Duo Melodika: This month one of our engineers aims to augment the natural beauty of a baroque‑music duo playing in a church.
For five years, Elisabeth Detre and Ulrike Kalchreuter have been playing classical music together professionally as Duo Melodika, and recently decided to upgrade their publicity material. To this end, they asked me to help create some recordings for their web site (www.duo-melodika.de), showcasing the range of music they offer. In particular, they wanted to capture two baroque line‑ups: solo voice with church organ, and recorder with cembalo (a harpsichord‑like instrument). Both keyboard instruments were housed in a local church where Elisabeth is organist, and she was able to book the best part of a day there for recording purposes.
Since the church wasn't familiar to me, I arranged a preliminary visit with Elisabeth to scout things out. The space was about 50x14m, with an 8.5m ceiling height. At one end was the 3.5m‑deep organ gallery, elevated 5m above the floor, and to the left of its central keyboard console were two rows of fitted choir pews, behind which stood the cembalo. The tuning of wood‑framed baroque keyboard instruments can be extremely temperamental, so Elisabeth had sensibly scheduled our session hot on the heels of the tuner's visit, but that also meant that we were forbidden from moving the instrument in any way, as this would upset the refreshed intonation.
With small classical ensembles, the most common recording approach is to try to pick up all the players in one go using a single stereo mic rig. However, while it's possible to achieve excellent results with such a simple setup when all the cards fall in your favour, it's an inherently risky undertaking, because if you misjudge the balance, timbre, or perceived distance of any of the performers in the heat of the moment, there's precious little you can do to remedy things later. And when you're recording unfamiliar musicians in an unfamiliar location without the benefit of a controlled monitoring environment, as I knew we would be in this instance, there are plenty of opportunities for misjudgment!
To reduce this risk, most real‑world classical recording setups supplement the main rig with additional 'spot' or 'accent' microphones on individual instruments, thereby giving the engineer a bit of a safety net at mixdown, as well as affording a certain degree of post‑production control over the recorded musical balance, which is something that's particularly useful when working with less experienced performers. The downside with spot mics, however, is that you need to place each one close enough to its target instrument that you don't get too much spill from neighbouring instruments, which means that spot‑mic signals usually sound a lot less natural than the main stereo rig. This is why it's still vital to get the best out of your main stereo rig, whatever supplementary mics you plan on using, because the more sparing you are with the spot‑mic signals at mixdown, the more faithful the mixed result is likely to be.
My primary technical concern while investigating this recording venue was where I'd place my main stereo rig in relation to the performers and instruments. I already had to work around the organ and cembalo positions, but there were other considerations to factor in. The vocal‑plus‑organ setup presented the greatest difficulties, because it was important that the musicians be physically close enough to allow direct musical communication, and Ulrike was accustomed to singing from a position towards the back of the gallery directly alongside the keyboard console for this reason. Unfortunately, many of the organ pipes Elisabeth was using were housed in an enclosure along the front of the gallery, and if I placed my main stereo array a little way into the church to pick up a well‑balanced organ sound, I knew that would make the voice sound very distant by comparison. Moreover, I was worried that any vocal spot mic I used at the rear of the gallery to redress this depth misrepresentation might pick up excessive mechanical noise from the organ console's stops and keyboard action.
After a little head‑scratching, I suggested a compromise whereby Ulrike would stand at the front of the gallery instead, facing away from the organ console and singing into the church. I figured that this would help bring her forward in the main stereo rig's depth perspective, such that I wouldn't have to lean as heavily on her spot mic at mixdown, and that we could restore the interrupted sight line between her and Elisabeth by using mirrors. I have a pair of large lorry wing‑mirrors I picked up for £20 that are great for such eventualities, because they slot rather neatly onto mic stands, which makes positioning easy. (I also made a mental note to have a pair of foldback headphones on standby in case Elisabeth had any trouble hearing Ulrike's singing well enough. We didn't actually need them in the end, but it's better to be safe than sorry!)
With the recorder‑plus‑cembalo line‑up, I felt I had enough scope to adjust the balance and depth of the instruments at source by repositioning them, despite the inconvenience of having to work around the fixed pews. However, I suspected that the church acoustic's heavy reverberance might overwhelm instruments designed more for chamber‑music environments, so although I was keen to use omni mics as my main stereo rig for the organ recording, I decided to use a pair of directional mics as the primary pickup system for the other instruments. Another alternative would have been just to move the omni mics closer, but I rejected that idea because I was pretty sure I'd have to get so close in that the timbral balance of the instruments would suffer as a result. Not that I was going to let those omnis go to waste, though: I planned to repurpose them within this context as room mics, providing me with some independent wet/dry control at mixdown. Given the limited depth of the organ gallery, I opted to face Ulrike diagonally across it towards the centre of the church, so that I could place the directional mics far enough away.
The day of the session arrived, and I'd already arranged with the musicians that I'd need an hour or so to set up before they needed to be on site. Given that no separate 'control room' space was available, I decided to set up my hard‑disk recording workstation in the organ gallery on the opposite side to the cembalo. This meant I could talk and signal to the players easily, but was far enough from any mics to avoid hard‑drive and fan noise sullying the recording. From this position, I ran one eight‑way XLR multicore cable across to the other side of the gallery, and carefully lowered a second multicore to the church floor to service any mic stands there.
One of the secrets to speedy recording setup is making reasonable guesses beforehand about how the mics should be configured. For the main omni pair, I decided to use some new DPA 2006C twin‑diaphragm mics that UK distributor Sounds Network had kindly lent me to try out. Estimating that I'd probably be miking around 3‑4m from the edge of the gallery, I plumped for a fairly conservative two‑feet spacing between them. This gives a stereo pickup angle of around 60 degrees in my experience and, given that the organ itself occupied around two‑thirds of this angle, I reckoned this would give its image a sensible width in the recorded panorama. Admittedly, preferences vary quite a lot between different engineers when it comes to image width, but personally I like to leave a little space around the edges of a classical ensemble in the stereo field so that the room sound can then spread to the edges of the stereo field to surround the players on either side.
In order to capture as much of Ulrike's direct sound as possible, I mounted both mics on my tallest stand, raising them around six metres off the floor, where they could clearly 'see' her over the gallery rail. Although I have no long stereo bar (in studio situations I normally just use separate stands for spaced‑pair rigs), I was able to jury‑rig a workmanlike substitute by using a regular mic stand's boom arm as a crossbar on the end of the tall stand's boom, to which I could then fix the mics using the handy little K&M 238 screw‑on mounts that cost me only about a tenner apiece.
For the directional pair, I decided to use my Shure KSM141 pair in an NOS‑style near‑coincident setup — with the mics 30cm apart and at a 90‑degree mutual angle. This offers a stereo pickup angle of roughly 80 degrees in practice, the suitability of which I again judged in terms of the kind of distance I imagined placing the mics from the performers — at about three metres away from the players, I reckoned they'd occupy roughly half‑left and half‑right positions respectively in the recorded panorama. Although the NOS configuration retains a good deal of inter‑channel time‑difference information (which gives recordings a pleasing spaciousness), it supplements that with inter‑channel level difference information for a sharper imaging illusion, and I felt this would benefit the chamber‑duo sound.
In addition to these two main mic pairs, I cabled up a couple of potential spot mics for Ulrike's singing (an AKG C414B XLS and a Superlux R102 ribbon mic) and two additional cardioid pencil mics (Avantone CK1s) to cater for any unforeseen miking requirements. I'd much rather have more mics line‑checked in advance than I need than have musicians waiting around on a session while I rig up additional mics, cables and stands from scratch.
We decided to concentrate on the vocal‑plus‑organ pieces first. I'd already warned the girls that we might need to spend 20 minutes or so doing some test recordings for this purpose, given that I was monitoring on open‑backed headphones in the performance space. Even so, I deliberately worked on Elisabeth's organ sound alone initially, in order to avoid unnecessarily tiring Ulrike's voice. As I'd expected they would, the DPA mics delivered a lovely clear, rounded tone right away, but when I compared a test recording against some snippets of commercial releases (which I'd imported into my system beforehand) I decided that we could do with more of the church's reverberant 'bloom'. Pulling the whole stand back roughly another metre did the trick, whereupon my only remaining concern was that the musical key's root note seemed over‑prominent in the balance. When you're working on location, you frequently find yourself at the mercy of the venue's room resonances, so I guessed that the problem might be that my miking height coincided with a room‑resonance hot‑spot. Sure enough, raising the mics two to three feet higher seemed to provide an adequate fix.
I then asked Ulrike to join in so that I could assess her contribution via the main mic pair. Fortunately, she seemed to come through with a respectable tone without any further changes to the mic positioning, so I proceeded directly to auditioning the two spot mics I'd already positioned a couple of feet above and in front of her. The inherent on‑axis high‑frequency boost of the large‑diaphragm AKG C414B XLS really didn't suit her at all, emphasising a slightly hard edge in her voice to the point of unpleasant harshness on some of the higher and louder notes, even after some hasty repositioning of the mic. I abandoned that pretty swiftly, in favour of the much smoother sound of the ribbon mic.
Although understated in the upper spectrum, as most ribbon mics are, the general quality of the R102's sound immediately felt much more appropriate, filling out the timbre from the main stereo pair in a very appealing manner without any undesirable phase cancellation. I fancied a touch less low midrange from the voice once it was in the mix, so cut a decibel out of the close‑mic signal at around 400Hz, and I applied a couple of decibels of high‑frequency shelving boost above 10Hz, to give it a fraction more upper‑spectrum detail. The level of organ spill worried me a little, though, so I re‑angled the mic slightly to place the front‑of‑gallery organ pipes better into its figure‑of-eight polar pattern's rejection plane, and I also asked Elisabeth if we could close a couple of wooden doors in the pipe‑housing cabinet in front of Ulrike that had hitherto been open.
I now had the recording rig up and running, so it was time to swap hats and assist the performers more as producer than engineer, listening for performance or tuning problems while they were laying down takes, and maintaining consistent tempo between takes to allow for later comping. We also had our work cut out avoiding unwanted background noises: trains, planes, automobiles, bell‑ringers, church flower arrangers — it got a bit ridiculous at times! Nevertheless, we eventually managed to get all the necessary takes in the bag, and after a quick pause for refreshments were able to press on with the second duo line‑up.
By comparison with the morning's activities, placing the musicians and main stereo rig for the recorder‑plus‑cembalo setup was pretty straightforward. It helps that the NOS setup captures something that sounds reasonably similar to what you actually hear when your head's at the miking position, so I was quickly able to find a good distance for it by ear. I could also identify a decent location for Ulrike in terms of the acoustic balance of her recorder against the cembalo, without having to go through the rigmarole of test recordings, which was a big time saver.
What wasn't apparent until listening to my first test recording, however, was that the captured image felt over-wide, such that the musicians sounded rather dislocated from each other. I've seen some home recordists try to compensate for this by narrowing the panning of their stereo mic pair, but this doesn't just narrow the perceived instrument spacing; it also narrows the recorded room sound. I always prefer to address stereo‑width issues while recording if possible, so that the room ambience remains as open‑sounding as possible. In this case, I could have narrowed either the spacing between the mics or their mutual angle, and I chose the latter simply because it was quickest to implement with my particular stereo bar. (I also applied a general low‑frequency shelving EQ boost at around 130Hz to compensate for the bass drop‑off that's inherent in the Shure KSM141's cardioid mode.)
By this point, I felt I'd attained a fairly creditable representation of what was happening in front of the mics, albeit erring deliberately on the dry side. I made that choice so that there'd be enough space for the wetter signal from the more distant omni pair (the main stereo array I'd used for the voice‑plus‑organ rig), which I'd rotated to face directly towards the players so that it wouldn't give a skewed stereo image. As such, there was scant technical reason for using any spot mics at all. Try as I might, though, I couldn't shake a niggling sense of dissatisfaction with the lightweight tone of the cembalo, which seemed strangely anaemic alongside the mellow timbre of Ulrike's wooden‑bodied alto recorder. Given that I'd not spent much time moving mics for this duo line‑up so far, I resolved to try using a spot mic to deliberately misrepresent the cembalo tone in the name of flattery — in other words, to make the instrument sound fuller and warmer than it actually did in the flesh!
I reached for my Superlux ribbon straight away, reasoning that its mellow tone and powerful proximity effect would work together in my favour if I used it fairly close up to the instrument, and that its unhyped transient reponse (which is typical of ribbon designs) might also prevent overbearing pick noise. Once again, the R102 delivered a star turn — it really is an amazingly useful little mic for the money! That said, I did have to work with its exact positioning for a few minutes to refine the degree of proximity bass boost, as well as shifting the mic over towards the tail of the instrument to reduce pickup of mechanical noises, and angling its null plane to reject spill from the recorder. A gentle high‑pass filter rolling off below 100Hz also proved necessary, because the proximity bass boost was over-emphasising the keyboard action's low‑frequency thudding in a rather distracting manner.
I still kept the spot mic fairly low in the mix, just adding enough to boost the cembalo's BMI to healthier proportions. Nonetheless, the keyboard's dry level was thus increased sufficiently to tamper with the balance and depth decisions I'd made while setting up the main stereo pair. One response to this might have been to ask Ulrike to shuffle a little further towards the main mic pair, but in the event I chose the more flexible option of giving her a spot mic too. For this I chose to use the C414B XLS, switching it to its hypercardioid mode to guarantee a usefully dry and spill‑free signal without having to bring the mic too close, and then calming down the mic's on‑axis HF boost with a few decibels of shelving EQ cut above 5kHz. For me, the most promising mic positions for woodwind are often to be found perpendicular to the instrument's body, and so it was here — I was lucky enough that the first option I tried seemed to work well.
With the sonics sorted, we shifted back into 'takes mode' as before, and managed to finish work by supper time. Back in the studio, later the same evening, I bounced out rough mixes of all the takes and sent them to Elisabeth and Ulrike, so that they could decide on edits. Once they'd made up their minds about everything, they sent me copies of the sheet music annotated with their desired take numbers, so that I could compile the final edited versions of each piece.
For the most part, a classical production's sound is decided during recording, so you shouldn't have to do much at mixdown. In this case, I was pretty happy with the sound of my rough mixes from the session, although I always make a point of re‑evaluating and finessing any in‑session EQ decisions with fresh ears. While doing that, I further processed the cembalo spot mic with iZotope's Alloy to warm it up a little more while keeping the keyboard's mechanical noises in check — combining a touch of low‑spectrum boost, some general transient‑reduction for the plectrum noises and a low‑frequency limiter to catch the structure‑borne thuds.
I felt no need to add artificial reverb to any of the main‑rig mics, but I did use some on the spot mics to help glue them into the overall picture better, especially the vocal close mic, since it was comparatively high in the mix. The algorithm I used was the 'Church' impulse response bundled with Christian Knufinke's SIR2 convolution engine, which seemed to match the venue we recorded in fairly well once I'd shortened its decay a little. The final step was to ride the soloist spot‑mic level in each case wherever I felt this benefited the musical balance.
Neither of the duo‑recording situations featured in this month's article were ideal, by any stretch of the imagination, but you won't get far on a budget if you can only reach a positive outcome when the going's easy! Fortunately, as I've tried to demonstrate here, there's plenty you can do to work around setup challenges if you plan carefully, think laterally and carry a few automotive spare parts...
Lady Luck wasn't on my side health‑wise for this session, because I managed to come down with a streaming head cold the day before we were due to record. By the morning of the job, my sinuses were well and truly bunged, so it felt like I was hearing the world through a blanket! I was reluctant to cancel the date, however, because it had been tricky to coordinate, so I resolved to soldier on regardless. This was where having some commercial reference tracks on hand in my recording system really paid off, because it helped me avoid the trap of over‑brightening things in response to my temporarily dulled hearing. I was also glad I'd managed to get hold of some cough drops on the way to the venue, because those helped me stave off a couple of potentially embarrassing mid‑take coughing fits...
On most sessions, I find at least one good reason to kick myself, and on this occasion it was because I forgot to ask the musicians to bring a spare copy of the sheet music to the session for me to use. If you can read notation, then it's much easier to annotate a score with comments about each take than it is to work in terms of minutes and seconds — putting a ring around a dodgy note is a whole lot quicker than looking at your recorder's time display and jotting down three or four figures! As it was, I think we probably ended up doing more takes than strictly necessary, simply because I wasn't as confident that we had all the material we needed to achieve a convincing comp.
Judge the recording techniques in this month's column with your own ears by listening to the audio examples I've provided, which include raw multitrack files from the session, the final mixes and some demonstrations of how my mic repositioning and processing decisions helped shape the sound during setup and mixdown.
The audio files available on this page accompany my article about recording Duo Melodika (Elisabeth Detre and Ulrike Kalchreuter) in SOS August 2014 (http://sosm.ag/aug14_sessionnotes). 'Duo1' in the filenames denotes the voice‑plus‑organ line‑up, while 'Duo2' denotes the recorder‑plus‑cembalo ensemble, and here are some additional notes to describe exactly what you're hearing in each case:
This is the first test recording I carried out while setting up the main stereo mic pair to capture the church organ. The mics are DPA's 2006C small‑diaphragm omni condensers, and were set up in an A-B configuration spaced 24-inches apart, to give a stereo recording angle of around 60 degrees. The mics were raised roughly 6m above the church floor on a tall mic stand, and were 3m away from the edge of the gallery rail.
My first‑guess mic placement turned out to be too close and dry sounding, so I moved the microphones another metre or so away from the gallery rail to give a more suitable reverbation pickup, as you can hear in this example. However, you can also hear that the root note of the music is being unduly over-emphasised in this position too.
Moving the microphone pair another 2‑3 feet higher into the air helped keep the music's root note more in balance, presumably by moving the mics out of a room‑resonance hot‑spot. This was the point at which I decided I could start introducing the singer into the equation.
Here's what the main stereo mic pair sounded like for a section of the Duo 1 recording, with both organ and voice in action. No processing has been applied, so all you're hearing is what came directly out of the DPA 2006C omni mics I used. As expected given Ulrike's singing position, the vocal feels rather distant compared with the organ, which necessitated the use of a vocal spot mic.
The first microphone I tried for the vocal spot mic was an AKG C414B XLS, but it really didn't suit her voice in this instance. In particular, the microphone's inherent on‑axis high‑frequency boost over-emphasised a hint of harshness in her vocal tone, for example when Ulrike sings the loud high notes at 0:24 and 0:37. (No processing has been applied to this mic signal.)
The second spot mic I tried on the vocal was much more successful: a Superlux R102 ribbon mic. In this example file you can hear its raw recorded sound, which, although a touch too cloudy tonally, is otherwise much smoother and more musical in this scenario. Notice also how low the organ spill level is, despite its pipes being only a few feet further away — a testament to the deep rejection null of the ribbon mic's figure-of‑eight polar pattern. (In case you're wondering, I also had the C414 switched to its figure‑of-eight polar pattern for the same reason.)
For the mix, I used a 70Hz high‑pass filter from Fabfilter's Pro Q plug‑in to remove some unwanted stand‑borne rumble (the ribbon mic had no suspension shockmount) and then also slightly clarified the tonality with a couple of other tweaks: a 1dB peaking cut at 410Hz (Q=1.4), and +3dB of high shelving boost (Q=0.4). Compare this example file with 06_Duo1_VocalSpotRibbon_Raw to highlight the differences this made.
Here's the same section of my final mix, so you can hear how everything fit together in context. The only further processing I applied was adding some church‑style convolution reverb to the spot mics and riding the spot mic level in the mix to help the musical balance where necessary.
In this example file you can hear what the mic signals from the recorder‑plus‑cembalo ensemble's main stereo pair sounded like, without any processing. The mics were Shure KSM141s in cardioid mode, set up in a quasi‑NOS near‑coincident setup. These particular mics have a low‑end 'suckout' in their cardioid mode, so the tonality feels rather thin, and they're deliberately placed for a sound that's slightly too dry, in order to allow some control over the wet/dry levels using room mics.
To combat the thinness of the Shure KSM141s' cardioid tonality, I applied a 5dB low‑frequency shelving boost at 130Hz with an extremely gentle slope (Q=0.5) using Fabfilter's Pro Q plug‑in. I still felt that the cembalo could be warmer‑sounding, though, but that was a problem with the instrument itself, so there was little I could do about that using my main stereo pair.
The DPA 2006C omni mics that had been the main stereo rig for the voice‑plus‑organ duo were repurposed for the recorder‑plus‑cembalo duo to serve as room mics, picking up a more reverberant sound in the room. Because both players were now situated to the left‑hand side of the organ gallery, I turned the stand to face them, so that we didn't get a lop‑sided room signal. These mic signals are completely unprocessed.
In order to fill out the cembalo sound, I placed the Superlux R102 ribbon mic at the tail end of the instrument, using the powerful proximity effect of its figure‑of-eight polar pattern to enhance the timbre's apparent warmth. However, as you can hear from this example (which is completely unprocessed), there was inevitably a certain amount of unwanted mechanical noise picked up from the instrument too from such a close mic position. Listen also for the amount of recorder spill, which is very low when you consider that the instrument was only about 3‑4 feet from the mic. Again, the figure‑of-eight polar pattern's deep rejection nulls are largely responsible for this.
The proximity effect of the cembalo's spot mic was emphasising the low‑frequency thudding of the keyboard action. I was able to deal with some of this by simply applying an 80Hz high‑pass filter from Fabfilter's Pro Q, as you can hear in this example.
Improving the cembalo spot mic's sound further at mixdown required more specialist processing from iZotope's Alloy. Firstly I used its multi‑band transient processing to recess the plectrum noise generally across the whole frequency spectrum, and then I compressed just the frequencies below 370Hz so that I could boost the instrument's warmth without making low‑frequency keyboard‑action thuds once more a problem.
Here's what the raw signal from the recorder spot mic sounded like. The mic in question was an AKG C414B XLS large‑diaphragm condenser, which I'd set to its hypercardioid polar pattern so as to reject both cembalo spill and general room reverberation, thereby allowing me to place it at a reasonable distance (3‑4 feet) from Ulrike's recorder.
Although the recorder spot mic was capturing a reasonable dry sound, I felt that the AKG C414B XLS's on‑axis high‑frequency boost was making the instrument's tone a touch abrasive, so at mixdown I rolled off 3.5dB of high end using a 4.4kHz high shelving cut from Fabfilter's Pro Q. Compare this example to the 15_RecorderSpot_Raw file to hear the extent of this change in practice.
Here's how the record‑plus‑cembalo ensemble sounded in my final mix. As with the other duo mixdown, the only additional work I did was to add a little convolution reverb to the spot‑mic signals and to ride the level of the recorder spot mic in the mix for musical reasons.
Here's a section of my final mix of the first duo, so you can hear how everything fit together in context. The only further processing I applied was adding some church‑style convolution reverb to the spot mics and riding the spot mic level in the mix to help the musical balance where necessary.
Here's how the record‑plus‑cembalo ensemble sounded in my final mix. As with the other duo mixdown, the only additional work I did was to add a little convolution reverb to the spot‑mic signals and to ride the level of the recorder spot mic in the mix for musical reasons.