, full, incorrectly formatted, or just needs a little blow on its connectors.

Note that when a patch is overwritten, a backup copy of the previous version is kept. You can use a SD-card reader on your personal computer to remove the latest version of the patch and restore the backup copy (Sorry… Ambika does not come with its own version of Norton Commander).

Clipboard and version manager.

Press S5 (versions command) to bring the version manager.

 pt 1 program         version   0
 copy|swap|paste      << | >> |keep|exit

The following commands are available for the select object (part 1 program in the example given above):

System preferences

From the library page:

 pt 1 program  .A000 junon
  |   init|send|save|versions  more|exit

Press S7 to display more commands:

 pt 1 program  .A000 junon
 pref|>ini|about               more|exit

Press S1 to bring the system preferences page:

 Inpt ....|outp thru
 help   on|snap  off|leds   on|auto   on

The following system preferences can be edited:

It is also possible to save the current multi so that it is reloaded every time Ambika starts up. This is very useful for saving your MIDI settings and favorite split/voice allocation parameters!

 pt 1 program  .A000 junon
 pref|>ini|about               more|exit

On this library page, press S2 to select the [ini]{style=“text-align:right;“} command.

 (?) use current multi as default?
                               yes | no

Press S7 to confirm. The unit will freeze for a few seconds to write the current multi to its non-volatile memory.

Firmware update

Firmware update through the SD-card

From the library page:

 pt 1 program  .A000 junon
  |   init|send|save|versions  more|exit

Press S7 to display more commands:

 pt 1 program  .A000 junon
 pref|>ini|about               more|exit

Press S3 to bring the about page:

 ambika    v1.0|port 1 device 1     v1.0
 upgrade       |upgrade             exit

The first half of the screen displays the version of the OS running on the motherboard. If a firmware update file named AMBIKA.BIN is present on the memory-card, pressing S1 (upgrade command) will load it to upgrade the main processor.

Ambika contains seven ports to which devices are attached. The first 6 ports (port 1 to port 6) are used for connecting the voicecards ; the last port (port 7) is used for connecting additional controllers or I/O connectors. The second half of the screen on the about page shows the device type and OS version number for a selected port. Use the encoder to scroll through the ports. When no device is attached to a port (or when the device firmware has been corrupted), a ‘?’ is shown in place of the device type. If a firmware update file named VOICE$.BIN (where \$ is the port number) is present on the memory card, pressing S4 (upgrade command) will load it to upgrade the voicecard processor.

Emergency firmware update

If for some reason Ambika is “bricked” – this can happen if the SD card is damaged, if the firmware update file has been incorrectly copied to the card, or if the power has been disconnected during the firmware upgrade procedure – there is still a way of initiating a firmware update:

Press S8 while powering up Ambika to force the firmware to be reloaded from the SD-card. During loading, the unit displays the number of bytes loaded (the firmware takes up to 60kbytes):

SD update...

Press S7 while powering up Ambika to put the unit in MIDI firmware upgrade mode. In this mode, the unit is waiting for the firmware to be sent as SysEx blocks over MIDI. A .syx or .mid file with the latest data is available on the Mutable Instruments website. While receiving blocks of MIDI data, the unit displays the number of bytes loaded (the firmware takes up to 60kbytes):

Syx update...

If you want to use a dedicated SysEx transfer tool, we recommend Elektron’s C6 tool available for both Windows and OS X, configured with a 250ms delay between blocks.

Resetting the unit to factory settings

Most of the data held by Ambika is stored on the SD card, with one exception: the boot configuration, which is stored on the main processor’s non-volatile memory. The boot configuration is the set of settings loaded when the unit is powered on. To restore the default factory boot configuration, keep the S1 key while powering the unit. The unit will freeze for a few seconds to retrieve and write the factory defaults.

Dirty technical details

MIDI implementation


A MIDI implementation chart is available here.

Program changes

Program change messages are used to load either:

The MIDI banks 0-25 are used to load program banks A to Z. The MIDI banks 26-51 are used to load multi banks A to Z. For example, to load the multi 002 in the B bank, send a control change 0 (bank select) with a value of 27 to select multi bank B, and a program change 2. To load the program 120 in the D bank, send a control change 0 (bank select) with a value of 3, and a program change 120.

External clock reception

Keep in mind that the tempo must be set to ext to receive an external MIDI clock! Otherwise Ambika continues running with its own internal clock.

Control changes

The Modulation Wheel, Breath Controller and Foot Pedal CCs ; along with aftertouch messages might not affect the sound unless they are patched to a modulation destination in the modulation matrix. If you want these sources to control the intensity of a LFO, it is recommended to do it through a modifier. Here is an example of patching:

 Modul.  8|srce mod1|dest   mix|amnt   63
 modif.  1|in1  lfo2|in2   whl2|oper attn

Mod1 is defined as “LFO2 attenuated by the modulation wheel 2”, and this modulation is applied to the oscillators balance.

Please refer to the CC page of the MIDI implementation chart given above to see the list of synthesis parameters controllable by CC. Editing a synthesis parameter by CC is fool-proof: values between 0 and 127 are translated into the full range of the parameter; so that 0 always means the minimum value taken by the parameter, and 127 its maximum value. This might cause a resolution loss for parameters with a wide range above 127.

NRPN patch editing

NRPN messages can be used to directly write into the Patch and PartData data structures.

The NRPN number indicates the address at which a value must be written ; and the data entry / data increment messages are used to set the value. Observe that you need to send both a Data Entry MSB and a Data Entry LSB message to send a signed value, or a value above 127. Signed values are represented in 2’s complement.

NRPN range Mapped to Data structure declaration
0 .. 111 Patch data structure, offsets 0 .. 111 struct Patch
112 .. 192 Part data structure, offsets 0 .. 80 struct PartData

Here is an example: you want to modify the part tuning to -13. The tuning byte is at offset 2 in the Part data structure, so its NRPN number is 114. In 2’s complement, the value 13 is represented by 128 13 = 115. Thus, the NPRN that needs to be sent is:

176 99 0 (NRPN MSB set to 0)`
176 98 114 (NRPN LSB set to 114)`
176 6 1 (Data Entry MSB set to 1 -- value above 127 or negative)`
176 38 115 (Data Entry LSB set to 115, because 115 - 128 = -13)`


Ambika can receive and transmit settings in system exclusive format. The messages all share the same format:

0xf0  (SysEx)
0x00 0x21 0x02  (Mutable Instruments Manufaturer ID)
0x00 0x04  (Product ID for Ambika)
0xf7  (End of SysEx)

Command indicates what the receiver should do with the data in the payload block, and argument is a byte that might contain an additional piece of information about what to do with the data.

Payload is a sequence of bytes in nibblized form. For example, the sequence 0x80 0x56 0x13 is transmitted as 0x08 0x00 0x05 0x06 0x01 0x03.

Checksum is equal to the nibblized-sum, modulo 256, of the bytes. In the example above, the checksum is equal to 0x80 + 0x56 + 0x13 = 0xe9 and is transmitted as 0x0e 0x09.

Data structure dump

When transfering patch/sequence/program/multi data to Ambika, command indicates the data structure being sent and argument indicates the part number the data must be loaded into.

The following table summarizes the different payload types.

Command Argument Affected data structure Payload size
1 Part number (0 for current part) Patch 112 bytes
2 Part number (0 for current part) PartData::sequence_data 72 bytes
4 Part number (0 for current part) PartData 84 bytes
5 Always 0 MultiData 56 bytes
Data structure request

Use this command to request one or several data blocks from Ambika.

Command Argument Reply
17 Part number (0 for current part) Patch bytes for selected part
18 Part number (0 for current part) Sequencer bytes for selected part
19 Part number (0 for current part) Patch bytes + PartData bytes for selected part
20 Part number (0 for current part) PartData bytes for selected part
21 Always 0 MultiData bytes + Patch Bytes and PartData bytes for the 6 parts
Random memory read (PEEK)

Use this command to request the contents of an arbitrary block of RAM, up to 32 bytes long. Command is equal to 0x1f (31). Argument is equal to the size of the requested block. The two bytes of payload indicate the address to read (16-bit, little endian address). For example, to retrieve the 4 bytes at addresses 0x3ac .. 0x3b0, the sysex string would be:

0xf0  (SysEx)
0x00 0x21 0x02  (Mutable Instruments Manufacturer ID)
0x00 0x04  (Product ID for Ambika)
0x1f  (PEEK command)
0x04  (Number of bytes to retrieve)
0x0a 0x0c  (LSB of the address, here, 0xac)
0x00 0x03  (MSB of the address, here 0x03)
0x0a 0x0f  (Checksum)
0xf7  (End of SysEx)
0x1f 04 0a 0c 00 03 0a 0f f7@
Random memory write (POKE)

Use this command to transfer arbitrary data into Ambika’s RAM. Command is equal to 0x0f (15). Argument is equal to the size of the block to transfer. The two first bytes of payload indicate the address to write to (16-bit, little endian address). The following bytes of payload contain the data.

With great power comes great responsibility - this feature has the potential to create some damage that would require a factory reset!

Developers can use the avr-nm tool to get a memory map and find out which datastructures are located at which addresses. For example: avr-nm build/ambika_controller/ambika_controller.elf -C -n. Addresses in RAM are prefixed by 0080. Mutable Instruments cannot guarantee that these addresses won’t be subject to changes from one firmware version to the other.

Memory card storage

Filesystem layout

The Ambika firmware supports both SD and SDHC memory cards, formatted with FAT16 and FAT32 filesystems. Memory cards with long file names are not supported. Congrats to ChaN for his amazing FAT library!

Firmware update files
 AMBIKA.BIN Firmware update for motherboard
 VOICE1.BIN Firmware update for voicecard 1
 VOICE6.BIN Firmware update for voicecard 6
Patches, programs and sequences
   /000.PAT Patch 000 in bank A
   /000.PA~ Backup version of Patch 000 in bank A
   /127.PAT Patch 127 in bank A
   /000.PAT Patch 000 in bank B
   /127.PAT Patch 127 in bank Z
  /CLIPBRD.PAT Contents of patch clipboard
  /CLIPBRD.PA~ Previous contents of patch clipboard
   /000.PAT Undo/redo sequence of part 1's patch
   /001.PAT Undo/redo sequence of part 1's patch
   /000.PAT Undo/redo sequence of part 6's patch

For programs and sequences, the directories are PROGRAM and SEQUENCE, and the file extensions are PRO and SEQ.

   /000.MUL Multi 000 in bank A
   /000.MU~ Backup version of Multi 000 in bank A
   /127.MUL Multi 127 in bank A
   /000.MUL Multi 000 in bank B
   /127.MUL Multi 127 in bank Z
  /CLIPBRD.MUL Contents of multi clipboard
  /CLIPBRD.MU~ Previous contents of multi clipboard
  /000.MUL Undo/redo sequence of multis
  /001.MUL Undo/redo sequence of multis

File format

All files are stored in RIFF format.

This is followed by one or many chunks. Each chunk consists of a 4 chars identifier, a 32-bit integer indicating the size N of the “payload” data, and N bytes of “payload” data. Two types of chunk are present in a file: