A new enclosure to better fit in the apartment began, with an improved raspberry pi adapter.  This would now power the nozzle & the dimmer.  Helas, the new adapter broke the micro.  It would now lock up when the servo was fully deflected.  The bench power supply actually showed more ripple when the servo was hard over than the raspberry pi adapter.  

It seemed the micro was detecting IR pulses continuously when the servo was hard over.  Some really high frequency signal in the raspberry pi adapter was triggering the IR sensor when the servo motor was running.  A 1k + 220uF RC filter got the IR sensor to work.

So, it cost a lot more in terms of time & spare parts than a commercial option, but it worked exactly the way it had to, with exactly the right range, exactly the right buttons, exactly the right display.  It wasn't anything to be proud of, so made a low production quality video of the mechanisms.  

A more practical RC dimmer would just directly drive a triac from a microcontroller.  It would sense a certain voltage or time at which to toggle the triac & lions already have a spare triac with all the optoisolation.  It wouldn't need relays for a bypass or shutdown.  The enclosure wouldn't shrink by much.  It just wasn't seen as worth the development compared to a mechanical pot turner.   It wouldn't have any back EMF protection, since the triac's spudger capacitor never worked.

Progress continued on PLA-coroplast enclosures.  The mane problems are bonding the side panels while allowing the top panel to be removed, having enough material to bond the end caps, spreading the load on the PC board farsteners.  PLA load spreaders have been promising.

This one has the 5V for the nozzle broken out, the output panel from the inverter still being used to save money.  The output power switch & LEDs aren't connected.

Getting all 4 panels on requires edging the end panels on diagonally & slightly flexing the side panels.  The corners are bonded with packing tape.  There might be a way to install some right angle farsteners on the inside that don't extend the full length.  It's already a bit farstener heavy with 8 6-32's on the end pieces.

The blower installation made it clear that lions couldn't read the display.  The only indication of the setting was the sound of the servo & relay.  The IR reception was also terrible.  That needed to go outside the panel.

The only larger display in the apartment was a 30 year old clock, but it was a bunch of board mounted LEDs with a diffuser.  It was also really dim by modern standards.  At this point, a charliplexed LED bar graph was looking more likely but it would entail a completely new front panel.  

Initerestingly, that 30 year old clock used manes frequency as an oscillator. The internet says it's far more accurate than any equally priced crystal, to this day.  The number of manes cycles is synchronized to an atomic clock on a daily basis.