Keeping clean signals throughout your DAW can be a challenge. Especially if you do not quite know where to look for distortion in the first place. In this article I’ll explain where to look for distortion in the first place.
The easy bit
Your DAW basically consists out of three parts. The input, the processing part and the output.
The input is usually an audio-interface of some sort. These interfaces have an analog to digital converter (AD-converter) which is responsible for translating your microphone or guitarsignal to a digital stream of numbers. This stream is then transported over a protocolbus (like usb, firewire or thunderbolt) into your computer, this bus is not directly of influence on your audio quality though.
The processing part is the software. The software on your computer records this datastream and puts it in a file. Also the software makes it possible to edit, filter or playback the recorded files. This is what makes your system a digital audio workstation, in short, a DAW.
The last part is the output. Again, most of the time this is the same audio-interface. The software sends a digital datastream to the audio-interface which in turn translates it to a analog signal. The interface uses a digital to analog converter (DA-converter) in your interface. Your monitors are connected to the analog outputs of your interface.
The digital bit
When you start a new project in your DAW you have to select which bitdepth you will be using. Most interfaces can sample up to 24 bit and by choosing so the software will set the AD/DA-converter to 24 bitdepth. Bitdepth is also called the resolution of the audiosample.
All recorded audio is stored on the harddisk with a depth of 24 bits. But since your interface is also your output device, it plays back audio at 24 bit. In other words, the same bitdepth is used for both the in- and output.
A bit more precise
24 bits actually means 24 bit integer. 24 bit means that the sampled values range from 0 to 16,777,215. There is no way to make a larger value than 16,777,215 or, to show the other end of the spectrum, a number lower than zero. Integer means that there are only whole numbers possible, e.g. 0, 1, 2, 3 and so on. A number like 1.6 is rounded to 2, the closest whole number. So any samplenumber will be rounded off. This is what is called quantization error.
Any analog signal that is too loud for the AD-converter will result in a maximum value (either top or bottom) and will sound like distorted audio, the ad-converter simply has no bigger number for louder than maximum than what fits in between those 24 bits. This shows up in your DAW as a clip, a signal that hits full scale for a period of time and shows a visual peak in your audiofile.
Distortion at the input is permanent, the DAW writes the distorted audio into a file directly to the harddisk. You can prevent this type of distortion by using your gainknob properly prior to recording. A good tip: try to record 10dB lower than the loudest part in the song. That is what is called ‘creating headroom’; making room for unexpected louder variations in level.
Now how about your software? Does it use the same bitdepth internally to process your audio? Well, simply put; no. Inside your computer the processing happens at a higher resolution. Depending on which system you are using it uses 32 bit float, 48 bit float or even higher.
The magic word here is ‘float’ which stands for floating point numbers. To be one step more precise, floating point calculations make it possible to create any number with a certain precision. The precision of the calculations increases when more bits are available to store the number in.
With floating point numbers it is possible to store 1,6 in a digital format in regards to integer numbers. As long as these numbers are inside your DAW the computer can calculate any number in the 32 bit floating point range.
The reason for some systems to use a higher floating point resolution is because the outcome of calculations they perform have a bigger precision. For example; 1.39578 is more precise than 1.4 due to numbers ‘behind the comma’. Now the numbers your DAW creates are bigger and certainly more than one; the more precision you can get the better your ‘depth’ your DAW can achieve sonically.
With 32 bit float resolution the DAW not only has more precision but can also calculate larger numbers than the 24 bit integer. In fact, it is nearly infinite. And that means 2 thing which are important to understand.
First of all, there is no such thing as clipping as long as you are INSIDE your DAW and using no plugins. As said before, the realtime calculated values are infinite in size and that also means there is no limit in level. Unless we add plugins or other external component which in turn might constrain the resolution. But until it leaves the DAW via the DA-converter there is nothing to worry about just jet.
Second, and here is the catch; there is a limit to your master output. The output resolution of the DA-converter is not 32 bit float, but is the same as your projectsetting: 24 bit. And so your mix must not hit 0dBFS at the master output because then you are hitting the roof of the 24 bits.
A bit of wisdom
Anything in between the in- and output of your DAW can be as soft or as loud as you want. If, for example your level of a track is way too loud, you could still have a clean output as long as you have a bus to take down the same amount of overload. That bus can be a subgroup or as in this example your masterbus.
In this picture you see a level coming in on ISA Mono In 1. This is the input of the interface, so coming straight from the AD-converter, hence the level; -6.0 dBFS. At channel Audio 01 this level is gained with the DAW gain of 48 dB. Way louder than the meter can show. The Stereo Out bus, the master connected to the DA-converter, is turned down 48 dB so the signal is still between the 24 bit resolution of the DA converter. And so the output of the DAW is still clean.
The difference of 3 dB between the AD and the master output comes from panning law. The mono input is delivered to a stereo master bus which has its panning law at 3 dB.
Whether it is wise to use such extremes, is hardly a question. Ofcourse you are better of using proportional levels, if only for your plugins that not always can handle such loud levels. If you mix within normal ranges it is unlikely having to adjust your masterfaders to keep the output clean.
When you render your mix to a file it you force your DAW to calculate it to 24 bit integer format. Although some DAWs allow you to mixdown your project to 32 bit float, you must be careful with that; it eats away space and not it is not a common filetype to deliver to a medium. Also you would have to playback the file with your DAW again to get proper levels.
The closing bit
So why would you try to keep signals in range of the meters presented in your DAW? For starters, it is good practice! Simple common sense tells us to keep between the meters, there is enough space for every sound to be there.
Secondly, your plugins do like levels that are ‘sane’ so they can treat your audio as they want. Most plugins do not like very hot signals, due to the structure of the plugin or due to the by programmers meant limitation of the emulation. Most of these plugins carry gain knobs and have output controls but the damage is already done as soon as it enters the plugin.
When keeping audiolevel between ‘the lines’ you are not likely to overload the output of your DAW and you do not have to perform any tricks to take down the level at the master bus. Also, you can mix with you masterfader at unity gain, which again is a true helper in getting your mixlevel right and helps you keeping an eye on the famous loudness war.