Class 8 · Hum Hunting and Scope Work

A meter tells you a number. A scope tells you a shape, and the shape is where the diagnosis is.

The oscilloscope: the first instrument that shows you time#

A meter gives you one number, averaged over a period. A scope draws voltage against time, and almost every problem in this course is a problem in time: a spike, a wrong frequency, a signal that is present but the wrong shape, a bounce, a reflection.

The four controls that matter#

Volts per division. Vertical scale. Each grid square is worth this much voltage. Set it so the waveform fills most of the screen: too small and you cannot see detail, too large and it clips off the top and you may not realise.

Time per division. Horizontal scale. Each square is worth this much time. This is the control that turns a scope from confusing to useful, because setting it right requires you to have an expectation of the frequency, which is exactly the prediction discipline from Class 2.

Trigger level and slope. The scope starts drawing when the signal crosses a chosen voltage going in a chosen direction. This is what makes a repeating waveform stand still on screen.

The four controls that matter#

Coupling: AC or DC. DC coupling shows everything, including the constant offset. AC coupling blocks the DC and shows only the variation. This is the control that finds 20 mV of ripple sitting on a 12 V rail: on DC coupling at 5 V per division you would never see it, and on AC coupling at 10 mV per division it is unmistakable.

Triggering, which is where beginners give up#

Triggering, which is where beginners give up#

An untriggered trace slides across the screen or looks like a blur. The instinct is to fiddle with the timebase. The fix is almost always the trigger.

The scope needs to know when to start each sweep. Set the trigger source to the channel with the signal, set the level to somewhere in the middle of that signal's amplitude, and it will lock.

The special case worth knowing now: single shot. For an event that happens once — a switch bouncing, a flyback spike, a start-up transient — set the scope to trigger once and stop. That is how you captured the flyback in Class 5, and it is how you will catch an intermittent fault.

Probes, and the mistake everybody makes once#

A ×10 probe divides the signal by ten and presents a much lighter load to the circuit. Most probes have a switch. If the scope does not know which setting you are on, every voltage it reports is wrong by a factor of ten.

Check the probe setting before you believe any number. Then compensate the probe on the scope's calibration output: a square wave with rounded or overshooting corners means the probe's compensation is wrong, and every measurement you take will have the same distortion baked in.

And the safety one: a standard scope probe's ground clip is connected to the scope's chassis, which is connected to mains earth. Clipping it onto anything that is not at earth potential creates a short circuit through the scope. Never put a standard probe's ground clip on a live mains conductor. Mains measurements need a differential probe or an isolation transformer, and in this class the technician does them.

What things look like#

Learning to recognise waveforms is most of the diagnostic skill. Here is the catalogue you need.

What things look like#

Hum versus buzz, and why the difference matters#

Hum is a fundamental at 50 or 60 Hz with few harmonics: a smooth, low tone. It points at magnetic induction from a transformer or a mains cable, or at a ground loop with a clean supply.

Buzz is rich in harmonics: harsher, and it extends much higher in frequency. It points at something chopping the waveform. In a theatre that is almost always a phase-control dimmer, or a switch-mode supply, or a bad connection arcing.

100 or 120 Hz — double the mains frequency — is the signature of a rectifier. It means a power supply, and usually a failing smoothing capacitor, which is one of the most common faults in old equipment and one you can now diagnose by ear and confirm on a scope in thirty seconds.

Being able to say "that is 100 Hz, not 50, so it is a power supply and not a ground loop" in the first minute of a problem is the difference between a ten minute fix and an evening.

A method for hunting hum#

The method matters more than the knowledge. Without one you swap cables until it goes away, learn nothing, and it comes back next week.

Step 1 — Characterise before you change anything#

Five questions, answered before a single cable is touched:

1. Hum or buzz? Which tells you induction and loops, or chopping and switching.
2. Does it change with the channel fader? If yes, it enters before that fader. If no, it enters after. This one question splits the system in half immediately.
3. Is it there with the input unplugged? If yes, the problem is inside the device or its power. If no, it comes in on the cable or from the source.
4. Does touching the connector shell change it? That is a screen or earth continuity problem.
5. Does it change when the lighting state changes? That is a dimmer, and now you are looking at a shared supply or a cable route rather than at the audio system.

Write the five answers down. They will exclude most of the possibilities before you move.

Step 2 — Half-split#

You have a chain: source, cable, preamp, processor, amplifier, speaker. Do not start at one end. Test in the middle, and you halve the problem with each test.

Six devices found by half-splitting takes three tests. Found by working along the chain, it takes an average of three and a worst case of six, and it feels like guessing because it is. On a twenty-device system, half-splitting is five tests and sequential searching is up to twenty.

This is the most transferable idea in the whole course. It applies to a DMX line, a network, a truss circuit and a piece of software.

Step 3 — Change one thing at a time, and put it back#

The rule that makes a technical rehearsal survivable. Every change is written down and reversed if it did not help. A system where four things were changed and the problem went away is a system that will do it again, because you do not know which change mattered.

Step 4 — Prove the fix#

Reproduce the fault deliberately after fixing it. If you cannot make it come back by undoing your fix, you did not find the cause, you found a coincidence.

This step is skipped almost universally and it is the one that separates a repair from a hope.

Measuring a frequency response#

A scope shows you one frequency at a time. To ask "what does this box do to everything" you sweep a generator across the band and record the output at each point.

The method, which takes about ten minutes once and answers a question that would otherwise be an argument:

1. Feed a sine at 1 kHz and set the level so the output is comfortably on screen. That is your 0 dB reference, and everything else is relative to it.
2. Change frequency without touching the level. Record the output amplitude at 20, 50, 100, 200, 500 Hz, 1, 2, 5, 10, 20 kHz. Ten points is enough to see the shape.
3. Convert each to dB relative to the reference: 20 × log10(V ÷ V_ref).
4. Plot it on a logarithmic frequency axis, because that is how hearing works and because a filter is a straight line on those axes and a curve on linear ones.

What you are looking for:

Measuring a frequency response#

At the bench#

Two hours and fifty minutes.

Block A — Learn the four controls (35 min)#

On a signal generator: get a stable 1 kHz sine on screen and measure its amplitude and period from the graticule, by hand, before reading the scope's automatic measurements. Then confirm with the automatic ones.

Change to 100 Hz and 10 kHz and set up each from scratch. Then square, then triangle. Then find the compensation output and check your probe.

Block B — The waveform zoo (35 min)#

Nine stations, each with a signal. Identify each one, sketch it, and write what it would mean if you found it in a real system. The list above is the answer key and you should not need it.

Station 7 is mains on a differential probe, and you will finally see the peak that Class 1 told you about.

Block C — Ripple hunting (30 min)#

Three DC power supplies, one healthy, one with a tired smoothing capacitor, one switch-mode.

Measure each on DC coupling first: all three read about 12 V and all three look identical. Then AC coupling at 20 mV per division and they become three completely different objects. Measure the ripple amplitude and frequency of each, and identify which is which from the frequency alone.

Block D — The hum hunt (60 min)#

A deliberately built system with four faults in it: a ground loop, an unbalanced connection where a balanced one belongs, a cable with an open screen, and a dimmer sharing a supply.

Work in pairs. Use the five questions, half-split, and change one thing at a time. Log every test and its result. Find all four.

The marking is on the log, not on the time. A pair who found three faults with a clear log has done better work than a pair who found four by swapping cables at random, because only one of those two methods works on a fault nobody has seen before.

Block E — Prove it (10 min)#

Put each fault back, one at a time, and confirm the symptom returns. Anything you cannot reproduce goes back on the list as unproven.

Numbers from this class#