Just how do you mic up and multitrack a large ensemble in a less‑than‑ideal recording space, and then produce an effective mix of the resulting recording?
While most of my studio time is spent working on pop music or small acoustic ensembles, I couldn't resist the challenge when David Etheridge (a regular contributor to our sister publication Performing Musician, and a moderator on the SOS on‑line forums) asked if I'd like to record the West Midlands Light Orchestra, of which he is conductor and musical arranger, with the intention of producing a promotional CD.
The recording session took place at a Kidderminster hotel's banqueting suite, where the band rehearse twice a month. The room was generously sized in terms of floor area, and sited away from the main hotel, which meant that noise wouldn't be a problem, but at only around nine feet the ceiling was relatively low, and there was no acoustic treatment other than the occasional curtain. A large ensemble such as this requires plenty of mics and plenty of mixer channels, so I called up SOS Technical Editor Hugh Robjohns, who has had a lot of experience with such projects, and who has an impressive collection of microphones and location recording equipment. As I'd anticipated, Hugh was as keen as I was to get involved.
When Hugh and I first discussed the project, we thought that the best way to go would be to use my Alesis HD24 hard disk recorder, along with a modest Alesis mixer to feed it. We'd used this setup on previous large‑scale recordings, the idea being to supplement a main stereo pair with some spot mics. However, when it became clear that the room acoustics and ensemble balance were unlikely to be ideal, the number of spot mics we might require started to rise rapidly, and quickly exceeded the 16 channels available in the Alesis mixer.
We could have used extra mic preamps to feed the 24‑track recorder, but monitoring would have been difficult, so Hugh suggested that we use his SADiE LRX2 DAW/recorder system instead. It's equipped with two 16-channel AES3 digital input cards and a 16-channel mic‑line input card, so all that was necessary was to hook up eight digital channels from a Focusrite ISA428 four‑channel preamp, linked with a GML 8304 four‑channel preamp. This gave us 24 mic channels in total and, more importantly, full monitoring facilities. Even better, the SADiE LRX2 allows a live stereo monitor mix to be made and recorded alongside the individual channel recordings, making it easy to check and play back takes during the session.
We agreed that Hugh would bring the monitoring, talkback, multicores and cables, as well as all his mics and stands, and I'd bring all mine: we'd figure out on the day what was best to rig in front of the band.
The orchestra, which is made up of keen amateur musicians, plays mainly MOR and light‑jazz material (in the BBC Radio 2 Friday Night Is Music Night style). The orchestra comprised a full rhythm section (drums, electronic piano, acoustic bass, electric guitar) plus a generous quantity of brass, woodwind and strings, numbering about 60 players in total.
Hugh and I arrived mid‑morning to start rigging about two hours before the rhythm section set up. The rest of the performers were scheduled to arrived an hour or so later at about 1pm, the idea being that we'd rig all the mics, cables and control‑room gear without having to work around the musicians.
David arranged the chairs for the players and drew us a map of who would sit where. He placed the rhythm section together at the rear‑left corner of the orchestra. The acoustic bass, keyboard and guitar all used their own small amplifiers that could be miked or DI'd as required. The rest of the orchestra was set up in a wide 'U' shape, with a brass section of three trumpets, a french horn and four trombones occupying the rear‑most row of seats facing forwards. In front of them was the woodwind section, which comprised three flutes, an oboe, a soprano sax, three clarinets and a bassoon. The violas set up directly in front of the woodwind, and the first and second violins were positioned along the left arm of the 'U' in front of four cellos, giving a total of about 20 string players. The right‑hand arm of the 'U' shape was occupied by the saxes: three altos, three tenors and a baritone. There were also three vocalists — although only one sang at any time — who we placed well in front of the orchestra to maximise the separation. David stood in the middle of the open end of the 'U' to conduct, facing the orchestra.
Ideally, we'd have used a separate room set up as a control room to provide us with the isolation necessary to hear exactly what was being recorded. However, there was nothing suitable available to us at this venue, and the best we could do was to set up the SADiE system and monitoring in a side annexe to the room, about 20 metres away.
A couple of tables were pushed together, the first supporting the LRX2 recorder, the controlling laptop and external hard drives, with PMC DB1 monitors (powered by Flying Mole amplifiers) set up for the monitoring either side. The second supported the Focusrite and GML preamps, along with the monitoring controller and D‑A converter.
The monitoring chain was driven via an AES3 digital output from the SADiE, feeding a Benchmark DAC1 converter, into Hugh's bespoke monitor-control unit, and on to the speakers. To make it easier to run the session, we decided to set up both a talkback system and a red/green cue‑light facility for communicating with David remotely, rather than shouting or running up and down the hall, and we set this up to run from the monitor‑control unit. A pair of Sony MDR7509 closed‑backed headphones was also used to check the recording.
For simplicity, the digital output from the Focusrite preamp was used as the clock master for the LRX2, and everything was recorded simultaneously to mirrored drives (the internal laptop drive and an external high‑speed drive). During the breaks, the files were also copied to a third 'pocket' drive, just to be sure. We decided to record the whole project at 24‑bit, 44.1kHz.
To minimise cable runs around the musicians, we used three of Hugh's shorter multicores to carry the mic feeds from the orchestra, with a 12‑way, 20‑metre multi in the middle at the back, and two eight‑way, 25‑metre multis, one on either side near the front.
While Hugh configured the control -oom end of things, I started setting up the initial mic placement, largely by guesswork, using David's floor plan. The drum kit was covered with an Audio‑Technica ATM250 kick‑drum mic, an Electrovoice ND468 snare mic, and a pair of Hugh's MKH40 cardioid‑pattern mics as overheads, both with SE Instrument Reflexion Filters fitted, to reduce the effect of reflections from the low ceiling. The bass was DI'd direct from the pickup, using one of Hugh's Radial J48 active DI boxes, and then looped through to the amp, as was the keyboard (another Radial, this time a passive Pro D2), but the guitar amp was close‑miked, using an SM57 placed off‑axis and very close to the grille cloth.
The first and second violins were covered using four Neumann KM184s, while each pair of cellos shared a Neumann TLM103. We were told that the woodwind and brass sections had a good internal balance, so we covered them with a slightly more distant spaced pair of Audio‑Technica AT4040s, along with a centrally positioned AKG C414 B‑ULS intended to favour the violas, which were just in front. The bassoon was spot‑miked with a Microtech Gefell M930, while the saxes were covered in pairs using two Audio‑Technica AT4021s and an AKG C414 B‑ULS on the baritone. Paul also set up an Audio‑Technica AT825 stereo mic just in front of David to provide an overall stereo image. The keyboard turned out to be mono, rather than stereo, and we ended up with 23 channels in total.
Once the players were seated and had started warming up, we fine‑tuned the mic positions slightly and retired to our side room to check the feeds and set the recording levels. We soon discovered that the guitar and bass were causing the snare-drum head to rattle, so we persuaded the players to move their amps further from the drum kit, which improved the situation. There was still some snare buzz audible at our end but that was mainly from the close snare mic, which could be gated, if necessary, at the mixing stage. Anticipating that the performances would get louder as the band settled in, we left generous headroom margins (a little too generous, as it turned out!), entered the track details for the first take and pushed the red button, capturing the first take at five past two.
One of the handy aspects of the LRX2 is that it incorporates nine motorised faders and a full set of hardware control buttons, and at the same time as recording the individual channel inputs it will also record a live stereo mixdown, using the faders, switched in banks of eight to address all the available channels. This meant that we were able to build up a rough mix as we went along, which would make test and demo playbacks easier, as well as providing confidence that we had enough 'on disk' to produce a reasonable end-result.
The first few hours were spent recording nine instrumental songs, most taking between two and four takes to get a version that David was happy with. As we were working with an amateur band, total perfection was an unlikely ideal, and David was very realistic and pragmatic in knowing when he had achieved the best take — and not repeating takes unnecessarily when he felt that there was little to be gained.
With the instrumental numbers in the bag, we took a break for refreshments, before setting up for the 11 songs that included vocal parts. These were done with the singer positioned well in front of the orchestra and to one side, facing David to maintain some eye contact, but with the mic's null side facing the orchestra to give us maximum separation, something that was further improved by a large foam screen behind the mic. However, the screen obscured the sight lines a little too much, so we eventually compromised by propping it up sideways‑on. As the singers had a curtain behind them and no nearby walls, we reckoned the sound would be reasonably dry and clean with this arrangement, and the playback of our first take seemed to confirm this. By eight o'clock (an hour earlier than we had scheduled!), we'd recorded several takes of all 20 songs, and everyone professed themselves very happy with the way the recording session had gone. An hour or so later, everything had been packed away and loaded back into the cars, and the plan was then to transfer the files to Logic Pro back at my home studio for mixing.
We'd set up SADiE to record standard WAV files, with each track having its own monophonic or stereo file, named according to the song title, channel name and take number. This made it all very simple and logical, and easy to copy from Hugh's hard drive into suitably named project folders in my Logic Pro‑based system. However, after a while we discovered a very odd bug in the SADiE (MTR) software, in which the file naming on a few of the takes was messed up. It turned out to happen only after very short abortive takes had been made, and resulted in the individual track names being truncated. Thankfully, it was fairly easy (although tedious and frustrating) to figure out which tracks were which and rename them — and I'm happy to report that SADiE have now fixed the problem!
Before mixing, Hugh and I checked the individual sources. Everything was clean and undistorted, as we'd left generous amounts of headroom during recording (peaks were generally between ‑20 and ‑12dBFS, which is a little lower than normal, but we were expecting the band to play louder than they did). Actually, on some later tracks the double‑bass channel gave us a clean and undistorted recording of a rattling mic/pickup (which, unfortunately, we'd failed to notice during the recording) so we rolled off the high end starting at 1kHz, added some 80Hz lift to fill out the sound, and then compressed the end result to even up the level. Auditioned on its own, this still sounded somewhat short of ideal, but when brought in under the main stereo mic mix it actually worked pretty well. The kick drum also needed some EQ and compression to get it sounding solid, and we gated both the kick and snare mics to cut down on snare rattle and spill. Rather than use the gates to mute the sound, we set them to attenuate by around 12dB, which reduced the spill usefully but without making the gate action too obvious.
Some of the tracks needed to be brought up in level so that we could mix with the faders in a sensible position, so we did that by means of destructive gain changes to the WAV file. After importing the first piece and setting up the necessary plug‑ins and reverb send, we saved the setup as a template, to give us a head start when working on subsequent songs, and this strategy saved us a lot of time. Our template also included creating separate fader groups (levels, solos and mutes) for the strings, the brass and the saxes. Most songs were mixed straight from the original live take, although a couple of the pieces were composites, edited together using the start of one take and the end of another before mixing.
As expected, with such a compact musical ensemble, there was spill on everything. It's amazing how loud the cymbals managed to be on everyone's mics, and how absolutely everything poured into the string-section mics! However, the stereo mic had picked up a good overall sound, which we used to form the basis of the balance. The plan was to use the close mics to do the fine balancing, and also to lift certain instruments for their solos, using level automation.
This strategy still left the strings struggling to be heard over the brass during the fuller pieces, but we managed an acceptable balance by using high‑ and low‑cut EQ on the string mics to reduce the spill, and help them cut through, and on some pieces we also boosted the 3kHz range very gently to provide a more defined sound. As the string section was relatively small and weak, we also patched in an instance of Antares' Avox Doubler (artificial double tracking) plug‑in on two of the four string mics to thicken the sound further. We had initially tried a Roland Dimension D plug‑in, but found the chorus modulation just a hint too obvious, and we ended up not using it. After adding a generous level of concert‑hall reverb, the resulting string sound gelled well with the rest of the mix and came close to the sound David was looking for.
Once a balance had been achieved and all the unnecessary noises had been trimmed from the starts and ends of tracks, the whole mix was treated to some very gentle compression, and a hint of high cut using the PSP Vintage Warmer plug‑in. At David's request, we also added some warm hall reverb to the entire mix, just to take away the dryness of close miking in a pretty dead hall, and to create the vintage vibe he was looking for. When bouncing the tracks, we had to add a few seconds to the end of each song, just to allow the added concert-hall reverb tail to fade in a natural way, after which the various mixes were assembled into a CD format using Roxio's excellent Jam software, which dithers to 16‑bit after any level changes have been made.
And the results? If you go to the West Midlands Light Orchestra web site, at www.wmlo.org.uk, you can hear a couple of the tracks recorded in this session.
For those of you who aren't already aware, a 'light' orchestra, despite the name, is not a small‑scale affair: the word 'light' refers to the style of music played, rather than its size, although the balance between sections can differ from a typical classical orchestra (more horns and fewer strings, for example). If you want to learn more about this, there's plenty more information at www.turnipnet.com/mom/index.htm.
We did 'cheat' in the mix on a couple of sections, where some of the string pitching was a little suspect: David played their lines using sampled strings in Logic's EXS24 sampler plug‑in, and we layered those parts with the real strings. The end result was surprisingly convincing, and although it wouldn't trouble Mr Mantovani's string section, we felt that the result was quite acceptable.
Other mixing challenges included balancing the sax parts, because one player was so much louder than the others, and bringing up the slightly reticent baritone sax and bassoon parts when they featured. In these cases, mix automation came to the rescue.
On the vocal tracks, the standard of singing was generally excellent, with adequate separation from the orchestra, but there were just a couple of places where the pitching drifted off briefly: we hadn't provided a headphone monitor feed for the vocalist, and singing into a foam bowl is hardly conducive to accurate pitching, to be fair. It was a simple matter to correct the very occasional tuning problem using Melodyne. So as not to upset any of the singers, we'll not say which one(s) we had to work on, but it was only a couple of times in 11 tracks, which wasn't bad, all things considered. By way of processing, we ended up using an LA2A compressor plug‑in to level the vocal dynamics, along with the Waves Maserati vocal processing plug‑in for a little tonal shaping, a little more compression and some ambience. We picked one of the softer‑sounding 'contours' and adjusted the settings to get closer to what I think of as a '50s tube‑radio sound.