Proposal: DSP Engine Extensions Without Increasing Voice RAM

[gospodindilly-ops]

I’d like to suggest a small DSP expansion pass for the existing Felucca engines. The main idea is to make the current synth engines deeper and more flexible while keeping the existing voice architecture and, where possible, avoiding any increase in per-voice RAM usage.

ANALOG

• In dsp.c, expose the separate SVF LP/BP/HP outputs.
• In eng_analog.c, add SUB, hard OSC sync, and selectable filter modes.
• Reuse the existing ph[] and s[] state without increasing the voice state unless necessary.

PHASE

• Add wavefold.
• Add controlled output feedback.
• Reuse the existing space in s[].

TRIO

• Add PWM modulation.
• Add wavefold.
• Leave audio-rate cross-FM for later, after measuring the CPU cost.

LOFI

• Use the currently unused ph[2] for a second oscillator phase.
• Add OSC2 and oscillator sync.
• Keep the existing parameter/preset format unchanged for the first stage.

FM6

• In fm6_core.c / eng_fm6.c, experiment with velocity → FM depth using the existing KVS/VMOD path.
• There is no need to implement a separate carrier-spread feature: the existing DTUN already provides carrier detuning/spread.
• Leave algorithm morphing for a separate experimental branch.

Approach

I’m not a C/DSP developer, so this is a proposal for the direction rather than a ready-to-implement patch.

If these ideas seem interesting, I’d be very interested in discussing which ones make the most sense given the current CPU/RAM budget of the FM-1.

In particular, I think it would be interesting to explore how much more synthesis depth can be achieved without increasing the voice RAM footprint or reducing the current polyphony.

If the direction seems reasonable, I can help work through the desired musical behaviour, parameter design and expected results, while leaving the actual implementation to the project

[kurogedelic]

Thanks for the detailed proposal. Several of these fit well, and I like that they keep the voice RAM as it is. I've noted them for 1.1. Each one will depend on what the FM-1's CPU allows with four tracks playing.

[kurogedelic]

A first piece of this is in Felucca 1.4: https://github.com/hugelton/Felucca/releases/tag/v1.4

ANALOG's WAVE now has SYNC (OSC 2 hard-synced to OSC 1, DTN sets the interval, and the envelope or LFO on SHP sweeps it) and SUB (a square an octave down, MIX sets its level).

The filter modes and the PHASE / TRIO / LOFI ideas aren't in yet, so I'll keep this open. Thanks @gospodindilly-ops!

[kurogedelic]

1.5 is out: https://github.com/hugelton/Felucca/releases/tag/v1.5

ANALOG's filter modes are in: an ANALOG track has a new EDIT > FILTER page (right after EDIT 2) where KNOB 1 TYPE picks LP (the low-pass as before), BP (band-pass) or HP (high-pass), with CUT, RES and KTR on KNOB 2–4. LP is the default, so every ANALOG sound you have stays exactly the same. The PHASE, TRIO, LOFI and FM6 ideas aren't in yet, so I'll keep this open. Thanks @gospodindilly-ops!

[gospodindilly-ops]

thx u too for awesome CFM!