Washington Trip
[ Home ] [ Up ]

Through some amazing quirk of fate, yesterday I had the immense good fortune of attending 10 hours of audio-engineering seminars at National Public Radio headquarters in Washington. Then there was the 3-hour train ride to get there and a 4-hour return trip, which explains why I feel hung over today even though I haven't touched a drop. I was the only independent producer there, the only non-engineer. Yet they were kind enough to let me join right in and even play with the microphones, slide the faders and twist the pots.

But I gotta say, one of the best moments was when I mentioned to a guy from one of the NPR affiliates that I had a Roland VS-1680. Immediately a small circle formed around me, with questions coming faster than I could answer them. These are people who use Protools and consoles the size of football fields in the day-to-day drudgery of their jobs; I had assumed they would snicker. They did not, or if anybody did I didn't notice because everybody was talking at once. It seemed most of them had heard good things about the 1680, not-so-good things about the 840 (don't take it personally, ANAYV -- I'm sure they haven't heard about the recent upgrade) and not much at all about the 880. Nobody mentioned Paris, either, come to think of it, except a guy from Pennsylvania who wants to find an engineer there to broadcast live symphony concerts via ISDN. But I digress.

I learned so much about so many of the issues that concern us and I'm going to do my best to pass it along for your consideration, with this disclaimer: I am not an engineer (exhibit A: SOTBL 4,4), and while the architects of the NPR sound were doing their best to impart their knowledge, I may not have retained it comprehensively or correctly.

Also, this is a lot of information, so I apologize in advance to anyone who feels it's off-topic, since you've already read the only mention here of the VS family of fine products. I did know some of this stuff already, and I expect many of you already know much more, so I hope you won't be bored. I'll break it into a series of posts to make it more manageable.

First session: Digital Recording Bootcamp, taught by Ken Pohlmann, author of "Principles of Digital Audio" (Third Edition, McGraw Hill, $44.95).

Thomas Edison invented the phonograph when, in a moment of frustration, he grabbed a piece of paper that had gotten caught up in some other invention of his and ripped it out. When he did, it made a noise, and that's what gave him the idea of the phonograph. Or so the legend goes.

Mr. Pohlmann then told us the story of Harry Nyquist, the Bell Labs engineer who developed the eponymous Nyquist Theorum. Mr. Nyquist discovered that we don't have to preserve the entire waveform of a sound. Instead, we can sample it, discard the rest of the information, and reconstruct the waveform. Nr. Nyquist found that you need a sampling rate at least twice the highest audio frequency. Since the human ear (ideally) has a range of 0-24 khz, a sampling rate of 48 would make the sampled information indiscernible from the original wave. If you're young, live in the quiet countryside, and female, you *might* actually have a hearing range of 0-24 khz, but if you're an aging male city dweller, you probably can't hear as well, maybe 0-20 khz. So why is the DVD sampling rate 192khz, for a frequency response of 96kh? Because there are things we don't actually hear that we may hear anyway, without realizing that we hear them, the sonic equivalent of the physicists' "dark matter" perhaps.

The biggest drawback with digital sampling is that the samples themselves aren't preserved. All that's left is a number that represents the amplitude. Digital sampling has a finite number of increments, while with analog the increments are infinite. So if you've got a wave amplitude that falls somewhere between the increments available, it gets rounded off to the nearest hash-mark on the ruler, so to speak. The fewer the bits, the bigger the space between the hash-marks, and the more problems you have with quantizing errors.

There is a formula for determining the signal-to-noise ratio for digital audio:

6.02n+1.76 dB

n= # of bits

So at 16 bits, the signal-to-noise ratio is about 98dB. With analog, about the best s/n ratio you'll get is 65, "on a good day," says Mr. Pohlmann.

But when CDs first came out in the mid-80s, the engineer-types were surprised when a bunch of musical-types insisted that CDs sounded harsh. Some of you fellow geezers may recall the "green felt pen" cure for that, which really made no difference except to the purveyors of green felt pens. The engineers were missing an important part of the puzzle. At full volume, a sound would get sampled with all 16 bits, but if the sound is at a lower amplitude, its sample would consist of fewer bits. For the softest sounds, only two or three bits were available. And the more subtle the sound, the more quantizing errors there were. That created some really ugly distortion. (BTW, did you know that on a phase scope, digital distortion shows up as a square? It's an ominous thing to hear, and see. After watching the scopes at NPR all day yesterday, I can't wait to get one. They say it'll be around $400.)

In 1984, Stanley Lipshitz and John Vanderkooy saved the day, sorta, by inventing dithering. At first, engineers were aghast. Get rid of noise by introducing noise? Isn't that kind of like using a vaccine to protect against disease, or fighting fire with fire? But the "noise" of dithering allows the encoding of amplitudes smaller than the least significant bit, magically eliminating the noise. Dither doesn't cover up distortion, it eliminates it, although it does raise the noise floor slightly. During the session, Mr. Pohlmann played a variety of examples on the studio speakers, speakers almost as big as our apartment. Some had an average of two bits, but with dithering they sounded pretty good. This bodes well for sound on the Internet, considering the bandwidth limitations.

We then learned about aliasing, another form of digital distortion. Aliasing creates redundant images of the samples, which makes the sound as bad as one of those movies where one actor plays several characters. The cure for that is oversampling, in which the CD player creates and interpolates new samples within the existing samples. The first-generation Phillips CD player used 4x oversampling, which seems to be just about right, at least until science advances further and we discover that is should be 16x or something.

Then there's jitter, a variation of the time axis caused by the mechanics of the machinery, just as the rotation of the old record player distorted the sound of the record. The cure for that is phase lock looping, which I'm glad I don't have to worry about.

The successors to Nyquist, Lipshitz and Vanderkooy are now working on creating a "software ear" that will emulate the real ear as much as possible, thus reducing the size of the pipeline needed to convey the sound information. They're also finding ways of masking noise and shifting it out of the audible range, thus moving what quantizing errors and other problems remain into a place where they can't bother anybody. Another approach is to make the bit rate variable, as it is on DVD -- the highest quality audio we have right now.

We also learned a little about the future of stereo (5.1 surround sound will make stereo seem as primitive as Edison's first phonograph) Then there was some stuff on digital and HDTV in its 18 different FCC-approved forms, and the fact that analog TV broadcasts will cease in 2006, but I can't see what that has to do with the VS.

That was the first session, which was followed by a six-hour intensive "Intermediate Recording Workshop," in which we recorded a salsa band and a string quartet -- not at the same time -- with more microphones than I've ever seen in one place, some of which aren't even made anymore. That began with a strip show by the instructor, who showed up in a suit and then announced that the first rule was, "Don't dress like this." He peeled off layer after layer until he was down to a comfortable T-shirt and jeans, and then we all got to work setting up for the salsa band. I got to mic the drums, which makes me want to move to a bigger apartment, or better yet, a house in the country, so we can have some of our own. We also learned what engineers should eat to keep their hearing at optimum levels, and a gazillion other handy hints. Stay tooned for the next installment, in which we play with coincident pairs, Blumleins, and spaced omnis. We recorded 21 tracks with a Yamaha digital mixer onto DAT.

It was great. After the strip show, our instructor, hereinafter referred to as Mark Greenhouse, gestured toward the enormous console in the studio and announced that we wouldn't be using it because "you won't find that out in the real world" and besides, the smaller digital workstations are the future of audio. On the subject of digital audio workstations in general, he said, "They have given us the keys to the car and told us to go have some fun. Power to the people!"

And BTW, NPR now has a (legitimate) copy of "Songs of the Big List(s)" for their edification and amusement. Maybe they'll listen, maybe they'll use the discs as coasters. Stay tooned.

Janet Dagley Dagley

one-half of the Bohemian Hillbillies

NY, NY

SOTBL 4, 4

Chapter Two: Setting up for a Live-to-DAT Remote Recording

This is the second installment in the saga, "Mrs. Dagley Dagley Goes to Washington," notes from the NPR training seminar I attended on May 12. My apologies to those who might consider this off-topic, since there is nothing about the Roland VS series in this post. The emphasis here is on recording live performances, understandable since our instructors were the engineers for NPR's Performance Today. I realize that most of us are not using our VSes in remote recording situations, but some are.

As you recall from our last episode, our instructor, NPR engineer Mark Greenhouse, had just stripped down to his working attire: a black T-shirt and jeans. He recommends always dressing in black, not for fashion-statement purposes but to minimize your presence -- there's a good reason it's standard theater attire. No matter how unobtrusive you might try to be, you are a complication to the music, a disruption, the enemy of spontaneity. This reminded me of the Heisenberg Uncertainty Principle in physics (and journalism), which is that the observer affects what is being observed.

The band Bio Ritmo was kind enough to let us experiment on them. They set up their equipment in NPR's legendary Studio 4A, which is about four times as big as our entire apartment (hence the name). Separated from 4A by a double glass wall is 4C. Picture a Star Trek set, with a few rows of seats set up behind the captain's swivel chair for visitors who happen to be aboard for a seminar. Instead of the big TV screen, the crew monitors the mesmerizing, dancing image of the sound on a phase scope. They cost about $400, and I know we don't really need one but I've got a bad case of GAS for one all the same. As I mentioned in the first chapter, instead of using the enormous console in the studio, we worked with two Yamaha 01V 24-track digital mixers, about $1500 each, synched together.

http://www.yamaha.com/cgi-win/webcgi.exe/DsplyModel/?gRMC0000801V

Instead of the refrigerator-sized monitors built into the studio walls, we used Event 20/20 bas monitors, and they did sound pretty good. Greenhouse says you'd have to spend thousands to find anything close; he really likes them. We mixed down about 30 tracks to stereo DAT, recording on two DATs simultaneously for safety's sake, with a third DAT standing by in case we ran out of tape on the other two. The idea was to give us a setup similar to one we might find out in the real world, rather than the wonderland of a network radio studio.

BTW, many radio stations, WNYC for example, sample only at 32 khz. Anybody know why? The first person to answer correctly may already be the winner of a copy of Dave Freden's "Lesser Humans" CD, but you'd really have to talk to him about that.

Another important consideration for radio engineers is mono. No matter how good it sounds to them in the studio or to listeners with good stereo receivers, they have to remember that some folks are tuning in on little clock-radios. Otherwise mild-mannered little old ladies call up and complain when the vocalist vanishes in the middle of an aria due to wave-phase cancellation.

Shortly after we assembled, Mr. Greenhouse asked if any of us were from Tennessee. "Any Volunteers?" he asked, and of course I proudly admitted my citizenship. The next thing I knew I was assigned to help mic the drums.

More on that in a bit. First, let me throw in some assorted handy hints:

Your hearing is affected by many things, including what you eat and drink. You can hear better if you're hydrated, so drink a lot of water. Don't eat anything heavy -- you'll hear better if you eat fruit. Don't eat anything greasy beforehand, because even if you wash your hands there will still be some residue, and it will get on your equipment. Greenhouse suggests packing a bottle of water and four oranges for each member of your crew, as well as a pillow and a sweater. Fatigue also affects hearing, and recording engineers tend to work long days, so he recommends taking a nap if you get the chance -- even a short one will make a big difference -- after the soundcheck and before the show. For those of us recording at home, that may not be such a big issue, but it's something to remember if you find yourself frustrated during a long session. Take a break, even a short one. And if you're working in a big-time professional situation with union members, you'll have no choice but to take regular breaks. All the engineers at the seminar, students and teachers alike, agreed that if you find yourself getting short-tempered and impatient with the people around you -- or they with you -- it's time for a break. Greenhouse recommended taking a cheap set of walkie-talkies to every gig, just in case nothing else is set up, so that the crew doesn't end up yelling at each other just to be heard. Even if you're not angry, shouting tends to make it seem that way, and soon everybody starts acting that way without understanding why. Michael and I learned that lesson when we set up our isolation booth. Before we installed a cue mic so that the "engineer" could speak in a normal tone to the "talent" in the booth, the "engineer" would have to shout, which caused the "talent" to shout in return even though the "engineer" could hear perfectly well through the microphone in the booth. That led to frayed nerves and unfortunate misunderstandings.

When you're deciding which input should go to which track, Greenhouse recommends starting from the middle rather than with track 1 so that you're working from the center of the console rather than from the side. That's not much of an issue with consoles the size of the VS, but it's a good habit in any case. If you're using two VSes synched together, think of them as one large console and follow the same principle. Someday when we're all using 128-track VS 16,800EXs, it may come in right handy. And use adjacent tracks for the sake of efficiency, since you've only got 10 fingers and may have to move more than 10 faders simultaneously.

It's also a good idea to attach a cassette recorder to your recording chain so that you can make a courtesy copy for the band. You can hand it to them immediately after the performance, so they have some idea of how they sounded while they wait for you to work you magic on the master. If you're making a cassette copy, Greenhouse says to use only TDK SA90 tapes, and NEVER,EVER USE DOLBY. Write "NON-DOLBY" "turn off Dolby," and similar instructions all over the cassette. Dolby has its uses, but if your Dolby isn't the same as their Dolby, it could sound really awful, so it's best not to take the chance.

Bring along a CD player of some sort and a CD that you know is good. This will help you get your bearings when you find yourself lost in the sound. Take a break, stop trying to get all the levels and positions just right, and listen to the CD for a minute or two. Then get back to work.

If you're working in a big-time, real-world situation, you might find yourself working with union crews, which means you'll have strict rules to follow regarding starting and stopping times, breaks, and what equipment you're allowed to touch. Follow these rules carefully. But even if you're just recording your child's school play, it's a good idea to clear things in advance with the people who are running the event, and "in advance" doesn't mean 20 minutes before the show. If you can, go to the site and look it over well before the day of the show. If you can't do that, ask for a floorplan of the theater, as well as a copy of the stage plot and specifications that the artist provided. Work out as much as possible in advance, including such issues as where your electricity will come from (ask nicely for permission, don't assume), who sets up the microphones, who provides phantom power to what. Find out what time you should arrive, where you should set up, when the soundcheck is, what time the show starts, and what time the crew quits. Pack your gear in two separate sets: one for the stage, the other for the control room. If you have an uninterruptible power supply, bring it and use it. Otherwise, if you're plugged into the house power supply, you may find your equipment rebooting just as the performance begins because your power got cut off just as the lights went down.

If you're lucky, the room will provide its own natural reverb, an effect better than any you could create artificially. You probably won't be that lucky -- the room may well have too much reverb. If so, use cardioid mics, placed as close as possible to the source, to get as dry a recording as possible -- you can always add reverb to build a space around the music. The NPR engineers played numerous examples of good and bad recordings, including one in which the mic was too close and too far away at the same time. Directly in front of the cello, the mic got too much of that and not enough of the rest of the instruments.

I'll go into more detail on mic placement in my next installment, featuring highlights of the "stereo mic techniques" session.

But in general, try to avoid the two things that are almost impossible to fix: too much distance between the mic and source, and too much reverb, natural or otherwise.

When setting up mics, be sure to tighten everything as securely as possible. Our instructor followed us around as we worked, making sure he heard some sort of noise (grunting, groaning, sighing, foul language) come out of us as we tightened the boom stands, etc. "If you don't make a noise, that means it needs to be tighter. It's an awful feeling to be broadcasting a live concert and see the boom mics slowly, slowly creeping down onto the performers."

Bio Ritmo is an 8-piece band, with drums, timbales, congas, bongos, bass, keyboard, sax, trumpet, and trombone. All the performers sing as well. Unfortunately, my notes aren't good enough for me to tell you which mic we used for which purpose, except that all the vocal mics were Shure 58s or 57s. We used a Neumann TLM170 on the trombone, an Earthworks Z30X on the trumpet, and AKG 414s as overheads above the drummer, along with a tiny Neumann KM84 on the snare, very close and nearly parallel to the drum head. We used direct inputs for the keyboards and bass, with a mic on the bass amp as well. The direct-in alone made the bass sound too electronic, the miked amp sounded kind of distorted and fuzzy, but together they made a nice, warm, well-rounded bass sound.

Once you've set up all mics and done a "scratch and sniff" to make sure they're all hooked up right, it's time to set individual levels. Always start with the rhythm section: between the bass drum and the cymbals, you've got the whole frequency range. Set levels for individual tracks at around -18dB, to give yourself some room as you add them all together. It's also a good idea to work with subgroups if you're dealing with a large band.

Use EQ only as necessary. "EQ is not your friend. It is a surgical tool, and should be used only as necessary. If you have to use it to fix something that's broken, then go ahead, but don't use it if you don't have to," our instructor said.

Compression, however, can be your friend, especially when you're among strangers. If you don't know the band and its music well enough to anticipate peaks, protect yourself with compression, starting at around a 4:1 ratio, and adjust to taste from there. And don't forget to set the pans so that the band sounds comfortably spread out across your listening space, leaving space in the middle for the vocals.

That's more than enough for now. Stay tooned for our next exciting installment, Stereo Microphone Techniques.

Janet Dagley Dagley

1/2 of the Bohemian Hillbillies

NY, NY

SOTBL 4, 4

Mrs. DD Goes 2 DC

Chapter 3

It's an Off-Axis World: Stereo Microphone Techniques

(This is the third in a series of reports on the audio-engineering seminars I attended May 12 at NPR in Washington. I am not an engineer or an expert on this stuff, but I was fortunate enough to spend the day with some of the best in radio, and I did take copious notes, most of which I can still decipher.)

First, a couple of things I should have included in Chapter 2. When setting up a boom stand with a tripod base, be sure to line up the boom directly over one of the legs, so it will be less likely to fall over. And when you get two booms set up for your overhead mics, pan one hard left and the other hard right. Pan all your other mics somewhere in between, with vocals in the center. And when you're recording horn players, ask them *not* to point the bell of the instrument straight at the microphone. You might need to say that more than once. When micing an amp, place the microphone off-center and aim at the voice coil dust cover. And when placing a pop screen between a vocalist and a mic, place it slightly off-axis, in hopes that the vocalist will orient him/herself to the screen rather than the mic. "Let's face it," the NPR engineers told us. "It's an off-axis world."

For our second experiment, we recorded a string quartet to learn about stereo microphone techniques. We set up a coincident pair, a spaced pair of omnidirectionals, a near-coincident pair, an ORTF configuration, an MS array, and used a giant, ancient Neumann stereo mic (a USM 69, I think) which had a built-in Blumlein configuration.

Our textbook, which we didn't have time to read before the class, was "Stereo Microphone Techniques" by Bruce Bartlett, ISBN 0-240-80076-1, Focal Press, 1991, $42.95. I recommend it, and hope you can find it at a discount somewhere.

The first stereo mic experiment was demonstrated in 1881 at the International Exhibition of Electricity in Paris, where a performance of the Paris Opera was picked up by several spaced pairs and transmitted across town to listeners with binaural earphones. Stereo was first used commercially by Walt Disney Productions in 1940 for the soundtrack of that acid-droppers' favorite, "Fantasia." After that, nobody tried it again until the 1950s.

Stereo images are created by differences in time arrival -- the closer sound arrives first -- and intensity -- the closer sound is likely to be louder, and the louder sound wins.

With the coincident pair, the difference is in intensity only, since both mics are the same distance from the source. Align a pair of directional mics so that they are nearly touching, with their diaphragms one above the other, angled apart, approximately in the direction of the left and right sides of the musical ensemble. Cardioid mics are often used, but other directional mics will work as well. The greater the angle between microphones, and the narrower the polar pattern, the wider the stereo spread.

The coincident pair has lots of advantages, including the fact that the angular accuracy of the stereo image is unaffected by the distance of the mics from the source. And it has disadvantages -- since the stereo image is determined by intensity only, its sense of space may not be as clearly defined. It also may be narrower than the spread available on your speakers.

When Michael and I experimented with the coincident pair technique the other night here in the privacy of our home, we discovered another disadvantage. Be sure to take your headphones off, turn them off, get them away from your ears when making last-minute adjustments to a coincident pair. The sound of those two mics bumping into each other can be quite a jolt to your ears.

The Blumlein array is two bidirectional mics angled 90 degrees apart and facing the left and right sides of the ensemble, so that you get a cluster of two figure eights. It's named after British researcher Alan Blumlein, one of those brave pioneers who patented stereo reproduction for disks back in the 1930s. It's also known as 'stereosonic." Blumlein is particularly good recording stereo that will hold up under mono playback, so that the clock-radio (or mono TV) listeners get their money's worth. It works well in churches or other places with a lot of reverb. Its disadvantages are some funky phase problems, and the sense of a hole in the middle. We really weren't able to accurately judge the Blumlein. It didn't sound all that good to us, but the NPR engineers seemed to think the problem was that the microphone itself was giving up the ghost.

The most popular stereo mic array is the Mid-Side (MS) technique, a variation on the coincident pair. It works like this: One directional mic is aimed at the middle, with a bidirectional mic pointed toward the side, so that:

M+S=L

M-S=R

(M+S) -(M-S)=2M

With the MS array, you can control the stereo spread by varying the ratio of the mid signal to the side signal. You need an MS matrix box for that. OR, with a mixer, you can pan the M signal to the center, then split the S and send it to two inputs, reversing the phase of one input. Pan the out-of-phase S signals hard left and right.

The spaced pair is fairly straightforward: just take two identical mics, usually but not necessarily omnidirectionals, place them several feet apart and pointed directly forward at the band. With this array, you get a stereo effect not only from differences in intensity, but also in time arrival. You can also get phase problems. One cure for that is to use a third mic, aimed slightly outward, with reversed phase. This array can leave the source of the sound diffused, so that you can't quite place it in your listening space. But it has a nice warm ambience.

If you're using 3 omnis, it helps to bring the middle one down 3-6 dB.

OK, who knows what ORTF stands for? The winner will have the zen satisfaction not only of knowing the answer, but of not receiving the same Dave Freden CD ("Lesser Humans") that Mike Jones didn't get in our last round, not to mention the pleasure of hearing one hand clapping through that virtual Avalon VT 737sp.

ORTF is the most commonly used near-coincident arrays. ORTF places two directional microphones angled apart about 110 degrees, with their grilles spaced a few inches apart, crossed like an asymmetrical x. The stereo effect is produced by both intensity and time-arrival differences. Mids and lows tend to disappear with ORTF, but it sounds very good also.

We took turns listening to the different arrays, went back into the studio and adjusted the mics and listened some more. We were unanimous in preferring the sound of the spaced omnis. I never knew they made so many different kinds of Neumanns! We were also in agreement that the Neumann Blumlein sounded not-quite-right somehow. But what did we know? A few hours earlier, we were unanimously wrong: Our teachers played us a sample with some scratchy distortion, and we all yelled, "Digital clipping!" Uh-uh. It was *analog* clipping. Then they gave us an example of digital clipping, an ominous sound indeed. More about that in our next exciting chapter, The Phase Scope.

Janet Dagley Dagley

1/2 of the Bohemian Hillbillies

NY, NY

SOTBL 4, 4

Janet Dagley Dagley can be reached at

daeggadaegga@earthlink.net