Advance warning: This is a really technical post, which I’m posting because I’d mentioned in an email that I could go into more detail about the process of recording and mastering a CD for anyone that was interested, and I received an email from one of our singers saying they were interested. If the technical details of mastering a CD aren’t thrilling to you, feel very free to skip this post!!
We recently recorded our first CD – “BeVox: The First Five Years”. The recording was done using several high-grade microphones, and with the backing track being played back in the room at the lowest level we could get away with. In order to allow the backing track to be heard without picking up too many bass frequencies, when it was played for us to sing to during the recording it was heavily EQed – virtually all the bass frequencies were stripped out, and the treble frequencies seriously boosted. As the higher frequencies carry the majority of the “pitch” information, and the bass frequencies carry the majority of the “rhythm” information (a simplification, but a convenient one), this allowed us to remain in tune with the track without overloading the microphones with the more energetic bass frequencies. The human ear can’t hear sounds below about 50 Hz, but these still carry audio energy, so the less of these frequencies we record, the less wasted energy there’ll be in the recording. (As an aside, we had a few problems staying in tune with the track on the first day of recording – I tweaked the EQ to boost the treble further on the second day and it made a big difference).
Jules and Robert at 4 Part Music then went away and mixed the recording. They took the outputs from all the mics in the room, and then offset them by a few milliseconds depending on the distance from each microphone to the “optimal listening point” – essentially, where I stood to conduct the choir. Sound travels at 330 metres per second, and this means that if someone is stood 33 metres from me when they sing, the sound will reach my ears 0.1 seconds later. If someone else in the choir is stood 16.5 metres from me (half the distance), the sound from them will reach my ears 0.05 seconds later (half the time). As we had several different microphones placed around the choir, the timings for each of them needed to be offset so that the sound they recorded all hits the listeners’ ears at the same time as each other. They also slightly reduced the stereo width of the recording – because we’d been spread quite wide in the recording room, the sound felt too separated initially, as though the sopranos were nowhere near the altos. Reducing the stereo width made the sound more compact, more “together”.
Next, Jules and Robert mixed in the backing tracks we’d sung to, and adjusted the volume of the track and the output from each microphone to create a good balance. They added a very small amount of artificial reverb to compensate for the fact that the room we recorded in had quite a “dead” acoustic. (Actually, this process of adding reverb is something which really appeals to the geeky side of me. Up until about 10-15 years ago, reverb was added purely artificially – various algorithms calculated the changes that needed to be made to a sound to make it appear more “live”. In the last decade or so, a new system has been introduced – convolution reverb. Acoustic engineers go into spaces with interesting reverb characteristics, and using highly sensitive and carefully set-up microphones, they record the sound of various different frequencies being played in that space. Highly sophisticated software then strips out the sound of the tones that were played, and captures the sound of just the reverberations from that source tone – essentially capturing the sound of the room. This can then be applied to any sound source you choose – making it sound like the recording you’ve done in a really dead space has been transported to the location where the reverb was recorded. I’ve got a great collection of convolution reverb sounds – everything from the Cathedral of Notre Dame in Budapest, to a tiled bathroom!).
Jules and Robert finished their part of the process by mastering the tracks – bringing them all to a consistent volume across the whole CD, and smoothing out any variations in sound from track to track. Their approach to mastering was very “hands off” – they didn’t do anything that particularly interfered with the source sound.
I then took their finished tracks, and applied my own set of mastering tools to it. First of all, I removed any frequencies below 50 Hz. These can’t be heard by the human ear, but still carry audio energy, so I needed to remove them before I started on the next part of the process. By pulling out the frequencies below 50 Hz, some of the low frequencies from 50 Hz to 100 Hz were diminished too. These don’t affect the sound of the choir particularly, as the lowest note I usually write for the basses is a low G, which is 98 Hz. Admittedly, there is a low D (73 Hz) for the basses in “Let me go”, so I adjusted the EQ curve for this song.
I then mixed in the original backing tracks again, but applied some very heavy EQ settings – only letting frequencies below about 120 Hz through. This puts back some of the bass end that was stripped out by the first set of EQing I did – but only the bass frequencies from the backing track, not from the microphones on the choir. This allows the kick drum, bass guitar and double bass sounds to retain their warmth without including any low rumbles from the live mics in the recording room. (For the recording equipment geeks out there, all the EQing was done using FabFilter Pro-Q 2).
Next up was some multi-band compression. Compressors essentially take the louder sounds and make them a little quieter. You can also boost the overall volume with most compressors, so essentially you’re making the quiet stuff louder and the loud stuff quieter – thereby compressing the overall variation in volume. The multi-band compression I applied at this stage applies a very small amount of compression to each of five overlapping frequency bands, and the idea is to smooth out any sudden loud noises. Applying it in different frequency bands means that if there’s a sudden loud noise in the higher frequencies (a cymbal crash, for example), the volume of this is reduced without affecting the volume of the lower frequencies. This is done with quite a light touch, so it doesn’t make a huge difference to the sound, but it makes the next step in the process a lot easier. (This was done with Cakewalk’s LP64 Multiband Compressor – cheap and cheerful, but it does the job).
The final stage comes in two parts – making subtle adjustments to the overall track level to compensate for sections that are particularly quiet (such as the introduction to “You raise me up”), in conjunction with applying a level-boosting tool that allows me to increase the overall volume without making the loud bits too loud! This applies some really clever compression algorithms, reducing the volume of “transients” – sudden spikes of sound that can occur at the onset of certain sounds, particularly short, high-frequency sounds like a hi-hat cymbal or a singer singing a word beginning with the letter T. By reducing the volume of these incredibly short but high-volume sounds, the maximum volume the track reaches is reduced without changing how loud the track sounds. This then enables me to turn the volume of the track up without getting any distortion on these brief peaks. (All the final mastering transient compression was done using the Slate Digital FG-X Mastering Processor – a truly miraculous piece of kit, used in top-flight studios around the world).
The overall effect of all of this processing is to reduce the difference in acoustic energy between the quietest and loudest bits of the track, whilst preserving the overall sound. In a typical classical orchestral recording, the difference between the quietest and loudest sections might be as much as 20 or 30 decibels. Most modern pop and rock recordings have a difference of around 6 decibels or less. The practical outcome of this is that it’s hard to listen to classical orchestral recordings if there’s much background noise, such as in the car whilst you’re driving. The road noise obliterates the quieter material, unless you turn the volume right up, and then the louder sections are deafening! This is the approach that the guys at 4 Part Music had applied to their initial mastering – it gives the recording a real sense of space and room to breathe, which sounds great when listening at home in a quiet room. I wanted people to be able to listen to our CD in any environment – I suspect a number of people will have it on in the car, and so I took a more “pop” approach to the mastering. The difference in overall volume from the quietest to the loudest sections of our CD is around 10 decibels – more than in a commercial pop record, but a lot less than a typical classical record. This seemed like a sensible compromise between the space and headroom of a less compressed recording, and the practicality of a more compressed sound.
The end results will be available on our CD “BeVox: The First Five Years”. This will be delivered to those who took part in the recording, and those who pre-ordered, in the second half of October, all being well. After that, the CD will be available for sale at all our concerts, priced £8.