Long (crossfaded)samples are not "locked" to the engine model. They are just adjusted in pitch to match the rpm. Also they usually sound bad at very low rpms.
I would not like to see the synthesized way of producing them abandoned. Very short "explosion" samples fired at proper sync with the engine model are the way to go IMO. As no doubt stated before, this way of creating the sound offers better feedback, and isn't the feel what LFS is all about?
My thoughts for improving the current model for better immersion:
- Multiple impulse sounds which could change according to load, rpm range etc.
- "Polyphonic" stream of explosions so that the sounds can overlap. Don't know for sure, but i get the feeling the current system is monophonic - at high rpms the explosion sound is interrupted by the next one. This is not the way it works IRL. Overlapping explosion samples would sound richer and less tinny, i assume. Maybe a few "channels" per cylinder?
- Some kind of reverberation processing could bring out resonances from intake, cockpit, etc.
- Separate impulse samples for header / collector / tail. Filtering a single sample does not offer enough flexibility.
- More dirt! Random variations firing when lifting throttle. (Like the popcorn effect.) Race-tuned engines should not idle so smoothly.
- Variable distortion
Having worked with software synthesizers and effects I get the feeling these improvements would not be a problem for newish cpus. Accuracy, high samplerates and high fidelity are not so important in this case, rich dynamic texture is.