Build and Break a Rig

Anybody can build a rig that works. Today you build one, then break it fourteen times, and learn what each failure looks like from the far end of an auditorium.

Why breaking things is the lesson#

You will spend your career diagnosing systems you did not build, under time pressure, with somebody standing next to you asking when it will be fixed.

The only way to be good at that is to have a library of symptoms. Not knowledge of how DMX works — you have that from Class 9 — but a memory of what each specific failure actually looks like from the auditorium.

Today builds that library deliberately, in four hours, instead of accidentally over five years.

The rig you are building#

Per team of four:

• A console or a laptop with a DMX output
• An isolated DMX splitter, one in, four out
• Three runs of different lengths, one deliberately long
• Six fixtures with settable addresses, of at least two different types
• A network switch, a node, and one universe of sACN feeding two of the fixtures
• Terminators, spares, and a labelling machine

Build it properly, because you will be marked on the drawing#

The build is not the hard part. The documentation is, and it is where the marks are.

Before the rig is powered, produce a signal flow drawing that shows:

• Every device, with its address, its universe and its channel footprint
• Every cable, with its length and its label
• Where the terminators are
• Where the isolation barriers are
• Which device is at the end of each run

Then power it up and correct the drawing, because it will be wrong. The drawing that matches reality is the deliverable, not the drawing you made first.

Labelling, which is not administration#

Every cable gets a label at both ends. Every fixture gets its address written where it can be read from the ground. Every universe gets a colour.

The test: could somebody who has never seen this rig find the third fixture on run two, in the dark, in ninety seconds? If not, the labelling has failed, and this is exactly the situation a production electrician is in at 22:00 during a technical rehearsal.

The fourteen faults#

Once the rig works and the drawing matches, the technician introduces faults, one at a time, while your team is out of the room. Your job is to find each one and, more importantly, to log the symptom before you find the cause.

The list#

The method, which is the same one as Class 8#

1. Characterise before touching. What is affected, what is not, and where is the boundary between them? The boundary is where the fault is.
2. Half-split. Test in the middle of the affected section. A rig with sixteen possible failure points is four tests, not sixteen.
3. One change at a time, written down, reversed if it did not help.
4. Prove the fix by putting the fault back and confirming the symptom returns.

For each of the fourteen, log: the symptom in a stage manager's words, the tests you ran in order, the cause, and the fix. That log is the assessed artefact.

The three that teach the most#

Fault 6, the open screen. It works perfectly until the dimmers come up, because with no screen the data pair has no shield and induced noise from the dimmer harmonics corrupts it. The lesson is that a fault can be dormant and be triggered by something in an entirely different department. If your notes say "it only happens during the ballroom scene", that is diagnostic information, not a complaint.

Fault 7, the passive star. Some branches work and others do not, and which ones changes when you touch things. It looks like a haunting. It is impedance: a passive split creates multiple reflection paths, and the pattern of constructive and destructive interference depends on the exact lengths. The only fix is an active splitter, and this is why splitters exist and why a "DMX Y-splitter" bought online is a device for creating intermittent faults.

Fault 13, the personality mode. The channels arrive correctly and the fixture does the wrong things with them, because the console thinks it is in 16-channel mode and the fixture is in 8-channel mode. Nothing electrical is wrong. No test on the cable will find it. It is found by comparing the console's patch to the fixture's own display, which is a step people skip because it is not electrical.

The network half#

Faults 10, 11 and 14 are network faults, and they behave differently from DMX faults in one important way: the network will tell you things, if you ask it.

A DMX line is silent. A network has link lights, switch statistics, IP configuration and packet counters. Three checks, in order, before anything else:

1. Link. Is the light on at both ends, and at the right speed? A link light that is on at 100 Mbit/s where you expect 1 Gbit/s often means a damaged cable using only two pairs.
2. Address. Are the console, the node and the switch on the same subnet? A node showing an address of 2.x.x.x when the console is on 10.x.x.x cannot receive anything, and the node's own display will look completely healthy.
3. Traffic. Is the node receiving packets at all? Most nodes have a data indicator. If it is dark, the problem is upstream of the node. If it is lit and the fixtures are dark, the problem is downstream, and you are back to a DMX fault.

Those three questions split the network half of the rig into thirds in under a minute, and the sibling module on computer systems and networking goes considerably deeper on all three.

Establishing the boundary#

Every method in this course begins with the same move, and it is the one people skip when they are in a hurry. Before touching anything: what works, what does not, and where is the line between them? The fault is on that line.

On a rig, the boundary questions are concrete:

The last one is the one nobody asks, and it is how fault 6 gets found. "It only happens during the ballroom scene" is not a complaint, it is a measurement.

Write the boundary down before you move. It takes fifteen seconds, and it stops the thing that actually wastes the evening: forgetting what you have already excluded, and testing it again.

The drawing, and what makes one good#

A lighting plan says where fixtures hang. A signal flow drawing says how data reaches them, and they are different documents answering different questions. The second is the one that matters at 22:00.

A signal flow drawing that works carries: every device with its address, universe and channel footprint; every cable with its length and its label; where the terminators are; where the isolation barriers are; and which device is at the end of each run.

Three properties separate a drawing that gets used from one that gets ignored:

It matches reality. The drawing you made before the build is a plan. The one you corrected after walking the rig is a drawing. Only the second is a deliverable.

Somebody else can read it. Get it signed off by another team before you power up. A drawing only its author understands is a private note.

It answers the question you will actually have. Not "what does this rig look like" but "which fixture is third on run two, and what is its address". If the drawing cannot answer that in ten seconds, it is decorative.

At the bench#

Three hours and twenty minutes, all of it hands-on.

Block A — Build and document (60 min)#

Build the rig. Label everything. Produce the drawing. Get the drawing signed off by another team before you power up, because a drawing that only its author can read is not a drawing.

Block B — Prove it (25 min)#

Every fixture responds, on the right channel, at the right address, in the right universe. Scope the far end of each run and confirm the signal is clean. Photograph the clean trace: you need it for comparison in Block C.

Block C — Fourteen faults (100 min)#

The technician introduces faults one at a time. Teams rotate so you meet all fourteen.

Log everything. Approximately seven minutes per fault, which is tight, and that is intentional: methodical is faster than clever under time pressure, and this is where you find that out.

Block D — Write the report (15 min)#

One page per team: the three faults that took longest, why, and what test would have found each one faster. That last column is the whole point of the class.

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. 1A fixture in the middle of a run flickers intermittently. Give the first three things you would check and why in that order.
2. 2Why does an open screen produce a fault that only appears during certain scenes?
3. 3What makes a passive DMX Y-splitter so much worse than no splitter at all?
4. 4A node’s display looks completely healthy and no fixtures respond. What is the most likely cause?
5. 5Why is the rig drawing the deliverable rather than the rig?

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.