Mixing the Sound

Perhaps the most influential processing tool you have at your command as a mixer is not in your recording chain at all, but rather is your monitor chain. As a system installer, it is important to observe good practice, too, so that your customer, whether a professional mixer or home theater enthusiast, is able to hear the program material correctly. Some of this burden lies on the electronics, and some on the “electroacoustics,” that is, the combination of loudspeakers and their application to a specific room environment.

The reason that monitoring is an essential part of the recording process is that it influences dramatically everything that you hear, and as a competent mixer or producer you take action based on what you hear, not what the meters tell you. This effect is so pronounced that I made it the first operational chapter of my book, 5.1 Surround Sound – Up and Running, whereas normally you might think monitoring comes last.

One reason to think that monitoring comes last is in looking over the many glossy articles in many magazines. Just what is the relative importance of monitoring when often we see loudspeakers worth a few hundred dollars sitting on consoles that cost hundreds of times as much? The idea behind this is that the console is in the recording chain, while the monitor is less important, because it is not. Several experiences, those of others and of my own, illustrate that the monitor can have just as much impact on the sound as the console. Here they are: Svein Erik Borja wrote what I think is a classic AES paper, “How to Fool the Ear and Make Bad Recordings” (J.A.E.S., Vol. 25, Number 7 pp. 482 (1977)). One of the beauties of this paper is its very direct title. What it refers to is that, if the monitor system is not flat, the author proves that a competent mixer will take measures to fix this, namely equalizing the program material to compensate for the monitor. Thus if the monitor is in error, the program material will reflect an inverse equalization to compensate for the error. That’s why the paper about monitoring, referred to in Relevant Research this month, is so important. Measuring hundreds of loudspeakers in control rooms and giving us the statistics on this shows how much the program material is being influenced by the control room (the loudspeaker models were similar, all three-way Genelec designs). At low frequencies, the errors exceed the standards in half the measured control rooms, and the standards themselves are pretty loose, one would have to say, permitting quite variable sound (all in the name of being “practical’). Also, consider that the authors probably measured the better control rooms, ones that would be subjects of such a study. So it seems that this factor alone could be causing a large part of the variation in recorded software.

An experience I had at the AES 19th conference on surround sound shows how much variation there is among professional control rooms. The Studer BRS (described in Relevant Research) was used to monitor a 5.1-channel recording of ambience of the glorious Bavarian Alps early on the last morning of the show. I was hearing the OCT microphone arrangement for the first time, over headphones using binaural models of various control rooms, with head tracking. (Whoever says there’s no progress in audio is crazy – here’s a new multichannel microphone method monitored over a new multichannel headphone method; the only thing completely recognizable in the chain was the Mackie mixer!)

My impression was somewhat different than the startling good presentation that Studer did in their demonstration room. Here I had no visual reference for the monitors, and I could switch among various professional control rooms. I found it hard to get out-of-head imaging until someone walked up to the back left side of the microphone array outdoors and talked to me; I was indoors. It just happened that there was someone standing in the same relative position as the outdoor talker indoors. I felt rather foolish when I replied to the wrong talker, the one indoors with me! It really did work. But what did not work so well was comparing the various control rooms: their sound was quite different, even though most of these probably meet all the standards in place.

Both of these experiences illustrate that we’ve got to get it right for the professional and for the end user – especially if we have any hope of translating the experience that we have with the program material on the production end into the myriad listeners’ homes.

One thing that came out well in the survey cited above is that the average of many professional installations was pretty flat. Applying the idea to home listening, we shouldn’t tailor home systems for the defects we find in one or two homes, because on the average of thousands of installations, well designed products in homes will be correct. What we’d like to see is the studio and home match, with the studio having tighter tolerances than home-by-home installations, but having the same average response. This may be accomplished down the road by having smarter speakers react better to their environment (see Legacy, pg. 50), both in the studio and the home. For now, room equalization is probably necessary in most cases. One home theater installer I talked to said, “I wouldn’t think of doing an install without equalization; that would be like serving cake without icing.” Yet EQ is routinely ignored because it has a bad reputation from years of being misapplied. Some principles of application are: measure the soundfield at multiple points and average them together correctly; use the correct bandwidth, center frequency, and time tradeoff for the analyzer in use; and temporally average the results. Any one of a number of devices can work at this. Apply equalization using your favorite method; 1/3-octave, parametric, or fully digital. It doesn’t matter much which, as I found in a double-blind test: all methods of equalization stand head and shoulders above the unequalized condition – and this in a room that met all the acoustical standards with a high-quality professional tri-amplified monitor (and subwoofer). The speaker manufacturer was steadfastly against equalization, but listeners preferred all the equalized conditions (from different equalization methods) strongly over the unequalized case in the double-blind experiment.

Bass Management

In these pages, I wrote about bass management in the first issue because it’s such an important topic, and then I’ve been generally quiet while others were interviewed here about their monitoring. They often said, in effect, “We don’t need no stinking bass management.” The implication was that their speakers were good enough, why should they need this added complication? Okay, okay, in the interest of full disclosure, my company makes and sells one of the several of these things on the market. But do you think I would do that if it weren’t necessary? We do it because we saw it as an essential need to develop multichannel sound, but we hear all the time that it’s not needed in a professional environment. “We just don’t need this extra problem in our lives,” they seem to be saying.

Without bass management, the home listener hears sound to a lower frequency than you will hear in the studio. That’s because virtually all home systems use it, and by extending the frequency range from what the satellites can do down to what the subwoofer can do (the principle of bass management), the range exceeds that of studio monitors. Most studio monitors are 40 to 50 Hz aligned boxes, and will be down significantly by 25 Hz. Most home subwoofers go down to the 25 to 31.5 Hz region with reasonable response, and, when low frequency “room gain” is added in (caused by the image woofers below the floor and beyond the walls caused by reflections off the surfaces), can be solidly flat to 25 Hz or below. Remember that a small difference at low frequencies is equivalent to a much larger change at higher frequencies, above 400 Hz. That’s because the equal loudness contours, named for Fletcher-Munson but better measured by others, are converging at low frequencies, making changes there more prominent than at higher frequencies.

Also, there already have been discs released that had to be recalled because there was low-frequency noise on the main channels that went unnoticed in the professional environment in which they were made, but was prominent at home. The noises included a conductor beating time with his foot on the podium, stage sets being moved about, and a subway. The “pro” monitoring simply didn’t go low enough to hear the problems.

There is one way around this: use full range monitors that go down to 20 Hz on all five channels. Even here you’ll get in trouble potentially, because bass management adds everything electrically in phase and loudspeakers in rooms add sound in a more complex way because the phase relationship among the parts is less certain. Someone who has done this missed the fact that the front-channel bass was recorded out-of-phase with the surround channels, and when they were summed at home in a bass manager, the bass canceled, virtually completely!

In the home, bass management is subject to being done correctly or incorrectly, with the correct filters to match the loudspeakers and with adequate headroom given the summing that occurs in the process.

Sometimes this isn’t the case because filter slopes are not what they should be to add up correctly, or are not steep enough to prevent localization of the subwoofer from higher frequency components getting through. The biggest difficulty of all is that DVD-A machines provide six discrete analog outputs (and no digital output) and do not incorporate bass management, whereas virtually all receivers connect their discrete analog inputs into the circuit after bass management (which is normally done digitally nowadays). This makes DVD-V and the AC-3 or DTS tracks of DVD-A work correctly with respect to bass management, but the main linear PCM channels would not be so treated. The result is that the data-reduced coded audio sounds better than the linear PCM tracks, which will sound bassless. One magazine article called this situation bass mis-management. The simple solution is to use an analog bass manager outboard, between the player and the receiver, but practically no one does.

Speaker Placement

By now the facts of speaker placement for surround sound are well known to readers of this magazine, so is there something new to say about it? Indeed there is, since so often we find that compromises in the letter of the standard have to be made to accommodate equipment, furniture, and other factors in both studios and homes. Perhaps the most egregious installation I know of was for a well-known film critic. His center channel loudspeaker was mounted in the ceiling above the screen, pointed down!

Elevated Mounting

We see a lot of crazy things looking at the consumer magazines, but are all studios great? By no means. In fact, the most popular cover photo in pro audio mags for years has been the conventional control-room-into-studio shot, showing the elevated monitors over the control room window. Such a position produces two major problems: [1] the frequency response is wrong, and [2] strong reflection occurs off the console top.

The frequency response is wrong because the speakers are elevated, and the reference is for speakers to be in the plane of listening. They often are at home: just look at the number of “tower” loudspeakers that put the tweeter at seated ear height in any consumer magazine. With control room speakers elevated, even if they match a front-ahead speaker perfectly, a different frequency response will be produced for the direct sound in your ear canal. That’s due to the difference in head-related transfer function that occurs due to an elevated source – see fig. 1. In fact, one enterprising company invented a “de-elevator” circuit to make television settop loudspeakers seem lower by putting in equalization for the difference (fig. 2 for the case of 0 degrees vs. 30 degrees elevated, straight ahead). Of course, such a simple correction is very easily fooled due to dynamic cues, so if you can move your head, it won’t work. Producing the wrong frequency response at your ears is one main reason not to elevate the monitors so much.

The second reason is the reflection off the console top. Fig. 3 shows the reflection, as measured by SYSid, in a “normal” music control room with an elevated monitor. The reflection is not late enough to be separately perceived from the direct sound. Its direction comes from the same general horizontal angle as the direct sound, albeit with different azimuth, so the reflection won’t change the perceived direction much, if at all.

However, the timbre is strongly altered. In fact, the threshold of hearing a timbre change due to reflections within the first 20 ms is -15 to -20 dB spectrum level vis-à-vie the direct sound, and this reflection averages about -5 dB regarding the direct sound between 1 and 5 kHz, and comes within just a couple of dB in the 10-12 kHz region (compare fig. 4, the direct sound spectrum and time response to Fig. 5, the reflected sound). Fig. 6 gives the summed response of the direct sound plus the one reflection. So the console reflection has a strong impact on the timbre that you hear. Seems those consoles will have to go, or else the monitors should be lowered. On the other hand, there seems to be little wrong with elevating surrounds. One test reported in an AES preprint done in Japan found little difference as surrounds were elevated from 0 to +45 degrees. Of course, if they get very much higher than this, the effect of two channels is reduced, as they can’t be separated as well. Also, there will be response variations due to the height, just as with the fronts, but these are probably smaller than the ones due to frontal versus rearward angles due to HRTFs. One complaint arises with elevated surrounds from our European colleagues: applause is elevated, which seems unnatural. Pragmatically speaking, surrounds can usually be elevated to clear equipment, furniture, doors, etc., but not be so elevated as to make the listener think they’re “underneath” the soundfield.

There’s another thought rampant in the land. If we can’t get a center loudspeaker in the correct position because there’s a picture in the way, why not form a top-bottom phantom image with two loudspeakers, to place sound in the center of the screen? Well we’ve already seen that the different angles of arrival have different responses, so that’s a big problem for this theory. Another problem is that adding sound up in space without comb filtering from differing arrival times is very difficult. Just varying heights of listener’s ears from person to person is enough to throw off the delicate balance in time needed to get things to add up properly.

Soffit Mounting

Soffit or cavity mounting is generally a good thing, but there are several caveats. Some years ago, well-known researcher Sigfried Linkwitz made a comparison of in-wall speakers with those on stands in front of the wall. For stereo, he found those out in the room produced a more realistic soundfield, likely due to the off-axis response of the speaker interacting with the room to make the sound seem “deeper.” Since Linkwitz is an engineer who well controlled his experiment, matching level and equalization for instance, his results are probably universal. However, he was working only in stereo at the time. Surround sound provides a means to produce the “missing” ingredient, envelopment, from 2-channel stereo by way of the surround loudspeakers. So the rules of the game are changed, and his result might be considered moot today, although this is not proved – it is a postulate of mine to be proved or disproved by some researcher who needs a hot topic of use to all of us.

With a cavity but without flush mounting, all bets are off. A cavity in front of a loudspeaker is a big no no. Think of it as an organ pipe, driven at one end by the loudspeaker, and open at the other. Such a structure is resonant, and noticeably so. You’ll even see woofers mounted from the front of the box as opposed to the back side of the mounting plane to overcome this effect, or sculpted edges around a woofer, and certainly a tweeter. The speaker designer knows that these things make a difference, so it is best to follow their “advice” and keep stuff out of the front of the loudspeaker.
Sometimes we see cavities that contain a flush loudspeaker, but which have not had the face extended to meet the loudspeaker face. This can be difficult in light of toe in and tilt needed to aim the loudspeakers.

Sheet lead or vinyl loaded with lead can be used to mold the loudspeaker into the cavity with minimum diffraction effects. The cavity should also be filled with sound absorbent material, because otherwise the speaker will excite the modes of the cavity, which can potentially be heard since delayed energy comes back from the cavity later than the direct sound, which might otherwise mask it.

Cloth Coverings

Often cloth must be used to conceal loudspeakers. I heard a story once that MIT professor Robert Newman of BBN reputedly had a method to test the acoustic transparency of cloth. He charged clients hundreds of dollars to blow smoke through the cloth and thus determine its acoustic impedance, and its propensity to pass sound! While today we use measurements, often there is a conflict between interior decorator and acoustical designer: they always seem to want heavy, thick cloth, while we want open light cloth. I have learned to blow air through cloth and get a pretty good guess as to its high-frequency transparency, and have found that some even medium cloths can be used, so long as we can equalize for them. But equalization is imperative. Even the much-more-acoustically transparent than normal for motion-picture screens, microperforated screens, need equalization. (Stewart Filmscreen supplies an equalizer to use with their microperforated screens.)

Surround Professional Magazine