← Back to context

Comment by Animats

9 days ago

It's about how radio control toy servos from the 1970s work. Annoyingly, those pre-computer dumb devices with no feedback output still dominate the low end of mechanical output devices.

Makes me wonder if the generic servos of the described kind are really close enough to the performance a cheap-class servo can have, or if modern advances in monolithic power stage ICs could allow a servo free of sliding movement (no brushes, no wiper potentiometer (maybe a capacitively coupled differential sensing of angle, or the tricks of the cheap digital calipers with their iirc nonius-like scale read through several parallel tracks of non-touching capacitive electrodes?), instead just a clever chip digitally controlling a brushless electric machine using the feedback sensing available to it).

Being able to run an even just very simple digital controller allows things like severely dropping negative feedback gain at a resonance frequency of the larger system. And so much more.

  • The nice thing about using a potentiometer for position sensing is that you don't have to home the thing.

    There are lots of alternative sensors, but most are bigger, heavier, or more expensive. If 1% precision is good enough, pots are fine. The next step up is Dynamixel servos, which have a nice daisy-chain digital interface, encoders, about the same form factor as toy-type servos, at about 10x the price.[1]

    [1] https://www.robotis.us/dynamixel/

Yeah, but they're cheap and basically trivial to use. Cheap enough and trivial enough that they can replace solenoids in a lot of use cases.

One of my most amusing applications was the client who put an R/C servo on the choke cable of a carbureted generator motor instead of spending more money to buy the fuel-injected version. Servo cost about $5 and we were already measuring air temperature and had a PWM output available.