Table of Contents
The project is an article about the best ways to resize images in photoshop, giving birth to the topic of How to resample in Photoshop. As part of this article, please include your personal preferences for each method and why they work well, as well as the benefits and drawbacks of each method.
By the way, the best ways to use Photoshop for resampling will be easy to understand.
What is resampling?
In short, it is a way of changing sampling frequencies to make a sample sound the same as another one. The general idea of resampling is that you can get a different output from an input using a different sampling rate. There are many ways of doing this – here I’ll discuss four methods for doing so:
3) Samplerate and
4) Fade In & Out. Resampling usually involves making one sample sound like another while retaining its original timing and amplitude information.
The word “resample” often also refers to the process of changing the sampling depth (number of bits that represent each sample, or quantization) of an audio file, which is usually done with software. For example, a signal with 8-bit quantization can be resampled to 16-bit when being converted from a digital audio format that supports it, such as DSD.
Resampling does also involve some kind of conversion; it is just the term used to refer to any form of conversion. It can be possible to convert an audio file without resampling that file at all (i.e., without altering the original time or amplitude information). An example would be converting between 16-bit and 24-bit samples, without any resampling.
For example, let us consider a WAV file called sound.wav. If you take this file and convert it to a new sample rate by multiplying all the samples by two, then you can say that you have resampled that sound to be at double its original speed. If you wanted to change the speed of the song without changing its pitch, however, you could use resampling so that it plays back at half speed after conversion – or at any other speed you want.
This relates to converting because there is often little difference between the two terms. Sometimes they are used interchangeably – especially when referring to digital audio workstation software. But in the case of computer software, there is often a slight difference between what you do when you convert an audio file and what you do when resampling that same file.
Converting doesn’t work quite the same way as resampling does. For example, if you convert an MP3 to another format by sampling at 256 samples per second, the result will be completely different than if you started by taking the same input and resampling it at twice that speed (i.e., 512 samples per second), because conversion changes time while resampling changes amplitude. So sometimes, “converting” is used to denote resampling an audio file to some other format, and sometimes it is used as a synonym for both.
While these methods can all achieve the same results, they are all different in some aspect or another. The differences amount to the ways in which the sampled sound is affected and the affect that it has on the final output. We can define resampling as: The process of resampling a digital signal in order to change its timbre or timbre characteristics for a given purpose; also: an audio sample that was resampled in this way. This definition helps explain how one could reverse the effects of each method by performing a conversion, i.e. converting from one sampling rate to another (from an original higher sampling rate to a lower one).
What is Converting?
Converting is done by dividing or multiplying each sample’s value by some factor (usually 2). The sound sample is not changed, however. So let us take a 16-bit audio file with a sample rate of 44100 samples per second, and convert it into an MP3. This is done by multiplying every sample by 4.541 – that means each sample’s value must be divided by 4.541 in order to keep the same amplitude as the original sound. In this case, converting does not change the sound at all, but merely makes it easier to compress the data into a small space for storage or transmission purposes.
When using resampling, a new file is made based on a different sampling rate than the original file was recorded at. The sample rate of the new file is usually greater than or equal to that of the original. Let us take a 16-bit MP3 at 44100 samples per second. If we resample that file at 96000 samples per second, then we can say that resampling has been done. This is not the only way to do this, however – it can be done in any number of ways; each method will have its own advantages and disadvantages.
One way to use resampling is to take an audio file and resample it by multiplying all its samples by a given number (usually 2), and making the original file sound like it was recorded at double it’s original sampling rate (i.e., 44100/2). This technique is also known as pre-reverb: the resampling is done to achieve a certain reverb property.
One advantage of a pre-reverb is that it can be used to achieve any desired reverb tailored for a particular sound. If the original file was recorded with no reverb, this method will generate an unprocessed signal, and change it through some form of reverberation. If the original file was recorded with ambiance or some other reverb at its beginning or end, then pre-reverb works by multiplying those effects by 2 and using that as part of the final output.
Pre-reverb can be used to make it seem as if the sound was recorded in a different environment, but to do this requires some form of echo or echo simulation. This is usually achieved by adding a small delay to the output of the file, with that delay being set by how far away the sound was recorded from.
Advantages of pre-reverb
The main disadvantage to pre-reverb is that it will not take into account how many reflections came before or after the original sound – because there are no reflections. This means that applying reverb intentionally this way will not match any other sounds included in the same audio file unless they are also processed with pre-reverb.
A related method is called looped reverb. This approach is used to smooth out the sound at the beginning or end of an audio file, by adding a short burst of reverberation that sounds natural with the rest of the output. The disadvantages are that it will only match any other sounds in the same file if they are also processed with looped reverb, and its method of matching is not as accurate as pre-reverb. Also, looping ambiance can make it seem as if there were too many echoes or reflections – which would not be good for most applications.
Finally, some resampling techniques are used to change the pitch without changing the tempo. This is often done using a VST plugin to change the pitch while preserving its tempo. By using resampling at a higher sampling rate than the original, it is possible to change the pitch of an audio file while keeping its original tempo and timing. This resampling method is used more often than pre-reverb and looped reverb, however, it has some disadvantages. Changing the sampling rate alone can cause problems with some instruments – especially percussion instruments such as drum kits.
There are many different methods for converting an audio file from certain sampling rates to others. Some of these methods are more complex than others, with some being far more accurate than others. High accuracy is something that is very important when converting between sampling rates, as one would not want to have the pitch changed by any significant amount.
There are many ways to achieve high accuracy without too many difficulties. Although there are many advantages to using resampling compared to converting, there are also some disadvantages to using resampling.
Disadvantages of Resampling
The main disadvantage of resampling is that it does not take into account how many reflections came before the original sound – meaning you cannot apply reverb intentionally this way unless they are also processed with pre-reverb. The same applies to looped reverb. Also, resampling will change the pitch of an audio file as well as its tempo. This can cause problems with some instruments – especially percussion instruments such as drum kits.
Methods for Converting between Sampling rates
There are many different methods for converting between sampling rates, and each method has its own advantages and disadvantages. To compare these methods we will look at four different scenarios:
Resampling is a process of changing the sampling rate and keeping the duration (or length) of the sound file unchanged. It can be done in three ways:
Converting is a process of changing either the sample rate or the number of bits used to represent each sample; this may involve changing both of them at once. This can be done in four ways:
A sampling rate converter (SRC) is a device or program that converts the sampling rate of an analog signal. It is called a “converter” because it both lowers and raises the sampling rate, unlike a decimator, which only lowers the sampling rate. A typical use of an SRC is to convert to and from CD-format audio (44100 Hz sample rate). The process can involve interpolation, resampling (multiplying), or neither.
One application of SRCs is to convert between different sampling rates. However, because of the amount of work involved in the interpolation step (see below), some people prefer to use them only for resampling and leave resampling and interpolation to stand-alone programs.
Interpolation is a method used when converting between sampling rates. The idea is that some instruments’ playback rate changes over time, so we sample at multiple points in time and then blend them together. It can be very hard to do accurately on real instruments, but it does work fairly well for synthesizers and other digital musical instruments. Some software (such as the popular Audacity program) does not support sample rate conversion via interpolation, while others (such as Sox) do.
Interpolation is typically only used on the input side; i.e., when converting to a new format. It may be used on output, but in practice, this is usually just implemented by resampling rather than interpolation.
Converting is typically done via resampling or interpolation, or both, with no real consensus which way is best.
One application of resampling is to convert the sample rate of an audio file so that it can be burned to an audio CD. In particular, a sampling rate of 44100 Hz is used for the Red Book standard – the physical medium on which all audio CDs are recorded.
Due to technical reasons, some software can be configured to use a different sampling rate than 44100 Hz. This should be done only when necessary and with care, as over-sampling or under-sampling can result in audible artifacts. Software that does this is known as a “resampler”. Some resampling drivers offer other options as well, such as interpolation for “upsampling” or dithering for “downsampling”. In this case, the driver may decide which method to use by trial and error or internal algorithm.
Note: That the implementation of resampling and interpolation is identical to normal conversion, except that it adds “somewhere” before or after the audio. Therefore, a resampler can be considered a kind of converter.
Converting may also be done during recording by simply storing samples at different sample rates in a file. This can be done either with a special file format or with a codec, depending on the operating system and hardware involved. The file must then afterward be converted to one format by audio software such as Audacity, which supports this feature. For example, recording from an audio CD at 44100 Hz using this method will result in a sampling rate of 96000 Hz after conversion.
In books and newspapers, it is common practice to have a note in parentheses after a number, indicating the resolution that the number represents. For example: “x = 4000/16 = 400” means that four times oversampling will achieve the same result as one full resampling. In reality, it is not quite exact, because when keeping the format, the number of samples in the new format is different from when keeping the sample rate.
In contrast to resampling, “interpolation” refers to a process of changing a sampling rate while avoiding changing the duration or pitch of a signal. In this case, for each sampling point, we always have a corresponding point in time. Interpolation is typically done with what is called “phase vocoders”. The most common phase vocoder implementation uses finite impulse response filters (FIR) and their inverse form called an IIR filter.
Disadvantage of Resampling
The main disadvantage of resampling is that, when interpolating, each sample is represented by two or more samples in the new format. This makes interpolation less precise than resampling. Each sample in an interpolated format is rather like a small mirror image (even if the original was very slightly distorted).
Interpolation can also be done during recording. For example, using oversampling will result in a higher sampling rate when recorded at 320 kHz or oversampling an already recorded recording at 96 kHz will result in an even higher resolution when recorded at 192 kHz.
Some digital audio editors such as Audacity, LAME, and CoolEdit Pro support oversampling and downsampling via the use of separate plug-ins.
The word “dither” describes a process that introduces random noise into a signal to increase its dynamic range. Dithering attempts to avoid aliasing which can occur when upscaling or downsampling an analog signal. When dither is used during recording, it can simplify conversion to digital data formats such as MP3 which require not only the original signal but also dithered noise for error correction. This may improve sound quality. In fact, there are several different types of dithering algorithms available, each with its advantages and disadvantages.
“Downsampling” is the reduction of a signal’s sampling rate. This can be done at any time during the process, and for any reason.
When downsampling is done during recording, it will result in a lower resolution (and therefore more bits per sample). Thus if you want to store a high-resolution recording in an audio file format that supports only low resolutions (such as MP3), you must apply downsampling at some point. This can be done using software such as Audacity. Although upsampling and downsampling share the term “resampling”, they are actually different processes, because they use different methods and thus affect the audio signal differently.
In the case of upsampling, the “new” sample rate is higher than the originally recorded sample rate. In downsampling, a lower rate is used. However, both processes may reduce the number of samples in a sample format.
Disadvantage of downsampling
The main disadvantage of downsampling is that it reduces the amount of information involved in each sample. This makes upsampling more precise and therefore better suited for some applications, such as converting from one sampling or quantization format to another.
When upsampling, each sample is represented by two or more samples in the new format. For example, if you have a recording at 44100 Hz and you are upsampling to 96000 Hz, each original sample will be represented by at least two “new” samples—one for each channel. Thus the audio signal will not be preserved during this process.
Upsampling also introduces noise as a result of adding extra samples. This noise may be undesirable in certain circumstances such as when it masks other sounds or when it interferes with the desired output frequency response (sound spectrum). In this case, dithering may be used to randomly place the noise samples to minimize their effect.
In some cases, upsampling can be useful for recording from an analog source—for example, from a vinyl record—and then converting to digital data. In this case, upsampling can be useful as a way of going from a lower-resolution format (such as 44 kHz) to a higher-resolution format (such as 96 kHz or above).
“Upsampling” is not the same thing as “downsampling”. The latter is a reduction of a signal’s number of samples while being recorded. The former is an increase of that same signal’s number of samples.
Drum recordings are commonly upsampled from 44.1 kHz using normalization for the stereo signal, which results in a 96 kHz or 192 kHz recording.
Some audio editors may upconvert to 44.1/48/96k, as well as downsample from 22050/44100/etc to a higher sample rate, such as 96k.
Most samples are generated by having the time-varying part of a signal at a given point in time divided into a series of numbers. These numbers are called samples and the process of dividing the time-varying part of a signal into these samples is called sampling.