---
Title: Headphone amplifier and USB DAC – AudioLab
---

# Headphone amplifier and USB DAC

![A tube headphone amplifier with a USB DAC on a desk, open-back headphones plugged in and a laptop playing music beside it](figures/headphone_amp.jpg)

**An analogue headphone amplifier** is wired like any line device – *Stereo, references first*, the reference at its input ([Wiring](wiring.html#which-wiring)) – with its outputs through the load box below. Every step with what to set and why, and the results of a real amplifier: [Headphone amplifier, step by step](step-headphone-amp.html). This page is the overview: what to measure, what a good value is, the load box, and a USB DAC.

## What to measure: headphone amplifier

**Why:** power into your headphones, output impedance (it changes how headphones sound), hiss with sensitive IEMs.

**Connect:** output → **load resistor** 32 Ω (later 300 Ω), the interface input across it (IN 3 / IN 4 with *Stereo, references first*, or through the load box below). A balanced 4.4 mm or XLR output goes balanced to a TRS input (tip +, ring −), the minus never to ground.

**Set once:** *On the device output* = the resistor (32 Ω) · inputs calibrated.

| What | Task | Set | Good value |
| --- | --- | --- | --- |
| Output impedance | Output Z | 1 kHz, low level; *On the device output* = *Nothing (open)*, then the resistor (32 Ω) | below 1 Ω excellent, below 2 Ω good (rule of 1/8 of the headphone impedance) |
| Power at 1 % THD+N into 32 and 300 Ω | THD vs level | 1 kHz, Load 32 / 300 Ω | as rated (e.g. 500 mW into 32 Ω) |
| THD+N at listening level | THD+N meter | level for about 10 mW | 0.001–0.01 % |
| Noise | Noise & S/N | volume at its usual position; read the noise in µV (Details) | below about 5 µV is silent with IEMs |
| Crosstalk | Crosstalk | stereo wiring | −60 to −80 dB; balanced better |
| Frequency response | Gain & phase / Stepped sine | about 0.25 V at the line input | ±0.5 dB |

## USB DAC, or a headphone amplifier with a converter

A device with its own converter is measured as a whole: AudioLab plays the test signal into it over USB, digitally, so its level is in the **DAC's own dBFS**. Step by step on a real card's output: [Measuring a USB audio card, part 2](step-usb-card.html#part-2-the-output-external-dac).

- **CONNECTIONS → Wiring: External DAC (USB), no reference**, then the **DAC** among the Windows playback devices (↻ lists them again). Its output L / R → **In 3 / In 4**, calibrated; a headphone output into a load resistor (32 Ω, 300 Ω). The interface's outputs stay silent.
- **Set to 0 dB** next to the DAC's volume: the Windows playback volume is a level control in the device even in exclusive mode. Allow exclusive mode in its Windows properties and let nothing else play on it. Keep the device's own volume knob at a fixed position (maximum for a line output) and note it.
- The DAC runs on **its own clock**: no reference, no phase. **Gain & phase** and **Tape response** are off – use **Stepped sine** – and *Subtract reference* is unavailable. Playback through Windows lags 100–200 ms, so stepped measurements wait longer at every step.
- **Its output voltage at 0 dBFS:** the calibration wizard plays the tone from the DAC; with the multimeter at its output it shows the DAC's voltage at 0 dBFS at the present volume (not stored).
- A **balanced output** (4.4 mm, XLR) goes balanced to TRS (tip +, ring −), **never its minus to ground**. Both on the same computer's USB, watch for a **ground loop**: 50 Hz hum in the spectrum; a USB isolator breaks it.
- **Headphone power:** load the output with a resistor rated for the power (32 Ω / 2 W) and choose the same value in **On the device output**: THD vs level then runs over power, and the bar gives the **power at 1 % THD+N**.
- In the demo the DAC is the *Simulated USB DAC* (+30 ppm, 120 ms latency).

| What | Task | Set | Good value |
| --- | --- | --- | --- |
| Output voltage at 0 dBFS | Level meter / calibration wizard | calibrate IN 3 and IN 4 with the DAC tone | line about 2 V, balanced about 4 V |
| THD+N at −6 (the generator maximum) and −10 dBFS | THD+N meter | 1 kHz | 0.001 % and better, often at the interface's limit |
| Linearity, dynamic range | THD vs level (down to −90 dBFS), Noise & S/N | *S/N against a test tone* off | DR (AES17) above 110 dB |
| Frequency response | **Stepped sine** (Gain & phase is not available) | 20 Hz – 20 kHz; at 96/192 kHz to see the filter | ±0.2 dB to 20 kHz |
| Power, output impedance | THD vs level, Output Z | Load 32 Ω | Z below 1 Ω |
| Crosstalk | Crosstalk | – | −70 dB and better |
| IMD | IMD | CCIF at 96 kHz, SMPTE | below 0.001 % |
| USB noise | Spectrum (and its Spectrogram) | – | no spikes, no 50 Hz from a ground loop, nothing that comes and goes |
| Jitter | Player (the test file played bit-perfect through the DAC) | its J-test part; a 44.1 or 48 kHz file | no sidebands above the floor, or below about −120 dBc |

With top DACs you mostly learn that they are better than the interface; comparing two devices is the useful part.

## A load box for headphone amplifiers

One box gives every load AudioLab offers and a safe tap for the interface:

![Headphone load box: the amplifier's headphone output (TRS jack) feeds a 2-pole 6-position switch with loads of 16, 32 (16 + 16), 300 and 600 (300 + 300) ohm from each channel to ground, off at both ends; each channel also feeds a 1k-1k-1k divider to ground, and the middle resistor is taken out balanced (tip and ring, sleeve not connected) to the interface's line input](wiring/headphone_load_box.png)

- **SW1** (2 poles, 6 positions) puts the same load on L and R: *off*, **16 Ω**, **32 Ω** (R4 + R5), **300 Ω**, **600 Ω** (R6 + R7), *off*. Choose the same value in **On the device output**. Use resistors rated for the power (16 Ω and 32 Ω: 2 W or more).
- **The tap to the interface** is a 1 kΩ + 1 kΩ + 1 kΩ divider on each channel, read **balanced across the middle resistor** (tip and ring of J2/J3, sleeve not connected): a third of the voltage, about −9.5 dB, and no ground shared through the tap – it helps against the hum of a ground loop. Connect J2 and J3 with **TRS** cables to the line inputs (In 3, In 4).
- **The divider's value:** the interface's input impedance lies across the middle resistor, so the real ratio is a little lower than −9.5 dB (about 10.3–10.6 dB with an M4's line inputs). Calibrate In 3 and In 4 straight, as usual, and **measure the divider** with the calibration wizard (*Dividers only ▸ → Measure dividers…*) at the start of every session, then check it with a multimeter at the headphone jack ([step by step](step-headphone-amp.html#step-2-the-load-boxs-divider)).
- The divider (3 kΩ, about 2.9 kΩ with the interface's input) loads the amplifier in parallel with the chosen load. Measure the loads at the headphone jack with the divider in (300 Ω reads about 272 Ω) and enter them in *Measured…*; the power in W is then read from the voltage across the load itself. In the *off* positions the amplifier still sees the 2.9 kΩ: enter it in the **Off** row of *Measured…* so that **Output Z** takes it into the open reading – left out, Z into 300 Ω reads about 10 % low, into 32 Ω about 1–2 % ([Output Z](output-z.html#more-accurate-z-measured-loads-and-path-resistance)).
- **The interface's floor rises by the divider's loss.** Its own noise (about 14 µV on an M4's line input) reads about 50 µV at the headphone jack through 10 dB of divider: low THD, IMD and noise through the box show the interface. Read the noise with the headphone output straight into the interface as well (the dividers set to 0 for that).
