The SOS team battles through snowdrifts to help Nick Redman and Mike Sinnott with two different vocal sounds.
This month's tale of intrepid studio rescue takes us back to the end of January, and sees Technical Editor Hugh Robjohns and myself battling blizzards and navigating the M25 in first gear to reach Nick Redman's bedroom studio near Croydon. Nick had called us asking for help with his vocal mixing, because, although he had good equipment and some decent vocalists, he was never happy with the vocal sound or the way it sat in the mix. Along with his partner Mike Sinnott (who looks after student welfare at the Alchemea recording school), Nick composes and records mainly dance music, but to gain experience and to help pay for his well-stocked studio, he also works with a number of paying clients. This means he has to deal with various vocalists and what he needed was a strategy to help him fine-tune each one to its best advantage.
Nick's studio was a very unfortunate shape, being an almost perfect cube around nine feet along each side. This was bad from an acoustic point of view, as the room modes all stacked up at the same few frequencies where ideally you need them to be as evenly distributed as possible. Only the bass trapping effect of the window saved it from low-frequency chaos, while a few strategically placed foam tiles calmed the mid-range and high-frequency reflections as well as providing a reasonably dead corner where the vocals were generally recorded. The studio's Mackie HR824 monitors were set with the bass switches in their middle positions and a pair of NS10s were available for alternative monitoring. A Soundtracs Project 8 mixer looked after mixing the outputs from Cubase SX running on a 2GHz Pentium 4 machine (Windows XP) with extensive silencing mods — Nick is an IT engineer, so this part of the system presented him with few problems. There was also a whole rack of well-chosen outboard including an Eventide Orville and an Avalon VT737 tube voice channel. The PC was also fitted with a Universal Audio UAD1 card and had a full bundle of Waves plug-ins.
The first step was to ask Nick how he went about recording the vocals. He explained that the singer stood with his or her back to the treated corner and sang into either a Neumann U87 or an AKG C414ULS through a pop shield. The house is right next to a busy commuter rail line, so Nick usually waited until a train had passed, then tried to get a take down before the next one turned up! He also used high-pass filtering to keep low-frequency rumble to a minimum, although he also used good-quality shockmounts with the mics.
Standing with their backs to the corner in the confined space of the room meant that the singers couldn't back off very far when singing loud passages, but otherwise it seemed to work OK. The mic was normally fed to Cubase via the Avalon unit, which was used to apply the bass roll-off, as well as a little gentle compression just to keep the peaks under control. When mixing, Nick fed the vocal track back through the Avalon so that he could use its EQ and compressor, adding vocal reverb from the Orville.
My only suggestion about the recording environment was to fit a curtain pole or shower curtain rail to the ceiling and then hang a duvet from it to divide the vocal corner from the rest of the room. This would further cut down any spill from the computer drives and fans and also reduce coloration caused by room reflections. Nick felt that this was a good idea that wouldn't cost much to try out, but in deference to our art department, I declined to be photographed holding up a duvet yet again!
As Nick and Mike work with a number of different clients, I suggested that we check out two mixes that were causing particular problems and then see what we could do to improve them. The first track featured a female singer with a very attractive voice, but as soon as she hit the chorus the dynamic of the vocal rose dramatically causing very high peaks and a shift in timbre from very sweet to very forceful. Although this was the intention for the mood of the song, the shift in tone introduced some harsher frequencies in the recording which Hugh and I felt created on over-strident timbre. Although there were plans to re-record this vocal part in the near future, I didn't feel that the timbre would be much different, and a comparison recording of the same vocal part made at the Alchemea studio showed the sound to be almost identical, so there was clearly nothing much wrong with the basic recording technique.
The dry recording was clean enough, but the Eventide Orville plate reverb setting Nick and Mike had chosen sounded too splashy to us. Also, while it was obvious at higher settings, it almost disappeared when it was turned down in level, leaving the vocal sounding dry and thin. As you can only really work on one thing at a time, we switched off the reverb and concentrated on getting the tonality of the voice right. After trying to tame the offending frequencies using the Avalon's EQ, it became apparent to me that it was the wrong tool for the job. While it is a lovely-sounding EQ that can be used to add warmth, air and polish to a sound, I felt it was too polite for the task in hand, where we needed to apply some pretty aggressive notching to those frequencies causing shrillness. Furthermore, because of the energy being put in by the vocalist during the choruses, the vocal character was very different between the verse and chorus, so I suggested to Nick that he shift the verses and choruses to separate sequencer tracks so that we could more easily treat them differently without having to mess around with plug-in automation.
As the chorus was the most difficult section, I asked Nick to load up his Waves Q4 equaliser so that we could address any problematic frequency peaks, then we looped around the offending section while making adjustments. I used the usual technique of applying maximum boost while sweeping through the frequency range searching for the problem frequencies, and it soon became apparent that there was a very aggressive peak at around 3.8kHz, with another peak an octave below that. Deploying two notch filters with pretty large amounts of cut tamed these quite effectively — the exact settings can be seen in the screenshot. To avoid dulling the sound, and to add a little 'air', we used some gentle boost at 16kHz, and we also tried to warm up the lower mid-range by adding around 1.5dB of boost at 250Hz with a very low Q value. The overall consensus was that this improved the chorus vocal dramatically, making it far less strident, adding warmth and body, and at the same time keeping the sense of air and detail. The verse vocal wasn't nearly so strident, so we copied the EQ setting to another Q4 plug-in and then reduced the depth of the cut notches until a musical sound was achieved without the verse and chorus sounding mismatched in any way.
That left reverb and compression to sort out. The vocal had been compressed to some extent while recording, so we used the modelled 1176 compressor plug-in that comes with the UAD1 card, setting a fast attack and a ratio of 8:1. We adjusted the threshold so that the compressor came in on the louder parts of the chorus (showing a gain reduction of 6-7dB) and set the release control to around 25 percent. This helped even out the levels, and also gave the sound a little of that 1176 character. A similar setting worked with the verse. However, given that there were fewer peaks, the amount of gain reduction could be reduced to 3-4dB.
Now it was Hugh's turn to try to find a better reverb sound. We wanted something with strong early reflections after a longish pre-delay to provide some character and support to the voice, but with a gentle and low-level reverb tail that wouldn't get in the way. The original plate program offered very few parameters for tweaking, and none of the quick tweaks made to alternative Orville programs provided the result we wanted. It may have been possible to find something suitable given enough time and familiarity with the Orville, but as we had neither in abundance I suggested that we try the UAD1's Realverb Pro plug-in, something Nick and Mike had shied away from as they'd always assumed outboard reverb would be better.
Nick set this up to return to two channels on the desk, giving us quick access to the return levels, then I had a go at creating something suitable — as I hadn't used a Realverb Pro before, I was keen to see what it could do. Manipulating the various elements of the reverb, I managed to create a reverb setting with a strong early reflection pattern and a decent amount of pre-delay to fatten and thicken the sound without making it too wet. On top of that, I added a reverb tail around three seconds long, but brought its level down with respect to the early reflections so that the overall result wasn't too washy.
With a little adjustment to the EQ and 'wall materials', we soon came up with a nice treatment that added real weight to the voice and gave it a sense of existing in a real space without the splashy wetness we'd suffered from originally. Bringing up the rest of the backing confirmed that we'd improved the situation significantly, with the vocals sitting much more comfortably in the mix. Hitting the Waves EQ bypass button brought back a stark reminder of just how strident the original vocal track had sounded in the chorus and everyone agreed that the sound we'd finally arrived at was a lot sweeter and much closer to what Nick and Mike had originally been aiming for.
The second track we heard featured a male vocal which was well pitched and carried plenty of expression, but again it didn't sit well in the track, even though the overall level wasn't bad. My immediate impression was that the voice sounded boxy and congested. Also, as with the previous track, the chosen reverb treatment didn't get it to sit naturally with the backing. The boxiness was tamed by using Nick's Waves Q10 equaliser to apply gentle notches at around 500Hz and 2kHz, and again this was balanced by a little low-end lift (this time right down at 63Hz, but with a very wide bandwidth) and the obligatory 'air' EQ at 16kHz. The difference was dramatic, and right away the vocal felt better in the mix, even before we started on the reverb treatment.
Compression was applied using the UAD1's 1176 emulation, again in a fairly gentle fashion with no more than 6-7dB of gain reduction on the loudest notes — this was quite enough to keep the vocal level stable in the track, and again you can see the exact EQ and compression settings in the screen shot. For reverb, we ended up settling on the Realverb Pro again, with just a few minor tweaks, the main point being to add strong early reflections, once again to give the vocal weight and space.
The main thing that came to light from this challenge was that not all high-end EQs are suitable for all jobs, and in this case the Q4 and Q10 were far better suited to surgical notching than the mellifluous Avalon. It also pays to keep in mind that you can get away with quite drastic EQ cutting without the processing itself becoming obvious. In most instances, though, boosting has to be gentle and done with high-bandwidth settings to avoid sounding phasey or nasal. Even when a voice is perfectly recorded, there may be physiological resonances that are best attenuated, so although EQ shouldn't be considered mandatory, you shouldn't be afraid to use it when it is needed.
Additionally, we demonstrated that you can't be over-casual with the choice of reverb and in many cases a little basic editing is necessary. Part of the secret is knowing what you want the reverb to achieve, and in this case we needed the early reflections to add weight to the voice as well as to provide a convincing sense of location. The reverb tail is also important, but it doesn't have to dominate the sound, and in modern styles where the vocal needs to have a less obviously reverberant sound, keeping the reverb tail down below the early reflections can help a lot.