---
Title: Player test file – AudioLab
---

# Player: measuring with a test file

![A CD player on a wooden shelf, its display showing track 1 at 0:30](figures/cd_player.jpg)

A player with its own converter – a CD player, streamer or portable player – plays AudioLab's **test file** while AudioLab records its output. There is no reference: the player runs on its own clock.

1. MEASUREMENT → **Player** → SIGNAL → the file's sample rate and bits (CD: **44,100 Hz, 16-bit**) → **Create test file…**.
2. Copy the file to the player; connect the player's output to the device-output inputs of the wiring (on an M4 *Stereo, references first*: L → In 3, R → In 4).
3. **RUN** in AudioLab, then play the file **from the start, to the end** (2 min 50 s).

- The file fits the **present sample rate and FFT length** (its tones sit on the bins); keep them while measuring, and create the file again after changing them. Files made before 12 October 2026 are not recognised – their "1 kHz" parts played at about 1365 Hz.
- WAV or lossless FLAC only; EQ and sound effects off, no resampling (bit-perfect), the player's digital volume at 100 % and no gain (*ReplayGain*, *boost*).
- The analysis uses the Blackman-Harris window; no phase.

## For a CD player

- Create the file as **44,100 Hz / 16-bit** – any other format would be resampled by the burning software.
- Burn it as an **audio CD (CD-DA)**, one track, without *normalise*, *level* or *crossfade*, and **without pre-emphasis** (the player would switch de-emphasis on and the treble would drop by about 10 dB). A good CD-R at a lower speed: read errors are concealed by interpolation, which upsets the quiet parts most.
- A CD player gives about **2 V** at 0 dBFS (some 2.2–2.5 V). An M4 line input without a divider takes about 2.2 V, so the −1 dBFS part is close: play the start with the **Level meter** first, and if In 3/4 come near 0 dBFS or show *CLIPPING*, use a divider box and enter its *Divider*.
- Only the analogue output is measured; a digital output (coaxial, optical) needs an interface with a digital input.
- A player with switchable filters, HDCD or upsampling: measure each setting – the differences show in the top of the response and in the J-test.

## Results – what is measured and how to read it

While the file plays, the **chart** shows the spectrum of the part being played (outputs L and R), and the **status line** says which part plays and how many are done (*Now: 1 kHz at −20 dBFS · 12 of 46 parts measured*). Results fill in as the parts pass; at the end the bar and Details have them all.

**The bar** (per channel): **THD+N** (or **SINAD**) at −1 dBFS, **DR** – dynamic range A-weighted, flatness **20 Hz – 20 kHz within … dB**, and the parts measured (*46/46*).

**Details**, line by line (the values are a rough guide for a good 16-bit CD player):

| Line | Meaning | Good player |
| --- | --- | --- |
| *0 dBFS in the file gives … dBFS at the input (= … V)* | the voltage at full scale – in volts with a calibrated input; the data sheet's *output level* | CD about **2 V** (1.8–2.5 V) |
| *THD+N* / *SINAD* at −1 … −80 dBFS | distortion + noise at each level. At −1 dBFS distortion decides, in quiet parts noise (SINAD falls as much as the tone does – that is normal) | −1 dBFS: SINAD **85–95 dB** (THD+N 0.006–0.002 %); 16 bits allow no more than about 96 dB |
| *Linearity* at −40, −60, −80, −90 dBFS | how much louder (+) or quieter (−) a quiet tone comes out than it should – the converter's lowest bits | within ±0.5 dB down to −80 dBFS, ±1–2 dB at −90 dBFS (a few LSB) |
| *Dynamic range … (AES17, A-weighted)* | full scale against the noise and distortion left beside a −60 dBFS tone, A-weighted. **The most reliable noise figure** – read with a signal playing, so a muted output cannot flatter it | **95–100 dB** (16 bits: about 98 dB in theory) |
| *Idle noise … · S/N …* | noise on digital silence and its distance from full scale. Many players **mute** on digital zeros; S/N then reads unrealistically good – trust the dynamic range | S/N A-weighted 95–110 dB |
| *Response* | 20 Hz – 20 kHz: gain at 1 kHz, −3 dB points, spread. The top shows the converter's digital filter | flat within ±0.5 dB; 0.1–0.5 dB down at 20 kHz. A converter without oversampling (NOS) up to −3 dB at 20 kHz |
| *Crosstalk R → L* | the leak from one channel into the other at 1 and 10 kHz ([Crosstalk](crosstalk.html)) | better than −80 dB at 1 kHz, −70 dB at 10 kHz |
| *IMD: SMPTE … · CCIF …* | intermodulation of two tones ([IMD](imd.html)): SMPTE 60 Hz + 7 kHz, CCIF 19 + 20 kHz (sensitive to the treble and the converter's filter) | below **0.01 %** |
| *J-test: largest sideband … dBc … ps* | jitter: sidebands around the 11.025 kHz tone that appear when the data move the converter's clock, and the jitter they stand for. *No jitter sidebands above the floor* = none visible | sidebands below **−110 dBc**, jitter below about 100 ps; hundreds of ps to ns = a weak clock |

**Warnings in red** (status line, the bar's *⚠ player alters the signal*, Details): the player changes the sound – a compressor or normaliser (quiet parts more than 1 dB louder or quieter at −40/−60 dBFS), an equaliser (the treble falling already below 10 kHz), spatial sound or crossfeed (the channels mixed above −30 dB or unevenly). The figures then describe the effects, not the player: switch them off (on Windows in the device's properties) and measure again.

**No signal · check the cable** (the bar, Details, status line): the file plays on the other channel but nothing comes through on this one – 0 dBFS in the file reads below −80 dBFS. The player's output is in another input than the wiring names (on *Stereo* the device outputs are In 2 and In 4, on *Stereo, references first* In 3 and In 4). The channel then gives no figures, and the other channel no crosstalk.

**Export** (CSV) and **Report** carry the same per part, with the conditions (file, sample rate, FFT).

### Example: a DAC with a tube output stage

*A real device, measured as described: what the measurement shows and how to read it.*

A USB DAC whose output stage is a tube, fed from Windows (a 48 kHz / 16-bit file, the sound enhancements off, *Stereo, references first*: L → In 3, R → In 4). No ⚠: the linearity within 0.12 dB down to −90 dBFS shows that nothing in the playback path changes the level.

![Details of the player test of a DAC with a tube output stage: SINAD best at −20 dBFS, linearity within 0.12 dB, response −3 dB at 55 and 57 Hz and at 17.8 kHz](figures/player_tube_dac.jpg)

- **0 dBFS = 0.92 V** – below the 2 V usual for a CD player; an amplifier needs more gain for the same loudness.
- **SINAD 42 dB at −1 dBFS** (0.8 % THD+N), **59 dB at −20 dBFS**: the distortion of the tube stage falls with the level, then the noise takes over. A transistor or IC output is best at the top instead. Mostly H2: during a response part the spectrum shows it about 50 dB below the tone (below).

![While the file plays: the 159 Hz response part at −10 dBFS, H2 at 318 Hz about 53 dB below the tone, H3 about 83 dB below, a row of higher harmonics near −100 dBV, and the supply's hum at 100 Hz](figures/player_tube_dac_spectrum.jpg)

While the file plays the chart shows the spectrum of the part being played and the status line names it. A clean player shows the tone alone above the noise; here the harmonics stand out, H2 the strongest – the tube stage's signature – and the hum at 50 and 100 Hz.

- **Linearity** within 0.12 dB down to −90 dBFS: the converter itself is fine.
- **Dynamic range 85 dB (A)**, idle noise −99 dBFS: the tube stage's noise and its supply's hum (100 Hz in the spectrum) set it, not the 16 bits (about 98 dB).
- **Response −3 dB at 55 Hz (L) and 57 Hz (R)**: a coupling capacitor at the tube output with the resistance it drives forms a high-pass; the M4's line input is a lower resistance than most amplifiers, so with an amplifier the bass may go deeper. **−3 dB at 17.8 kHz**: the treble rolls off before 20 kHz.
- **Crosstalk −59 dB at 1 kHz, −38 dB at 10 kHz**: rising with frequency, a capacitive leak – usual for a tube stage with a shared supply.
- **IMD SMPTE 1.2 %**: high, as SMPTE's loud 60 Hz tone drives the tube furthest.
- **J-test: the largest sideband −73 dBc at ±250 Hz**, about 5.96 ns peak jitter, in both channels, over a floor of −93 dBc. At 48 kHz the J-test is a 12 kHz tone with a one-LSB square wave at 250 Hz; sidebands at ±250 Hz from the tone mean that the data reach the converter's clock (data-correlated jitter), as with a USB DAC that takes its clock from the stream. The figure follows from the sideband: a sideband ratio of 10^(−73/20) = 2.24 · 10⁻⁴ is a phase modulation of 4.5 · 10⁻⁴ rad peak, which at 12 kHz is 4.5 · 10⁻⁴ / (2π · 12 000 Hz) ≈ 5.9 ns. A good converter keeps the J-test sidebands below about −120 dBc, under 30 ps.

A file at 44.1 kHz played to a DAC set to 48 kHz in Windows is resampled: the treble then falls from about 16 kHz (−23 dB at 20 kHz), CCIF is missing (45/46) and the J-test means nothing. Create the file at the rate the player's output runs at.

### The player's clock

The player runs on its own clock and moves every tone by its error – a CD player typically 50–100 ppm, 1–2 Hz at 20 kHz. AudioLab estimates that error from the tones it has already recognised and looks for the next parts at the moved frequencies, so a player several hundred ppm off is followed through the whole file. **Speed variations** (wow and flutter – a tape, or the simulator's *Tape recorder*) smear the tones and move them in time instead: AudioLab then misses the higher parts and the status says *Unknown tone at …*. The test is meant for digital players; in the simulator try it with the *Headphone amplifier*.

## What is in the test file

A stereo WAV of **46 parts** separated by 0.5 s of digital silence, **169.5 s** in all. Every part is a pure sine (two for IMD) with a 5 ms fade in and out against clicks. **Every tone has a frequency of its own**, by which AudioLab recognises the part being played, even when playback starts in the middle. So the "1 kHz" tones are not exactly 1000 Hz but neighbouring bins around it (2.2 Hz apart), and the response steps keep clear of 50/60 Hz and their harmonics (hum would look like a part of the file).

The exact frequencies follow the analysis sample rate and FFT length (each tone sits on a bin centre). The table is for an analysis at **48 kHz, FFT 65,536** (bin 0.732 Hz) and a **44.1 kHz / 16-bit** file; other settings move them by a few Hz.

| From the start | Length | Part | Frequency | Level | Channels | What it gives |
| --- | --- | --- | --- | --- | --- | --- |
| 0:00.5 | 8 s | 1 kHz −1 dBFS | 999.76 Hz | −1 dBFS | L + R | level at full scale (V at 0 dBFS), THD+N |
| 0:09.0 | 5 s | 1 kHz −10 dBFS | 1001.95 Hz | −10 dBFS | L + R | THD+N, linearity |
| 0:14.5 | 5 s | 1 kHz −20 dBFS | 997.56 Hz | −20 dBFS | L + R | THD+N, linearity |
| 0:20.0 | 5 s | 1 kHz −40 dBFS | 1004.15 Hz | −40 dBFS | L + R | THD+N, linearity, compressor |
| 0:25.5 | 5 s | 1 kHz −60 dBFS | 995.36 Hz | −60 dBFS | L + R | **dynamic range (AES17)**, linearity, compressor |
| 0:31.0 | 5 s | 1 kHz −80 dBFS | 1006.35 Hz | −80 dBFS | L + R | linearity |
| 0:36.5 | 5 s | 1 kHz −90 dBFS | 993.16 Hz | −90 dBFS | L + R | linearity (converter resolution) |
| 0:42.0 | 8 s | digital silence | – | zeros | L + R | **idle noise** |
| 0:50.5 | 5 s | 1 kHz L only | 1008.54 Hz | −1 dBFS | L only | **crosstalk** L → R |
| 0:56.0 | 5 s | 1 kHz R only | 990.97 Hz | −1 dBFS | R only | crosstalk R → L |
| 1:01.5 | 5 s | 10 kHz L only | 9999.76 Hz | −1 dBFS | L only | crosstalk L → R in the treble |
| 1:07.0 | 5 s | 10 kHz R only | 10001.95 Hz | −1 dBFS | R only | crosstalk R → L in the treble |
| 1:12.5 – 2:29.5 | 31 × 2 s | response | 20 Hz – 20 kHz (below) | −10 dBFS | L + R | **frequency response**, equaliser |
| 2:30.0 | 6 s | IMD SMPTE | 60.06 Hz + 6999.76 Hz, 4 : 1 | sum −1 dBFS | L + R | **IMD SMPTE** |
| 2:36.5 | 6 s | IMD CCIF | 18999.76 Hz + 20000.24 Hz, 1 : 1 | sum −1 dBFS | L + R | **IMD CCIF** (1 kHz difference tone) |
| 2:43.0 | 6 s | J-test | 11025 Hz + 1 LSB square at 229.7 Hz | −3 dBFS | L + R | **jitter** |
| 2:49.5 | – | end | | | | |

**Response steps** (2 s each, −10 dBFS, 31 points a third of an octave apart): 19.78 · 24.90 · 31.49 · 39.55 · **52.00** · 62.99 · 79.83 · **102.54** · 125.98 · 158.94 · 202.15 · 251.95 · 317.14 · 399.17 · 502.44 · 632.81 · 796.14 · **1010.74** · 1261.96 · 1588.62 · 2000.24 · 2518.07 · 3169.92 · 3990.23 · 5023.68 · 6324.46 · 7962.16 · 10023.93 · 12618.90 · 15886.23 · 20002.44 Hz. The bold steps are moved away from 50/100 Hz (hum) and from 1 kHz (taken by the level parts). With an analysis that does not reach 20 kHz the response ends at 0.45 × the analysis sample rate.

**What the parts look like** (each panel zoomed to show the waveform; the times are from the start of the file, for a 44.1 kHz file):

![The parts of the test file: the levels from −1 to −90 dBFS and 8 s of silence (0.5–50 s); crosstalk with the tone in L only, then R only, at 1 kHz and 10 kHz (50.5–72 s); 31 response tones all at −10 dBFS from 20 Hz to 20 kHz; IMD SMPTE, a 60 Hz wave carrying 7 kHz at 4 : 1; IMD CCIF, 19 and 20 kHz beating at 1 kHz; the J-test, an 11.025 kHz sine sampled four times a period at ±0.5](wiring/player_test_file.jpg)

- **Gaps** of 0.5 s are digital zeros (also at the start and the end).
- **Dither:** 16 and 24 bits get triangular dither of ±1 LSB (TPDF), so quiet tones show the player, not the file's rounding; 32-bit float has none. The **J-test** is written as exact codes **without dither** (in 16 bits alternating 0x4000 ↔ 0x3FFF), as the test requires.
- **In the J-test panel** the dots are the samples actually stored in the file, one every 22.7 µs at 44.1 kHz; the curve is what the player's converter makes of them. Four samples a period, placed at 45°, 135°, 225° and 315° of the −3 dBFS sine (peak 0.707), all have the value ±0.5 – in 16 bits the exact code 0x4000 – so the file needs no rounding. The 1-LSB square moves them by one step (1/32768 of full scale), too little to see.
- **The J-test** (Julian Dunn): a tone at a quarter of the file's sample rate at −3 dBFS and a square wave of one LSB at 1/192 of the rate. Toggling 0x4000 ↔ 0x3FFF turns all the low bits over at once; a converter whose clock follows the data shows sidebands at odd multiples of Fs/192 around the tone. The square's own harmonics fall on the same places (about −124 dBc in 16 bits) and are not counted. The J-test is in the file only when the analysis can hear its tone (a quarter of the file rate below 0.45 × the analysis rate): a 96 kHz file (24 kHz tone) with a 48 kHz analysis leaves it out and is 6.5 s shorter.
