There’s a lot about producing recorded music that isn’t very obvious, especially for someone who’s still working through the basics. Some questions crop-up over and over again, and the basic fallback answers of “try it and see (and hear) what it does”, and “if it sounds good then it IS good” can sometimes be seriously un-helpful; so I thought I’d lend Making Tracks to a Q&A format for a while to see if we can set some of the more common puzzles to rest.
One incredibly common question is about how loud to record; and as this is pretty much at the beginning of the process, it’s a great place to start. “How Loud?” can be a pretty big question, so I’ll break it down into a couple of parts: analogue and digital.
First off, how loud should the actual source that we’re recording be? Basically, as loud as it needs to be to sound good: so let the guitar player turn up her amp until it explodes, let the drummer split the skins and break the sticks, and let the singer work those big notes until he faints. If you need to record very quiet (or distant) sources, be aware that every dB of gain we later add to the signal also raises the background noise that surrounds it, so healthy levels and/or low background noise will give cleaner recordings so long as we don’t overload the signal chain. Use mics that will cope with the SPL (Sound Pressure Level) involved, listen to the sound in the room, and carry-on (oh, do watch out for the neighbours though!).
For the analogue part of the signal chain, we can read the spec sheet for recommended levels, but in reality we’ll just watch the meters and overload lights; it’s generally OK to abuse these levels a bit to get character out of a device; just keep an ear on creeping noise and distortion levels and you won’t go far wrong.
Where it really can go wrong is when we move from the analogue to the digital world. Our electrical signal is now going to be sampled and converted into digital data by an ADC (Analogue to Digital Converter). The number of times a second we sample is called the “sample rate” (this is the 44.1, 48, 96, 192 kHz part of the spec), and each time we sample we create a “word” which is made-up of a number of BITS (BInary digiTS) which is the 16, 24 bits part of the spec. Each bit can have a value of 1 or 0, and if all of the bits in the word are equal to 0, then it’s mighty quiet, if they’re all 1, then it’s as loud as its going to get.
Now back in the analogue days, tape machines would be calibrated around average levels of 0 dBVU, which basically gives us around 20 dB of recording headroom above 0. Tape is a nice sounding but quite noisy medium, so it became a “thing” to use-up some of that headroom and push your recording “into the red” to get a better signal to noise ratio, as well as desirable machine and tape saturation.
Our digital system is calibrated around a peak level of 0 dBFS (deci-Bells relative to Full Scale) which is produced by a word full of 1s. Thinking about that for a moment, 0dBFS is the loudest thing that our digital system can know about, and if we hit the front end of our converter with a signal that’s bigger than expected, then we’re simply losing all the level data above that point. If even one peak of our signal converts to 0dBFS then we have no way of knowing if that is its correct level, or if it was actually meant to be higher than that. If we get a number of 0dBFS peaks, then we have no idea how they are supposed to relate to each other or to the rest of the samples, and they can create an audible and rather unpleasant effect known as digital clipping. Unfortunately, old habits do die hard, and for many years the myth has stayed with us that we need to record close to the “0” level which puts us at risk of overloading the converters.
As with many things, the answer is stunningly simple: don’t ever hit 0 dBFS and you’re fine. If your meter allows for things called Inter-Sample-Peaks then just pull down the level hitting the converter until the highest peak you ever see is at least -0.1 dB below FS and you’re safe; if it doesn’t then allow an additional 1 or 2 dB safety margin.
That said, the only reason to not leave lots of headroom in a digital recording is to keep the quiet parts above the noise floor of the digital system, and seeing as every part of the analogue world, including the performer, the air con, the mic, the preamp, even the cable, is creating noise at some level, that digital noise floor is most likely to be buried anyway. A good modern audio interface recording in 24 bit mode can have a dynamic range of over 100 dB, so make life a bit easier; allow a real-world 6 to 10 dBFS above your loudest peaks and don’t worry if you need to record with -20dBFS or lower average levels on the louder passages; you won’t run out of headroom, you won’t hear the noise from your converters and you will probably find that everything sounds better because the analogue circuits around the converter chips in your audio interface are operating in their optimum zone.
The single biggest improvement ever in my recorded sound, far, far more than any piece of hardware or plugin I’ve ever bought, was after I read about optimising levels into my converters. I grabbed a software VU meter for the sake of a bit of old time vibe, calibrated it for 0dBVU to be the same as -20dBFS, set my levels and then never looked at it again because everything just worked.
..and you can find more stuff about recording and music at