---
Title: Measuring a USB audio card – AudioLab
---

# Measuring a USB audio card

A USB audio card is two devices: an **input** that records into the computer and an **output** that plays from it. Neither shares a clock with your measuring interface, so AudioLab measures them in two wirings: **External ADC** for the input, **External DAC** for the output. This page goes through both, step by step, and then through the pitfalls – most of them we met on a small USB card with a phono/line input, a line output and a headphone output ([its full protocol](measured-behringer-ufo202.html)).

Measure your interface first ([Measure your interface first](step-interface.html)): its floor tells you which of the card's figures are the card's.

## What you need

- Your measuring interface, calibrated (outputs 3–4 and inputs 3–4 in volts).
- A **USB isolator** between the computer and the card (see pitfall 1).
- Cables from the interface's outputs to the card's input, and from the card's output to the interface's inputs. For a phono input, a −40 dB attenuator ([Phono preamplifier](phono.html)).
- For a headphone output: a stereo-jack-to-two-jacks cable, and 32 Ω resistors or a load box.

## Levels in this guide

- **The input (Part 1)** gets volts from your interface: they are given in volts at the card's input, with the dBFS they were on our interface (2.45 V at full scale) in brackets. **Calibrate the interface's outputs first** ([step 1 of the interface page](step-interface.html#step-1-calibrate)): only then does the line under *Level* say the volts and do the keys 100 mV – 2 V play what they say. A sweep (*From* / *To*) is set in dBFS: dBFS = 20 · log₁₀(volts / full scale).
- **The output (Part 2)** gets numbers, not volts: AudioLab plays into the card digitally, so the level is in the **card's own dBFS**, the same on every interface; the interface's calibrated inputs then read what comes out in volts.

More on this: [Levels](levels.html#volts-for-the-device-dbfs-for-the-interface).

## Windows, once

The card's input and output are ordinary Windows sound devices. AudioLab opens them in **exclusive mode**: no Windows mixing, no resampling, no volume in the path – once Windows allows it.

![Windows sound settings, the card's recording device: format 2 channels, 16 bit, 48000 Hz; the input volume here still at 100, which on this card is about +30 dB – AudioLab's Set to 0 dB sets it right](figures/step_usb_windows_input.png)

1. *Settings → System → Sound → More sound settings*: on the **Recording** tab the card's input ("Microphone (USB Audio CODEC)" or similar), on the **Playback** tab its output ("Speakers (…)").
2. *Properties → Advanced*: tick **Allow applications to take exclusive control** and **Give exclusive mode applications priority**; untick **audio enhancements**. Note the *Default Format* list: it shows the sample rates the card really offers.
3. Close the dialogs before measuring – applying them resets the card while AudioLab holds it.

![Windows device properties, Advanced: both exclusive-mode options ticked, audio enhancements off](figures/step_usb_windows_exclusive.png)

## Part 1 – The input: External ADC

![External ADC: AudioLab plays through the measuring interface (ASIO), OUT 3 / OUT 4 feed the card's input L / R – through a −40 dB attenuator for phono, straight for line – and the card records over USB through an isolator (WASAPI, exclusive mode); two clocks, so no reference](wiring/usb_card_external_adc.svg)

**Set:**

- CONNECTIONS → *Wiring* **External ADC (USB recording), no reference**; *ADC (recording device)* = the card's input (↻ lists the devices again).
- **Set to 0 dB** next to the volume line: on a USB card the Windows input volume is a **gain in the card**, also in exclusive mode – ours at 100 added about 30 dB.
- *Sample rate* **48 kHz**, *Generator outputs* **3–4**, *Generator output limit* −6 dBFS, **997 Hz**.

**Measure, in this order:**

| Step | Task | Set | What it gives |
| --- | --- | --- | --- |
| 1 | **THD+N meter** | **245 mV** (−20 dBFS), LINE | the level the card reads for a known voltage – hence its **full scale in volts** – and THD |
| 2 | **THD vs level** | −60 to −2 dBFS, 2.45 mV to 1.95 V (limit raised for the sweep) | where it clips, and where the distortion turns up |
| 3 | **Stepped sine** | **245 mV** (−20 dBFS), 20 Hz – 20 kHz | response; the tone against the generator level |
| 4 | **Noise & S/N** | the card's input on the interface's silent outputs | noise in the card's dBFS, S/N against its full scale |
| 5 | **Crosstalk** | **245 mV** (−20 dBFS) | channel separation |
| 6 | **IMD vs level** | SMPTE and CCIF, −40 to −6 dBFS (peak 25 mV to 1.23 V) | intermodulation |
| 7 | Phono: **Stepped sine** with *Compare with phono curve* = RIAA | *Attenuator before the device* −40 dB, then about **0.8 mV** at the card's input | deviation from the RIAA curve |

![THD+N of the card's line input at 997 Hz, −20 dBFS through the isolator: the tone at −15.5 dBFS, H2 at −100 dBc, the noise near −87.5 dBFS](figures/step_usb_thdn_line.png)

**Read:** levels are in the **card's own dBFS** – it is not calibrated, and need not be. 245 mV that reads −15.5 dBFS puts the card's full scale at 1.45 V. The noise in dBFS against 0 dBFS is the input's dynamic range; a 16-bit card can reach about 98 dB.

![THD vs level of the card's line input: THD+N falling 2 dB per 2 dB to −80.7 dB at 1.23 V; THD at −98 dB up to 0.3 V, then the right channel rising to −63 dB above 0.4 V; clipping at +0.4 dBFS](figures/step_usb_thd_level_line.png)

**Why no Gain & phase:** the card runs on its own clock, so there is no reference: no phase, Gain & phase is off. **Stepped sine** reads the tone's level at each step; AudioLab measures the card's delay (ours: 250 ms) and waits that long at every step.

![Stepped sine of the card's line input: within 0.9 dB from 20 Hz to 20 kHz, +0.5 dB in the bass, −0.4 dB at 20 kHz](figures/step_usb_response_line.png)

## Part 2 – The output: External DAC

![External DAC: AudioLab plays through the card over USB, through the isolator (WASAPI, exclusive mode); the card's output L / R feeds the interface's line inputs IN 3 / IN 4, calibrated, recorded over ASIO; two clocks, no reference](wiring/usb_card_external_dac.svg)

**Set:**

- CONNECTIONS → *Wiring* **External DAC (USB), no reference**; *DAC* = the card's output.
- **Set to 0 dB** next to the DAC's volume: Windows playback volume 100. Ours at 54 played 5 dB lower.
- The card's output → **In 3 / In 4**, which are calibrated: the output is read in volts.
- **Leave the card's input open** while measuring the output (pitfall 6).

**Measure** (levels in the card's dBFS): THD+N at −20 dBFS (the output's level in volts for a known digital level, hence its full scale), **THD vs level** −60 to 0 dBFS (limit raised), **Stepped sine**, **Noise & S/N**, **Crosstalk**, **IMD vs level**.

![THD vs level of the card's output, −60 to 0 dBFS: THD lowest at −10 dBFS (−89.8 dB), THD+N best at −6 dBFS (−82.6 dB), then a jump at 0 dBFS](figures/step_usb_thd_level_output.png)

**Read the top step carefully** (pitfall 8): ours read a 25 dB jump in THD+N at exactly 0 dBFS, with lines that are not harmonics of the tone.

**A headphone output** is measured the same way, the jack through a stereo-to-two-jacks cable to In 3 / In 4, the headphone volume at maximum (and noted). Then **Output Z**: *On the device output* **Nothing**, **RUN**, choose **32 Ω** in AudioLab first, then connect the resistors – see [Headphone amplifier, step by step](step-headphone-amp.html#step-6-output-impedance).

![Output Z of the card's headphone output: the level falls 8.2 dB into 32 Ω; Z about 50 Ω](figures/step_usb_headphone_z.png)

**Other sample rates:** CONNECTIONS → *Sample rate* changes the interface and the card together. Repeat THD+N and Stepped sine at 44.1 kHz – the rate of CDs and most files.

## The pitfalls

**1 – A ground loop through USB.** The card and the interface share their ground over the cables and the computer. Without an isolator ours read mains hum, a row of lines at whole kilohertz, 11 dB more noise on the line input, 27 dB more on phono, and false THD: 0.030 % instead of 0.0016 %. A **USB isolator** between the computer and the card breaks the loop. It brought small spurs of its own (near 220 Hz and 7 kHz) – far below what it removed.

![Noise of the card's line input without the isolator: mains hum at 50 Hz and its harmonics and lines at whole kilohertz, the floor 11 dB higher](figures/step_usb_noise_ground_loop.png)

![The same through the isolator: a flat floor at −87.6 dBFS, only small spurs near 220 Hz and 7 kHz](figures/step_usb_noise_isolator.png)

**2 – 997 Hz, not 1 kHz.** The card had lines of its own at whole kilohertz; a 1 kHz tone puts its harmonics exactly on them, and they count as distortion. At **997 Hz** (AES17) the harmonics lie beside them. The *997 Hz* key sets it.

![THD+N at 1 kHz without the isolator: the harmonics land on the card's own lines at whole kilohertz and the hum is high; THD 0.030 %](figures/step_usb_ground_loop_1khz.png)

**3 – The Windows volume is a gain.** On the input it is an amplifier in the card (+30 dB at 100), on the output a level control in the card even in exclusive mode. Use **Set to 0 dB** on both, and check the line under it before every session.

**4 – Exclusive mode and enhancements.** Without exclusive mode Windows mixes and resamples; with enhancements on it processes. Allow exclusive mode, switch enhancements off, keep other programs off the card.

**5 – The interface's buffer.** With the interface's ASIO buffer at 128 samples the recording had gaps; at **1024** none. AudioLab watches every step for a gap and reads it again (*1 step(s) read again … a gap in the recording*) – one now and then is fine, many mean a larger buffer.

![The interface's driver panel: ASIO buffer size 1024 samples](figures/step_usb_asio_buffer.png)

**6 – What else is connected to the card.** Measuring the output with the phono input still connected through the attenuator, ours read the noise 20 dB higher, with hum. Leave the unused input open.

**7 – The rates a card really records.** Windows may call a rate supported and then refuse to start at it in exclusive mode: ours records only 16 and 48 kHz, and at 44.1 kHz failed (error 0x88890008). AudioLab now tries the card's other formats and then says so plainly; the *Default Format* list in the card's Windows properties shows its true rates.

**8 – Digital full scale.** At exactly 0 dBFS our card clipped the sine in its digital path: the 47th to 49th harmonics folded back below 1.2 kHz at 48 kHz, and THD+N jumped by 25 dB. At −1 dBFS all of it was gone. Read the output's maximum at −1 to −3 dBFS, and treat a step at 0 dBFS as a test of the card's digital headroom, not of its analog stage.

![THD+N of the output at 0 dBFS: lines at 844.5, 1149 and 152 Hz – the 49th, 47th and 48th harmonics folded at 48 kHz – besides the tone's harmonics](figures/step_usb_output_0dbfs.png)

**9 – The interface's own generator near full scale.** For the input sweeps the interface plays up to −2 dBFS with the limit raised; near 0 dBFS an interface's output distorts more (on an M4 its 7th harmonic). Keep it below 0 dBFS, and the limit back at −6 dBFS afterwards.

**10 – A card near the floor.** Compare every result with the interface loop. Ours read CCIF IMD of 0.0009 % on its input at 0.31 V – only 1–2 dB above the interface's own loop at that level: there the reading belongs to the measuring path as much as to the card.

**11 – Settings left over.** A phono curve left on in *Compare with phono curve*, a load left in *On the device output*, a limit left at 0 dBFS: each happened to us once. Glance at ANALYSIS and CONNECTIONS before each step; the export header lists them, so a wrong one can be found afterwards.

## What our card showed

| What | Result |
| --- | --- |
| Line input full scale, THD at 245 mV | 1.45 V, 0.0016 % (−95.7 dB) |
| Line input S/N, response | 90.5 dB(A); within 0.9 dB, 20 Hz – 20 kHz |
| Phono input against RIAA | −6.3 dB at 20 Hz, +1.3 dB at 10–20 kHz |
| Output full scale, best THD+N | 1.17 V; 0.0074 % (−82.6 dB) at 0.59 V |
| Output S/N, response | 95.0 dB(A); within 0.3 dB |
| Headphone output | 50 Ω, 0.15 mW clean into 32 Ω |

A clean line input and output; a phono input that tilts the RIAA curve, and a headphone output meant for sensitive, higher-impedance headphones.
