Reply to thread

Message: [QUOTE="Bombastus, post: 90927, member: 2178"] Roland's SuperSaw stacks seven sawtooth waves, and the mix control that thickens them also drops the level of the center saw to avoid clipping. That detail is missing from most recreations. It is also why a maxed-out unison stack can sound wide and weirdly hollow at the same time. The original oscillator is one waveform, not a chorus effect. A master saw sits at the played pitch and stays put, while six slave saws move around it under the detune and mix controls. Everything sums to a single output before it reaches the filter. Newer synths pull those saws apart and place them across the stereo field, which the summed original never did. [B][URL='https://goldmidi.com/community/threads/discodsp-introduces-corona-7-synth.76250/']The Super 7+ unison mode added in Corona 7[/URL][/B] pans each saw and gives you a spread control over the result. Useful, but it changes the shape of the sound before you touch a single tuning parameter. [HEADING=2]The detune spread bends instead of stepping evenly[/HEADING] Analysis of the hardware found the detune amount follows a parabolic curve rather than a straight line. Small movements near the bottom of the range barely separate the saws. Push toward the top, and the spacing opens up fast, which is where the classic wide trance lead lives. At full detune, the six slaves sit in three symmetrical pairs around the master. The outer pair lands at the full detune distance, the middle pair at three quarters of it, and the inner pair at seven eighths. So the cluster is lopsided by design, tight in the middle and stretched at the edges. That matters when you rebuild the sound in a synth with plain unison. Most unison engines space voices evenly and scale them linearly, so [B]the curved detune spacing between those saws[/B] is the first thing you lose. If your synth lets you tune unison voices by hand, uneven offsets get you closer than any amount of detune knob twisting. Odd voice counts help too, since an even count leaves nothing sitting at the played pitch. [HEADING=2]The mix knob trades center weight for side sizzle[/HEADING] Measurements of the hardware put the center saw at roughly 1 minus 0.555 times the mix amount, and each slave saw at seven seventeenths of it. Run the numbers at maximum. The master drops to about 45 percent of its full level while each of the six slaves climbs to around 41 percent. So the mix control is not a wet and dry blend, and it is not a simple layer count either. It is a level trade that keeps the summed output from clipping while the crowd of saws grows. The old assumption was that mix raised the six slaves and left the master alone, and spectrum analysis of the hardware showed otherwise. Copy that behavior and your stack stops sounding louder every time you add width. Push the side voices up in a modern synth without pulling the center down, and you get the same buzz with a fatter middle, which sits differently in a mix. Neither version is wrong. They just occupy different amounts of space once a bass line and a kick arrive. [HEADING=2]Free phase and a filter you never see[/HEADING] The seven oscillators run free, so each new note starts at whatever point in the cycle the oscillators happen to be. Two identical keypresses give slightly different results. That drifting quality is a large part of why the sound feels alive rather than static, and it disappears the moment you enable phase reset on a unison stack. There is also a high-pass filter in the chain that nobody mentions. It runs at 24 dB per octave and tracks the key, so it clears the rumble that detuned saws generate below the fundamental without thinning the top of the range. Reverse engineering of the original code backs this up. The saws are high-pass filtered and run at a doubled sample rate of 88.2 kHz, and the person who dug through it summed up the design as no modulation, no phase tricks, no chorus, no magic. Aliasing that the doubled rate does not remove stays in the sound and reads as air rather than as a fault, which is why cleaner modern saws often sound smaller than the thing they are copying. Try the same trick on your own patches. A high-pass parked at a fixed 120 Hz cuts the low end out of your high leads and leaves your low chords muddy, while one that follows the played note clears exactly the same proportion of every voice you play. [/QUOTE]

Name