Showing posts with label mixing. Show all posts
Showing posts with label mixing. Show all posts

Saturday, December 31, 2011

Hans Zimmer & Co’s DAW methods are far from ‘elementary’

AURO-3D
I ’ve continued exploring the Hans Zimmer ‘Sherlock Holmes’ soundtrack described in the previous post. There are a variety of spatial effects—things that add considerable dramatic excitement to the score—that can be “deduced” from exploring the recorded cues with tools like Cubase® or Nuance®. Some deductions are, at best, “educated guesses” based on inspection of the digitized waveforms.

F or example, you can get instruments to sound like they are coming from outside the stereo image by inverting the relative phase angle of a channel. Start with a single channel (say, an explosion in this Sherlock Holmes film). Pan it hard-left. Duplicate the channel, and pan it hard-right. Then invert the phase angle (180 degrees). Now start with the volume on the duplicate at zero, and bring up the level on the duplicate/phase-inverted channel. You now hear the explosion on the left move out wider than it was when panned hard left. Sounds like a paradoxically larger space than what appeared to your eyes.

AES surround standards
A dmittedly, some of these “film” spatial sonic effects might only be implemented in live chamber music performance by the likes of Maya Beizer or Tod Machover, or other devoted electroacoustic artists. But the prospect of producing (and hearing) live ‘augmented reality’ through DAW processing and sound reinforcement becomes more appealing/exciting the more we understand how it works, even for artists who are today squarely in the analog/acoustic world.

Vier, p. 297

Saturday, March 29, 2008

SR: Chamber Musicians Plugged and Unplugged

 Ahnert & Steffen book, Berlin concerthall
O  ur string quartet is planning to acquire some sound reinforcement (SR) gear, to enable us to better control the sound levels and acoustics in some of the venues where we perform. I feel like the sales guys are genuinely trying to help us, but the fact that we’re classical musicians leads them to try to recommend all sorts of exotic things that they think would please us, like really low total harmonic distortion (THD) specs and so on. Invariably, these seem to us to be over-spec’ed and over-priced. We’re a classical string quartet and, yes, of course the SR requirements are different from what they’d sell to a rock band or to a symphony. But this is after all just a commercial sound system, right? Aren’t there engineering equations that determine what’s really necessary? I feel like the fact that we are acoustic musicians and classically trained and somewhat finicky ... brings out the worst ad hocness in the salesmen’s nature.”
  —  Anonymous.
I  wrap a piece of masking tape around the top three strings (most violinists don’t seem to mind) and clip the mic with a windscreen on the high side between the bridge and tailpiece to that tape.”
  —  Brian Frost, 15-MAR-2007, SR Forum.
W  here precisely do you wrap that tape? I have one young charge who is the keeper of an incalculably expensive Guarneri. I may not touch it.”
  —  Tony Tissot, 15-MAR-2007, SR Forum.
I  f you have four condenser mikes, just position one each on a tall boom above each violin and the viola and another in front of the cello on a short stand with a boom. If the quartet has decent instruments you do not need to have the mikes right on top of the instruments. A distance instrument-to-mike of 25 to 50 cm usually works for me. If you don’t have four condensers, use a dynamic on the cello. If you don’t have three but do have two, try one over the mid-point between the two violins, and one on the viola. "Open up the gain" on the violin pair 3 to 6 dB more than the viola mic.”
  —  Lee Brenkman, 24-FEB-2007, SR Forum.
It does seem that there’s a shortage of reliable guidance for sound engineering for small classical ensembles. Yes, there’s plenty of information on SR for symphonies, and there are quite a few acoustics engineering and architect firms out there that take on those large and well-paying civic projects. But there are not many resources that have detailed information or services for string quartets or other small groups. Some links provided below may be of interest, though.

 Ahnert & Steffen book, Fig. 3.14, required/available peak SPL, as a function of ratio of room volume to reverb time; “symphony” is approx. 10 W curve [conductor’s podium], “string quartet” is approx. 0.1 W curve
It is true that commercial sound work for a full symphony orchestra in an air-conditioned hall imposes fewer demands on an SR system than a commercial sound system for a small ensemble. After all, the peak sound pressure levels (SPL) for an orchestral fortissimo at the conductor’s podium are equivalent to 10 W of power or higher. A single violin’s peak SPL is less than 1/100th of that.

 Carnegie Hall, Stern Auditorium, Yamaha PM1D mixer
Let’s assume that your SR loudspeakers can deliver a level Lp = 110 dB at the ears of listener who is at the furthest distance from the stage—equivalent to the fortissimo for a full orchestra at the conductor’s podium. If the diffuse ambient noise level from a typical concert hall’s air-conditioning system is 32 dB at such a listener’s seat, then Lpdiff = Lp2 – Lp1 = (110 – 32) = 78 dB. If you add 6 dB to avoid undesired addition of levels, you get 84 dB. So you calculate that the THD target for the hall and those peak SPLs is 100 x 10 ( -84 / 20 ) = 0.006%. You can confer with an acoustics engineer or a sound designer at an architect firm for a more expert and precise answer than the crude suggestion I’m making here, but you can imagine from the calculation above that you probably won’t be needing amplifier gear whose THD specs are better than about 0.01%.

 Sokol book
If the air-handling HVAC is turned off, though, and if the audience is really attentive and really quiet and if there’s zero street noise coming in through the concert hall walls and roof, then the diffuse ambient noise level will typically be closer to about 18 dB instead of 32 dB, in which case Lpdiff = 92 dB, or, with an extra margin of 6 dB, make that 98 dB. With that you get is 100 x 10 ( -98 / 20 ) = 0.001% THD, about an order of magnitude better than the figure we calculated before. So you naturally wonder whether a sound system could be that good, to take advantage of such a quiet space to provide those dynamics, with a sound so clean that it’s indistinguishable from unreinforced acoustics.

Part of the answer to this question is that you need to think about every element in the audio chain—from the mics, to mixer, to amplifiers, to loudspeakers. The loudspeakers that are available for practical SR systems are, at best, not better than about 1% THD. And, in general, carefully thought-out designs for contemporary commercial sound work (including symphonic venues) have THD around 0.8%. And a THD of 0.8% represents 20 log10 (0.8 / 100) = –42 dB in terms of signal-to-noise ratio (SNR).

For a harmonic to equal this amplitude, you’d have 42 + 32 = 74 dB with the air conditioning “on”, or 42 + 18 = 60 dB with the air conditioning “off”.

 Stark book, Fig 3.3, typical Sound Pressure Levels (SPL), in decibels
In other words, if you’re buying SR gear with fabulously better THD specs (lower than, say, 0.8%) then you wouldn’t actually be deriving any benefit from those specs unless (a) your venues are tremendously quiet and (b) the musical passages include delicate pianissimos below 70 dB SPL.

 Central Studios, Utrecht Incidentally, one of the venues where I’ve witnessed heavy-duty SR and mixing gear deployed pretty unsuccessfully was the Oude Muziek Festival in Utrecht in 2007. The Sampson-Bezuidenhout concert was held in the CentralStudios facility, in Studio-1. There were about 3,000 persons packed into that cavernous 1,000 m2 (volume = 10,000 m3) hall, which is about twice as deep as it is wide. The miking of the performers was scanty. The reverb from front-of-house to rear was far too large, and the delays and phasing of the reinforcement loudspeaker arrays were wrong. Much of the ensemble’s sound was never captured by the mics and what was captured was piped and amplified in such a way that confusing delayed reflections arrived at your ears hundreds of milliseconds after the reinforced sound. The vocal parts tolerated this somewhat better than Kristian Bezuidenhout’s fortepiano.

 Ahnert & Steffen book, Fig. 3.14, required loudspeaker power, as a function of level difference L - Lk in dB and distance from source to listener in m; design for ‘L – Lk = 0 dB’ curve
According to Ahnert’s and Frank’s book (p. 285) the required amp power in Watts is a function of the ratio of the volume of the room and the mean reverberation time:

           P = 0.10 x V / T,

where V is in m3 and T is in sec. (See also top graph, above, the one at the very top of this post.)

 Ahnert & Steffen book, Fig. 3.15, reverb required (in seconds, referenced to 500 Hz) by volume of room (in m3); “chamber music” is curve 3, curve 2 is “symphony”, curve 1 is “organ”; CentralStudios should’ve been SR-tuned to 2.0 sec according to this …
CentralStudios’ equipment has the requisite power. But on this occasion the power available was under-utilized. Possibly the sound-check with the empty hall was not subsequently amended when the hall filled with people.

 Central Studios, Sampson & Bezuidenhout concert, Oude Muziek Utrecht 2007
Even in Row 11 where I sat, the sound was weak—it was difficult to hear the direct sound from the performers on-stage, and it was difficult to make sense of the disparate SR and reflected sounds impinging on us from further to the rear.

 Central Studios, Utrecht
 Central Studios, Utrecht, Studio 2, setup for refreshments at interval


Thursday, March 27, 2008

Dave Moulton: Golden Ears and Just-in-Time Spectral Management for Chamber Musicians

 Dave Moulton
M  astering [engineering recordings] is an art, not a science. It involves a lot of technical craft and control, plus practice.”
  —  Dave Moulton.
Your plane is delayed by bad weather, and you land in late afternoon at the airport of the city where your performance tonight is to occur. The taxi whisks you to the performance hall where you greet the presenter and observe the production people busily getting the stage ready. You have less than 2 hours before the performance begins. You should be focusing, centering, warming up, getting your game-face on. You should be taking deep breaths.

But wait! The house has a much taller ceiling than you thought it would have. The acoustic reflective panels on-stage are adjustable, but at the moment they seem to be in a perverse flat configuration that lets a lot of sound leak out into the wings. What was the last gig played in here anyway, some break-away faction from Cirque du Soleil? The whole acoustic ‘feel’ of the place is horrifically drier than your ensemble is used to. And there’s an odd acoustic ‘sweet spot’ slightly to the left of center-stage. What the hell! Where in this catastrophe of a venue is the best location for the harpsichord? Precisely where should the other ensemble members be placed, so that the solo passages of each will be heard and so that the balance of the group will have some hope of realizing your artistic vision?

Very few musicians receive any formal training in how to cope with this part of performance. As a result, it’s an incredibly stressful and agonizing thing for most ensembles and individual musicians. And yet it’s possible for you to systematically acquire skills that will help you make accurate acoustics decisions confidently and quickly. How do you train your ears to immediately recognize what you and your ensemble members need to do to deliver your best?

Golden Ears is one way—Dave Moulton’s CD-based audio ear-training course. It’s not perfect pitch or interval training. Instead, it aims to teach your ears to hear the frequencies, the signal processing, the compression, the left-right stereo imaging, the distortion, and the amplitudes in acoustic media. The course is self-paced and available as a set of eight audio CDs. To use it properly, you’ll need a set of good speakers or headphones. I use the Bose noise-cancelling headphones, to make sure that ambient noise around my house doesn’t confuse my ears when I’m practicing with the Golden Ears CDs.

 Moulton book
Moulton basically puts forty years of his recording studio, performance and teaching experience on these CDs. There are hundreds of exercises on the CDs that systematically take you through different dimensions of acoustic spectrum and timbre and electronic equalization (EQ) and processing of sound. While the main purpose of the course is to help recording engineers and producers to understand and be able to describe the elements that recorded tracks contain—their spectrum, dynamics, reverb and other audio qualities—with the goal of insightfully adjusting them in recording gigs or in post-production, the course is just as relevant for sound-reinforcement engineering in live performances and for classical musicians and presenters who work in traditional non-reinforced acoustic settings.

In a way, Moulton’s educational process is a little like a culinary course for chefs—helping them to quickly assess deficiencies in a dish, and equipping them with the ability to know instantly how to salvage a dish that suffers from imbalances in flavor.

Once you get a feel for the acoustic ‘ingredients’, you can take better control of the weird performance hall you find yourself in, take control of the hall’s sound-reinforcement control board, take charge with the presenter’s or the hall’s sound engineering staff, or take control and better utilize your own recording gear and approach the process of tracking, mixing and engineering with more confidence and less time consumed in trial-and-error. You can consistently ‘cook’ a better ‘dish’.

 Golden Ears audio course
How does the course work? There are four volumes, with two CDs per volume, each covering different components of the recording process. In Volume 2, after you’ve followed instructions for optimizing your listening set-up, you go through a series of A/B drills using excerpts of recorded music. The first recording (A) is the reference piece and the second (B) is a clone of the first with an applied amount of as-yet-undisclosed signal processing, frequency boost, delays, etc. Your task is to learn to identify the difference between the two recordings. To assist you, Moulton limits the number of options and groups them into six families of effects: amplitude change, distortion, compression, equalization, stereophony, time-delay, and reverberation. The drills are progressively more difficult. After you’ve completed your first pass through them, you can do the drills in random order to practice and improve your skills. The course materials are automatically randomized so you can’t identify any drill until it’s over. Golden Ears teaches you:

  • to recognize the effects of compression on a various music signals;
  • to identify fast and slow compressor attack and release times;
  • to correctly identify musically important EQ problems;
  • to recognize when loudness is the only difference between two signals;
  • to distinguish ranges of 1 - 10% and 10 - 30% Total Harmonic Distortion (THD) in recorded music or in sound reinforcement rigs used for live performances in larger chamber music venues;
  • to recognize abnormalities in stereo imaging (reverse image, mono summation, polarity reversal, pseudo-stereo, etc.) ;
  • to identify channel-to-channel time differences over a 1 to 50 msec range; and
  • to correctly recognize and manage gated and ungated reverb.
Moulton is a pro engineer and producer. He holds degrees from Bard College and The Juilliard School of Music. In the 1960s and 70s, he owned and operated his own commercial recording studio in New York. Previously at SUNY and the University of Massachusetts at Lowell, he was in the 1980s and early 1990s Chairman of the Music Production and Engineering Dept. at Berklee College of Music in Boston. He is now teaching privately and at the Museum of Fine Arts in Boston. He has worked as a studio designer and has designed playback room configurations for recording studios and wide-dispersion acoustic lens arrays for performance venues. In 2000 he (with Curt Wittig) received a Grammy nomination for the CD recording of music by George Crumb performed by the Philadelphia-based ensemble, Orchestra 2001. Dave wrote and produced Golden Ears in 1994, based on traditional music school ear-training techniques.

 Orchestra 2001 CD
The first two CDs each have a series of 14 EQ drills, each drill consisting of ten short bursts of sound with EQ applied and cancelled as you listen. First there is a warm-up for each drill, where Moulton describes what’s coming. Then the drill begins. For example, drillset_1/example_1 has ten seconds of pink noise, and for a few seconds an octave band centered around 500 Hz is boosted by 12 dB. Drillset_1/example_2 does the same thing to an octave centered around 63 Hz, and so on, until all of the octaves of the human hearing range from 31 Hz to 16 KHz have been covered—all ten octaves in random order. The manual is well-written. The comprehensiveness of the drills enables you to discover and make note of your own hearing deficits in a systematic way. If you do this with other members of your ensemble, you can identify certain aspects where one or another ensemble member has particular strengths. More likely, doing this together will confirm in an objective way which one of you is actually the most skillful and accurate acoustician (for purposes of reconnoitering with venue engineers and production people), rather than who merely has the strongest opinions and the most insistent expression of them.

After the pink noise drills, Moulton conducts more exercises involving applications of EQ, except this time they are done to musical excerpts. Next, there are more reps of EQ apps, this time with octaves being cut instead of boosted. Vol. 2 also has two CDs, this time with a wide variety of signal processing applied to a vast number of musical selections and styles. You can assess and practice your perceptions of things like amplitude changes, distortion, compression (including recognizing slow and quick attack and release times), some different examples of EQ, stereo/mono/pseudo-mono switching, time delays (including time differences over a 1 – 50 msec range), and reverbs. Vols. 3 and 4 extend the skillset into progressively more detailed and exotic effects that are important for performance and for recording/engineering.

Some effects are simple and easy to recognize, like a 3 dB amplitude boost. Yet, depending on the texture and timbre of the music, even something so simple as this may be difficult to hear. It may be even more difficult in high or low registers. Other effects are more obvious—clipping distortion, long reberbs, etc. But most are somewhere in the middling range of subtlety, like ‘left channel mids (1 KHz) boosted by 6 dB while right channel highs (10 KHz) cut by 6 dB.’ You can refine your ears to hear subtle EQ cuts and boosts within narrow intervals of a fourth or a fifth, and to readily hear multiple frequency bands at once. After 20 or so hours with these CDs, you’ll be able to quickly diagnose what’s going on acoustically in any performance venue and intelligently strategize what to do about it—to make the best of the situation you’re in.

In summary, all the hundreds of drills on these CDs train you to hear and to recognize problems in all these areas and across the whole span of registers—before you go on-stage or before you exit the recording studio. By revisiting the CDs before you travel to give a concert in an unfamiliar hall, or by revisiting the CDs a few days before you start a recording gig, you can refresh your familiarity and refresh your confidence to be able to quickly amend and improve the acoustics that will affect the music you make.

In effect, for those of us who are not recording engineers or conductors in our day-jobs, rehearsing our ears to recognize and thoughtfully manage these ‘macro’ acoustics effects is a very important yet under-recognized aspect of our art. The Moulton CDs are an efficient way to establish and maintain a mastery of this.

 iZotope sound 2hrs ago

 Prosser book


Tuesday, January 1, 2008

Miking a Violin

White, Fig. 14.1
B asic technique:
1. Place a mic slightly above and about 2 meters in front of the violin;
2. Place a ribbon mic over the player’s [right] shoulder;
3. Position a mic [5 cm to 15 cm] underneath the violin;
4. Position a mic behind the violinist so that the head and body of the violinist are partially obstructing the direct path between the mic and the instrument.”
  —  Owsinski, p. 150.
To this, consider adding #5: a ‘near-coincident’ pair—two cardioid-pattern mics spaced about 15 cm apart horizontally (about as far apart as your ears) and with their axes angling away from each other at about 120 degrees—can improve depth and phasing without adding reverb confusion. You can set this up on a normal mic stand or mic boom with a Sabra Som ST4 Mic Bar or equivalent. Depending on the orientation of the two mics with respect to the instrument, the physical (acoustic) time-delay associated with the 15 cm between them (i.e., between the centers of the two mics’ cardioids) will be as much as 0.5 msec for normal chamber music concert hall sound velocities, room temperatures and humidity levels.

Some might think that adding more mics like this is too complex, too many degrees-of-freedom, too many choices in mixing, too much bother.

But I think it’s not overly complex. And there’s physiology and acoustics underlying a rationale for doing it. The results can be worth the effort if you’ve got the time to do it. In fact, in comparison to so many other aspects of sound engineering that have changed dramatically over the decades, it’s a bit surprising that miking practice has changed so little.

To figure out how ‘near-coincident mic pair’ and other mic array techniques work, it’s useful to consider binaural physiology of hearing—specifically, the psychoacoustical phenomenon known as the ‘precedence effect’. The precedence effect is when reflected signals are neurologically ‘inhibited’ for a period ranging from hundreds of microseconds to a few milliseconds after a direct signal is received by a human listener in a reverberant room.

Multi-microphone digital processing schemes have of course been used for years in connection with adaptive noise canceling (ANC) and other sound engineering techniques to remove noise and reverberation distortion. The individual microphone signals are divided into frequency bands whose corresponding outputs are co-phased (delayed differences are compensated) and summed. The gain of each band is set according to the degree of correlation between microphone signals in that band. ANC operations are equivalent to a time-varying linear filter whose properties depend on the short-term spectra of the two (or more) input channels. This approach can help improve coloration (early echoes that contribute spectral distortion) and reverberant tails (late echoes). But ANC isn’t what I’m talking about.

Human binaural hearing’s ability to localize a sound when there are delayed reflections of the original sound that interfere with the localization process has been termed the ‘precedence effect’ or the ‘first wavefront’ effect. The precedence effect is how human binaural hearing tends to base the judgments of localization and pitch predominantly on inter-aural cues carried by earlier, direct sound—and it contributes quite a bit to our perception of the depth and timbre of stringed instruments’ sounds. In general, the precedence effect affects pairs of coherent acoustic wavefronts that differ in arrival time at the ear by from less than 1 millisecond up to about 10 milliseconds. Conventional wisdom is that the precedence effect arises because of both ipsilateral (same-side) and contralateral (opposite-sides) neural inhibition in the conduction pathways for each ear. Colburn and Durlach (1978) gave a good, early summary of this.

Sayers and Cherry (1957) were among the first to quantitatively describe binaural hearing directly in terms of interaural correlation in terms of a running crosscorrelation function:
Running Cross-correlation
where xl and xr are the left and right signals respectively and τ is the time delay between the signals at time t.

Lindemann also used a model of binaural hearing based on a running crosscorrelation function. Lindemann’s model, which provides a quantitative basis for the precendence effect, proposed two different criteria that are associated with accurately perceiving the lateral displacement of auditory events: location of the centroid, and location of the maximum. These criteria both relied on the information in the running inhibited crosscorrelation function Ψ.

While violin and other string sounds contain continuous excitations, they aren’t flute-like; they aren’t uniform excitation of pure sinusoids—which is why somewhat esoteric miking techniques might be helpful for strings but offer no noticeable advantage for other instruments.

Basically, I suggest that the natural correlation and inhibition mechanism can be profitably leveraged by thoughtfully placing ‘near-coincident’ dual mics, to supplement the more conventional schemes described in the blockquote above. Use your normal mics just as you routinely do. But add a ‘near-coincident’ dual-mic pair as well. For recording, the dual-mic signals should ideally be sampled at high KHz rates, to be consistent with capturing the sub-millisecond precedence effect.

(In terms of post-processing and mixing, a set of parallel independent delay lines with a 0.5 millisec-or-better resolution of delay times may be helpful to tweak the high-bandwidth-sampled multi-mic signals—but I have not experimented with that myself.)

Just a thought… Try it and see what you think. Happy New Year.