Loads, Switching and Isolation

A 5 volt thought has to become a 230 volt action. Everything in this class is about the boundary between those two worlds, and about making sure neither one can reach across it.

The problem this class solves#

A microcontroller pin can supply about 20 milliamps at 3.3 or 5 volts. That is a fortieth of a watt. A stage practical, a solenoid, a smoke machine or a motor wants somewhere between ten watts and three kilowatts.

Between those two numbers sits every piece of show automation ever built. The bridge has three jobs, and confusing them is the source of most home-built show electronics failures:

1. Amplification. A small signal controls a large current.
2. Protection. The large side must not be able to destroy the small side, ever, including during a fault.
3. Isolation. In many cases the two sides must share no electrical connection at all, so that a mains fault cannot appear on a control desk that a person is touching.

Relays: switching by moving metal#

A relay is an electromagnet that physically pulls a contact closed. It is the oldest solution and still the right one more often than people expect.

What it gives you. Genuine galvanic isolation between coil and contacts, typically rated for several kilovolts. It switches AC or DC equally well. It has essentially zero on-resistance, so it dissipates almost nothing while closed. And it is obvious: you can hear it, and you can see whether it has operated.

What it costs you. It is slow, five to fifteen milliseconds, which rules it out for anything that needs to be fast or silent. It wears out mechanically, typically after tens of millions of operations, and electrically far sooner if it is switching an arc. It clicks, which matters enormously in a quiet auditorium. And the coil is an inductor, so it kicks.

The flyback diode, and why it is not optional#

When you interrupt current through a coil, the collapsing magnetic field drives the voltage as high as it needs to go to maintain that current. Hundreds of volts, from a 12 volt supply, in microseconds.

A diode across the coil, reverse biased in normal operation, gives that current a loop to circulate in until it decays. It costs two cents.

Without it: the transistor driving the coil dies, sometimes immediately and sometimes after two hundred operations. Mechanical contacts driving the coil erode and weld. And the spike radiates, resetting microcontrollers elsewhere in the same box.

Every inductive load gets a flyback diode. Relays, solenoids, valves, motors, contactors. In Class 6 you will build a board with one and then remove it and watch the scope, which is the only way this becomes permanent knowledge.

Contact ratings, read properly#

A relay marked "10 A 250 VAC" is not a 10 A relay. It is a relay rated 10 A for resistive loads at 250 V AC. Three derating factors apply and all three matter in a theatre:

• Inductive loads — motors, transformers, solenoids — draw an inrush and produce an arc on break. Derate substantially, often to a third.
• DC is much harder than AC because there is no zero crossing to extinguish the arc. A relay rated 10 A at 250 V AC may be rated 0.5 A at 30 V DC. This surprises people constantly.
• Lamp loads have a cold filament inrush of ten to fifteen times the running current for a few milliseconds, which welds contacts over time.

Transistors and MOSFETs: switching without moving#

For DC loads, for anything that must be silent, and for anything that must be fast or dimmable, you switch with a semiconductor.

A MOSFET is a voltage-controlled switch. Put enough voltage on the gate relative to the source and the channel between drain and source becomes a low resistance. Take it away and it becomes an open circuit.

Three numbers decide whether a given MOSFET will work for you:

V_GS(th), the threshold voltage. The gate voltage at which it starts to conduct — not where it is fully on. This is the single most common mistake: a standard MOSFET has a threshold around 4 V and needs 10 V to turn on properly, so a 3.3 V microcontroller pin leaves it half on. Half on means high resistance, which means heat, which means a dead part and a scorched board. You want a logic-level MOSFET, which is what the L in IRLZ44N is telling you.

R_DS(on), the on-resistance. Typically 10 to 50 milliohms. The dissipation is I² × R_DS(on), so 5 A through 25 mΩ is 0.6 W: warm, fine, no heatsink. But if the gate is under-driven and R_DS(on) is effectively 2 Ω, the same 5 A dissipates 50 W and the part is destroyed in seconds.

Continuous drain current. Read the small print: it is quoted at a case temperature you will never achieve. Halve it and you are being realistic.

Why switching beats varying#

You could control brightness by putting a variable resistance in series with a lamp. Nobody does, because that resistance dissipates the power it removes. Dim a 100 W lamp to half by resistance and you are heating a resistor with tens of watts.

Instead you switch fully on and fully off, fast, and vary the ratio. A fully-on switch has almost no voltage across it and a fully-off switch has almost no current through it, and power is voltage times current, so both states dissipate almost nothing. All the loss is in the brief transitions.

That is pulse width modulation, it is why every modern dimmer and motor drive works this way, and Class 11 builds one.

Wire it before you wire it#

Delete the flyback diode and watch the rail. That spike is the energy that was in the motor's magnetic field, and it has to go somewhere; without the diode it goes through your transistor. This is the one to run twice.

Isolation: the part that matters most to you#

Amplification is engineering. Isolation is safety, and it is the part of this class that a technical director signs their name to.

An optocoupler is an LED and a photosensitive transistor in one package with no electrical connection between them, only light across a millimetre of transparent plastic. The control side lights the LED; the load side sees a transistor turn on. Isolation voltages of 2.5 to 5 kV are routine.

The consequence: a fault on the mains side cannot put voltage onto the control side. A performer touching a prop cannot be connected, through any single failure, to the switched circuit.

Where you must insist on isolation:

• Anywhere a control system touches mains.
• Anywhere a control cable leaves an enclosure and runs to another building or another power source.
• Anywhere the control side is touched by a person, which in a theatre is nearly everywhere.
• On DMX inputs and outputs, which is why good DMX splitters are described as "optically isolated" and why they cost more.

Isolation is defeated by a single wire. Two isolated halves connected by a shared ground for convenience are not isolated any more. This is subtle, it happens constantly in home-built systems, and you should be able to spot it on a schematic by Class 6.

Solid state relays#

An SSR combines an optocoupler and a triac or MOSFET output in one package. Isolated, silent, no moving parts, no contact wear, and fast.

Two properties to know. A zero-crossing SSR waits for the AC waveform to pass through zero before switching, which greatly reduces electrical noise and makes it unsuitable for phase-control dimming. And an SSR dissipates real power while conducting, about 1 to 1.5 V times the current, so a 10 A SSR dissipates 10 to 15 W and needs a heatsink, which people forget because it is small and looks like a relay.

Inrush, and why the breaker trips at the top of the show#

Three loads in a theatre draw enormously more current at switch-on than in normal running:

The switch-mode case is the one that bites modern rigs. Every LED fixture, every media server, every network switch has an input capacitor that looks like a short circuit for the first few milliseconds. One fixture is nothing. Forty fixtures on one circuit, all energised by the same switch at the same instant, is a very large momentary current, and the breaker sees it.

This is why the answer is often sequencing rather than a bigger breaker: bring circuits up in groups a second apart, and the peaks never coincide. It is why powered devices in a rack are switched by a sequencer, and it is a design decision that belongs to you, not to the electrician.

Dimmers: how a lamp is actually dimmed#

Everything above is switching a load on and off. Dimming is the same problem with a number attached, and the theatre industry has solved it three different ways, which is why "it does not dim properly" is such a common complaint.

Phase control, which is what a dimmer rack does#

A triac or a pair of thyristors waits some fraction of each mains half-cycle and then switches on, conducting for the rest of it. Delay the firing angle and the lamp receives less energy.

It is cheap, it handles kilowatts, and it has three consequences you will meet constantly:

It chops the waveform, so it generates harmonics far up the audio band. That is the buzz in Class 7 and Class 8, and it is why dimmer feeds and audio cables are routed separately.

It needs a choke. The sudden switch-on would otherwise produce a current edge steep enough to make the lamp filament sing audibly. The choke slows the edge, which is what the rise time figure on a dimmer's specification means, and it is why a dimmer rack is heavy.

It assumes a resistive load. A tungsten filament is happy with a chopped waveform because it integrates the energy thermally. An LED driver, a fan or a transformer is not, and connecting one to a phase-control dimmer produces flicker, buzz, or a dead driver. This is the whole technical content of "you cannot dim that on a dimmer".

Sine wave dimming#

An IGBT-based dimmer that reconstructs a genuine reduced-amplitude sine rather than chopping. Silent, harmonically clean, kind to non-resistive loads, and several times the price. It exists because opera houses and broadcast studios could not live with the first option.

Constant-current drivers, which is what LED fixtures do#

An LED fixture is not dimmed by varying its supply at all. Its internal driver holds the current and dims by PWM, exactly as in Class 11, at whatever frequency the manufacturer chose. The fixture is fed a constant mains supply and told what to do over DMX.

This is why an LED fixture must go on a relay circuit or a permanently live feed, never a dimmer way, and why patching one to a dimmer is one of the fourteen faults in Class 10.

At the bench#

Two hours and ten minutes. Low voltage. The mains work is demonstration only.

Block A — Drive a relay properly (40 min)#

Build a transistor relay driver on breadboard: signal in, base resistor, transistor, relay coil, flyback diode. Prove it switches.

Then remove the flyback diode and put the scope across the transistor. Photograph the trace. Measure the peak. Put the diode back and photograph again.

That pair of photographs goes in your notebook and it is the answer to any future argument about whether the diode matters.

Block B — MOSFET, driven well and driven badly (40 min)#

Same load, switched by a logic-level MOSFET at 5 V on the gate. Measure V_DS while on, calculate R_DS(on) from the current, and calculate the dissipation. Touch the tab: cool.

Now drive the same MOSFET's gate through a 100 kΩ resistor so it switches slowly, and repeat. Measure the temperature rise with the IR thermometer. Explain, in writing, using P = I² × R, why a slower gate makes a hotter part.

Block C — Isolation, proved (30 min)#

Build an optocoupler stage. Then prove the isolation with a meter: measure resistance between every input pin and every output pin. It should be open in every combination.

Then introduce the classic error — connect the two grounds together "for convenience" — and repeat the measurement. Write one sentence on what you have just destroyed.

Block D — Inrush, watched (20 min)#

With a current probe on the scope, capture switch-on of a tungsten lamp, a small switch-mode supply, and four switch-mode supplies together. Measure the peak and the duration of each. Then capture the four supplies switched a second apart and compare the peaks.

Dimming by chopping#

Watch the waveform rather than the LED. The eye integrates and tells you a comfortable lie; the scope does not, and a camera shutter agrees with the scope.

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 relay is marked "10 A 250 VAC". Why can it not switch 10 A of a 24 V DC motor?
2. 2Why is a MOSFET driven at 5 V from a 3.3 V-tolerant standard part the worst case rather than simply a weaker one?
3. 3How is optical isolation defeated without touching the optocoupler?
4. 4Forty LED fixtures on one switch trip the breaker at switch-on but run at 8 A. Explain, and give the engineering answer.
5. 5Why is a 10 A solid state relay not a drop-in replacement for a 10 A mechanical relay?

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.