Before we sample an instrument (see Sample Store), we characterize it. This article covers interesting details of the Matrix6 series (Matrix-6 Matrix-6R Matrix-1000) architecture, learned during the characterization process.
While at first glance, the Matrix-6 family's dual oscillators waveforms appear limited (only pulse and saw, although DCO2 adds noise), they are very versatile, since both waveshapes are variable (e.g. the saw can become a triangle, the pulse can become a square and be modulated) and each oscillator can output both at once.
The Matrix-6 and Matrix-6r both have an auto-tune function, which implies they use VCOs. In reality, the pitch is controlled by a digital pulse, but it still drifts out of tune for about an hour after it's powered up. Interestingly, Osc2 detune affects the sync function when in normal mode, but not when in Split mode.
The pulse waveform does show a DC shift. At 0, the pulse is completely positive (i.e. above the X axis, while completely negative at 63 (see figures 1-5). It's not surprising that the waves round off as the frequency increases (see figure 6). At low frequencies, the waveforms look odd (see figures 7-9). While C0 (frequency = 0) 0% looks ok, at 25% it looks like a 50% pulse and at 50% it looks like a 75% pulse (the Matrix had difficulty outputing a C0 pulse at 75% or 100% width). BTW, the Matrix-6 series seems to have difficulty outputing the pulse unless the width is below 62.
Per the manual, the "waveform" setting is a variable saw/triangle. It appears the Matrix morphs the waveform from saw to triangle by just changing the almost linear decay to an instant decay (i.e. discharging a cap) at the variable switch point (see figures 1-3). Once again, the low and high frequency waves diverge from the manual description. At C0 (frequency = 0), the saw looks similar for most shape settings (saw with a rounded rise, see figure 4-6). However, the shape settings gradually take effect until the intended waveform is output at A0. As the frequency increases, the shapes get more rounded (i.e. closer to a sine, see figure 7).
Each oscillator can output one or both waves (i.e. pulse and/or saw/triangle, see figures 1-3). In addition the two oscillators can be synchronized (soft/medium/hard, see figures 5-7). As figures 3-7 show, each method of mixing the waveforms produces different results. In summary, the availability of variable saw and pulse waves, along with the ability for each oscillator to mix these waves makes the Matrix6 series one of the most versatile oscillator based synth. BTW, DCO2 can also produce the obligatory noise, instead of a pulse or saw wave. One could argue a sine wave would complete the waveform set, but sine waves are of little use to subtractive synths and the triangle does a reasonable job of approximating a sine in the higher keys.
We characterized the 4 pole filter by routing noise from DCO2 through it and cranking up the resonance to make the cutoff point more obvious. Without key scaling, the cutoff starts at around 12kHz (set at 127), but by the time it's set to "100", the cutoff is already down to 2.5kHz (see figure 2). By the "84" setting, the cutoff is at 1kHz (see figure 3). Below a setting of "60" on the Matrix1000, the filter cutoff is reduced to two frequencies (155Hz for settings around "50", 80Hz for a setting of "44" and below). The Matrix6r shows a more gradual frequency change from 230 to 30Hz. So, the filter cut-off range is relatively compressed.
However, with key scaling, the filter cut-off range becomes more useful. The key scaling doubles the filter frequency every octave. Filter keyboard tracking implies the actual range starts at C-1 (i.e. tracked C0 is 2x no tracking). What started as an unimpressive filter range (i.e. 12kHz max with no key scaling), becomes quite useful with key scaling, as the filter ranges from about 28Hz-28kHz.
The Matrix6 series has one of the most versatile modulation schemes available on a non-modular analog synth, giving it similar capabilities to a modular synth. The "matrix" modulation proved the flexibility of software based modulation, which has been adopted by many later synths. The matrix consists of 10 programmable source/destination pairs, along with 18 hardwired source/destination pairs. Each of the matrix settings allows access to 20 sources and 32 destinations, each with a +-63 depth range. Sources include 3 envelope generators, 2 LFOs, 2 ramp generators, a tracking generator, a lag processor (portamento) and various performance controls (e.g. 2 levers, 2 footswitches, aftertouch, attack and release velocity). Destinations include the DCOs, VCF, VCA, LFOs, EGs. Additionally, the tracking generator and the S/H (sample and hold) LFO waveforms can use any of the matrix sources as their source. Triggering the envelope generators, LFOs and ramp generators is also very flexible, allowing single, multiple or external (via pedal jacks) settings. The envelope generators add LFO1 to their trigger source list too.
In application, the tricky part is rewiring a semi-hardwired paradigm to take advantage of the flexible modulation routings. For example, try routing the LFO (50% rate, triangle wave, keyboard input) to VCA2 (+ or - depth). Instead of the hearing a tremelo effect, you'll here an arpeggio. To get the tremelo effect, you'll need to route the LFO to ENV2 Amplitude. That example shows the subtle routing difference normally hard-wired in other non-modular synths and the paradigm shift needed when dealing with modular routings. For another example, route the LFO (same settings) to the Tracking Generator input and route the Tracking Generator to VCA2. Now the Tracking Generator controls the arpeggio note duration (i.e. legato/staccato for low/high point values, respectively)