For five years I’d been struggling to record a tricky small drum called the dumbek. You’ve probably heard it—while it has always been common in countries near Turkey and around the Mediterranean, these days it pops up in mixes of all kinds, from “world music” to smooth jazz and all manner of dance/trance/samplemania.
Most of those appearances are samples, but to us recording musicians the recording of a real performance is way more fun and challenging than programming a static sample. I had tried half a dozen techniques, without success—until, at last, I had a good recording. Elated, I told a brief version of the story in my review of the BLUE Kiwi microphone (Recording, May 2002).
Our editor, Lorenz Rychner, suggested that readers might want to know the rest of the story. “And, while you’re at it,” he added, “you might talk about how to record other ethnic percussion instruments found in world music—things like the djembe, congas, bodhran, guiro, tablas. 4000 words, please.” Gulp.
Okay, let’s start with the dumbek story, because it illustrates some important issues about recording percussion.
I seldom use condenser microphones to record ethnic percussion instruments.
Why? First off, many condenser mics, particularly those with smallish diaphragms like the Oktava MC-012 or the Neumann KM 84, have resonances in their upper octaves that can be excited by high ringing frequencies on instruments like bells, tambourines and the like. These frequencies can interact badly with the A/D converters on a digital recorder.
Second, sometimes condensers are a bit too literal for recording percussion. The undiminished dynamic range of percussion instruments can be huge, with a peak-to-average ratio greater than 20 dB. Combine that with the condenser microphone’s ringing at high frequencies, and you can get a signal that’s all peaks and ringing, with no oomph remaining.
Instead of condensers, dynamic microphones of good quality are usually my first choice. Sometimes I go for ribbons—modern, tough ribbons like the Beyer M260 that has done yeoman service over the years (one of these days I mean to try out the M160). Not the fragile RCA-type ribbons of days gone by—those would twist and die in an instant if you put them under, say, a djembe.
My first choice among dynamics is often the Electro-Voice RE15, now discontinued (and deeply lamented), or its surviving sibling the RE16. Other winners include the Sennheiser MD 421 (Mark I, not Mark II), the less-common but excellent MD 441, the Beyer M201, and the wonderful little Sennheiser MD 109.
By using dynamic microphones I cut down on ringing and take advantage of a certain amount of nonlinearity in the dynamic mics. They add some useful compression to the signal, with very fast attack time, squeezing the peak-to-average ratio enough to bring out the guts of the sound rather than the tink-tink.
Of course, I sometimes break my own rules; we’ll talk about that later. First, though, let’s look at that dumbek.
Dumbek—five trying years
The dumbek usually has a ceramic or metal body; traditionally, the head is a goatskin. It makes a remarkable variety of sounds. Hit it one way, and you get a sharp “bek;” hit it another way, and the whole room resonates with the low “douuuum.” That variety is the delight of the instrument, of course, but it’s also the nightmare—over the last five years or so I’ve tried several methods of recording this drum, with mixed success.
The “bek” is easy. Put an RE15 about 8–12" from the head of the drum, facing a little off-center and angled at about 45 degrees, and you’re there. Normally no eq is necessary, but a small boost at around 8 kHz may help if you need extra “skin” sound (usually, on a dumbek, you don’t).
Unfortunately, this doesn’t do squat to capture the “douuuum.” That sound originates inside the drum that’s acting as a resonant chamber, it exits mostly through the hole on the slightly flared bottom piece, and it resounds through the room. Indeed, the room’s response is an integral part of the sound. And did I sweat to catch it!
My first attempts came in a large church with live acoustics. I was recording a Sephardic band with voice, guitar, flute, and dumbek. To preserve the integration of the room and the musicians I miked the guitar/voice and flute using an ORTF stereo pair, about 3 feet in front of the lead singer, with the flute sitting close on his left. The singer/guitarist and the flautist were up by the altar; the dumbek I placed out in the congregation, as otherwise it leaked badly into the main pair.
I miked the top of the dumbek with my usual RE15 and tried placing another RE15 about 6" below the drum, aiming up, in the hope of capturing the “douuuum.” I failed completely. I caught an inchoate boom that had nothing to do with the sound I was hearing in the room. I moved the mic around, always keeping it within a foot of the bottom opening, but no go—I got nothing but mush.
I finally punted, deciding the room resonance leaking into the main pair was adequate for the session. I axed the bottom mic and we rolled. I was never satisfied, though, and indeed we never used those tracks.
Had I but known that I had come very close to the solution!
Under the canopy of douuuum
Fast forward a few years. I was now working with a partner, Brad Sarno. This time we recorded the Sephardic band in the lead singer’s study. We close-miked the players, adding an ORTF pair for the room sound. We set up a wedding canopy (chuppah) over the main musicians to control a bad floor-to-ceiling bounce.
I decided to keep the dumbek player separate from everyone else; we put him in the living room, with headphones. I still miked the top with a single dynamic mic, this time a Beyer M88, but I took a different tack for the douuuum. I decided to capture this with another ORTF pair of room mikes, two Oktava MC-012s with cardioid capsules, mounted on a stereo spreader bar from AKG (Beyerdynamic sells the same one). We recorded on a pair of 20-bit ADATs (the earlier recording was direct to 2-track DAT), giving us the chance to tweak in mixdown.
The results were better—much better. It was now possible to get a blend of room and drum top that at least suggested the richness of the dumbek’s sound. But it wasn’t quite there, not yet.
Another day, another church
Time passed. We convened for another recording session, this time in a church with lovely, warm acoustics, non-parallel walls, and plenty of diffusion. An acoustic dream.
This time, we recorded the dumbek (now played by the guitarist/singer) as an overdub, rather than deal with the problems of bleed inherent in recording it live. I remembered Wes Maluk of the Indiana/Michigan band Merriweather, an excellent sound guy as well as a good player. His dumbek sounded great in concert. In addition to miking the top, he had inserted a dynamic microphone inside the body of the dumbek, about 3–4" in, and slightly off-center.
I decided to try it. For our overdub session, I set up my usual top microphone, the RE15 again. The Beyer M88 I put on a long boom, extending it to the floor, and angling the clip to put the mic inside the drum, facing up toward the head. (Our player holds the dumbek between his knees, vertically, rather than under his arm.) This required some fancy maneuvering, and I wished for a right-angled XLR connector on the end of the cable, but we managed to avoid hitting the floor or the drum body.
Why the M88 inside? First, it’s a hypercardioid microphone, with a whopper of a proximity effect. This means that the bass response tilts up as you get closer to a sound source, rather like turning up a shelving eq. I suspected that would be good for the douuuum.
Second, the M88 puts up with extravagantly high sound levels without audibly distorting, but it adds some nice, smooth compression, keeping the dynamic range within reasonable limits.
Finally, once again I miked the room using an ORTF pair of Oktavas, but this time I faced them away from the musicians, allowing the mics to capture a lot of room sound and not much direct sound.
For microphone preamps I used the Project r DIY solid-state preamp (a three-parter, Oct.–Dec. 1998 Recording) for the M88 bottom mic, a tubed DIY preamp for the RE15 on top, and Brad’s Manley preamp for the room mics. Again, we recorded to 20-bit ADAT, each mic on its own track.
Quickly it was apparent that we’d hit the jackpot. The sound was rich, sharp and resonant, all at once, just like a real dumbek. In doing preliminary mixes I had expected it would be necessary to flip the polarity of the bottom microphone, since it was catching sound from the back side of the drum. To my surprise, it wasn’t necessary, the bottom sounded better with the track mixed normally.
Likewise, we didn’t slide the room-sound tracks to make the drum hits line up with the close mics; they sounded better mixed straight, incorporating the natural delay. They also sounded better than any artificial reverb I’ve heard. A good room still beats everything.
So, at long last, I had a good recording of the dumbek—using four tracks, and doing it by overdub rather than simultaneously with the rest of the band. Well, hard instruments sometimes need complex solutions.
Let’s look at what worked. (1) I miked the top with a dynamic mic, for minimal ringing. (2) I miked the inside with another dynamic, for compression; it had a strong proximity effect, which helped capture the deep bass sound. (3) I miked a real room with a true stereo pair, aimed away.
While this worked for me, it’s not necessarily a universal solution. I had a luxury that you might not have: enough open tracks to devote four of them to a single drum overdub. And I had the rare super-luxury of a wonderful large room.
How might I handle the same drum with fewer available tracks, or while stuck with mediocre room acoustics? I think I would mike the drum itself as above, but do away with the room mics. I prefer to mix in the computer, so I would try to create a good room with a program such as Soundstage. This plug-in differs from most software reverbs in that it allows you to create a “room” and adjust its dimensions, then place musical sources and microphones at any spot in that room.
It’s possible to more or less recreate a room like the one I found at the church, non-parallel, with lots of diffusion. I’d place the virtual mics similarly, ORTF, facing away from the musician, and I would generate a pair of “room sound” tracks to be mixed with the direct ones. (I am presuming here that you can handle more extra tracks in a computer mixdown setup than are available on, say, an ADAT recorder or the equivalent. Obviously, your mileage may vary.)
Now let’s consider some other instruments.
The djembe can be thought of as the dumbek’s big brother. I suspect that what worked for the dumbek would also work for the djembe, with one possible caution: the djembe develops some hellacious bass, and placing a microphone with strong proximity effect inside might well give you too much of a good thing. I’d be ready to substitute another RE15, or perhaps an RE20, for the M88. These mics are especially designed for minimal proximity effect but still can give good, clean compression. Some of the mics that are especially intended for use with kick drum might also work well. Experiment.
Congas present a similar problem to the djembe, and I’d suggest handling them the same way. The hard part is that you’ll need extra mics and preamp channels, as congas tend to come in twos and threes. Although I usually prefer to work with separate mic preamps, connected directly to the recorder, this is one place where you might opt to use a mixer, panning all of the top mics to one channel, all the bottoms to another.
Walking the dogma
A different problem is posed by the bodhran, a small, open-framed Irish drum. Often the problem with the bodhran is too much boom, rather than not enough. The bodhran is often played at a fast tempo, and if not miked properly it can woof out to the point that it muddies up an entire song.
For years I recorded the bodhran with my good ol’ RE15 or RE16, using an X/Y crossed pair when I could afford the tracks (giving the instrument some three-dimensionality). The RE15’s lack of proximity effect kept the boom more or less under control, and the slight compression inherent in dynamic mics helped tame the dynamic range. It’s still a good combination; the RE20 would also be good here.
Recently, though, I’ve taken a new tack. I bought a pair of Electro- Voice RE200s—it’s a small condenser microphone with a sharp peak up top and a thorough bass rolloff. It breaks most of my rules for recording percussion—but I love it on the bodhran. The peak brings out the “skin” sound of the drum, giving definition to the beats, and the bass rolloff counteracts the mic’s proximity effect, yielding a frequency response that sounds natural without being overbearingly heavy. I’m not getting compression any more—the RE200 is more linear than its dynamic cousins—but it doesn’t feel like I need it, the drum sits well in the mix, holding its own without going muddy.
I guess all of this goes to show that rules are meant to be broken, or at least bent, and that it’s a bad idea to be too dogmatic about any one technique.
Tablas and timbales
The tablas are an exceptionally beautiful-sounding pair of drums (the large bayan and smaller dayan), carefully tuned to particular notes. They have a lot of bottom end which can overwhelm if badly recorded.
My friend and colleague Brad Sarno plays tablas, and he clued me in to a mic that does wonders for them: the Sennheiser MD 109. Place it about 2" in front of the drums’ rims, and about 6" above the heads; aim smack between the instruments, then adjust the balance by moving the drums (or microphone) back and forth. And prepare for nirvana: the sound is sweet and tuneful.
Another place where the MD 109 shines is on timbales. Here I have to make a confession—I’ve never actually had the occasion to record timbales. There simply aren’t many of them in St. Louis, salsa isn’t that common here. Brad has done it, though, and he reports that a smooth, sweet mic like the MD 109 tames the timbales as well as he’s heard. Be prepared for them to bleed into everything else, though—and angle the mic away from the cowbell up top.
Things that go “tink” in the night
Percussion instruments that jingle can be nightmares. For many years engineers have used a quick-and-dirty test of audio gear: they jingle a ring of keys in front of a microphone, then listen to the recorded result. It separates the sheep from the goats most effectively.
The fashionable upper-frequency boost that can make vocals cut through a mix often creates nightmares when recording percussion. As I said before, the jingles excite capsule and body resonances in precisely the frequency ranges that interact most unpleasantly with many analog-to-digital converters, not to mention the electronics in all too many mic preamps.
Take the tambourine, for example. It’s actually a drum, with jingles hung on the side, but listen to many recordings and you’ll hear nothing but jingle, no beat. And the jingles will sound terrible, almost distorted, a wash of “fffff” sound that has no definition, no metallic quality; you might as well use gated white noise and be done with it.
What are your options? First, if the performer will let you, tame the instrument—gaffer tape is your friend. Most tambourines will record much better if you tape about half of the jingles so they don’t ring.
That done, you face another problem: most tambourine players are kinesthetic spirits, prone to waving the thing around in front of their faces, down by their waists, etc.
Since handcuffs are probably not an option, your interpersonal skills will get a workout. Anything you can do to keep that bloody tambourine within a reasonably confined area will pay off—it sometimes helps to put the player on a high stool, for example.
Even so, you’ll probably have to mike the tambourine from farther away than you’d like, and use a mic with a wide cardioid pattern. A crossed pair of cardioids or hypercardioids mixed down to mono will simulate that nicely. If at all possible you should record the tambourine as an overdub or isolated from the other musicians—those jingles leak everywhere.
I mostly like my usual suspects, dynamic mics with un-hyped high ends, like the RE15s. But a recent experience in another part of the forest has me wondering...
I recorded a set of finger cymbals as part of the same Sephardic project mentioned before. I normally mike these with Beyer M260 ribbon mics, their superb transient response and lack of ringing complement the jingle beautifully. This time it wasn’t happening, the ribbons didn’t cut it, and the RE15s sounded dull and dead. In desperation I pulled the last mic out of my suitcase, an Alesis/Groove Tubes AM51. This is a large-diaphragm cardioid-only solid-state microphone sourced from China, but with better-than-average electronics. I placed it about 18" from the musician, centered on the area she covered with her hands (finger cymbal players are as mobile as tambourine players, if not more so—many belly dancers use them to keep time).
I didn’t really expect good results. While the AM51 has an unusually uncolored midrange, it’s a bit sharp on top, and that usually spells trouble with jingly things. Well, I was dead wrong. The AM51 gave me a most natural, realistic rendition of those pesky finger cymbals. They were sharp and clear, the ring came through without losing the clash of their impact (or the impact of their clash), and the wide-ish pattern let the player move around a bit.
It didn’t hurt, of course, that I was using a good mic preamp, the tubed home-brew I mentioned earlier, without a hint of edginess on top. Lesser preamps might have had a problem. They also might have created more problems down the line. Distortion is a cumulative thing, and any high-frequency intermodulation distortion (“crud”) will be intensified by the analog-to-digital converters, particularly lower-grade ones such as those found on older 16-bit recorders and low-end sound cards. I hate to say it, but sometimes there’s no substitute for using better gear.
All this leaves me wondering how that large-diaphragm AM51 would do on a tambourine—worth a try, anyway. Some of the less-expensive large-diaphragm mics of Chinese origin might sound more raunchy on jingles (they often have a nasty intermodulation in the top couple of octaves), but I suspect installing Scott Dorsey’s modification board from our January 2002 issue would help a lot. (Okay, Scott, you can send the check now.)
Triangles present similar problems as finger cymbals and tambourines, but worse. It’s far too easy to get nothing but a “ting” and none of the “clank” that provides the rhythmic pulse, especially in Cajun or Zydeco music where the triangle plays a crucial role. Here I would stick with dynamic microphones most of the time.
Rattle and plock
I mentioned before that I haven’t done many salsa bands, but I’ve had experience with some of the smaller instruments they use—the claves (small sticks of hardwood that make a satisfying “plock” when struck together), the guiro (an instrument that can best be compared in appearance to a serrated eggplant, scraped with a stick to make a rasping noise), gourds (rattles) and bongos.
All of them can be lumped into a single category: “Instruments That Really Need Flat Microphones.” The basic sound of these small guys is down in the midrange, not up top, and mics with ringy high ends will make them sound thin and awful. Forgive me for sounding like a broken record—err, a sample loop—but these instruments sound best with something like an RE15 or a Beyer ribbon, although I once used a Shure SM81 on a guiro and gourds with surprisingly good results. (Engage the 10 dB pad!)
Keep a level head
Speaking of that pad brings me to the subject of levels. I mentioned earlier that many percussion instruments have huge peak-to-average levels, which means that something which sounds to your ear like 85 dB may actually be peaking out at 105 dB. They also have waveforms with very fast attack times—well, what did you expect? You’re hitting it.
This makes percussion instruments difficult to record. It also makes them hard to meter. Most mechanical meters (VU-type) are close to useless, they simply don’t respond quickly enough to tell you anything about the real level you’re putting out of a mixer or recording onto a tape. Analog recording, with its soft overload, is pretty forgiving under these circumstances (in fact, a bit of analog squash can sound downright nice); digital recording is not.
Therefore, if you’re recording digitally (as most of you are), you need to keep an eagle eye on levels. Most digital tape recorders are good about this; the LED meters on DATs, ADATs and DTRS recorders read the actual digital level as recorded on tape. These meters respond quickly enough to give an honest picture of what’s going on, even with difficult instruments like claves.
Computer recording systems are a different story. Theoretically they should work just as well as their tape-based cousins, since they meter the actual digital signal. Unfortunately, metering takes up CPU resources, and often a multitrack program that’s running out of CPU power will opt to concentrate on the recording itself (good) by starving the metering (bad). I’ve seen meter bars lag behind the signal by as much as ten seconds, and often short peaks will be passed over completely.
Different programs vary greatly in how well they meter, and, of course, the more resources your computer has to begin with, the less difficulty you’ll have. I mentioned in my review of the M Audio Delta 66 card (Recording, March 2002) that the meters in the card’s own monitor mixer responded with far more alacrity than those in my multitrack program (Fasoft n-Track Studio). In my computer, at least, the program’s own meters were effectively useless when I was recording four or more tracks at a time, while the Delta’s showed me everything.
There’s no simple solution to this problem except, perhaps, faster computers. On occasion I have hooked up a DAT recorder to my soundcard’s S/PDIF output and used its very fast LEDs to keep an eye on levels. I’ve even resorted to the utterly dorky method of making a test recording, then going back and looking at its waveform to make sure it didn’t overload. (Well, it works. Just allow an extra 3 dB for adrenaline when you’re rolling.)
Speaking of levels, you could use a compressor on the way into the recorder—many recordists do, of course. I’d rather not. While on many percussion instruments some compression is necessary, I prefer to do it when mixing. Good 24-bit digital recorders can hold the entire dynamic range of even the toughest percussion, and it’s very hard to guess how the compressor should be set until you hear it in the mix. That’s a personal preference, of course.
A world of possibilities
I’m glad Lorenz asked me to write this article, because it lets me talk about instruments from all kinds of places. Listen, for too long we’ve been tied to a small set of instrumental flavors. This last decade’s explosion in “world music” has opened up our ears to a far broader audible menu.
Percussion, of course, is a huge part of that—maybe even the center. It’s been suggested that the first musical instrument outside of the human voice came about by someone idly knocking a couple of sticks together and noticing that they made a neat sound. People got up and danced. They still do.
Enjoy! And pass the satay chicken, please.
Paul J. Stamler (talkback@recording mag.com) records just about every type of music you can think of—except maybe salsa—in the St. Louis area.
Dumbek, djembe, and tabla photos courtesy Mid-East Mfg., Inc., on the web at www.mid-east.com.