Layering Two Hybrid Voices
One hybrid voice sounds real. Two layered -- one for body, one for shimmer -- sound expensive. That is the secret behind lush patches.
The Layer Trick
A single LA voice is convincing for one timbre. But real instruments have layers of complexity: a piano has hammer noise AND resonant body AND sympathetic strings. No single attack-plus-sustain voice captures all of that.
The fix is simple: use two hybrid voices. Voice A provides the body -- maybe a piano attack into a warm sawtooth. Voice B adds shimmer -- perhaps a bowed attack into a soft triangle an octave up. Mixed together and run through reverb, the result sounds rich and expensive. This is how the D-50's famous "Digital Native Dance" patch works.
Try it -- press Play and use the presets to hear different two-voice combinations. Adjust the Mix slider to blend between Voice A (body) and Voice B (shimmer). The reverb adds space and glues the two voices together.
Why Layering Works
Each hybrid voice fills a different frequency range and temporal role. Voice A might carry the fundamental with a warm saw sustain, while Voice B adds high-frequency sparkle with a triangle an octave up. The attack samples give each voice its own character -- a hammer plus a bow is a combination that does not exist in nature, but your ear accepts it because each component is individually realistic.
Reverb is the glue. It blends the two voices into a single acoustic space, smoothing over any seams. Without reverb the two voices sound like two separate instruments. With reverb they fuse into one rich timbre.
What the Controls Do
Voice A and Voice B each have their own attack source and sustain waveform. Voice A is typically the "body" (lower, warmer) and Voice B is the "shimmer" (brighter, higher harmonics).
Mix controls the balance between the two voices. At 0% you hear only Voice A, at 100% only Voice B. The sweet spot is usually somewhere around 40-60%.
Reverb controls the wet/dry mix of a convolution reverb. Even a little reverb helps glue the two voices together.
Xfade A and Xfade B control the crossfade point for each voice independently. Different attack types work best with different crossfade times.
References
- Wikipedia: Linear Arithmetic synthesis
- Roland D-50 Owner's Manual -- "Upper" and "Lower" partials in each tone