DAC output stages: what the chip doesn't tell you

By Marcus Vale · October 6, 2026 · 9 min read
Cambridge Audio CXN100

I had a reader email me last month asking why a $250 DAC and a $1,800 DAC built around the exact same conversion chip sounded nothing alike through his system. He'd seen the datasheet, matched the part number, and figured he'd found a giant-killer. He hadn't, not quite, and the reason why is the subject of more confusion in this hobby than almost anything else I write about.

The conversion chip — the actual digital-to-analogue converter, the bit of silicon from ESS, AKM, Texas Instruments or Cirrus Logic that does the maths — is maybe fifteen percent of what determines how a DAC sounds. The rest is the output stage: the analogue circuitry that takes the chip's raw current or voltage output and turns it into something that can drive an amplifier input without falling apart. That circuitry doesn't appear on the spec sheet brands quote at you, and it's where almost all the money in a good DAC actually goes.

What the chip actually does, and where it stops

A delta-sigma DAC chip, which is what you'll find in the overwhelming majority of consumer gear, takes an oversampled, noise-shaped digital signal and converts it into a low-level analogue current or voltage. That's genuinely the hard engineering problem, and it's one the industry solved well a long time ago. Modern DAC chips, even budget ones, measure extraordinarily well in isolation. Total harmonic distortion plus noise figures that would have been laboratory curiosities fifteen years ago are now standard in parts costing a few dollars.

But that raw output from the chip is tiny, often just a few hundred millivolts, and it's riddled with high-frequency noise from the oversampling and noise-shaping process — energy well above 20kHz that needs filtering out before it reaches an amplifier. The chip's job finishes there. Everything after that point — filtering, amplification to line level, buffering so the output can drive real-world cable capacitance and amplifier input impedance without sagging — is down to discrete analogue design. That's the output stage, and it's where a designer's choices actually show up in what you hear.

The reconstruction filter is doing more than you think

Every delta-sigma DAC needs an analogue low-pass filter after the chip to remove ultrasonic noise before amplification. Cheap designs use the chip's internal digital filter and a minimal, often single-pole, analogue stage after it — good enough to pass a distortion measurement, not necessarily good enough to avoid phase shift and ringing in the audible band edge.

Better designs spend real component cost here: matched capacitors, low-noise op-amps selected for the specific gain and bandwidth required, sometimes discrete transistor stages instead of an off-the-shelf op-amp entirely. I've measured units where the difference between the cheap and expensive approach shows up clearly in a square-wave response test — ringing and overshoot on one, a clean settle on the other — even though both pass a standard 1kHz THD+N test with near-identical numbers. That's the trap with quoting a single distortion figure: it tells you almost nothing about transient behaviour, and transient behaviour is closer to what you're actually hearing on a well-recorded cymbal hit.

Voltage output versus current output chips change the whole design

Some DAC chips output a voltage signal directly. Others — and this includes some well-regarded chips from Texas Instruments and certain AKM parts — output current, which then needs an I/V (current-to-voltage) conversion stage before anything useful happens with it. That I/V stage is itself a significant piece of analogue design, often the single biggest sonic lever in the whole circuit, and it's entirely down to the engineer building around the chip, not the chip itself.

This is why two DACs using the identical chip can measure similarly on a basic THD test and still sound different in ways that are genuinely audible, not imagined. One manufacturer might use a simple resistor-based I/V stage because it's cheap and the numbers still look fine. Another might spend real engineering time on a discrete, carefully biased current-to-voltage conversion because they know that's where the character of the final output actually gets set. Neither approach is dishonest. But only one of them is telling you the full story when it quotes "same DAC chip as the flagship" in its marketing copy.

Output impedance: the spec that quietly decides how your DAC behaves downstream

A DAC's output impedance, usually a few hundred ohms for cheap designs and well under 100 ohms for better ones, interacts with whatever you plug it into. Drive a long interconnect into a power amp with a high input impedance and a mediocre output stage might cope fine. Drive a headphone amp with a lowish input impedance, or run a long cable through a busy equipment rack, and a poorly designed output stage can start rolling off treble or picking up noise it wouldn't otherwise have.

This matters even more for DAC/amp combo units feeding headphones directly, where the output stage has to do double duty — line level plus enough current drive for real headphone loads. I've written before about how the best desktop DAC/amps on the market earn their keep specifically in this area: not in the conversion chip, which by now is close to a commodity, but in how much current the output stage can deliver without the frequency response sagging or distortion climbing as the load gets demanding. The RME ADI-2 DAC FS (check price) is a good reference point here — RME publishes genuinely detailed measurements of its own output stage behaviour under load, which is rarer than it should be in this industry and, frankly, more useful to a buyer than another digital-domain THD number quoted in isolation.

Balanced output stages aren't just about longer cable runs

A fully balanced output stage, with differential signal paths all the way from the chip to the output connector, isn't simply a cable-run convenience for professional studios. Done properly, it doubles the available output voltage swing for the same supply rails and rejects common-mode noise picked up along the signal path, which matters more in a rack with a power amp, a streamer, and a scattering of wall warts all sharing the same mains circuit than most buyers assume. Done lazily — and I've seen this inside more than one mid-priced unit — a balanced output is just a single-ended circuit with an inverting buffer bolted on afterward to generate the second phase. That gets you a balanced connector on the back panel. It does not get you the actual benefits of a genuinely differential signal path, and the two are not the same thing even though they'll both wear XLR sockets.

Chord Electronics has built a reputation partly on doing unusual, FPGA-based digital filtering work before the analogue stage even starts, which is a different approach again — their published technical notes on pulse-array DAC architecture are worth a read if you want the engineering detail straight from a manufacturer rather than filtered through a reviewer. You can see the contrast in something compact like the Chord Electronics Mojo 2 (check price), where a genuinely unusual digital front end still has to hand off to a conventional analogue output stage, and that handoff is exactly where a lot of the unit's actual character gets decided.

The power supply is part of the output stage, not a separate line item

I'll go out on a limb here: the single most underrated line item in any DAC's bill of materials is the power supply for the analogue section, and it's the one buyers never ask about because it doesn't show up as a headline spec. A noisy, poorly regulated supply feeding an otherwise well-designed output stage will dump ripple and switching noise straight into the audio path, and no amount of clever filter topology downstream fully recovers from that. This is a real reason some budget DACs with genuinely excellent conversion chips still sound flat and slightly grey next to pricier units running the same silicon — not snake oil, just a cheaper switch-mode supply doing a mediocre job and an output stage with less supply rejection built into its design.

It's also, in my view, where some flagship pricing stops making engineering sense and starts chasing diminishing returns. A well-executed linear supply with decent regulation gets you most of the way there for not much outlay. Beyond that point you're paying for marginal noise-floor improvements that are real but genuinely hard to hear in anything but a very quiet room with very revealing headphones.

What this actually means when you're shopping

None of this means ignore the DAC chip entirely, or that chip choice is irrelevant. It means stop treating the chip part number as a proxy for the finished product's sound, because it isn't one. Two units sharing a chip can behave very differently once you look past the digital-domain spec sheet and into how the output stage, the reconstruction filter, and the power supply were actually engineered around it.

Practically, this is why reading a full measurement set matters more than reading a single distortion number, and why a unit's behaviour into a real load — not just a 100k ohm test bench input — tells you more than most marketing copy ever will. If you're comparing options, our rundown of the best DACs and network streamers digs into exactly this kind of output-stage detail rather than just chip-spotting, and it's worth reading before you let a datasheet make the decision for you. The iFi Audio ZEN DAC 3 (check price) is a decent case study at the budget end of doing the output stage properly rather than cheaply, which is a different thing from doing it expensively.

The AES has published technical papers on DAC output stage design and measurement methodology for anyone who wants to go deeper than a review can take you, and it's genuinely worth the read if this stuff interests you the way it interests me. Failing that, ask a dealer what the output stage topology actually is before you ask what chip is inside. Most won't know off the top of their head. The ones who do are usually selling the better product.

— Marcus Vale, Editor — Electronics & Measurement, Sound Technology

Tagged

Common questions

If two DACs use the same chip, will they measure the same?
Not necessarily. Basic distortion figures might look similar, but things like square-wave response, output impedance under load, and noise floor depend on the analogue output stage design, not just the chip.
Does a more expensive DAC chip always mean better sound?
No. Modern DAC chips, even budget parts, measure extremely well in isolation. The output stage, filtering and power supply built around the chip usually matter more than which specific chip was chosen.
Is a balanced XLR output always better than single-ended RCA?
Only if it's a genuinely differential circuit from end to end. Some units add XLR connectors onto what's still functionally a single-ended design, which gets you the connector without the real benefit.
Why do two DACs with identical specs sound different through headphones?
Output impedance and current delivery under real headphone loads vary a lot between designs and rarely appear on a basic spec sheet, which is where a lot of the audible difference comes from.
About the author
Marcus Vale
Marcus Vale
Editor · Electronics & Measurement · Sydney, NSW

I'm Marcus, and I'll be honest up front: I trust a measurement before I trust my own ears, because my ears lie to me daily. I spent fifteen years designing audio electronics before I started writing about them, so when a brand tells me a number, I want to see the graph. That doesn't make me cold about this hobby — I love a system that disappears as much as anyone — it just means I'll tell you when an expensive box is selling you confidence rather than performance.

Former audio electronics engineer; objectivist; runs the test bench

More from Marcus Vale