The Analog vs digital debate has been going on for what feels like an embarrassing amount of time, with neither side showing any sign of relenting. Should there even be an opportunity for a winner is a question that will perhaps never be answered, but what is certain is that without proper knowledge, a proper debate is unable to ensue.
So what is the difference between analog and digital audio? It is possible to summarise analog as a continuous warm signal, and digital as a stepped, exact, and clinical signal. When copied, analog audio degrades with every bounce, whereas digital audio does not. With ‘equal’ playback systems, there is no discernible difference between the two.
There is obviously a lot more to this complex and controversial topic as you will see below. The details of the differences become very technical, very quickly, but can be resolved and undertood with ease.
The following article is written by our guest producer Andy Bell. Andy is a producer from Newcastle, UK. After receiving his Master's degree in composition from Newcastle University, Andy has worked as a freelance engineer, lending his ears and tastes to production and mixing work of all genres, from Indie Rock, to EDM, to Rap. He currently plays in the synthpop duo, Drive, in which he co-writes, produces and mixes all of their material out of Mono Studio in Newcastle.
I hope that this article will enlighten those who are somewhat in the dark about the differences between the analog and digital worlds. Both intertwine and work together in many ways, and while I hope you be a little clearer about the technicalities of recording in each domain, you can also see the power to use the specifics as a creative tool that you can use to enhance your productions.
Production is always about serving the song and ensuring that it can sound the best it can, and therefore projecting the most emotion that it possibly can. You can place microphones and dial EQs and compressors until you are wholly content, but without what might be seen as the proper ‘safeguarding’ of your audio, you might be causing more headaches than you would like – all of which can be avoided with the proper knowledge of the differences and the idiosyncrasies of the analog and digital domains.
Analog recording relates to signals or information represented by a continuously variable physical quantity such as spatial position, voltage, and more. The information is stored in or on the media being used.
Magnetic tape is the best example, and perhaps the pinnacle of analog recording; the physical image of recorded sound is stored as magnetic particles on the tape itself, and the physical characteristics of the sound are directly proportional to the soundwave.
Similarly, the grooves in the lacquer of vinyl records constitute the sound, inside which a stylus is placed that is attached to a magnet that moves, enabling the electric current generated to be converted into sound through an amplifier.
Figure 1: Analog Tape
In the digital domain, signals are quantized, represented as discrete numbers, and stored. These figures represent changes in the environment. To reproduce these sounds, the information must be converted back into the original analog waveform through the use of a D to A converter, as we will touch on later in this article.
How these sounds are reproduced have several variables that affect the efficiency, and ultimately the fidelity, of the recordings, the first of which is the sample rate at which the music is recorded and reproduced.
The sample rate is the resolution of audio in time – it measures frequency in ‘samples per second’ and is generally expressed in kilohertz (kHz).
The Nyquist-Shannon theorem states that when the sampling frequency (the sample rate at which you’re recording) is greater than twice the maximum frequency of the signal being recorded, the original signal can be faithfully reconstructed. This is in effect during both the recording and playback stages.
The range of human hearing is roughly from 20Hz to 20kHz, and even then, the higher ranges are almost inaudible to those who have near-perfect hearing. What this does mean, however, is the standard 44.1kHz sample rate used in most Digital Audio Workstations can accurately produce the highest of signals, with the highest frequency able to be reproduced being 22.05kHz.
So why might a producer want to change this? The sampling rate determines the frequency range able to be represented because of the Nyquist limit, and it is literally a band-pass limit. If a signal contains frequencies greater than the Nyquist frequency, they are mapped to frequencies less than the Nyquist frequency in place set by the sample rate. This produces a distortion that can thin out recordings, and is generally regarded as unpleasant. The distortion occurs at different frequencies relative to the sampling rate.
A sampling rate of 12kHz will have a Nyquist frequency of 6kHz, at which point distortion will sound. A frequency of 8kHz, 2kHz above the Nyquist limit will have its frequencies remapped, causing a distortion around 4kHz, 2kHz below the limit.
This can be rectified with an anti-aliasing filter; a bandpass filter that blocks frequencies above the Nyquist frequency from being converted, thus preventing the realignment of said frequencies during the conversion process.
That is not to say that this distortion is unpleasant in all situations. You might have implemented this effect in your productions previously on purpose; plugins such as Avid’s Lo-Fi, Goodhertz’s Lossy and many more all have options to reduce the sampling rate for a chosen sound, and manipulate the anti-aliasing filter if used at all.
This takes the inaccuracies created in the digital conversion of a waveform and use them creatively à la glitch music that celebrates these digital malfunctions, or in styles reminiscent of early Hip-Hop that used samplers such as MPC6000 or the E-MU SP-1200, whose sampling rates and concurrent anti-aliasing filters became famous for their sound and the way they made records feel.
Figure 2: Digital Sampling and Anti-aliasing
On a macro level, you might have noticed when setting up a session that your audio interface (read: converter) has the ability to perform at many other sample rates, including the more extreme figures up in the 96kHz to 192kHz range.
These are impressive looking numbers that have ultimately caused debate upon debate in online communities over recent years as to what sounds better and which ones we should use and when.
What it does ensure is a minimal chance of aliasing when recording high-frequency content, and therefore an extremely low level of distortion, especially at lower frequencies. But it does beg the question; who can hear that high up, and what can produce such frequencies anyway?
The very fact you’re listening back to a recording implies that at some point you’ve gone from the analog domain into the digital domain and then back again. To do so, you need a converter. With this, it is crucial to keep in mind the physical limitations of a higher sample rate on your technology, also.
Higher sample rates take their toll on recording systems, necessitating higher CPU usage. More samples = more information, so it’s also imperative to have more disk space to be able to retain and run the larger files the recordings create. Also, if you’re a Universal Audio user, you will be aware that a higher sample rate will take up more DSP, meaningless DSP-based plugins in your session.
While some plugin developers state that their software works better at higher sampling rates, the reality is that not all of them fare so well, and pushing technology like this might cause unwanted distortion and inaccuracies of replication of your audio.
The frequency content might merely be working too fast for the technology to keep up. Every piece of equipment has its own sweet spot, and you might find that you just prefer the sound of one sampling rate on your convertors than another. But again, these inaccuracies can be pleasing and kept at the producer’s discretion.
Bit-depth is the alternative side to the digital domain. It is frequently paired up with sampling rate, but where sampling rate refers to the resolution of each sample, bit-depth is the resolution in the amplitude of a signal. Its purpose is to limit the signal-to-noise ratio in a recorded piece of information.
There are millions of amplitude points in a recording, so much so that it impossible to be plotted onto a whole, resolved number, as is the definition of analog. The signal is rounded down, and some of the signal is ‘shaved off,’ presenting the opportunity for fallibility of said quantization, known as the ‘quantization error.’
This error introduces a very quiet distortion that has an effect on the smallest aspects of a recording; attack points, the end of reverb tails and more. This is barely noticeable but can become more obvious and more of a risk at a lower specified bit-depth.
The quantization error is measured in decibels (dB) and effectively limits how much headroom there can be in a recording.
Headroom is where the recorded signal’s dynamic level is quieter than what is allowed by the bit-depth set by the converters. Standard 44.1kHz, 16-bit CD recordings have a theoretical maximum of 96dBs, whereas a professional 24-bit digital audio recording has 144dBs of headroom. This applies to individual channels in a production, as well as the overall master buss.
As noted, more headroom means increased perceptibility of transients, and ultimately can make a mix feel louder when mastered. Everything adds to headroom – and an important point might be to check your low end if you feel like you’re having difficulty preserving headroom.
Changing bit-depth introduces noise. Dithering is a way of stopping this, in which a small amount of random noise is placed in the audio to mitigate the non-linearities of the quantization. In general, it is recommended that 16-bit recordings use dithering. Dithering can be applied to increase effective dynamic range, allowing the dynamic range of a 16-bit recording to be 120dBs.
This can become tricky while combining the two worlds. Analog mixing consoles are generally calibrated for a +4dB analog signal level (line level), that will show as 0VU on a similarly calibrated VU meter. Professional quality recording gear is designed to be capable of reproducing +24dB, meaning that a signal that registers at 0VU will have +20dB of headroom. This allows for the anticipation of an increased level and the ability to reproduce transients correctly.
Just as with bit-depth, when the signal is too loud for a piece of equipment, the peaks of the waveform distorts, resulting in a form of distortion, also known as clipping. It is generally advised to avoid this, although sometimes the overdrive occurring through the driving of a mixing console, or overdriving magnetic tape can be pleasing to the ear. This is known as soft-clipping, and is a form of limiting as the peaks of the waveform are shaved off at the expense of a slight distortion that colors the sound, and has the effect of ‘rounding’ out the audio.
It is imperative to get the level right when recording (-18dBFS is typically known as the analog ‘sweet spot’), as too high a level will clip your audio, and too low a level will have a higher signal-to-noise ratio, and therefore will end in a nasty hiss when the audio level is increased, akin to the bit depth in digital recording.
It is of course quite rare to keep a recording in a purely analog realm, and converting the analog recording to digital can be slightly trickier. As stated, recording to -18dBFS is a guideline, but you can record at a louder level – I like my drums, for example, to come in around -10dBFS.
In mixing or mastering, the signal can get even louder. When a signal hits 0dBFS, it has run out of headroom and can cause digital clipping. Any data above the threshold of 0dBFS cannot be accurately converted into digital information and is therefore discarded.
A common example of this is clipping a sub-bass made up of a sine wave that changes to a square wave when clipped. This can sound harsh and hurt your production, but to discern this, it really is best to use leave it up to your ears and personal taste.
Different bit-depths have different amounts of headroom; a 32-bit float bit-depth has the promise of reaching up to 1680dBs of gain. With this, it is incredibly difficult to clip your audio, so long as the audio stays in the DAW. Should the audio be converted to 16 or 24-bit, it is a different question and is likely to clip (hence the process of dithering).
Hopefully, it is evident at this point that at whatever stage in the audio production process you are at, there is always a stage of conversion at play. Whether you’re recording or consuming, the audio files in question have been converted into the correct format to be heard. And one converter does not necessarily sound as good as another.
Figure 3: Analog to Digital Conversion
For example, a smartphone has an entirely different, basic converter when compared to a Cranesong HEDD or a Burl converter – two incredibly popular and well sought-after analog converters. A smartphone’s converter allows you to carry out your telephone conversation but is limiting when it comes to working with high-resolution audio.
An outboard converter also takes into account jitter, which occurs when the digital information is not successfully changed into a consistent timing sequence, therefore reducing the fidelity of the audio. Such converters are equipped to handle these types of issues where an internal soundcard might not.
Whether the consumer uses Focusrite, Universal Audio, Apogee, Antelope Audio, or any other choice of interface, is down to the tastes of the consumer. They all have different qualities of sound and various limitations as to what they can produce.
The increase in quality of these converters naturally works in conjunction with the cost of the item. More expensive converters are expected to handle the more extreme sample rate choices with more ease, utilize more inputs, and provide what some may call a more ‘focussed’ sound, although the latter is again down to taste. This allows the producer to make better informed critical decisions, making for a more straightforward production process.
Naturally, a foray into the differences between analog and digital recording has ended up as a discussion of sample rates, bit-depths, and an insight into the debate as to what is best. The discussion of what to record and mix at is a never-ending debate and one that I will not attempt to answer definitively.
I always work by whatever sounds best is best, and have found specific ways of working that I believe get the best results. If I am doing a full production, I will tend to record and edit at 96kHz/24-bit, and downsample to 48kHz/24-bit during the mix. I deliver the mix to the mastering engineer at 48kHz/24-bit and let them downsample it themselves to 44.1kHz/16-bit through their own preferred method of doing so.
The studio in which I work has the tools to be able to handle the
higher rates – and I encourage anyone who has the power to utilize a
higher rate to try it out. If I get a session at 44.1kHz/16-bit, I will
retain its qualities and mix at that level – there’s no point up-sampling
when the information isn’t there in the first place.
-- Andy Bell, Mono Studio, Newcastle, UK