If you work with digital audio or are just starting to explore the world of recording and mixing, you will have come across terms such as sampling, oversampling, aliasing and bit depth<\/b>.<\/p>\n
You can find them in your settings. DAW (Digital Audio Workstation)<\/b>, plugins, sound cards and AD\/DA (analogue-to-digital and digital-to-analogue) converters. But how do they actually work?<\/p>\n
This guide, written in a simple and straightforward manner, yet sufficiently comprehensive, will assist you in:<\/p>\n
We will not use complex formulas, but clear concepts and many practical examples<\/b> to enable you to understand and apply this information immediately in your work.<\/p>\n
1. Sampling Frequency: what it is and how to choose the right value<\/b><\/p>\n1 \u2013 What is sampling?<\/b><\/h2>\n
Imagine you want to transform an analogue sound (i.e. the electroacoustic signal produced by a voice, an instrument or any other sound source) into a numerical sequence.<\/p>\n
Since a computer can only handle discrete values (fixed, non-continuous numbers), it must \u201cphotograph\u201d the value of the sound signal at precise moments and at regular intervals. Each photograph represents a sample<\/b> (or sample<\/i>), which is a measurement of the intensity of the waveform at a given moment.<\/p>\n
The more often we take these \u201cphotographs\u201d, the more detailed and fluid the digital representation of the sound will be, and the more faithful the result will be to the original.<\/p>\n
Practical example:<\/b><\/b><\/p>\n In the digital world, this acquisition frequency is called sampling frequency<\/b> and is measured in kHz (kilohertz)<\/b>, which indicates how many samples are recorded per second<\/b>. For example, a sampling frequency of 44.1 kHz<\/b> means that the system records 44,100 samples per second<\/b>.<\/p>\n Sampling frequency<\/b><\/b><\/span><\/p>\n<\/td>\n Where it is used<\/b><\/b><\/span><\/p>\n<\/td>\n Pro<\/b><\/b><\/span><\/p>\n<\/td>\n Against<\/b><\/b><\/span><\/p>\n<\/td>\n<\/tr>\n \u2705 Use 44.1 kHz for pop, electronic and rock music. It is the standard, compatible everywhere, no conversion is needed when exporting (bouncing) for streaming platforms or CDs, and it already has optimal quality, which is more than sufficient. \u274c It makes NO sense to record at 192 kHz for pop song tracks.<\/b> The improvement will be imperceptible, but the load on the CPU and the space occupied by the files will increase dramatically, putting the system under strain after just a few tracks of recording or certainly during mixing, as you add equalizers, compressors and other effects.<\/p>\n L'oversampling<\/strong> is a technique that consists of temporarily processing the audio signal at a higher sampling frequency<\/strong> compared to the one set in your recording or mixing session.<\/p>\n Practical example:<\/strong> Oversampling does the same thing with sound: it increases the number of samples to improve the accuracy of plugins. equalisation, compression, distortion and other effects<\/strong>, reducing the risk of aliasing<\/strong> (an issue we will look at in the next point).<\/p>\n Many modern plugins offer the option to enable oversampling with a simple button. In other cases, you can select the desired value from a drop-down menu, expressed as multipliers of the session sampling frequency (e.g. 2x, 4x, 8x, etc.<\/strong>).<\/p>\n What do 2x, 4x, 8x mean?<\/strong> When the plugin applies its processing (e.g. distortion or saturation), it does so on the version oversampled<\/strong> of the signal. After processing, the signal is returned to its original frequency (downsampling) so that it blends seamlessly with the rest of the mix.<\/strong><\/p>\n Oversampling It must NOT be activated randomly.<\/strong>. Using it excessively can unnecessarily overload the CPU, so it is important to choose the right value based on the project.<\/p>\n Here is a small practical diagram<\/strong><\/p>\n Session frequency<\/b><\/b><\/span><\/p>\n<\/td>\n Recommended oversampling<\/b><\/b><\/span><\/p>\n<\/td>\n Notes<\/b><\/b><\/span><\/p>\n<\/td>\n<\/tr>\n Summary:<\/strong><\/p>\n Oversampling increases the load on the CPU exponentially:<\/p>\n \u274c If you enable oversampling on too many plugins, your DAW may slow down, crash, or introduce latency.<\/strong>. Practical example:<\/strong> \u2705 Use it sparingly.<\/strong>: only in plugins where you hear aliasing or digital artefacts. If you learn to use it wisely, oversampling can greatly improve the quality of your mix without putting too much strain on the system.<\/p>\n L'aliasing<\/strong> This is a problem that occurs when an audio signal contains frequencies that are too high for the sampling capacity of the digital system. When these frequencies exceed the Nyquist frequency<\/strong>, cannot be recorded correctly and are distorted, transforming into spurious frequencies<\/strong> that did not exist in the original signal.<\/p>\n The sampling theorem of Nyquist\u2013Shannon<\/strong> establishes that in order to accurately represent a signal, it must be sampled at a frequency of at least double<\/strong> relative to the maximum frequency present in the signal itself. Half the sampling frequency is called Nyquist frequency<\/strong> and represents the maximum limit beyond which the system can no longer correctly represent the original frequencies.<\/p>\n When the signal exceeds this limit, the excess frequencies are not deleted<\/strong>, but yes reflect<\/strong> in the audible spectrum by generating aliasing<\/strong>. This phenomenon is called folding<\/strong>, as unwanted frequencies are \u201cfolded\u201d downwards, producing unpredictable and irreversible distortions<\/strong>.<\/p>\n Aliasing appears as spurious frequencies<\/strong> which may sound like:<\/p>\n This is a particularly common problem when working with: Example: 40 kHz frequency in a 48 kHz session.<\/strong><\/p>\n The Nyquist frequency is always half<\/strong> of the sampling frequency:<\/p>\n 48 divided by 2 = <\/span>24<\/span>kHz<\/span><\/span><\/span><\/span><\/span><\/span><\/p>\n So the Nyquist frequency is 24 kHz<\/strong>.<\/p>\n Yes! Our original sound is 40 kHz<\/strong>, which is highest<\/strong> Of 24 kHz<\/strong> \u2192 so there will be aliasing<\/strong>.<\/p>\n We must subtract the twice the Nyquist frequency<\/strong> from the original frequency:<\/p>\n 24 kHz x 2 = 48 kHz Result<\/strong>: What does this mean in practice?<\/strong> The higher the original sound is compared to Nyquist, the more spurious aliases appear at lower, audible frequencies.<\/strong><\/p>\n If the original sound is much higher than the Nyquist frequency, the process repeats cyclically, generating multiple aliases<\/strong>.<\/p>\n Using anti-aliasing filters in AD\/DA converters<\/strong><\/p>\n Professional sound card converters apply a low-pass filter to eliminate frequencies above Nyquist before the signal is digitised. This prevents aliasing during recording, but does not solve the problem in digital processes during mixing. If aliasing is subsequently generated by a plugin, this type of filter cannot intervene.<\/p>\n Enable oversampling in plugins<\/strong><\/p>\n Some distortion, saturation, synthesis, and compression plugins can generate high harmonics that exceed the Nyquist frequency, creating aliasing. Oversampling allows them to operate at a higher frequency, reducing the risk before downsampling. This feature should be used judiciously, as it increases the load on the CPU. If a plugin offers multiple oversampling levels (2x, 4x, 8x), in most cases 2x or 4x are sufficient.<\/p>\n Working at a higher sampling frequency (only if necessary)<\/strong><\/p>\n If a project uses a lot of digital effects and software synthesizers, working at 88.2kHz or 96kHz can reduce the risk of aliasing without the need for oversampling. For standard music productions, 44.1kHz or 48kHz are generally sufficient, especially if the plugins are well designed and oversampling is only enabled where necessary. Frequencies above 96kHz rarely bring audible benefits and unnecessarily increase the system workload.<\/p>\n Use equalizers and filters to attenuate problematic frequencies.<\/strong><\/p>\n Some precise equalizers and low-pass filters can reduce aliasing by drastically cutting ultrasonic frequencies (i.e., above 20,000 Hz, which could be reflected in the audible range). This technique is useful when a plugin generates aliasing and does not have an oversampling option. However, this process can sometimes produce phase errors near the cutoff frequency, making the super-high range less crystal clear.<\/p>\n Conclusion<\/strong><\/p>\n Anti-aliasing filters on sound cards only solve the problem during recording. To avoid aliasing in digital processes, you can enable oversampling in the most critical plugins or work at a higher sampling rate if necessary. In some cases, a low-pass filter can help eliminate problematic frequencies. If you don't hear aliasing in your mix, you probably don't need to change anything.<\/p>\n Are you using a software synthesiser and notice that when you play very high notes, the sound becomes metallic and unnatural? Practical summary:<\/strong> Aliasing is a predictable problem that can be solved by carefully managing sampling settings and plugins. Avoiding it will help you achieve a cleaner, more natural sound<\/strong>, without unwanted digital artefacts.<\/p>\n If the sampling frequency<\/strong> determines how many times per second is sound measured<\/strong>, the bit depth <\/strong>establishes how accurately each measurement of each individual sample is recorded<\/strong><\/p>\n The greater the bit depth, the greater the dynamic range<\/strong>, i.e. the difference between the weakest and strongest sounds that the system can record without distortion.<\/p>\n\n
Sampling frequencies in practical use<\/b><\/h3>\n
\n\n
\n \n \n \n \n \n 44.1 kHz<\/span><\/td>\n Audio CDs, streaming, radio, multi-track recording<\/span><\/td>\n Good quality standards, universal compatibility<\/span><\/td>\n May lose detail in high frequencies<\/span><\/td>\n<\/tr>\n \n 48 kHz<\/span><\/td>\n Professional audiovisual production, multitrack recording<\/span><\/td>\n Greater definition in high frequencies compared to 44.1 kHz<\/span><\/td>\n Increased CPU and memory usage<\/span><\/td>\n<\/tr>\n \n 88.2 \u2013 96 kHz<\/span><\/td>\n Professional studios, film and video post-production, mastering<\/span><\/td>\n Greater detail fidelity, lower risk of aliasing<\/span><\/td>\n Little audible difference compared to 48 kHz, very large files, high CPU load<\/span><\/td>\n<\/tr>\n \n 176.4 \u2013 192 kHz and above<\/span><\/td>\n Professional firms with enormous computing power resources<\/span><\/td>\n Greater fidelity, useful for extreme editing, mastering, lower risk of aliasing<\/span><\/td>\n No audible difference compared to 96 kHz, huge files, extremely high CPU load<\/span><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n How to choose the right sampling frequency for your home studio work?<\/b><\/h3>\n
\n\u2705 Use 48 kHz if you are working with audio for video. The improvement over 44.1 is practically inaudible, but it is the standard for films, television, and podcasts, so it may be preferable to avoid conversions if you are creating a soundtrack or working on a DVD or Blu-ray disc.
\n\u2705 Only use 88.2 or 96 kHz if you are recording very high-quality acoustic sources that require few recording tracks. It is a good solution to use it for mastering, strictly using a frequency that is exactly double that used for recording the tracks (to avoid uneven truncation of the samples).
\n<\/b>\u2705 Use 176.4\u2013192 kHz or higher only if you are doing very high-quality sound design (a professional activity that is rarely done in a home studio).<\/p>\n2 \u2013 What is oversampling?<\/strong><\/h2>\n
Imagine you have a video at 30 frames per second (fps)<\/strong>. If you convert it to 120 frames per second<\/strong> Before applying graphic effects and then returning to 30 fps, the effects will be smoother and more precise, avoiding visual artefacts.<\/p>\nHow do you activate oversampling?<\/strong><\/h3>\n
If your session is set to 44.1 kHz<\/strong>, oversampling 2x<\/strong> will make the plugin work at 88.2 kHz<\/strong>, a 4x<\/strong> a 176.4 kHz<\/strong>, and so on.<\/p>\nWhich oversampling values should be used?<\/strong><\/h3>\n
\n\n
\n \n \n \n \n 44.1 kHz \/ 48 kHz<\/span><\/td>\n 2x (recommended for plugins that introduce distortion), 4x (in rare cases)<\/span><\/td>\n To reduce aliasing in distortions, saturations, digital synths<\/span><\/td>\n<\/tr>\n \n 88.2 kHz \/ 96 kHz<\/span><\/td>\n 2x only in extreme cases<\/span><\/td>\n Usually not necessary, the sampling frequency is already high, you could only use it for distortion plugins.<\/span><\/td>\n<\/tr>\n \n 176.4\/192kHz or higher<\/span><\/td>\n No oversampling<\/span><\/td>\n Completely useless, just a waste of resources<\/span><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n \n
Disadvantages of oversampling: watch out for the CPU!<\/strong><\/h3>\n
\n
\u2705 If you notice performance issues, try reducing oversampling or freezing\/rendering the heaviest tracks.<\/strong><\/p>\n
Do you have 10 distortion plugins on different tracks, all with 8x oversampling enabled?<\/strong> Your computer will probably start to suffer.
Do you only have 1 or 2 plugins that cause aliasing?<\/strong> Enable oversampling 2x or 4x only on those<\/strong>.<\/p>\nConclusion: when and how to use oversampling?<\/strong><\/h3>\n
\u2705 2x is often sufficient<\/strong> to improve sound quality.
\u2705 More than 4x rarely brings audible benefits.<\/strong>
\u274c Avoid activating it on all plugins at random!<\/strong> You may slow down your computer unnecessarily without any real gains in quality.<\/p>\n3 \u2013 Aliasing: the invisible enemy of digital technology<\/strong><\/h2>\n
What is aliasing?<\/strong><\/h3>\n
\nHow does aliasing manifest itself in audio?<\/strong><\/h3>\n
\n
\u2705 Digital synthesizers<\/strong>, which can generate frequencies above Nyquist.
\u2705 Digital distortions and saturations<\/strong>, which create high harmonics that can become aliasing.
\u2705 Pitch shifting and time-stretching effects<\/strong>, which manipulate frequencies in ways that can generate aliasing.<\/p>\nThe aliasing formula: how is the spurious frequency added to the audible sound calculated?<\/strong><\/h3>\n
Step 1: Calculate the Nyquist frequency<\/strong><\/h3>\n
\nStep 2: Is the sound above the Nyquist frequency?<\/strong><\/h3>\n
\nStep 3: We calculate the aliasing frequency<\/strong><\/h3>\n
\n<\/span><\/span><\/span>40 kHz \u2212 48 kHz = \u2212 8 kHz, which, converted to positive, brings us back to 8 kHz.<\/span><\/span><\/span><\/p>\n
The original 40 kHz sound cannot be recorded correctly and will be transformed into a spurious 8 kHz alias that will be added to the clean signal, contaminating it.<\/p>\n
\n
If you have a sound source that emits a component at 40 kHz<\/strong>, but you are recording at 48 kHz<\/strong>, your system will not record the actual sound at 40 kHz<\/strong>, but will turn it into a false 8 kHz sound<\/strong>, which was not present in the original signal.<\/p>\nHow to avoid aliasing?<\/strong><\/h3>\n
\n Practical example: aliasing in a synthesiser<\/strong><\/h3>\n
Solution<\/strong>:<\/p>\n\n
\nConclusion: aliasing can be avoided!<\/strong><\/h3>\n
\u2705 If you use digital synthesizers or distortions<\/strong>, activate oversampling in plugins<\/strong>.
\u2705 If you work with actual audio recordings<\/strong>, the anti-aliasing filters on the sound card are already doing their job<\/strong>.
\u2705 Not necessary always work at 96 kHz or higher<\/strong>, If you do not have any specific requirements, but if you can do so, you can ensure that you prevent it almost entirely from the outset.<\/p>\n
\n4 \u2013 Bit depth: 16, 24 or 32?<\/strong><\/h2>\n
What is bit depth?<\/strong><\/h3>\n
\n