Motion-detector doorbell hack

After the renovation was done I still had an empty light block sitting above the side door outside, and an ugly hole in the kitchen ceiling with wires hanging out of it. The wiring hole going out through the wall did get temporarily sealed up in an easily removable way, as I intended to eventually put *some* kind of light outside and feed the power out to it. What I wanted was something like a typical motion-detector outdoor light, but also a way to utilize the sensor to alert me inside whenever it triggered on. I also wanted better control over the light itself, since there's no point in for example turning it on when someone comes to the door in the daytime. A few products had entered the home-improvement market that came close to the functionality I wanted, but none of them quite fit the bill.

Parts of motion-detector light/alert design Besides, in theory I already had all the components of a simple but solid system, including a trashpicked infrared-detector outdoor light and my old doorbell parts. The doobell hadn't been functional for years anyway and its wiring through the basement was long since history, but the "dinger" was still on the wall in the hallway and the little electrical box with the transformer had come off the fusebox during the upgrade. Why not put those pieces back together? I would just have to figure out how to tap into the detector trigger output and fire the doorbell briefly, and it would be totally appropriate.
After thrashing around with tentative designs involving relays and timing circuits and whatever else, I decided the easiest way to go was use the AC output to the light to also power the doorbell transformer for some small time interval. Conveniently, the original light system had the sensor module and the output triac separated inside with a simple three-wire interface between them -- to keep AC high voltage out of the sensor electronics and the small wiring going to it, presumably. I would just need a longer wire between them and to bring the AC module to the indoor side of the wall where I'd have control over its output.

I anticipated four possible states:

  • Completely off (controlled by existing wall switch for light)
  • Detector enabled, signal, but light stays off
  • Detector enabled, signal and light both go on together
  • Light continuously on regardless of detector
so I needed to be able to select any of the latter three modes in the circuitry.

The required routing and switching would actually fit inside the old fuse-box for the doorbell transformer, so that seemed like an elegant solution to work toward. A bit of rooting around in junkboxes turned up a rotary switch that could do three positions and some appropriate wire to connect everything up. This would go through a new connector in the transformer box. The initial alert idea was some kind of passive time-delay circuit with a rectifier and large cap to power the doorbell only momentarily, but ultimately that didn't really work out.

The doorbell sounder elements are basically steel bars like a xylophone, mounted on loose rubber grommets next to resonance tubes tuned to roughly the same frequencies of the "ding" and the "dong". Those actually do help amplify the sound a little. For this purpose I only really wanted the initial "ding" and replaced the larger bell-bar with a small metal bracket for the solenoid to return against.

Before being put back together the detector module got an LED added to shine through the lens and show when it was actually seeing movement, so I could diagnose the pattern when walking around in front of it. That required a bit of digging into its electronics, but there seemed to be a convenient couple of spare open holes in the circuit board right about where an LED and a series resistor would connect -- probably for the purpose, although the original unit didn't come with any indicator. The circuit appears very similar to the one detailed here [PDF, 143K] on page 9, but sans relay output.

It also needed a photosensor bypassed inside which would otherwise disable triggering entirely if it was light outside. Sensible if it was only driving a light, but not what I needed as I wanted the motion signal to always be active. Some modules don't have that feature, such as whatever some neighbors down the road have; their driveway light pops on any time someone passes by. That was kind of hilarious during all the road construction.

Outdoor light installation w/ rain shroud First order of business was to reassemble and mount the light itself. This was shot a little after the whole thing was done, and shows the little rain shroud added on top and flashed into the light block. One of the original two light sockets was also eliminated, as I only really needed one to aim roughly down the steps toward where the car usually sits. This bulb is one of the nice LED spot type PAR 38 replacements, with a single emitter at the middle which looks more "dazzling" when viewed than a broad face and is visually a bit more of a surprise when it pops on in the dark. It also works way better than a CFL in the cold. Part of the idea is a deterrent for anyone messing around in the driveway, of course, and its beam is still narrow enough to not spill too much light toward the neighbors. Some screening was also added over the installed lamp with a small hole cut for the output, screwed in around the outside to discourage casual removal.

Soldering the connector I could fish wires through the hole in the wall but not any sort of suitable connector without significantly enlarging it, so I had to get up here and attach the connector pins after the wires were through and foam-sealed at the outside. I used one of the 9-pin circular ones like in the wheel speed hack for the car, the male side of which could nicely chassis-mount in the top of the box. It carries the sensor interface in, and the lamp AC back out.

In the "mighty have fallen" department, yeah, the generic reading-glasses made this a lot easier. Kind of ironic to think that I used to work on small mechanical watches by naked eye when I was a kid...

Quick lashup for testing After installing the bracket onto the solid wood of the joist and getting the box mostly wired up, I hung the whole mess temporarily in place and connected power with alligator clips for a quick smoke test.

A mild concern was how the triac module would react when driving the inductive load of the doorbell transformer, but in this case it's just full-on or full-off rather than a more iffy scenario like trying to drive a wall-wart supply from a lighting dimmer. It seemed fine.

Light control box open Ready to install for real, and here we sort of see what went inside the box. The triac module is tucked against the right side and most of the remaining space is taken up by AC wiring. The lamp output goes through the rotary switch at the bottom to select the operation mode, and the incoming line goes through a low-current inline fuse for extra safety. With the lamp and transformer drawing all of maybe 25 watts together, it doesn't need much.

Light and doorbell rig done Done! The box tucked neatly up against the ceiling with a small plate of white coil stock above to close up the rest of the hole, and the old doorbell could basically mount anywhere. I can just reach up to the rotary switch knob to change the mode, usually at night or morning to enable or disable the light. The wire between the components did get prettied up a little later on, tucked up against the molding.

It took a little fiddling to aim the sensor and and get the sensitivity figured out, but on average it seems to sense bodies moving around in the driveway fairly well. I say "on average" because it's far less sensitive to human body temperature on a warm day, and someone really bundled up in the winter won't be seen at longer range because they're insulated. A motorcycle going by way out at the street with an exposed hot engine, on the other hand, can easily kick it from that far away. All depends on the delta. Scattered clouds moving past the sun can have the thing falsing all over the place, as can shadows of wind-blown trees at certain times, and there are some mornings I have to just turn the whole thing off.

A bit of fooling around with the doorbell itself ensued; the proposed passive one-shot-pulse circuit via charging a capacitor through the solenoid was a loss as it apparently needed a much healthier kick to fire hard enough to ding. Eventually I simply wired the transformer direct to the bell like its original design and just let the solenoid sit there and buzz while the sensor output is on. That's a little outside the original design expectation that its usual duty would be momentary; it gets a little warm but not in any hazardous way to worry about for the minute or less that it stays powered.

The LED indicator also got clipped later on as warmer weather arrived and some subtle thermal characteristic of the circuit changed. It turned out that the modest current to light the LED was loading the tiny sensor-head power supply down enough to throw off its internal thresholds and make the whole thing go "stuck on" braindead. That was okay and simple to do; I already had a good enough idea of the coverage pattern.

Below the doorbell are some of the other whiteboard doodles about heat flow, likening the thermal characteristics of various substances to tanks of water with different size holes in and out. Brick and stone, for example, have high heat *capacity* but aren't a very good insulator, meaning that changes in delta take longer to propagate through it but once a gradient is established, it won't resist the net steady-state flow very well. It's good for averaging out large swings such as in adobe walls in desert environments, keeping interiors more comfortable by simple time delay of the mass. Wood-frame walls with no insulation just plain leak with almost no reserve capacity, thus represented by a much smaller "tank" that fills up quickly and spills over. Polyiso itself doesn't have much inherent specific heat capacity but presents a very tiny leak against a delta, whatever it is. Although word on the building-science street is that its effective leak orifice ironically gets a little bigger when the stuff is really cold, giving less R-value. At single-digit temps I think I was seeing some of that effect, but fortunately nights like that are relatively rare around here.

_H*   140616