Iron in the path: what a DI box actually does

The humble direct box is the most misunderstood object on your stage. Inside that steel brick is a hundred years of transformer physics, a war over input impedance, and the answer to why the acoustic act sounds like a wasp in a tin. This week we open it up.

It's 4:40 on a festival Saturday and you have a twenty-minute changeover. The next act is a singer-songwriter with an acoustic guitar, an upright bass, and a keyboard player who has brought, for reasons known only to him, a vintage electric piano with a quarter-inch output and a hum you can hear from the monitor desk. Your stage tech holds up two direct boxes, one in each hand. One is a passive unit that weighs as much as a brick. The other is an active box with a phantom power LED and a battery door held on with gaff tape.

Which one goes on the guitar? Which one goes on the piano? And when the upright bass with the under-saddle pickup comes up sounding like someone strangling a kazoo, which switch do you reach for first?

If your answer is "whichever one is on top of the workbox," this piece is for you. The DI box is the least glamorous item in the input list and the one most likely to quietly wreck a channel before the signal ever reaches your preamp. It's also, once you look inside, one of the most elegant. A good passive DI is a single component doing four jobs at once, and the physics of how it does them explains half the mysteries of live audio input: why long unbalanced cables lose their top end, why piezo pickups squawk, why the ground lift sometimes fixes the hum and sometimes does nothing, and why one transformer costs eleven dollars and another costs a hundred and eighty.

So let's take the lid off.

Two jobs, one steel box

Strip away the marketing and a DI box does exactly two things. It converts an unbalanced, high-impedance, instrument-level signal into a balanced, low-impedance, mic-level signal. And it electrically separates the instrument's ground from the console's ground.

That first sentence is doing a lot of work, so let's unpack the three conversions hiding in it.

Unbalanced to balanced. Your guitar cable is a single conductor with a shield. Any interference the cable picks up along its run lands directly on the signal, and there's nothing at the far end to distinguish noise from music. A balanced line carries the signal on two conductors that are treated identically, and the receiving end responds only to the difference between them. Noise that lands equally on both conductors gets subtracted out. This is why you can run a mic line 300 feet across a festival site through a snake lying next to the lighting distro, while a 20-foot unbalanced guitar cable next to a dimmer pack sounds like a Geiger counter.

High impedance to low impedance. A passive electric guitar pickup wants to see a very light load, something like 1 megohm, or it starts to lose level and top end in ways we'll compute shortly. A console mic input has an input impedance somewhere between 1.5k and 10k ohms. Plug the guitar straight into the mic input and you've hung a load on the pickup that is a hundred to a thousand times heavier than it was designed for. The DI sits in the middle and makes each side see what it wants to see.

Instrument level to mic level. Instrument signals are all over the map, from tens of millivolts off a single-coil pickup to several volts off an active bass. Mic inputs expect mic level. A passive DI's step-down ratio conveniently drops instrument level into the range a mic preamp is built for, which is why you can plug the DI's XLR output into the same input, on the same gain range, as an SM58.

The second job, ground separation, is the one that saves your changeover. The instrument's electronics are referenced to whatever ground exists on stage, through its amp's power cord. The console is referenced to its own ground, possibly on a different leg of the power distro, possibly in a different building. When you connect the two with a piece of wire, any voltage difference between those grounds drives a current around the loop, and that current has power-line hum riding on it. A transformer passes the audio magnetically, with no wire connecting one side to the other, so the loop never forms. This is galvanic isolation, and it's the passive DI's quiet superpower: the input side and the output side of the box are not electrically connected at all.

Hold onto that thought, because everything else in this piece is really about how one lump of iron and copper pulls off all of the above simultaneously.

What a transformer actually does

A transformer is two coils of wire wound around a shared magnetic core. Audio current flowing in the primary coil creates a changing magnetic field in the core; the changing field induces a voltage in the secondary coil. That's it. No power supply, no transistors, nothing to fail. Faraday worked out the principle in 1831 and we have not improved on the concept since, only on the materials.

The magic is in the turns ratio. If the primary has twelve times as many turns as the secondary, the output voltage is one twelfth of the input voltage. Write that as N:1 and two rules follow.

Voltage scales by N. A 12:1 transformer turns 1 volt into 83 millivolts. In decibels that's a 21.6 dB drop, which is exactly the neighborhood you want for turning a hot instrument signal into something a mic preamp greets politely.

Impedance scales by N squared. This is the one that separates people who own DI boxes from people who understand them. Impedance is the ratio of voltage to current, and the transformer scales voltage down by N while scaling current up by N. So any impedance viewed through the transformer changes by N times N, which for our 12:1 example is a factor of 144.

This squared relationship is why a transformer is an impedance lens. Look through it one way and small things appear large; look the other way and large things appear small. The load on the secondary is reflected back to the primary multiplied by 144. The source impedance on the primary appears at the secondary divided by 144.

There's a third, subtler property. The transformer only responds to current flowing around its primary loop, which means it only responds to the voltage difference across its winding. It has no idea what either terminal's voltage is relative to ground, because the winding isn't connected to ground. Engineers call this common-mode rejection, and a good input transformer is spectacular at it. Noise that lifts both conductors of a balanced line together, exactly the kind that ground differentials and induced interference create, produces no differential current in the winding and simply doesn't come out the other side.

So one passive component gives you voltage step-down, impedance transformation, balanced-to-unbalanced conversion, and galvanic isolation. The catch, and there's always a catch, is that all four of those jobs depend on the quality of the magnetic core and the geometry of the windings, which is where the eleven-dollar transformer and the hundred-and-eighty-dollar transformer part ways. We'll get there. First, the worked example.

The 12:1 worked example

Numbers make this concrete, so let's push a real signal through a real ratio. Take a passive DI built around a 12:1 transformer, which is the classic direct-box ratio; Jensen's famous DI transformer presents roughly 200k ohms on the input side when the output side sees a standard mic preamp.

Start at the console end. Your mic preamp has an input impedance of 1.5k ohms. The transformer reflects that impedance back through itself multiplied by 144. From the guitar's point of view, the DI input looks like 1.5k times 144, which is 216k ohms. Not the megohm an electric guitar would ideally like, but a workable load, and this is precisely why passive DIs are recommended for active instruments and keyboards, which have low output impedances and don't care, while passive single-coil guitars sometimes sound slightly dulled through them.

Now go the other way. Suppose the source is an active bass with a 1k output impedance. Through the transformer, the console sees that 1k divided by 144, about 7 ohms, in series with the transformer's own winding resistance. Add the windings and the output impedance lands near the 150 to 200 ohm range, which is to say: it looks like a microphone. That's the entire trick. The DI's job is to disguise an instrument as a mic so convincingly that the console never suspects.

Level works out just as neatly. A keyboard sending a healthy line-level signal of 1.5 volts hits the primary; the secondary delivers 125 millivolts, which is hot mic level, comfortably inside the range your preamp expects. The 21.6 dB of free attenuation is not a bug, it's the point.

Two practical consequences fall out of this arithmetic, and both are worth pinning to the workbox lid.

First, a passive DI is a passive device, so it can only trade voltage for current. It cannot add gain. Whatever comes out is quieter than what went in, always. If the source is weak to begin with, a passive DI makes it weaker, and you make up the difference with preamp gain, which also raises the noise floor. Weak sources want active DIs.

Second, the impedance lens works on everything, including things you'd rather it didn't. If someone plugs a passive DI's input into a source with a big DC offset, or a speaker output without a pad, the transformer faithfully transforms whatever it's given, and the results range from ugly to smoky. The box has no judgment. That's your job.

Why the cheap one sounds cheap

Here is where transformer physics stops being tidy. An ideal transformer works identically at all frequencies and all levels. Real iron does not, and the ways it fails are exactly the ways cheap DI boxes sound bad.

Start at the bottom of the spectrum. A real transformer's primary winding is, among other things, an inductor, called the magnetizing inductance. It sits across the signal like a shunt. At high frequencies its impedance is enormous and it might as well not exist. As frequency falls, its impedance falls with it, and it starts to load down the source. The source impedance and the magnetizing inductance form a high-pass filter: the corner frequency is the source impedance divided by two pi times the inductance. A premium DI transformer has a huge magnetizing inductance, hundreds of henries, precisely so that this corner lands below the audio band even with a high-impedance source. A budget transformer with a small core and fewer turns might have a tenth of the inductance, and its low-frequency corner climbs right up into the bass guitar's fundamental range. Same schematic, same ratio, and one of them quietly filters out the low B string.

Notice the dependency: the low-frequency response of a passive DI depends on the source impedance driving it. A synth with a 100 ohm output will get flat bass out of almost any transformer. A passive bass with a 10k source impedance driving the same cheap transformer loses real low end. When someone tells you a particular DI "sounds thin with passive basses but fine with keyboards," they are describing this equation, whether they know it or not.

Then there's saturation, the failure mode with a fan club. The magnetic flux in the core is proportional to the applied voltage divided by the frequency. Double the level, double the flux. Halve the frequency, double the flux again. Iron can only carry so much flux before it saturates, and when it does, the transformer stops transferring the extra signal linearly and starts generating odd-order harmonic distortion, mostly third harmonic, concentrated at low frequencies where the flux is highest. Hit a small core with a hot, low bass note and you get a distinctive thickening blur. In a vintage console transformer at moderate levels, tape-adjacent and flattering. In a cheap DI catching the full output of an active bass on a festival stage, mud.

This is why transformer cores are the expensive part. High-nickel alloy cores saturate at lower flux than steel but are far more linear below saturation, so premium designs use generous nickel cores that keep flux density low at working levels. Bigger core, more turns, more nickel, more dollars. The eleven-dollar transformer isn't a scam; it's simply built for a smaller flux budget, and the price of exceeding the budget is paid in third harmonic.

The top of the spectrum has its own villains. Not every line of magnetic flux from the primary actually passes through the secondary; the strays are modeled as leakage inductance in series with the signal, and together with the winding capacitances it forms a resonant low-pass filter at the top of the band. Good designs wind primary and secondary in interleaved layers to minimize leakage, then damp the resonance so the response shelves off gracefully above the audio band. Sloppy winding gives you either an early high-frequency roll-off or an ultrasonic resonant peak that rings on transients. When people describe a transformer as sounding "slow" or "veiled," the honest physics underneath is usually leakage inductance and capacitance doing their thing.

The takeaway isn't that transformers are flawed. It's that a passive DI's frequency response is a negotiation between the source and the iron, and the expensive boxes are expensive because they hold up their end of the negotiation regardless of what you plug into them.

When the battery earns its keep

An active DI replaces the transformer's impedance transformation with a powered buffer amplifier: a high-impedance input stage, a bit of electronics, and a balanced low-impedance output, running off a battery or phantom power. Some active boxes keep a transformer on the output for isolation; many don't. The active box gives up galvanic isolation on the cheap models and gains two things no passive box can offer: an input impedance as high as you like, and the ability to buffer, or even add, level.

To see why that matters, meet the piezo pickup, the squawk factory under the saddle of nearly every acoustic instrument on your stage. A piezo element is electrically a capacitor with a voltage source behind it. Under-saddle elements commonly measure around 10 nanofarads. That capacitance in series with whatever input impedance it drives forms a high-pass filter, and the corner frequency is set by one over two pi R C.

Run the numbers and the entire mystery of bad acoustic tone dissolves. Into a 100k input, a 10 nF piezo rolls off below 159 Hz. Everything below the open A string is fading away, and what's left is the honky midrange squawk we all know. Into 1 megohm, the guitar-amp standard, the corner drops to 16 Hz, flat across the band on paper, though the element's response near the corner still benefits from a lighter load. Into 10 megohms, which is why specialist acoustic DIs advertise exactly that figure, the corner is 1.6 Hz and the pickup finally behaves like the wideband transducer it always was. Same guitar, same pickup, three different instruments, and the only variable is the input impedance.

A passive transformer can't get you there. To present 10 megohms through a step-down transformer, the reflected ratios and the winding practicalities run away from you; the transformer's own losses and parasitics eat the benefit. Ultra-high input impedance is a job for an active buffer, full stop.

Magnetic guitar pickups have a subtler version of the same disease. A passive pickup is a big inductor, several henries of it, with thousands of turns of fine wire. That inductance against the capacitance of the guitar cable forms a second-order low-pass filter with a resonant peak, and that peak is the pickup's voice: its frequency and height decide whether the pickup reads as glassy, honky, or dark. The load impedance damps the peak. Into 1 megohm the peak stands proud and the top end sparkles. Into 200k the peak flattens and the guitar dulls noticeably. Add thirty feet of cable and the extra capacitance drags the resonance down several hundred hertz on top of that. So the guitarist who swears his guitar sounds different through different DI boxes is not imagining it; he's hearing the load impedance retuning his pickup's resonance.

So the field rule writes itself. Passive source, active DI: the buffer's high input impedance leaves the pickup's resonance alone and the piezo's corner frequency below the band. Active source, passive DI: the instrument's own electronics have already done the buffering, its low output impedance drives the transformer's magnetizing inductance without bass loss, and you pocket the passive box's isolation and indifference to phantom-power accidents. There are exceptions, there always are, but you'll win with that rule far more often than you lose.

One more active-DI virtue that gets no spec-sheet glory: headroom management. Active instruments can put out several volts, enough to overload an active DI's input stage, which is why the good ones carry a pad switch. The pad on a DI is not decoration. An active bass with fresh batteries into an unpadded active DI is one slap away from square waves, and no amount of console gain-staging downstream will un-square them.

Balance is about impedance, not symmetry

Time to kill the most durable myth in audio. A balanced line does not work because it carries two copies of the signal with opposite polarity. Plenty of balanced outputs do that, but it is not what makes the line reject noise. What makes the line reject noise is that both conductors have equal impedance to ground. Bill Whitlock has been patiently explaining this in AES papers since 1995, and the industry has been patiently ignoring him for the same thirty years.

Here's the mechanism. Interference couples onto both conductors of the cable roughly equally, as a common-mode voltage. The receiving input responds only to the difference between the conductors. If both conductors have identical impedances to ground, the noise divides identically on each, the difference is zero, and the receiver hears nothing. If the impedances are unbalanced, the same common-mode noise produces slightly different voltages on each conductor, the difference is no longer zero, and the noise has just converted itself into signal. How well an input rejects common-mode noise is its common-mode rejection ratio, and real-world CMRR is set less by the input's laboratory spec than by how impedance-balanced the whole interface is, source and cable included.

This is where the transformer quietly flexes. A transformer input's winding floats; its impedance to ground is enormous and essentially identical for both terminals, which makes it forgiving of imbalances in whatever drives it. Cheap electronically balanced inputs with 10k resistors to ground are far touchier: a small imbalance in the source, even a few ohms of contact resistance difference, measurably degrades rejection. It's why a transformer-isolated DI so often fixes a hum problem that swapping cables did not.

Now the ground lift, the most pressed and least understood switch on the box. On a passive DI, the input circuit and output circuit are already galvanically isolated by the transformer; the only remaining connection between stage ground and console ground is the XLR shield on pin 1. The lift switch disconnects it. If a hum current was circulating through that shield, driven by a voltage difference between the instrument's ground and the console's ground, opening the switch breaks the loop and the hum stops. That's all it does. It is not a filter, not magic, and on some units it does nothing audible because the loop your hum rides on is somewhere else entirely.

Which brings us to the pin 1 problem, named by Neil Muncy and documented at length by Whitlock. A distressing amount of audio equipment connects pin 1 not to the chassis at the connector, but to a trace that wanders across the circuit board's ground plane before finding chassis. Shield current then flows through the guts of the device, and the tiny voltage drops it creates along shared ground traces get amplified along with the audio. The shield connection, which should be a drain pipe bolted to the outside of the building, has been plumbed through the living room. When you meet a piece of equipment that hums no matter which lift you throw, you've probably met a pin 1 problem, and the fix lives inside the offending gear, not on your DI.

The practical hierarchy for hum, then: first, isolate the loop with a transformer if one isn't already in the path. Second, lift the shield at one end, ideally via the DI's switch, never by defeating a power cord's safety ground, which is how people die. Third, if neither works, suspect the equipment itself and start reading Whitlock instead of buying more cables.

The phantom in the box

Phantom power deserves its own stop on this tour, because a DI box is where 48 volts and instruments meet, and the meeting is not always cordial.

A quick refresher on the plumbing. Phantom power places 48 volts DC on both signal pins of the XLR, pins 2 and 3, through matched 6.8k resistors, with the return path on the shield, pin 1. Because both signal conductors sit at the same DC potential, a differential input sees no difference, hence "phantom": equipment that doesn't need it shouldn't notice it. Condenser mics and active DIs steal their operating current from it. Everything else is supposed to shrug.

A passive DI does better than shrug. The transformer's secondary winding connects pin 2 to pin 3 through a length of copper, and both ends of that winding rise to 48 volts together. No current flows through the winding, because there's no potential difference across it, and the instrument side never learns that phantom exists. Galvanic isolation earns its keep again: you can leave phantom on across a whole stagebox of passive DIs and nothing whatsoever happens. This indifference is one of the strongest arguments for passive boxes on festival stages where the phantom switching is whatever the last engineer left behind.

Active DIs live off that same 48 volts, and here the details start to matter. The box draws its current through those 6.8k source resistors, which means the voltage that actually arrives at the electronics sags below 48 depending on the draw. A design that needs lots of current has to work with whatever survives the resistors, which limits the input headroom the buffer can offer. This is why some active DIs specify better maximum input level on battery than on phantom, and why a hot active bass that sounded clean at soundcheck through a battery-powered box can edge into grit when someone swaps in phantom power and the rail sags. If an active DI has both options and the source is hot, a fresh battery is not nostalgia, it's headroom.

The genuinely dangerous corner case is phantom applied where it doesn't belong. The 48 volt rail itself is current-limited and polite, but switching it on or off sends a DC transient down the line, and whatever is connected unbalanced downstream of a patch error gets that transient asymmetrically. Ribbon mics patched to the wrong line, consumer gear adapted onto an XLR with pin 3 grounded, vintage equipment with transformer secondaries center-tapped to ground for reasons lost to history: these are the patients. The discipline that protects them is boring and effective, mute first, phantom second, wait, then unmute. The DI itself is rarely the victim; it's usually the innocent bystander that made the miswire possible by accepting whatever was plugged into it.

One habit worth stealing from broadcast people: treat phantom as a per-stage decision, not a per-channel afterthought. A stagebox feeding passive DIs and dynamic mics can run phantom globally without a second thought. The moment ribbons or adapted consumer gear enter the input list, phantom becomes something you switch deliberately, with the channel muted, like a safety on a tool. Thirty seconds of discipline versus a rewound ribbon; the economics aren't close.

A field guide for the input rack

Theory earns its rack space at changeover. Here's how the physics maps onto the things people actually hand you.

Electric bass, passive. High source impedance, precious low end. An active DI with a 1 megohm input, or a premium passive with a serious core if you want isolation. A cheap passive DI is the worst option on the stage for exactly the magnetizing-inductance reasons computed above: high source impedance into small iron equals vanishing low B.

Electric bass, active, and all keyboards and synths. Low output impedance, healthy level, and this is the passive DI's home game. The transformer gets an easy source to drive, you get bulletproof isolation, no batteries die mid-set, and phantom power mishaps pass through without drama. Engage the pad for hot stage keyboards.

Acoustic instruments with piezo pickups. If the instrument has an onboard preamp with a fresh battery, it's an active source; treat it like a keyboard and use whatever's clean. If the piezo is passive, wired straight to the jack, nothing less than a multi-megohm active input will sound right, and the difference is not subtle. This is the one input where a specialist box genuinely pays for itself.

Amplifier outputs. The guitarist wants "his amp sound" in the PA and hands you a speaker cable. Speaker outputs carry tens of volts; a DI rated for speaker-level input has a heavy pad ahead of the transformer for exactly this. Check the legend printed on the box before, not after. And never leave a tube amp's output unloaded while you do it; the speaker stays connected, the DI bridges across it.

Laptops and playback rigs. Electrically trivial, low impedance, line level. But they arrive riding on the nastiest grounds in the building, USB power, wall-wart chargers, video systems. This is pure isolation work, the impedance conversion is almost incidental, and it's why a stereo passive isolation box lives in every playback engineer's bag next to the adapters.

The pattern across all five: identify whether the source is active or passive, guess its output impedance within an order of magnitude, and the right box picks itself. The label on the instrument matters less than the impedance behind the jack.

What you can take back to the desk

Strip this week down to what fits on a channel strip's worth of memory.

A DI box does two jobs: it disguises an instrument as a microphone, and it breaks the electrical connection between stage and console. Impedance conversion and galvanic isolation. Everything else is refinement.

The turns ratio moves voltage; the square of the turns ratio moves impedance. A 12:1 box drops level by 21.6 dB and multiplies the console's input impedance by 144 as seen from the instrument. If you remember one equation from this piece, make it N squared.

Passive DIs subtract level, never add it. Weak or high-impedance sources come up noisy through them because you replace the missing level with preamp gain. Active sources with low output impedance are what passive DIs are for.

A passive DI's bass response depends on the source driving it. Big magnetizing inductance tolerates high source impedances; small cores lose low end and saturate into third harmonic on hot bass notes. That's the eleven-dollar difference.

Piezo pickups are capacitors. One over two pi R C decides how much instrument survives. 100k input, bass gone by 159 Hz; 10 megohms, flat to DC for practical purposes. Squawk is a loading problem before it is a tone problem.

Balance is impedance symmetry, not signal symmetry. Hum rejection lives or dies on how equally the two conductors face ground, which is why floating transformer inputs forgive what cheap differential inputs punish.

The ground lift disconnects pin 1 and breaks shield loops. It does nothing else. If lifting both ends of everything changes nothing, the hum is inside somebody's chassis, and no switch on your DI reaches in there.

And the changeover answer, for the record: passive DI on the vintage electric piano, because it's a hummy active source begging for isolation. Active DI on the acoustic guitar and the upright, the highest input impedance you own on whichever of them has the passive piezo. Twenty minutes is plenty.

Above the Fader

Sauce

• DI Boxes & Why We Need Them — Sound on Sound
• Using DI Boxes — Sound on Sound
• Design of High-Performance Balanced Audio Interfaces — Bill Whitlock (hosted at Elliott Sound Products)
• Transformers, Part 2 — Elliott Sound Products
• Guitar Pickups — Elliott Sound Products
• Direct Box Transformers — Jensen Transformers
• JDI Stereo user guide — Radial Engineering
• PZ-DI piezo-optimized DI — Radial Engineering
• Finding Problems in Balanced Audio Equipment Interfaces — ProSoundWeb

Featured photo by Les Taylor on Unsplash.