Miles and I have been working on *ams* (acoustic modular system) a lot of late. As you can see on this site it's a varied affair, spanning many musical instrument types. We've even made some music with it at times.
Part of what you battle with when making musical machines is the signal-to-noise ratio. How much machine noise is there vs "desirable" sound. We discussed this the other day and Miles found me to be very sanguine about it. I liken the machine noise to the many extraneous sounds other musical instruments or musicians make when playing. Whether it's the dense reed hiss of a sax, the clonking of valves on a brass instrument, bow noise, the groaning of a jazz pianist or the whirring of a motor, I am content that this is part of the texture of the output/performance.
In fact, I would go beyond this to suggest that in machine music the sound of the machine itself is part of the embodiment of the sound. Part of its expressive potential. It's something I like to harness sonically in my work. Maybe it's due to my background in experimental music, or the fact that my machines are not entirely self-playing, which means I can use it more deliberately but I certainly celebrate the machine sound, rather than trying to hide it.
This and the "expressiveness" of musical machines is taking up a lot of my thinking as my PhD develops. You can expect more rambling about it all over on my PhD development website in due course.