2.0 idea: install engines and effects from a catalog (plugins)

[kurogedelic]

Felucca has a lot of engines now, and the FM-1's flash is nearly full. For 2.0 I'd like to change how engines and effects get onto the FM-1: instead of everything being built into the firmware, you'd pick the ones you want from a catalog and install them, like apps.

How it could work

• Two kinds of package: instruments (an engine, like ANALOG or FM6) and effects.
• A catalog of packages in the web editor: name, author, what it does, how much CPU and memory it needs. Ideally you can try one in the browser emulator before installing it.
• Install, update, remove from the editor, over USB, without reflashing the firmware. The editor tells you when a package has an update, or when it needs a newer Felucca.
• On the FM-1: a list of installed packages in the MENU, where you can turn one off. A project that uses an engine you've removed plays that track silent ("NO ENGINE") and sounds as before once it's installed again.
• The firmware gets smaller: a core set of engines stays built in, and the less used ones can move to the catalog.

Writing your own

• Anyone can write a package in their own repository. To get it into the catalog, you open a pull request to the official catalog repository that adds your package (a fixed release of your repository).
• For safety, each package is reviewed, then built and signed by the catalog, and the FM-1 only installs signed packages from there. Every new version goes through the same review.
• Developer mode: to try your own builds while you work on them, you'll be able to load unsigned packages after confirming on the FM-1 itself.
• Licence: the built-in engines are part of Felucca and stay GPL-3.0-only. Packages are not part of Felucca: their authors can choose their own open-source licence.

Not decided yet

This is an idea, not a promise: it's a big change (2.0), and the details will change as I work on it. The sample packs planned for 1.4 could use the same catalog.

What would you want to install? And if you'd like to write an engine or an effect, what would you need from the plugin interface?

[jsmonzani]

That's exactly how I think it should evolve! This would prevent too many forks for simple features, and would resemble open programmable playgrounds (I remember creating modules for VCV Rack back in the day).

What is missing most for me (ambient composer) is a proper "non metallic" reverb. Cheap way is to add modulation to it but a better implementation would make it really usable. I've seen someone working on a FDN algorithm https://www.reddit.com/r/MVaveFM1/comments/1x02re2/enough_verb/ and another one using a hall reverb https://www.reddit.com/r/MVaveFM1/comments/1x00i7m/presenting_ghoulbox_a_free_dungeon_synth_firmware/

So a proper reverb (if not already there) would be my first implementation.

Besides synths and FX, there are some deeper tweaks that are desirable but seem difficult to add as modules. I'm thinking of off-grid recording (i.e. no quantization). This has been created already here and might be added to the v2.0 agenda:
https://www.reddit.com/r/MVaveFM1/comments/1x05njg/melodee_012_is_out_big_improvements/

Finally maybe being able to extend sequencing functionalities through a module would please some people? I'm thinking generative sequencers or advanced sequence editing functions that might only appeal to a few.

What I particularly enjoy with Felucca is how logical and easy the firmware is. While sloop keeps adding fun features, it's way to complex to navigate for its own good.

Modularity is definitively a way to add just the complexity someone wants. Much like vst plugins.

Hope this helps!

[etyrnal]

That's exactly how I think it should evolve! This would prevent too many forks for simple features, and would resemble open programmable playgrounds (I remember creating modules for VCV Rack back in the day).

What is missing most for me (ambient composer) is a proper "non metallic" reverb. Cheap way is to add modulation to it but a better implementation would make it really usable. I've seen someone working on a FDN algorithm https://www.reddit.com/r/MVaveFM1/comments/1x02re2/enough_verb/ and another one using a hall reverb https://www.reddit.com/r/MVaveFM1/comments/1x00i7m/presenting_ghoulbox_a_free_dungeon_synth_firmware/

So a proper reverb (if not already there) would be my first implementation.

yeah, ditch the 'spring' which is nothing more than a short time, high rolloff delay.

replace that junk spring verb (which sounds literally NOTHING like a spring verb), and replace it with a long decay plate or 'cathedral'

i also want the ambient people to be included which is why i put in an idea about changing the granularity of the knobs from 'milliseconds', to a look up table of musical subdivisions of the current tempo... like 8 bars, 4, bars, 2, bars, 1 bar, 1/2, 1/4, 1/8, 1/16, 1/32, 1/64, 1/128, 1/256 and include the in between triplets and dotted notes. It's only like 31 values or so, and below 1/256 you can do ms , and above 8 bar you could do seconds.

[kurogedelic]

Thanks for the ideas! HALL reverb was added in v1.4, so I'd like to keep this discussion focused on the plugin system.
Sequencer extensions are interesting too, but I'd like to start with engines and effects first.

[firstnevyn]

I wish it'd go a little further.. so that it could run on the SMK-37 family as well (same mcu same midi sysex same screen same usb. different buttons and knobs.

perhaps even build config system?

[gospodindilly-ops]

#163 (comment) "And Felucca isn't just for the FM-1. We're preparing the architecture for other devices, including the SMK-37. The FM-1 is simply where it started."

[kurogedelic]

@firstnevyn I'm looking into the SMK-37. Thanks, this would be a big step forward. I'm going to buy an SMK-37 myself, so it will probably come soon.

[firstnevyn]

Did you see #174 (I don't have a windows or mac machine)

This confirms ota works with the python tool.

[gospodindilly-ops]

@firstnevyn I'm looking into the SMK-37. Thanks, this would be a big step forward. I'm going to buy an SMK-37 myself, so it will probably come soon.

I have too if u need some test - feel me)

[firstnevyn]

In an Ideal world I'd like a web pick a basket where I could pick and choose features to meet my requirements (think kconfig in a browser)

• fixed core (hardware support for whatever hardware we're on plus base chip bring up and ota functions)
  • USB midi and ota functions,
  • depending on hardware capabilities and requirements
    • trs midi,
    • screen,
    • keyboard and controller elements
• a device mode/overall model
  • 4 ch like felucca is now
  • alternatives (midi rompler/synth mode) or 4 midi channel synth no sequencer
• usb (midi is a given but options for extra virtual ports possibly)
  • Audio profile and number of channels
  • Midi virtual ports
  • serial control
• bluetooth (I know we're not there yet.)
  • midi definitly
  • audio ? maybe latency is a challenge but fm1 playing back to wireless headphones might be useful. or a wireless bt speaker
  • host mode?
    • bt pair a smk-37 to a fm1
• Synths
  • selection of synths
    • assign virtual midi ports? (allows native librarians for emulated synths?)
  • midi learn for synths?
• Midi data managment generator and export features
  • Midi maps (switchable?)
  • Arpegiators
  • Generative midi functions/mutators
  • Tunings and scales
  • standard control schemas (MCU/HUI)

looking at the list above.. you can't have all of that stuff on all the time. so some sort of configuration/build system is necessary.
The stuff other than the core would be user selectable via a web form.. so you build a custom firmware for them.. if they don't want bluetooth and want more synths they can have that. if they want more space for samples and less features they can have that within the firmware budget

Once it's built.. they can test it in an emulator.. then flash that fmwc to their fm-1

[kurogedelic]

The initial plugin experiments are complete. All three tests succeeded on a real FM-1:

1. Execution: Ran code directly from a dedicated flash slot using a function pointer.
2. Relocation: Built a plugin for address 0, then relocated it at install time. It ran successfully from either of two flash slots.
3. Engine plugin: Moved the LOFI engine out of the main firmware and got it running through the first version of the plugin API (v0).

One interesting finding: most engine code is surprisingly small, around 1–4 KB. Samples actually consume much more flash.
The bigger challenge is RAM. Heavy engines like PHYS need around 50 KB, so plugins will need a way to share memory with the core.
That's why I'm reorganizing engine state into a shared RAM region in v1.5. This is the groundwork for 2.0.
Custom firmware builds are an interesting idea, but I see them as an intermediate step. The goal is still to install, update, and remove engines and effects without rebuilding or reflashing the entire firmware.
There's still plenty of work ahead, but the initial hardware tests are looking promising.

Once the design is finalized, I'll move the 2.0 work to a dedicated development branch.