Studio SOS: Violin Recordings

The SOS team visit the sleepy Herefordshire town of Bromyard to help violinist Mike Burnham with his folk recordings.

A graduate of the Birmingham Conservatoire and London's Trinity College of Music, Mike Burnham is a seriously good violinist! He plays both orchestral concerts and Celtic folk concerts, the former with the New English Orchestra and the latter as half of Aranmore — his musical partner Meg Lawrie sings beautifully and plays electric piano along with numerous other instruments. Mike also builds and deals in violins so is very critical of tone, and while his recordings of himself and Meg (made in his lounge using his Roland VS1680 digital multi tracker) would have sounded quite acceptable to most people, he wasn't entirely happy with the violin sound and was even less satisfied with the overall sound of his mixes. Our challenge therefore was twofold: firstly, to come up with a better way of recording the violin and, secondly, to recommend ways in which his existing mixes could be made to sound better.

Miking The Violin

As is now becoming the norm for such visits, we discovered a number of problems in areas that were not obvious from the description of the original problem. The first step was to ask Mike how he normally recorded the violin — and the answer that we got was quite unexpected. It turns out that he has two capacitor microphones, an AKG C3000 and an Audio Technica AT3525. However, the C3000 was producing the top end Mike felt he wanted, but wasn't doing it for the low end. The Audio Technica, on the other hand, seemed less effective at the high end, but had a nice warm bottom end. Mike figured that if he could mix the characteristics of the two mics by using both at once, he'd get the sound he was after. This he set out to do by spacing the mics a few inches apart, set in front of and above the playing position, then (in most cases) he combined the outputs of both mics onto a single VS1680 track. To be fair, much of the time this produced a very usable sound, but Mike hadn't been aware of the phase implications of working this way.

Because the mics are so close to each other, both pick up sound from the same source, and unless the mic spacing and source position is absolutely fixed the results tend not to be repeatable, as changing the mic spacing only slightly will change the way the signals combine. Furthermore, if the source (in this case the violin itself) moves relative to the mics, as it must to some extent, the differing path lengths to the two mics will result in small phase differences between the two mic outputs, and when the signals are combined, colorations will be introduced as some frequencies reinforce each other while others cancel. As the violin moves, these colorations will change slightly, so it makes it very difficult to get a repeatable sound at different recording sessions, as well as making it harder to maintain a consistent sound throughout a recording.

Mike also told us that he'd borrowed a Neumann TLM103 from his friend and fellow violinist Dave Swarbrick (previously of Fairport Convention — Mike played in his band Smiddyburn with Mike and Kevin Dempsey during the late '90s), and that this had produced more of the sound he wanted. However, the TLM103 was currently outside his budget, so we had to find some way to work with the mics he already had.

Our first experiment was to use a single mic and to try to find a good position that would give acceptable results. The first tests were carried out in Mike's lounge where he'd made his original recordings, and it soon became clear that the AKG mic was tonally better suited to the task. The optimum mic position was decided using the old trick of monitoring via headphones while moving the mic. After both Hugh and I had taken turns at moving the mic around while listening, we settled on a suitable position with the mic a little behind and above the player. We also experimented briefly with placing the mic very close to the ceiling to gain a little bass boost from the boundary effect, but found we didn't really need it.

However, while the top end of the sound came together quite well, Mike felt there was a slight boxiness to the tone. This turned out to be quite difficult to track down, because Mike was using a pair of Wharfdale Diamond IV bookshelf speakers for monitoring, driven from an elderly JVC amplifier, and these gave radically different results to his headphones. The speakers were also set up on a small table along with the recording gear, so they were fairly close together and also very close to the wall, which I felt might be affecting their low end. Moving them a foot or so from the wall seemed to alleviate some of their low end coloration problems, but it soon became evident that there was a further problem insomuch as the tweeter in the left hand speaker wasn't working. This meant that all further evaluations had to be done while listening to the good speaker. With these limitations in mind we continued.

A Change Of Scene

Both Hugh and I felt that a carpeted lounge wasn't the best place to record a violin, especially as the room came very close to being twice as long as it was wide, which would tend to exaggerate any modal problems at corresponding frequencies. I'd already spotted the vinyl flooring in the kitchen on the way in, and as the kitchen was reasonably large and irregularly shaped, I suggested we try the same tests in there. Immediately the sound was less congested and more lively, so what we ended up with was the same mic arrangement we'd arrived at in the lounge (around a couple of feet above and slightly behind the player) using only the C3000. Initially we experienced a rush of background noise which we thought was due to condensation in the microphone, so we changed the cable to eliminate it from our enquiries and the problem vanished.

While the problem may indeed have been the cable in this case, it does however serve to remind us that capacitor mics like a little pampering, and leaving them in the car overnight in British winter weather is not to be recommended. If condensation is suspected, the mic should be stored somewhere warm for a few hours to make sure it is properly dried out, otherwise sporadic bursts of noise and drops in sensitivity are to be expected.

With no EQ or effects, we were now getting a very acceptable basic sound, though both Hugh and I were surprised to find that the preamps on the VS1680 had to be turned almost flat out to get enough level, even with such a sensitive mic as the C3000. In an ideal world, we would have suggested Mike think about buying a separate preamp, but as he'd just spent his budget for the foreseeable future on a Mackie PA for his live work, further expenditure was out of the question at this time beyond replacing the defective monitors.

Fixing The Mix

Like many classically trained musicians, Mike is a little wary of technology, so while he'd tried some of the preset reverb treatments to find which worked best for his mixes, he hadn't found one he was completely happy with, and he hadn't known enough about the adjustable parameters to feel comfortable creating his own. So, we sat down and tweaked while Mike listened and pretty soon we came up with a hall setting that flattered and warmed the sound without making it sound too washy. This was based on the Hall 1 patch, where we used a two-second decay time with 60ms of pre-delay and a density setting of 57. The high-pass filter was set at 160Hz to dry up the very low frequencies without robbing the main sound of warmth, and the low-pass filter was set at 9kHz to take out any unnatural high-end 'steam'. By making small adjustments to the decay time and low-pass filter settings only, this reverb could be customised to suit a number of different mixes.

After settling on a suitable method of recording the violin, and creating a sympathetic reverb treatment for it, Mike played us some of the recordings he'd already made. He felt the violin sound might be too aggressive, but what we actually discovered was that the electric piano sound, which came from a Roland general-purpose keyboard, was way too bass heavy, and also lacking in high-frequency content. As the keyboard wasn't available at the time, we couldn't figure out why this should be, as even the more basic Roland keyboards are generally able to produce a reasonably balanced piano sound, so instead we set about trying to fix it using equalisation.

Because of the digital filter algorithms used in the VS1680, quite extreme EQ settings were required to sort out a dull piano sound.

This turned out to be an interesting experience, because it taught us not only about the sound we were dealing with, but also about the characteristics of the Roland VS1680's EQ, which, in common with many digital designs, needs laying on with a trowel before it has any real effect.

Task number one was to get rid of that excessive low end, so we used the three-band EQ set to give 5dB of shelving cut at 160Hz, 5dB of mid-band cut at 200 Hz (with a Q value of 0.5) and 8dB of high-frequency boost at 1.8kHz. We tried EQ'ing higher up, but found that there was nothing much above 2kHz to boost — maximum boost at 4kHz had almost no audible effect on the sound! Such savage EQ is best reserved for real salvage jobs and is not recommended as normal practice, but in this case it produced a markedly better piano sound that sat nicely in the mix without filling up the low end with sonic 'wool'. Furthermore, when the new reverb treatment was used on the violin, and a little on the piano, the two instruments worked well together and only a few of the tracks needed any EQ on the violin at all.

I showed Mike the usual trick of finding areas of the spectrum that need cutting by first boosting the mid-band to maximum, then sweeping the frequency to find the area of maximum irritation. Once this has been identified, a little gentle cut at the same frequency is often all that is needed and, indeed, this worked well enough on the few violin tracks that needed correction.

Pad Problems

Some of the mixes included a pad synth or synthetic string sound, and in most cases these again had far too much bottom end, resulting in the mix becoming muddy long before the pad was loud enough to actually be audible. A degree of low cut solved this very easily, but in one case I had to set the low EQ frequency right up at 600Hz and cut by 12dB to get the desired result, which just reinforces the fact that EQ has to be set using your ears, not just by watching the numbers!

My philosophy on pad sounds is that, in most cases, they serve as a kind of musical glue that holds the various elements of a song together. My old woodwork teacher always used to remind us that the strength of a joint isn't down to how much glue you can get into it but rather how much you can squeeze out of it — and please, no gags about the glue just making it harder to smoke! I feel the same is often true of pad parts — you only need enough to get the job done. Put in too much and it leaks into all the cracks and oozes out of the edges!

That left only the vocals to attend to and, as with the violin, Mike had hedged his bets and used both mics at once, though in most cases he'd recorded them to separate tracks. Overall the vocal sound was good, though we could hear the room intruding on some of the recordings and the only solution for that, other than bringing out the family duvet collection, would be to work a little closer to the microphone, not forgetting to use a pop shield at all times. In an untreated room such as Mike's lounge, a working distance of six to nine inches from the mic would usually be OK, though surrounding the singing position with duvets or other absorbers (at the rear and sides) would definitely help to clean up the sound and reduce boxiness.

Although the vocals recorded via the C3000 sounded fine, the Audio Technica was turning in a much more coloured sound, but as we weren't present during the recording we can't say why this was. We decided to continue using only the C3000 tracks where these were available separately. The basic sound was good with little or no EQ (some tracks needed -1dB at 140Hz and +4dB at 19kHz to add a little air) and the performances were reasonably consistent in level, though both Hugh and I felt some compression would help them sit more stably in the mix.

To this end, we used the VS1680's compressor via the vocal channel insert point and then set it entirely by ear, as there appeared to be no gain reduction read-out. What we arrived at was a threshold setting of -21dB (though this is of course dependent on the absolute recorded level of the track and these weren't recorded particularly hot), a 4:1 ratio, an attack setting of 30, a release setting of 50 and a gain setting of around +3dB to make up for level lost during compression. As expected, the vocals firmed up nicely and remained more uniform in those areas where the style of delivery changed.

To finish the day, we opened a number of Mike's existing songs and applied the same treatments to the piano and vocal sounds, and the general consensus was that the mixes worked much better, the most significant change being in the piano sound, the very dullness of which had appeared to make the violin sound aggressive when it was actually very close to being OK.

Summary

Because no additional budget is available for recording equipment, largely due to the needs of Mike's live setup, we didn't do our usual trick of trying external mic preamps or alternative microphones, but instead made the best of what we had. The only items that will have to be replaced are the speakers, as it would probably cost more to get a tweeter replaced than to buy different (and hopefully more accurate) monitors, especially if suitable second-hand models can be found. Moving the speakers away from the wall helps reduce boominess and placing them on stands rather than on a table would improve matters further.

Recording the violin in a room with reflective and irregular surfaces produced demonstrably better results than using the same mic technique in a carpeted lounge, and of course the woolly piano sound can probably be sorted out at source now Mike knows what to look for. In all, it was a very successful day all round and it was an added bonus that we got to hear some of Mike's fantastic playing.

Coda: Finishing Off The Album

Mike called me up a couple of days after our visit to seek help in mixing his tracks, as he needed to produce a CD quickly and, in addition to his damaged monitors being unsuitable for mixing, he also needed some editing done.

He brought his VS1680 round to my studio and we plugged it directly into my Yamaha 03D mixer, playing it back over my Mackie HR824 monitors. Everything sounded so much cleaner and, where EQ was needed, it was much easier to locate the problem frequencies. We used the direct outs on the VS1680, so I had to duplicate all the EQ and dynamics settings in the 03D, which worked out extremely well. The main task was to roll off low end from the piano and pad parts and to try to add some presence to the piano. Conventional compression was used to hold the vocal levels steady, again using the 03D's compressors. Because no two digital equalisers sound quite the same, I had to make a few tweaks to the settings we'd originally arrived at on the VS1680 and, in the case of the piano, I found that I could EQ higher up the spectrum and still get an audible effect — something we were unable to do with the damaged monitors.

Mike's finished album, available from his web site.

For vocal and violin reverb, I tweaked one of the Medium Hall programs on my Lexicon MPX500, which Mike immediately picked up on and said that it was the character he'd been looking for. We used variations on this throughout the album, changing only the decay time to between 1.6 seconds and 2.2 seconds. So with Mike doing manual mutes where necessary and me riding the one or two faders that needed adjusting, we mixed the whole album manually in a long afternoon

We then transferred the tracks to my Mac, where I compiled the album using the now ancient Sound Designer II. I would have used TC Works Spark or Bias Peak, but I didn't need any processing and I am still an order of magnitude faster using Sound Designer than any of its modern counterparts. No mastering processing was used other than the lightest multi-band compression (1.1:1, threshold -35dB) courtesy of my Drawmer DC2476. The result was a truly delightful album and, what's more, I think we managed to make all the mixes sound as though they belonged together.

Mike's finished album is available from his web site: www.folkicons.co.uk/burnham.htm.