Audio Aliasing Test: Check Your Sound Card Online
Overview
Audio Aliasing Test. Set the Volume slider, play the descending high-frequency sweep, and listen for whether it stays a smooth, steady descent or turns into warbling, radio-tuning-like artifacts — then click whichever one you actually heard to run the Audio Aliasing Test. Picking the artifact response flags real aliasing in your sound card or DAC, a fault a lot of people wrongly blame on their headphones. Watch the free spectrum analyzer update in real time as you test.
Use this audio aliasing test to hear whether your sound card turns frequencies it cannot represent into false ones. A sine tone glides down from beyond your hearing range, and a healthy digital system produces a steadily falling pitch, so any rise is your cue that something in the playback chain is mixing the signal with a mirror image of itself.
How the Audio Aliasing Test Works
The test file starts with three short beeps, then plays a sweeping sine wave at -3 dB FS that falls from 22 kHz to 1 kHz. For the first stretch the tone sits above your audible range, so you hear only silence. When the sweep enters the band your hearing covers, you pick up a high tone that keeps dropping in pitch until it ends at 1 kHz. Use the online speaker placement angle calculator any time you need a fast, accurate figure without doing it by hand.
That falling line is what correct playback sounds like. The test tone never rises, because the generator only ever moves downward. Whatever your ears hear going up must have been invented later in the chain, by a converter or a driver, and that invention is the aliasing you are listening for. Start at a modest volume, because the first audible moment arrives suddenly.
Sweeping Sine Tone: Reading a Digital Aliasing Sound Test Result
A clean result gives you one tone that descends smoothly and nothing else. A failed digital aliasing sound test sounds like a second voice climbing while the main tone falls, or like a tone that jumps to the wrong place. Each of these is the wrong frequency being rendered, a mirrored image of the real tone rather than the tone itself.
- Single falling tone: the converter and filters are behaving, and the audio path passes the test.
- Tone that rises while it should fall: the signal is being mirrored, which means frequencies increasing instead of decreasing.
- Extra whistles or noise: images of the sweep are leaking through, often from a poor sample rate converter.
Play the same file in a second app, or in your DAW, before blaming the hardware. If the rising pitch only shows up in the browser, the browser or the operating system mixer is resampling the file badly.
Aliasing Check at 48 kHz and 44.1 kHz
Most files are produced at 44.1 kHz, but many sound devices run at 48 kHz, so the system quietly resamples. Run the aliasing check at both rates: a card that passes at 48 kHz but fails at 44.1 kHz is converting the sample rate badly. Higher-rate versions of the file, up to 192 kHz, show whether the device behaves differently when it does not need to convert, and whether the sweep frequency you hear matches the frequency the file claims.
Nyquist Frequency Fold-Back Explained
A digital system measures a wave a fixed number of times each second. That count is the sampling rate, and it sets a hard ceiling called the Nyquist frequency, which is half the sampling rate. A 48 kHz system tops out at 24 kHz, and 44,100 samples per second tops out at 22,050 Hz. Run the online 3d directional sound check to see exactly what's working and what isn't.
Below that ceiling each cycle gets at least two samples, enough for the digital-to-analog converter to rebuild the wave. Above it, the samples describe a lower tone instead, so the converter plays the wrong one. This reflection around the Nyquist limit is why people call it fold-back: the part of the tone over the limit folds down under it. In digital audio that mirrored tone is permanent once it is recorded, because the samples can no longer tell the real tone from its image.
Aliased Frequency Formula by Sampling Rate
You can predict where any tone lands. For an input signal of frequency \(f\) and a sampling rate \(f_s\), the aliased output is
$$f_{alias} = \left| f - f_s \times \mathrm{round}\left(\frac{f}{f_s}\right) \right|$$
Take a 44,100 Hz system fed a 27 kHz tone. The ratio \(27 / 44.1\) rounds to 1, so the alias is \(|27 - 44.1| = 17.1\) kHz, and your ears hear 17.1 kHz where no such tone was ever played. The table shows how the alias moves as the input climbs, using the same 44,100 Hz rate. The mirrored value falls as the input rises, which is exactly the downward-to-upward reversal you listen for in the sweep.
| Input tone (kHz) | Above the limit by (kHz) | Heard as (kHz) |
|---|---|---|
| 23 | 0.95 | 21.1 |
| 27 | 4.95 | 17.1 |
| 33 | 10.95 | 11.1 |
| 40 | 17.95 | 4.1 |
| 46 | 23.95 | 1.9 |
Past the sampling rate itself, the pattern keeps folding back and forth, so a 46 kHz input lands near 1.9 kHz, well inside the range you hear best. Notice that the fundamental frequency of the input is lost: the output is not related to it by any simple multiple.
Low-Pass Filter and Reconstruction Filter Limits
Follow the audio signal through a typical chain to see where the protection lives. An analog signal leaves the microphone, passes the input filter and gets sampled. From then on every stage is digital signal processing: the signal is gain-staged, equalised, compressed and finally rebuilt as an analog voltage. A filter can protect only the first stage. A 48 kHz converter whose Nyquist frequency is 24 kHz must remove everything above 24 kHz before sampling, because after that point the signal is just numbers. If a later DSP stage creates frequency content above the Nyquist frequency, such as distortion plugins reshaping square waves, the numbers cannot represent it and it wraps around. The digital audio signal keeps no memory of what was above the limit.
Converters do carry a low-pass filter at both ends. The one before an analogue-to-digital converter should stop tones above the limit from entering. The reconstruction filter on the output smooths the staircase wave, but it cannot remove an alias, because by then the alias is an ordinary in-band tone. Only prevention works.
Where Aliasing Artefacts Come From
Recording through a proper input filter keeps aliasing out, yet artefacts still appear when you process audio in software. Any non-linear process creates new frequencies, and when those land above the limit they fold down as an unwanted digital artifact. Because these artefacts are not related to the original note, they are easier to hear than ordinary distortion.
Sweep the fundamental frequency upward and the two families of artefacts part ways: harmonic artefacts rise with it, while alias artefacts fall. That split is how audio engineers tell the two kinds of artefacts apart by ear. Plugins that generate harmonics without protection, especially cheap saturation plugins, are the usual suspects.
Clipping Harmonics and Inharmonic Distortion
Clipping a sine wave adds odd-order harmonics at three, five, seven times the base pitch and onward. Clip a 3,100 Hz tone and the 13th harmonics already reach 40.3 kHz, over any common limit, so it returns as a 3.8 kHz image in a 44,100 Hz session. That image is inharmonic, it sits between the real overtones, and it adds a gritty distortion that no later filter can take out. Saturation behaves the same way, which is why saturated tracks can pick up digital fizz at high levels.
Synthesizers, Compression and Limiting
A raw sawtooth or square wave contains endless overtones, so a synth that does not use band-limiting folds them down and plays ghost notes that follow the wrong pitch. Hard compression and brickwall limiters reshape the waveform quickly, so pushing limiting hard produces the same artifacts. Your DSP chain, whether hardware or software, is only as clean as its worst stage, and the falling tone from this page's sweep is the quickest way to hear a stage that fails.
Hearing Fold-Back from a 32 kHz USB Sound Card
I am deciding whether a 14-dollar USB sound card is worth keeping in my podcast bag. At 41 my own hearing stops somewhere near 15 kHz, which turns out to be useful: anything I hear in the first seconds of the sweep, when the tone is still above 15 kHz, cannot be the real tone.
The card's spec sheet says its converter runs at 32,000 Hz internally, so I work out the Nyquist limit first: 32,000 / 2 = 16,000 Hz. I plug it in, set the output to 44,100 Hz, and play the sweep that starts at 22 kHz. Instead of silence I hear a tone almost at once, and it climbs. Here is what the formula \(|f - f_s|\) predicts at 32,000 Hz for the points I can check:
| Sweep tone (Hz) | Predicted alias (Hz) |
|---|---|
| 22,000 | 10,000 |
| 20,000 | 12,000 |
| 18,000 | 14,000 |
| 16,000 | 16,000 |
That matches what my ears report: a tone near 10 kHz that rises to about 16 kHz while the real sweep falls, then turns around and descends normally. The card resamples to 32,000 Hz without a proper filter, so everything between 16 kHz and 22 kHz is mirrored into the audible band. I cross-check by running the 48 kHz file; the climbing tone shifts, as it should if the converter is the cause, and the original disk file plays the same way.
The number to watch is the 16,000 Hz limit, because ordinary music carries real content above it, such as cymbal overtones, and every bit of it lands below 16 kHz as noise. My decision: the card stays at home, and I rerun the test through a second interface with the rate forced to 44,100 Hz to confirm the fault is the card and not the driver.
One caveat applies to anyone repeating this. Laptop speakers and many earbuds barely reproduce tones above 18 kHz, so a silent start proves little about the card; judge by whether a rising tone appears at all, and use headphones.
Oversampling to Remove Aliased Artifacts
Oversampling raises the working rate inside a plugin before non-linear work, which lifts the ceiling and gives new harmonics room to exist without folding. The process runs in three steps: upsampling to a multiple of the project rate, running the effect there, then downsampling back. Four-times oversampling in a 48 kHz session means processing at 192 kHz, which moves the limit to 96 kHz. The cost is CPU load, and most plugins offer it as a switch rather than leaving it on.
Anti-Aliasing Filter Trade-Offs
The anti-aliasing filter that precedes downsampling decides how well the trick works. A gentle filter lets leftovers through; a steep one adds phase shift or pre-ringing, a faint smear ahead of transients. Good filter design balances these, which is why two plugins with the same oversampling setting can sound different.
Plugin Oversampling Versus 96 kHz Sessions
Working at 96 kHz lowers audible aliasing everywhere but doubles file size and processing for the whole project, mostly to protect high frequencies that only some plugin chains put at risk. For most music production, plugin oversampling at the normal rate is cheaper. Systems built on LPCM simply carry the samples they are given, so the fix has to happen before the rate is fixed.
Aliasing Tests with a Spectrum Analyzer and Sine Sweep
You can run your own aliasing tests inside a session. Load an oscillator set to sine, place the plugin after it, and put a spectrum analyzer last. Raise the oscillator frequency toward the limit with a slow sine sweep: if lines on the display move downward while the fundamental moves up, the plugin is aliasing. Switch on its oversampling and repeat; the stray lines should shrink or vanish.
For a listening-only approach, try a blind test: ask a friend to switch an effect between oversampled and normal settings without telling you which is which. If you cannot pick the difference over several rounds, the noise from folding is below what your recording and monitors can reveal.
Cutting Alias Distortion in Mixing and Mastering
Keep real-time oversampling low during mixing to protect the CPU, then raise it for offline rendering before you export. In mastering, turn on the highest quality setting for every limiter and saturator, since they push the hardest. Replace any unmaintained effect that still shows alias distortion after oversampling, and play this page's sweep again after every change to your interface, driver or plugin settings, since a falling tone with no rising partner confirms the fix worked and the playback side is clean too.
Frequently Asked Questions
- What does an audio aliasing test check?
- It plays a sine tone sweeping downward from beyond the audible range. A healthy system gives one falling tone; a rising tone means high frequencies are being folded back into the audible band.
- What is the Nyquist frequency?
- It is half the sampling rate, the highest frequency a digital system can represent. At 48,000 Hz it is 24,000 Hz; at 44,100 Hz it is 22,050 Hz.
- How do you calculate the aliased frequency?
- Subtract the nearest multiple of the sampling rate from the tone frequency and take the absolute value. For a 27,000 Hz tone at 44,100 Hz the result is 17,100 Hz.
- Why do plugins cause aliasing?
- Non-linear effects such as saturation, clipping and hard limiting create harmonics. Those above the Nyquist limit fold down as inharmonic tones unless the plugin oversamples.
- Does oversampling remove aliasing?
- It greatly reduces it. The plugin processes at a higher rate, then a steep anti-aliasing filter removes content above the original limit before returning to the session rate.
- Is a higher sample rate like 96 kHz better?
- It raises the Nyquist limit and lowers audible aliasing, but doubles file size and CPU load for the whole project, so plugin oversampling is often the cheaper fix.
- Why can I not hear anything at the start of the sweep?
- The sweep starts above the range most people hear, around 22 kHz. Only when it reaches your audible range, or when aliasing creates a lower tone, do you hear anything.