Build a Triggered Effect

Sensor, decision, actuator, on a cue, eighty times, with a defined behaviour when somebody pulls the plug halfway through.

The brief#

In teams of two, build a triggered effect that:

• Is triggered by a sensor and by a cue, and behaves sensibly if both arrive at once
• Drives a real actuator through your Class 6 driver board
• Has a defined and documented power-up state
• Detects loss of control and does something you decided in advance
• Detects an out-of-range sensor reading and treats it as a fault
• Can be reset by one documented physical action
• Shows its state on an indicator readable from across the room
• Comes with a one-page sheet somebody else can operate it from

Choose from: a triggered practical lamp that flares on an actor's arrival, a door latch released on a cue, a triggered smoke or fan burst, a flickering fire effect that responds to proximity, a mechanical reveal on a limit-switched slide.

The effect is not the assessment. The failure behaviour and the handover are the assessment.

Designing the chain#

Four questions, in this order, before a wire is stripped.

1. What does the audience see, and how tightly is it timed?

A flare that must land on a word needs the whole chain under about 50 ms. A slow reveal has seconds of slack. That number decides everything downstream, including whether you can afford a network hop and how much debounce you can spend.

2. What is the trigger, and what fools it?

Write down three things that would falsely trigger your sensor. For a pressure mat: a dropped prop, a stagehand crossing, a piece of scenery resting on it. For a beam: a cable swinging, a costume, a haze effect.

Then decide, for each, whether you prevent it, detect it, or accept it. All three are legitimate. Not deciding is not.

3. What is the actuator, and what does it do while it is not being asked to do anything?

A solenoid holding is heating. A servo holding is buzzing and drawing current. A motor stopped on a slope is being held by friction alone. The idle state is where the run-length failures live, and it is the question nobody asks in a workshop because the workshop is twenty minutes long.

4. What happens between the cue and the movement, and what if the cue arrives twice?

The double-tap is not an edge case. An operator who is unsure whether a cue fired will press again, every time, on every production, forever. Decide now: ignore, restart, or queue.

Cue and sensor together#

Two ways in, one effect. Four combinations, and all four need an answer:

The middle two rows are the design. Three common patterns:

Cue arms, sensor fires. The desk arms the effect during the right scene; the sensor triggers the exact moment. The best of both: the operator retains control of whether, the sensor controls when. This is the pattern to reach for first.

Either fires it. Simple, and it means a false sensor trigger fires the effect in the wrong scene. Only acceptable when a spurious fire is harmless.

Both required within a window. Sensor within two seconds of the cue. Robust against both false triggers and mistimed cues, and it fails silently when they do not coincide, so it needs an indicator that shows you missed rather than pretending nothing was asked for.

Choose one and write down why. In the handover exercise, another team will ask.

The build discipline#

Same as Class 6 and Class 12, because it is the same discipline everywhere:

1. Skeleton first. State machine, indicator, timing. No load. Prove every transition on the indicator alone.
2. Then inputs. Debounced, range-checked, with the fault path proved by shorting the input to both rails.
3. Then the output, on a lamp, not the real actuator. A logic error that fires a lamp is a lesson; the same error firing a solenoid at somebody's hand is an incident.
4. Then the real actuator, with the current measured and the temperature checked after four minutes of realistic duty.
5. Then break it, on purpose, in every way in your failure table.
6. Then write the sheet, and hand it over.

Nobody skips straight to step 4 twice. They skip it once, and then they have a story.

Testing for a run, not for a demo#

A demo is one operation with everybody watching. A run is eighty operations over six weeks with nobody watching. Four tests that find run failures and that a demo never will:

The hundred cycle test. Run the effect a hundred times, automatically, and watch what drifts. Solenoid temperature, servo position, stepper count, contact resistance. Log the first, fiftieth and hundredth, and if anything has moved, you have found a run failure in twenty minutes instead of in week three.

The thermal test. Run it at the worst realistic duty for fifteen minutes and measure everything that could get warm: the actuator, the MOSFET, the regulator, the connector. Anything you cannot hold your finger on is a question that needs an answer, and the answer must be a number.

The interruption test. Cut power mid-operation. Cut it at three different points in the cycle. Restore it. Does it come up where you said it would, every time? This is the test that catches the prop that resumes mid-flare.

The stranger test. Somebody who has not seen it, given only your sheet, must be able to power it up, trigger it, cause a fault, identify the fault from the indicator, and reset it. If they cannot, the sheet is wrong and the sheet is what goes on the production.

Joining the show#

An effect that only fires from its own button is a demonstration. An effect on a production is fired by whoever is running the show, from wherever they are running it, and the way you connect it decides who is responsible when it does not happen.

The question is not technical, it is about who is responsible. Putting the effect in the lighting cue stack means the lighting operator can see it, hold it, and take it out if the scene changes. Putting it on a sensor with no override means nobody can stop it, which is fine for a flickering lamp and unacceptable for anything that moves.

Whatever you choose, provide an override. A physical way to disable the effect that does not involve unplugging it, labelled, and on the documentation sheet. On the night something is wrong with your prop, somebody needs to be able to take it out of the show in five seconds without finding you.

The pattern worth reaching for, from earlier in this class: the cue arms it and the sensor fires it. The operator keeps the decision about whether; the sensor keeps the decision about when.

At the bench#

Three hours and twenty minutes. Continuous build.

Block A — Design review (25 min)#

Present your state machine and failure table to another team in five minutes. They ask three questions. You answer, or you change the design.

The question that will be asked and that most teams have not answered: what happens if the cue arrives twice?

Block B — Skeleton and inputs (55 min)#

Steps 1 and 2 above. Checkpoint: another team must be able to drive your state machine through every state using only the indicator to see where it is.

Block C — Output on a lamp, then the real actuator (60 min)#

Steps 3 and 4. Measure the actuator current. Measure the MOSFET temperature after four minutes. Note both.

Block D — Break it (40 min)#

Your whole failure table, plus the four run tests. The hundred cycle test runs while you do the others.

Log everything. Behaviours you did not predict are the most valuable output of today.

Block E — Handover (20 min)#

Write the sheet. Swap with another team. Operate theirs from the sheet alone, in silence, and write down every moment you had to guess.

Give the notes back. That exchange is the assessed artefact, and the notes you receive are worth more than the marks.

Numbers from this class#

Self test with model answers

Answer it yourself first, out loud or on paper, then open the answer. Reading the answer without attempting the question teaches you almost nothing.

1. 1Why test the logic on a lamp before connecting the real actuator?
2. 2Describe the hundred cycle test and name three things it finds that a demonstration does not.
3. 3What is the stranger test and why is it the assessed part?
4. 4Compare the three trigger patterns and say which you would reach for first.
5. 5"It fails safe, so a failure does not matter." What is wrong with that?

Note. These are model answers, not the only correct ones. If yours differs and you can defend it with the arithmetic and a measurement, that is worth more than matching the wording.