The buzz is coming from inside the rig

Ground loops, leakage current, and the pin 1 problem: why systems hum, why the ground pin is innocent, and how to get a quiet rig without electrocuting anybody.

It's 6:40 pm and the room is finally empty. Doors are in twenty minutes. You mute every input on the desk, push the masters up to show level, and stand in the middle of the floor to listen to your PA doing absolutely nothing. Which is when you hear it. Not hum exactly. Buzz. That fizzy, angry, edge-of-the-teeth thing sitting under the room like a wasp in a jar, and it was not there at soundcheck.

Nothing on your side changed. What changed is that lighting brought their dimmers up for the walk-in look, video powered the wall, and catering plugged a coffee urn into the same wall as your amp racks, because of course they did. Someone behind you says the words you've heard on every gig since your first one: "must be a grounding problem."

Here's the thing we want to spend this week convincing you of: it almost never is. The grounding in most venues is fine. Legal, even. The buzz is not a fault. It's physics doing exactly what physics does, through a circuit you built by accident when you plugged everything together. And because so few people can name the actual mechanism, the industry has spent decades treating the symptom with folklore: lift this, cheat that, wrap the snake in ferrite beads and hope. Some of the folklore wastes your time. One piece of it kills people.

So this week in the Woodshed: where the noise actually comes from, why it gets into your audio, what balanced interfaces really do (hint: it's not what the marketing says), and how to find and fix the problem interface in minutes instead of hours. Grab a coffee from catering's urn. It's their fault anyway.

Electricity always wants to go home

Start with the part of the system everyone uses and nobody thinks about: the power itself.

Every AC circuit is a round trip. Current leaves the transformer on the hot conductor, does its work in your amplifier, and goes home on the neutral. The third pin, the safety ground, carries nothing at all in normal operation. It exists for exactly one moment in the life of a piece of gear: the day something inside breaks and the chassis goes live. On that day, the safety ground is the low-impedance path that dumps enough fault current to trip the breaker before the chassis has a chance to run the fault through the next person who touches it.

That's the whole job. It's a fireman. It sits around doing nothing for years so it can save a life once.

It matters to be specific about what "save a life" means, because the numbers are smaller than most people think. Skin resistance is the only thing standing between you and the mains, and it's not much. Around 1 mA of current through the body is a tingle. At roughly 10 mA, your muscles contract involuntarily and you may not be able to let go of the thing that's killing you. Somewhere between 50 and 100 mA across the chest, the heart loses the plot entirely and goes into fibrillation. That's a tenth of an amp. Your console's power supply moves more current than that before breakfast. This is also why a Class A GFCI trips at about 5 mA: it's set to open the circuit below the level where you can't let go.

Now connect this to the oldest, worst piece of folklore in our trade: the ground lift adapter. The gray three-to-two cheater plug, snapped onto an amp's power cord because the system hummed and lifting the ground made the hum stop. It does often make the hum stop, which is exactly what makes it dangerous. The hum stops because you've broken the loop the noise current was flowing around. But you've also disconnected the fireman. If that amp ever develops a fault, the chassis floats up toward line voltage, and the path to ground becomes whatever touches it next. A guitar string, say. Connected to a mouth.

There's no nuance coming later in this piece that walks this back. Signal grounds are ours to manage; the AC safety ground is not. Never lift it. Every legitimate fix for hum, and there are several good ones coming, works with all three pins in the wall.

Every chassis leaks a little

If the grounding is fine, where does the noise current come from in the first place? Answer: from inside the gear, and it's supposed to be there.

Every piece of AC-powered equipment has a power transformer, and every transformer has capacitance between its primary and secondary windings. Nobody draws it on the schematic, but it's there, the same way there's capacitance between any two conductors near each other. Anything with a switching supply or digital circuitry adds an EMI filter, which deliberately places capacitors from line to chassis to keep the box's own hash out of the mains. All of that capacitance forms a path, and a small AC current flows through it from the power line to the chassis of every single device in your rack. It's called leakage current, it's unavoidable, and it's regulated rather than forbidden: UL allows up to 0.75 mA for double-insulated two-prong gear and up to 5 mA for grounded three-prong gear.

On grounded equipment that leakage flows harmlessly down the green wire. But it's not the only current in the building's ground system, and here's the fact that reframes the whole subject: a fully code-compliant, properly installed electrical system will still show small voltage differences between the safety ground pins of different outlets. Millivolts between adjacent outlets on the same branch. Up to a volt or more between distant outlets on different branches, thanks to voltage drops in the wiring and magnetic induction from the load currents flowing in adjacent conductors. Nothing is broken. Nothing is "improperly grounded." An electrician with a meter will find nothing wrong, because nothing is wrong.

So the picture before you've plugged in a single XLR is this: every chassis in the building sits at a slightly different AC voltage, each leaking a little current, all of it legal. Then you show up with a truck full of cable and connect all those chassis together through your signal grounds. Console to amp rack through the snake shield. Amp rack to video through an HDMI shell. Video to the lighting desk through a DMX drop somebody grounded at both ends in 2011 and nobody has checked since.

Each of those connections closes a loop: out through the building ground, back through your cable shield. And a loop with a voltage difference around it circulates current. That's all a ground loop is. Not a wiring mistake. Not an electrician's sin. Just the inevitable consequence of interconnecting equipment that was legally leaking before you arrived. The loop itself is even harmless; nothing about current flowing in a shield is audible by magic. It becomes audible through one specific mechanism, and we'll get there in a minute. First, a short detour into what the noise actually sounds like, because your ears are the best diagnostic tool you own.

Hum is 60 Hz. Buzz is its angry relatives

Close your eyes at front of house and the noise will tell you what it is before your analyzer does.

Pure hum, the low smooth woody one, is the line frequency itself: 60 Hz here, 50 Hz for our European readers, who at least get to suffer at a slightly more musical pitch. Hum that's mostly fundamental usually points at magnetic induction, a transformer field getting into something with too much loop area, or serious current at line frequency flowing where it shouldn't.

Buzz is different, and buzz is what you'll meet far more often. Buzz is the harmonics. The AC line in a real building is not a clean sine wave; every rectifier-fed power supply in the place gulps its current in short pulses at the crests of the waveform, which flattens the tops of the sine. Flat-topping generates almost entirely odd-order harmonics: 180 Hz, 300 Hz, 420 Hz, marching up the spectrum with energy that stays annoying well past where hum stops being a pitch and starts being a texture. The leakage current we just talked about flows through capacitance, and a capacitor's impedance falls as frequency rises, so leakage paths actually favor the harmonics over the fundamental. That's why leakage-driven noise buzzes instead of hums, and why there's measurable junk in a leakage spectrum all the way up past 100 kHz, into AM radio territory.

Then there's the aggravated buzz that shows up when the walk-in look comes up, and here we can finally point a finger at our colleagues in lighting with scientific confidence. A phase-control dimmer works by waiting partway through each half-cycle of the mains and then slamming the load on. That edge, delivered 120 times a second, is a broadband transient generator wired directly into the building. At those frequencies the building's wiring stops behaving like wiring and starts behaving like a badly terminated transmission line, reflecting the hash back and forth until it's absorbed or radiated into the nearest thing with an input jack.

The diagnostic value is real. Smooth low hum: think magnetic coupling, think loop area, think about what's physically near what. Edgy buzz that tracks the house lights: leakage current and dimmer hash riding a ground loop. Buzz with a fizzy top that comes and goes with the video wall: same mechanism, different perpetrator. The ear narrows the suspect list before you've unplugged anything.

The noise that never touches a wire

Before we get to the main event, a word about the couplings that don't need a ground loop at all, because they're the ones people misdiagnose as ground loops constantly.

Magnetic induction is the honest one. Any conductor carrying current wraps itself in a magnetic field, and any circuit sitting in a changing magnetic field has a voltage induced around it. The induced voltage scales with the area of the loop the circuit encloses, which is why the classic victims are things with big loops: an unbalanced cable and its return path splayed apart, a guitar pickup, the summing bus of a console parked on top of its own power supply. The classic aggressors are things with concentrated fields: power transformers, especially cheap ones with generous stray flux, wall warts, and feeder cable carrying serious current. The tell is that the noise is dominated by the line fundamental, that smooth low hum, and that it changes when you physically move things. That last bit is the free diagnostic. If sliding the offending box thirty centimeters changes the hum, no amount of cable swapping was ever going to fix it, because the noise was arriving through the air.

The countermeasures are geometric, which offends people who wanted to buy something. Reduce loop area: keep send and return together, which twisted pairs do automatically and which is most of the reason balanced cable is twisted at all. Increase distance: field strength from a transformer falls off fast, and half a meter of separation routinely beats fifty dollars of copper. Cross at right angles: a signal cable crossing a power run perpendicular links almost none of its flux, while one lounging in parallel alongside it for ten meters links all of it. Every veteran who lays snake across feeder at ninety degrees is doing electromagnetics, whether or not they'd put it that way.

Electric-field coupling is the other one: capacitive, high-impedance, fizzy rather than smooth. A nearby mains conductor couples through the air into any high-impedance point, which is why an unterminated input turns into a 60-cycle theremin when you touch the jack. Shields actually solve this one completely, which is convenient, because it means a shielded cable that still hums is telling you the problem is one of the other mechanisms. The shield did its job. Something else didn't.

Keep these two filed separately from everything that follows, because the treatments don't transfer. No isolation transformer fixes induction into a big loop, and no amount of re-routing fixes a genuine ground loop. Ear first: smooth hum that moves when gear moves is a field problem. Buzz that appears when systems interconnect is a current problem. The rest of this piece is about the current problem.

The loop nobody wired on purpose

Here's the mechanism that turns circulating shield current into something you can hear, and it's worth slowing down for because once you see it, half the folklore in this industry evaporates.

The mechanism is called common impedance coupling, and it needs only Ohm's law. Take an ordinary unbalanced connection, a DJ mixer into a powered speaker with an RCA-to-quarter-inch cable, say. The signal travels on the center conductor and returns on the shield. But the shield is also the connection between the two chassis, which means it's also the path that ground loop current uses. Two currents, one conductor. The shield has resistance, maybe an ohm for a long cheap cable, and the noise current flowing through that resistance drops a small voltage across it. The receiver can't tell that voltage from signal, because it appears in series with the signal. The shield's impedance is "common" to both the signal circuit and the noise circuit. Hence the name, and hence the hum.

Run the numbers, because they're humbling. A 25-foot cable with a foil shield and a thin drain wire has about 1 ohm of shield resistance. Push a thoroughly ordinary 316 microamps of leakage current through it and you drop 316 microvolts of noise. Against consumer line level, which is 316 millivolts, that's noise just 60 dB below your program. A 60 dB signal-to-noise ratio was acceptable on cassette. On a modern rig it's a complaint, and on a quiet stage during the acoustic number it's a talkback from the client.

Notice what the math tells you about fixing it. The noise is I times R. You usually can't do much about I, the leakage current; it's coming from inside everybody's gear, legally. But R is the shield resistance of a cable you chose. Swap that foil-and-drain-wire cable for one with a fat braided copper shield at 0.065 ohms and the same current now drops 24 dB less noise. Same rig, same loop, same current, sixteen times quieter, and no soldering. This is, incidentally, the one situation in audio where an expensive cable purchase is defensible with arithmetic instead of adjectives. Savor it.

But also notice the ceiling. Lowering R shrinks the noise; it can't eliminate it, because the mechanism is baked into the unbalanced interface itself. One conductor doing two jobs. If you want the noise actually gone, you need an interface where the noise current doesn't share a conductor with the signal at all. You need to either break the loop with a transformer, which we'll come back to, or go balanced.

Which brings us to the most misunderstood word in professional audio.

Balanced means impedance, not symmetry

Ask ten engineers what a balanced line is and nine will tell you some version of this: the signal is sent twice, once normal and once flipped, and the differential input at the far end subtracts them, so the noise that hit both wires equally cancels and the signal, being opposite on the two wires, adds.

It's a lovely story. The subtraction part is true. The part about the signal needing to be there twice is flat wrong, and believing it leads to real mistakes.

What actually makes an interface balanced is in the name, and it isn't the signal. It's the impedances. A balanced interface is one where both conductors have equal impedance to ground, at the driver, along the cable, and into the receiver. IEC 60268-3, the standard that governs this, spells it out in so many words: only the common-mode impedance balance plays a role in noise rejection, and that rejection is completely independent of whether a symmetrical signal is present. The desired signal can live entirely on one conductor and the interface will reject interference exactly as well. Signal symmetry buys you a few dB of headroom inside the gear and a little less crosstalk. It buys you zero noise rejection. Zero. This is why an "impedance balanced" output, with signal on the hot leg and just a matched resistor on the cold, rejects noise as well as the fancy symmetrical kind, and why gear reviewers who check for signal on both pins are measuring the wrong thing.

Why do the impedances matter so much? Because the ground noise voltage between your two chassis, the volt or so we established earlier, appears identically on both conductors. It's common-mode. The receiver rejects what's common. But "identically" only survives the trip if both conductors respond to that voltage identically, and each conductor's response is set by the voltage divider formed by its impedance to ground at the source and its impedance to ground at the receiver. Four impedances, two per leg. If you've ever balanced a Wheatstone bridge, this is that, exactly. Null the bridge and the common-mode voltage produces no differential output at all. Unbalance any arm and a slice of that ground noise converts into a differential signal that is now, for all purposes, welded to your audio.

And here's where real-world gear falls off the spec sheet. The bridge is most sensitive to imbalance when its arms are similar in size. A typical electronically balanced input has common-mode input impedances in the 10k to 50k range, and typical output stages set their impedance with 5-percent resistors, so a few ohms of imbalance between the legs of a source is normal, not defective. Feed that normal, slightly lopsided source into a classic differential receiver IC and the datasheet's proud 90 dB CMRR collapses. The industry-standard test now inserts a deliberate 10-ohm imbalance for exactly this reason, and under it, that textbook 90 dB input measures 65 dB. One popular receiver chip loses 25 dB of rejection from a single ohm of source imbalance. One ohm. A dirty patch point can do that.

The fix follows from the bridge math: make the receiver's common-mode impedance enormous and the bridge stops caring about small imbalances in the source. That's what a good input transformer does at around 50 megohms, which is why transformer inputs shrug off hundreds of ohms of imbalance, and why the better modern input ICs use a bootstrap trick to fake the same trick in silicon. It's not magic and it's not mojo. It's just an unbalanced bridge being hard to unbalance further.

The practical takeaway: when a balanced line hums, stop interrogating the cable and start asking what the driver and receiver impedances are doing to the bridge. And if the answer is "nothing good," a quality isolation transformer at the input end buys you back 40 to 60 dB of rejection at hum frequencies, which is the difference between a complaint and silence.

Pin 1, and other wounds we inflict on ourselves

There's a second way balanced gear hums, and this one isn't physics being subtle. It's manufacturers making a mistake so common it has a name, a test, and an AES standard aimed squarely at it.

In 1994, Neil Muncy gave it the name: the pin 1 problem. Pin 1 of an XLR is the shield. In a properly designed box, pin 1 ties to the chassis right at the connector, so shield current, which we've established is always flowing, goes around the outside of the metal box and back out to the building ground without ever meeting the electronics. In an improperly designed box, pin 1 takes a scenic route: onto the circuit board, along a trace that audio circuitry also uses as its ground reference, and then to chassis. Shield current through that shared trace drops microvolts along it, the amplifier dutifully amplifies the drops, and now your balanced, professional, expensive signal chain hums for reasons no cable swap will ever fix. As Muncy put it, balancing acquired "a tarnished reputation, which it does not deserve." The problem was never balancing. The problem was a ground pin doing double duty inside the box, common impedance coupling again, just relocated indoors.

You can find it with ten dollars of parts. John Windt's "hummer" is a wall wart and a resistor arranged to push about 50 mA of AC through any two points you clip it to. Clip from chassis to pin 1 of each connector on a suspect device, with everything else disconnected, and listen to the output. A well-designed box stays silent, because that current runs around the chassis and never touches gain. A pin 1 offender buzzes on cue, and now you know, and you can stop blaming the snake. AES48 exists because enough people got tired of this that it became a standard: shields terminate to the shell or chassis, not to circuit ground.

While we're on shields: for a balanced line, the shield isn't a signal conductor at all, it's a drain for capacitively coupled junk and a bond between chassis. Where you connect it matters more than what it's made of. The counterintuitive rule, worked out in the same body of research: the shield should always be grounded at the driver end. Ground it only at the receiver and the cable's slightly mismatched capacitances, together with the driver's slightly mismatched output impedances, form a pair of low-pass filters that don't quite track, and the mis-tracking converts common-mode noise into differential signal at the top of the band. Ground at both ends, which is standard practice and fine, and you give up a little of that high-frequency rejection in exchange for RF immunity. The refined version, hybrid grounding, ties the receiver end to ground through a small capacitor: open circuit at audio, short at RF, best of both. If you've ever wondered why some snake manufacturers bond the shield hard at the stage box and softly at the tail, now you know it wasn't superstition.

Feeding the whole rig without feeding the noise

Zoom out from one cable to the whole show, because the same physics scales up, and load-in is where you either set yourself up for a quiet night or spend the changeover crawling behind amp racks.

Touring-scale power arrives as three-phase: three hot legs, each a sine wave displaced 120 degrees from its neighbors, sharing one neutral and one ground. In North America that's typically 120/208 wye; leg to neutral gives you 120 volts for your racks, leg to leg gives 208. The elegant part is the neutral: with the three phases loaded equally, their return currents overlap out of phase and largely cancel, which is why one neutral can serve three hots. The less elegant part is what happens when loading is sloppy. Imbalance between legs means real current on the neutral, voltage drop along it, and more magnetic field around the feeder, all of which raises the noise floor of the ground environment your shields live in. Balancing your amp racks across the three legs isn't just an electrician's tidiness fetish. It's noise hygiene.

The oldest distro question is whether audio should get its own transformer, its own service, its own island, away from lighting and video. Ideal-world answer: audio, lighting, and video fed from the same service, grounds bonded at one point, everything referenced to one potential, loops carrying millivolts instead of volts. The really important part is what you must not do, and this one still gets specified by people who should know better: do not drive a separate ground rod for audio and call it a "clean ground." Earth is a terrible conductor. Code allows a single ground rod up to 25 ohms to earth, and 25 ohms will never trip a breaker during a fault; your "quiet" rod is a shock hazard with a mystique. Worse, two rods at different potentials, and during a nearby lightning strike the soil between them can develop thousands of volts, which your snake will happily deliver from one end of the system to the other. If separate rods exist, code requires them bonded to the main electrode anyway, which quietly deletes the imagined benefit. There is no audio in the dirt. Stop digging.

The legitimate big-iron tools do exist. An isolation transformer feeding your distro breaks the leakage path from everything upstream and gives your system one clean locally derived neutral-ground bond. Balanced technical power, 60 volts on each leg in opposition instead of 120 and neutral, lives in NEC Article 647 and works by making the leakage currents from the two legs cancel; it's a real thing for studios and broadcast plants, installed by electricians, permitted, inspected. Notice the shape of every legitimate solution: they reorganize the power side with the safety system intact. The illegitimate solution snips the green wire. The line between the two is the line between engineering and folklore, and also, occasionally, between an incident report and a funeral.

What you can take back to the desk

Strip the physics down to what you do at half past six with the buzz already in the room.

First, think in loops, not in grounds. The noise is a current going around a circle you can draw: out one cable's shield, back through the building. Your job is to find which circle, and then either break it somewhere legal or stop it from sharing a conductor with your audio.

Second, find the problem interface methodically instead of by vibes. Mute upstream and the noise stays: it's entering downstream of the mute. Unplug a source and the buzz dies with it: the loop runs through that cable. A pair of shield-lifted test dummies and a lunch break will localize almost any noise to a single interface, and the hummer will tell you in thirty seconds whether a box has a pin 1 problem before you tear apart a rig looking for a bad cable that doesn't exist.

Third, fix at the interface, with the right tool. Unbalanced link that has to stay unbalanced: heavy braided shield, short as possible, and if the loop's still audible, a transformer isolator in the line. Balanced line into a hum: suspect the impedance bridge before the cable, and reach for an input-transformer isolator at the receiving end, which fixes lopsided sources and pin 1 problems in one move. It's a passive lump that costs less than one hour of the fee you'll burn chasing the buzz any other way.

Fourth, at load-in, spend the ten minutes. Same service for audio and lighting where you have any say. Legs balanced. Signal runs crossing power runs at right angles, not lounging alongside them in the same cable tray. Ask the house electrician where the neutral-ground bond is, partly because the answer matters and partly because the question instantly identifies you as someone not to be lied to about the power.

Fifth, and permanently: the safety ground is not yours. Every fix above works with three pins in the wall. The engineer who lifts a ground to save ten minutes is gambling someone else's life on the continued good health of a power supply built to a price. Break signal loops. Never break the fault path.

The buzz was never a mystery and it was never the ground's fault. It's leakage current, legally emitted, circulating through loops you closed, coupling through impedances you can measure. Everything in this piece fits on one index card: I times R, balance the bridge, bond don't float, ground the shield at the driver, and leave the green wire alone. Tape it inside the doghouse lid. See you next week in the Woodshed.

Above the Fader

Sauce

• Bill Whitlock, "Understanding, Finding, & Eliminating Ground Loops" (Jensen Transformers seminar)
• Bill Whitlock, "An Overview of Audio System Grounding and Interfacing" (AES seminar, with notes)
• Bill Whitlock, "Grounding and System Interfacing," audioXpress
• Rane Note 165, "Pin 1 Revisited"
• Rane Note 166, "Shield Current Induced Noise"
• "The Pin 1 Problem Revisited," ProSoundWeb
• "Getting Grounded: Power Distribution For Sound Reinforcement Systems," ProSoundWeb
• "Typical Temporary Power Distribution For Live Entertainment," Motion Labs
• Rod Elliott, "Earthing (Grounding) Your Hi-Fi," Elliott Sound Products
• Hugh Robjohns, "Understanding & Solving Ground Loops," Sound on Sound
• NEC Article 647, Sensitive Electronic Equipment (balanced technical power)

Featured photo by Markus Spiske on Unsplash.