The little steel box that saves the gig

Everything on your stage that isn't a microphone gets to the console through a DI. Here's what's actually inside that scratched-up steel box, why the transformer is the best part, and how to stop piezo quack and ground hum with physics instead of luck.

It's 4:40 on a festival Sunday and the changeover is going fine, which should have been your first warning. The headliner's bass tech hands you a pair of XLRs like he's dealing cards. The support act is a duo: one acoustic guitar with an under-saddle pickup, one keyboard player who has already told you, twice, that his rig "never hums." You line-check the guitar and it sounds like someone stretched cellophane over a banjo. All quack, no body. Then the keys come up and there it is, the 60-cycle drone, sitting under everything like a truck idling behind the stage. The keyboard player looks at you. You look at the input rack. Somewhere in there, bolted to a shelf and covered in board tape from four tours ago, is a passive DI that costs less than one of his patch cables and is about to fix both problems.

Here's how the next six minutes go. The guitar gets moved off the passive box someone patched by habit and onto the active DI with the 10 meg input, and the body of the instrument reappears like it had been waiting in the wings. The keyboard rig goes through the passive box the guitar just vacated, ground lift engaged, and the drone stops mid-cycle. The keyboard player says "it never does that," and you agree, pleasantly, because the changeover clock is running and this argument has no trophy. Total cost of both fixes: zero dollars, ninety seconds, and one small act of physics.

We spend a lot of ink on the glamorous parts of this job. Line arrays, RF coordination, whatever plugin emulates whatever console we can't afford. Meanwhile the DI box does its work in the dark under a keyboard stand, gets stepped on, gets beer poured into it, and never once gets soundchecked by name. This week we're giving it the full teardown: what direct injection actually is, how the transformer inside works at the level of magnetics and physics, why passive and active boxes fail differently, what piezo pickups actually want, and why the isolation, not the impedance conversion, is the real product.

What a DI actually does

Strip away the marketing and a DI box has two jobs. It converts an unbalanced, high-impedance, instrument-level signal into a balanced, low-impedance, roughly mic-level signal your console input wants. And it electrically separates the instrument's ground from the console's ground so the two systems can disagree about what "zero volts" means without you hearing the argument.

That second job matters more than most people think. Paul White put it plainly in Sound on Sound almost thirty years ago: DI boxes "provide electrical ground isolation between input and output, and they match the impedance of the source signal to that of the load." Note the order he put those in.

Why can't you just wire a guitar into a mic input with an adapter cable? Three separate mismatches, and each one bites differently.

Level is the obvious one. An instrument output sits somewhere between mic level and line level depending on how hard the player digs in, and a console mic input can handle that with the gain trimmed back. Not the real problem.

Impedance is the sneaky one. A passive guitar pickup is a coil with thousands of turns of very fine wire. Its source impedance isn't a nice fixed number; it's a few kilohms of DC resistance in series with a couple of henries of inductance, which means the impedance climbs with frequency and peaks hard at the pickup's resonance. Feed that into a 1.5k mic input and you've built a voltage divider that eats the top octaves of the instrument first. The player hears their guitar with a blanket over it and blames your console. They're not entirely wrong.

Balance is the third. An instrument cable is one signal conductor and a shield, and every foot of it is an antenna. A balanced line rejects interference because both conductors pick up the same junk and the receiving end subtracts it. Fifteen feet of unbalanced cable across a stage floor full of dimmer wiring is a science experiment. A hundred and fifty feet to front of house is a career decision.

So the DI takes a signal that only survives short trips and repackages it for the long haul. The question is how, and that's where the two families split.

Before we split them, a word on the name. "Direct injection" dates from studio practice: injecting the instrument's signal directly into the console instead of miking an amp. Live, we mostly use DIs for the opposite reason. We're not chasing purity; we're chasing survival. The signal has to cross a stage full of motors, dimmers, video walls, and wireless racks, then travel a snake to a console on a different power feed, and arrive without carrying the electrical biography of everything it passed. Direct injection was born a recording nicety. It grew up into a transport safety device.

The transformer, without the mythology

A passive DI is a transformer in a box. That's the whole parts list, minus a couple of switches. And because transformers are wrapped in more audio mythology than any component except the vacuum tube, it's worth being precise about what this one does.

A transformer is two coils of wire sharing a magnetic core. Signal current in the primary creates a changing magnetic field; the field induces a voltage in the secondary. No electrical connection between the two sides at all. The signal crosses as magnetism, which is the source of every good thing a transformer does and most of the bad things too.

The turns ratio sets the voltage ratio. The standard for a DI transformer is about 12:1, so the signal comes out 12 times smaller, about 21.5 dB down. That sounds like a loss until you remember it's going into a mic preamp with 60 dB of gain on tap. The trade is deliberate, because impedance transforms as the square of the turns ratio. A 12:1 transformer is a 144:1 impedance converter. Load the secondary with a console mic input and the instrument sees that load multiplied by 144. This is the entire trick. You give up voltage you didn't need to buy impedance you desperately did.

The physics also sets the limits, and this is the part the mythology skips. A real transformer is the ideal one plus a swarm of parasites. The primary winding has a finite magnetizing inductance, and at low frequencies that inductance starts to look like a short across your signal. How low depends on the source driving it: the corner frequency lands where the source impedance equals the inductive reactance. Run the numbers on a 100-henry primary driven from a 10k source and you get about 16 Hz. Fine. Drive the same transformer from a 50k source and your corner moves up to 80 Hz, and now the bass player is asking questions. Cheap DI transformers have less inductance and lose bass from friendlier sources than that, which is one honest reason the good ones cost what they cost.

At the other end of the band, the leakage inductance (flux that links one winding but not the other) forms a low-pass filter with the winding and cable capacitance. Bill Whitlock's chapter on audio transformers goes deep on this: good designs interleave or multi-filar-wind the coils specifically to drive leakage down so that filter's corner lands well past 20 kHz even with a long capacitive cable hanging off the output. Bad designs ring. If you've ever swept a mystery DI and found a 3 dB bump at 15 kHz followed by a cliff, you've met leakage inductance and its friend, resonance.

Good transformers also carry a detail you can't see from outside: a Faraday shield, a grounded foil layer between primary and secondary. Capacitance between the windings would otherwise pass high-frequency garbage straight across the galvanic gap, which is exactly the stuff you bought the transformer to block. The shield intercepts it and drains it to ground. Whitlock's chapter notes the design tension here: input transformers use Faraday shields because blocking common-mode junk is their whole job, while output transformers often skip the shield and go multi-filar instead, because their enemy is leakage inductance into a long capacitive line. Same species, different survival strategies. When a cheap DI lets a lighting dimmer's hash ride in on an otherwise clean keyboard channel, the missing shield is often the culprit, and no spec sheet will tell you it's missing.

And then there's the core. Magnetic core material saturates: past a certain flux density it simply can't carry more field, and the output flat-tops. Flux is highest at low frequencies for a given voltage, which is why transformer distortion always shows up first as hair on the bottom octaves at high level. A passive DI fed hot synth bass will distort in a way an active DI won't, and whether that's a defect or a feature is a religious question we're not settling today. What we will say: it's low-order, it's level-dependent, and some very expensive studio hardware exists mostly to do on purpose what a small DI transformer does by accident.

Passive boxes and the loading problem

Time for the worked example. Take the classic passive DI recipe: a Jensen JT-DB-E transformer wired 12:1, the same part Radial builds the JDI around. Your console mic input is around 1.5k. Reflected through the 144:1 impedance ratio, the instrument looks into roughly 216k. Radial's own spec for the JDI's input impedance is 130k, a bit lower once you account for the transformer's real-world losses and loading, but same neighborhood.

Is 130k enough? Depends entirely on the source, and this is the decision tree that should live in your head.

For anything with an output buffer, yes, easily. Active bass, keyboard, laptop interface, sampler, DJ mixer: these have op-amp outputs with source impedances of a few hundred ohms. They'll drive 130k without noticing it exists. This is why the standing advice is passive DI for active sources, and it's good advice. The transformer adds no noise, needs no power, and galvanically isolates a laptop's grubby switch-mode ground from your console. For keyboards, a passive box isn't the budget option. It's the correct one.

For a passive magnetic pickup, the honest answer is "mostly." Model a typical single coil as a couple of henries and 6k of wire resistance with some winding capacitance, and its impedance near the resonant peak reaches into the hundreds of kilohms. Load that with 130k and you damp the resonance and pull the peak down; load it with 1 megohm and the pickup keeps most of its sparkle. That resonant peak is a real part of the sound of an electric guitar. This is a case where the graph and the guitarist's complaint agree with each other, which is rarer than we'd like. It's also why the amp input the pickup was designed against has been 1 megohm since the 1950s.

The practical read: bass through a good passive DI is a time-honored and defensible move, since the top octave of a P-bass is not where the money lives. A vintage Strat straight into a 130k passive box will play, but it comes back a little polite, a little rounded off. If the guitarist notices, they're right, and the fix costs you one channel of active DI.

One more passive virtue while we're here: headroom. A transformer has no rails. Where an active DI's buffer will clip when a synth or DJ mixer sends it a line-level signal near the top of its supply, a decent passive box just steps it down 21 dB and hands it over. The loudest thing on your stage should probably meet a transformer first.

Active boxes: a buffer, a promise, and a phantom supply

An active DI replaces the transformer's impedance trick with electronics: a high-impedance buffer at the input, usually a FET, then circuitry to drive a balanced output. Because active circuitry has gain available, the input impedance and the output level are decoupled. You can present 1 megohm, or 10, to the instrument and still send the console whatever level you like. The Countryman Type 85, the boxy little FET classic that has survived four decades of stage abuse, presents 10 megohms. Pickups treat that like an open circuit, which is exactly the point.

The price is power. Batteries, or phantom off the console, and each has its gig failure mode. The battery dies during the set, always during the quiet song. Phantom depends on someone at front of house not un-pressing the button, and on the split or patch actually passing 48 volts, which on festival infrastructure you did not personally wire is an assumption worth checking at line check, not at downbeat.

The subtler price is what happens on the output side. A transformerless active DI floats its output with op-amps, and a clever differential stage does a fair impression of isolation. But an impression is what it is. There's no galvanic break; the grounds are still connected through the electronics. When the backline ground and the console ground genuinely disagree, at a festival with distro on opposite sides of the stage, or in a club whose wiring predates your parents, the op-amp trick runs out of common-mode range and the transformer does not. This is why the grown-up spec, boxes like the BSS AR-133 that have ridden in touring racks for thirty years, is an active input stage and a transformer output. Buffer for the pickup, iron for the isolation. You pay for both because both are doing a job.

So the folk rule "passive for active sources, active for passive sources" is real, but it's worth knowing why it's real: it's two different problems. Active sources don't need the buffer but benefit from the isolation and headroom. Passive pickups need the buffer or they get loaded. Piezos need the buffer so badly they get their own section.

Piezos: a capacitor pretending to be an instrument

The under-saddle pickup from our changeover deserves its own physics lesson, because a piezo is not a smaller, weaker magnetic pickup. It's a different device with a different source impedance, and treating them alike is where the quack comes from.

A piezo element is a crystal wafer that generates voltage when squeezed. Electrically, it looks like a small capacitor, a nanofarad or a few, in series with your signal. And a capacitive source into a resistive input is a textbook first-order high-pass filter. The corner frequency is set by the product of the two: f = 1/(2πRC).

Run the numbers. A 2.2 nF piezo into a 1 megohm active DI puts the corner at about 72 Hz. Everything below that rolls off at 6 dB per octave, and the response is already sagging for an octave above the corner. The low E on that acoustic guitar is 82 Hz. You've built a filter that specifically removes the body of the instrument and leaves the stringy, brittle top, and that is the quack: not a mysterious piezo character flaw, just a high-pass filter nobody asked for. Feed the same pickup into 10 megohms and the corner drops to about 7 Hz. The whole instrument comes through, warm end included.

This is why "1 meg ought to be enough for anybody" fails for piezos while being perfectly fine for magnetic pickups, and why boxes built for the job, Radial's PZ-DI being the obvious example, offer a 10 megohm position. Sound on Sound's review of that box says it straight: piezos "in general, need to see impedances of this level to give a linear frequency response and to avoid the dreaded piezo 'quack'."

The desk-side version of this lesson: when the acoustic act sounds thin, check what their pickup is plugged into before you reach for the low shelf. EQ can flatter a quacky signal; it cannot un-filter it. Boosting 100 Hz into a signal where 100 Hz never arrived just amplifies mud and stage rumble. Ten seconds swapping the passive box for a high-impedance active one fixes at the source what forty minutes of EQ will not fix at the desk. If the guitar has an onboard preamp, mind you, the piezo is already buffered inside the instrument and any DI will do; the battery in the guitar becomes your failure point instead. There's always a battery. It's batteries all the way down.

Hum, ground lifts, and why isolation is the real product

Back to the keyboard player and his drone. Hum on a DI'd source is almost never the instrument. It's the loop: his rig grounds through its power supply on backline distro, your console grounds through its own supply at front of house, and those two "grounds" sit at slightly different potentials with 60 Hz and its harmonics riding between them. Connect the two systems with a cable shield and current flows through it. The loop area is an antenna tuned to everything ugly in the building.

A transformer breaks the loop by construction. No conductive path from primary to secondary; the signal crosses as a magnetic field, so the two grounds can differ and no current flows between them. That's galvanic isolation, and it's the single strongest argument for iron in the box. The ground lift switch is the supporting act: it disconnects the shield bond, usually pin 1, so you can break the shield path while the transformer breaks the signal path. Standard operating procedure stands: start lifted on anything with its own power supply, un-lift if a battery-powered source needs the ground reference. When the hum changes as you flick the switch, you've learned something about the venue's wiring. Write it on the input list for the next crew.

While we're being precise about balanced lines, one Whitlock point worth engraving somewhere: what makes a balanced line reject interference is impedance balance, not signal symmetry. His AES work on balanced interfaces is blunt about it: matched impedances to ground on the two legs is what turns interference into a common-mode signal the receiver can cancel. Equal-and-opposite signal voltage is nice for headroom but irrelevant to noise rejection. A transformer secondary is inherently well balanced in exactly the way that matters, which is one more quiet reason transformer-isolated outputs behave so well on long, hostile lines. It also explains the occasional mystery of an impedance-balanced output (one leg driven, one leg just a resistor) rejecting hum as well as a fancy symmetrical driver: it's supposed to. The symmetry was never the active ingredient.

The same iron solves a problem one layer up the signal chain, too. Any time one source has to feed two consoles on different power, front of house and monitors, or your desk and the broadcast truck, you've built a potential loop between systems that have never met each other's electricians. This is why proper mic splits put transformers on the isolated legs, and why a drawer of one-to-one line isolation transformers belongs in every festival workbox. Same physics as the DI, minus the level conversion: a 1:1 winding passes the signal and breaks the ground. The day the OB truck shows up with hum on all forty-eight channels at once, no ground lift ritual performed one channel at a time will save you. The transformer split will, in the time it takes to repatch.

One safety note that belongs in every one of these discussions: a DI's isolation is a noise tool, not a life-safety device. White's article says it plainly, and it bears repeating, because a stage where the "the PA hums unless we lift everything" conversation is happening is sometimes a stage with a genuine wiring fault. Hum is occasionally a symptom of a shock hazard. If a performer reports a tingle from a mic or strings, stop the line check and find the fault. No transformer makes a hot chassis safe.

The switches, the thru jack, and other fine print

Every DI wears a small vocabulary of switches, and each one is a story about a gig that went wrong for somebody once.

The thru jack is the underrated one. It's a hardwired parallel tap: the instrument's signal continues to the player's amp untouched while the DI takes its copy. This is what lets the bass player keep his rig and his monitor feel while you take a clean pre-amp signal, and it's why the DI goes between instrument and amp, not on some spare output the backline tech found. Two things to know about it. On a passive box, the thru is truly passive, so it works with the battery dead, because there is no battery. On some active boxes the thru is buffered, which is lovely until the power dies and the player's amp goes silent along with your channel. Know which type is in your rack before the show teaches you.

The pad, usually 15 or 20 dB and sometimes stacked twice, exists because "instrument level" spans about 40 dB of reality. A vintage Jazz Bass and a modern synth both arrive on quarter-inch jacks, and one of them is twenty times louder. On an active DI the pad protects the input buffer from clipping; engage it for anything with a wall wart. On a passive DI the pad does something subtler: it also protects the transformer core from low-frequency saturation on hot sources, trading a little noise for a cleaner bottom octave.

Polarity flip is there for the day you're blending the DI with a mic on the same source, bass DI plus miked cab being the everyday case. The two arrive with different signs and different delays, and the flip gets you to the better of the two combs. It's a coin toss with a preference; flip it, keep whichever has more low end, then nudge the delay if the console gives you the tools.

The mono merge switch, on boxes that have it, sums a stereo source to one channel. Useful for the laptop act on a festival undercard when you have one channel to spare. Just remember it's a sum: anything the producer panned hard and out of phase disappears. When the backing tracks sound hollow, that switch is your first suspect.

None of these switches is exotic. Together they're why a good DI has outlived every format war in this industry: it's a small collection of correct answers to problems that haven't changed since the 1960s.

Speaker-level feeds and other ways to hurt yourself

Sometimes the signal you need isn't at the instrument. The bass player's entire sound is his amp's front end, or the harmonica player's rig is a '59 Bassman and a prayer. So you take the feed after the amplifier, and now the arithmetic gets spicy. A 100-watt amp into 8 ohms is about 28 volts RMS, north of +31 dBu. Your mic input wants signals ten thousand times smaller. DI boxes with a speaker-level input handle this with a heavy resistive pad before the transformer, typically 30 to 50 dB of attenuation, and the input is labeled for it. The label is not decoration. A speaker feed into an instrument input will slam any active DI into its rails instantly, and White's warning is worth taking literally: at these voltages you're risking damage, not just distortion.

Two rules keep this trick safe. First, the amp still needs its load. Tap the speaker line with the DI's thru carrying on to the cabinet, or use a proper dummy load; a tube amp driven into an open circuit can destroy its own output transformer, which turns your convenient DI feed into a very expensive lesson. Second, remember what you're capturing. A speaker-level DI feed contains everything the amp does and nothing the speaker does, and with guitar amps the speaker is half the sound. A 12-inch guitar driver is a brutal low-pass filter, rolling off steeply above 4 or 5 kHz, and the raw amp feed without that filter is a wasp in a can. That's what speaker simulators are for: passive or active networks that approximate the cabinet's rolloff. Or, in the current decade, a modeler ahead of the PA and the whole question dissolves. But festivals run on whatever walks off the bus, so the DI with the speaker pad stays in the rack.

While we're on things in the rack: buy DIs like you buy shovels. The stereo keyboard rig needs two channels or a stereo box; the quiet acoustic act needs the high-impedance active; the DJ needs the passive with the pad, because a DJ mixer at showtime outputs a signal best measured in volts and attitude. None of this is exotic. All of it is the difference between a changeover that takes four minutes and one that takes fifteen while the stage manager stares at you.

What you can take back to the desk

The DI decision tree, compressed to fit on the back of an input list. Active source (keys, laptop, active bass, anything with its own preamp): passive transformer DI. It can't clip, it can't run out of battery, and the isolation is real. Passive magnetic pickup: active DI, 1 megohm or better, so the pickup keeps its resonance. Piezo with no onboard preamp: active DI with a 10 megohm setting, no substitutes, because f = 1/(2πRC) doesn't care about your channel EQ. Amp-dependent tone: speaker-level pad, keep the load connected, expect to tame the top end.

Thin, quacky acoustic guitar is an impedance problem, not an EQ problem. Fix it at the box. Hum on a powered source is a loop problem: transformer isolation first, ground lift second, EQ never.

Spend your money on iron. The transformer is the one component in the chain where cheap has a measurable cost: less magnetizing inductance means bass rolloff into real sources, more leakage means a flabby or ringing top, and a small core saturates earlier. A good passive DI is a twenty-year purchase. Some of ours are older than the interns.

And when you flick a ground lift and the hum vanishes, take the extra ten seconds to wonder why the loop was there. Sometimes the answer is "festival power, two distros, nothing to see." Sometimes it's the answer that keeps a singer from getting bitten. The little steel box will cover for a lot of bad wiring. It shouldn't have to cover for yours.

Above the Fader

Sauce

• DI Boxes & Why We Need Them - Paul White, Sound on Sound
• Using DI Boxes - Sound on Sound
• Audio Transformers - Bill Whitlock, Jensen Transformers (handbook chapter, PDF)
• Design of High-Performance Balanced Audio Interfaces - Bill Whitlock (PDF)
• Radial JDI - product page and specifications
• Radial JDI Stereo user guide (Jensen JT-DB-E, 12:1)
• Radial Engineering PZ-DI review - Sound on Sound
• Piezo Pickup Preamplifiers - Elliott Sound Products

Featured photo by Les Taylor on Unsplash.