5point1

Setting Up Multichannel Systems

You’ve got questions, we’ve got answers….

By Tomlinson Holman

You’ve heard the debates: What about the center channel? What angle should the surrounds be at in plan view, and what do I do with the subwoofer? We’ve tried to give solid information here about all of these, and surround is well launched as a medium. But when it comes to having made those choices and the actual setting up of the studio, there are still some questions to answer.

How do I set the level?
Like home systems, the simplest systems are set up with no test equipment at all, are set when they are installed following the directions for relative balance among the five channels by ear, using the “haystack” pink noise (peaked up around the midrange) that every processor has. Interestingly, it is possible to set the level balance quite exactly if the listener is sensitive at all, such as to within ±0.5 dB. So, systems installed with just a little care can be well balanced. Note that the listener sits in his favorite easy chair and listens to each channel in turn, setting them to sound equal in loudness — making the reference position to get the levels equal is at the listening position.

As professionals, we have to be at least as good as a decent home installation, and hopefully better. What we want is not just a relative balance, but an absolute one, making the program material reasonably interchangeable for level with other program material for the medium in use. This means that the user doesn’t have to constantly fiddle with his volume control. To do this we need an absolute level reference. For that, we need test materials, or at least generators.

While measuring the level of sine wave tones is pretty easy (although, even there, there can be distinctions among peak, peak-to-peak, and rms levels) with most metering equipment, noise is more complicated. And we need noise to set up a room, and not sine wave tones. The reason is that, even at 1 kHz, standing waves will cause spatial variations of many decibels moving just a little in a space. You can easily prove this to yourself by playing a sine wave tone and moving your head around your principal listening area; usually you will hear large level differences as you move. That’s the problem with sine waves, originating from standing waves. So you might want to move on to broadband pink noise. Here all frequencies are stimulated, with equal energy in each octave (by definition). But one problem with pink noise is measuring it on meters. First of all, the meter bounces around, so what is the correct reading? Second, pink noise is random, called stochastic, and this means that instantaneously the various component frequencies can add up to quite a high level, or not much level at all. Usually the peak level of pink noise is some 10 dB above its rms level, which leads to all kinds of complications.

For instance, I once received a tape from a prominent engineer that said that it contained pink noise at –20 dBFS. Indeed it did: the maximum peaks were at –20 dBFS. But this is wrong, because the reference in use (SMPTE) was –20 dBFSrms. So the pink noise recorded was some 10 dB low, and turning the monitor level up to correct for this resulted in the program material being, you guessed it, 10 dB too loud. The most unequivocal level is that of sine waves; pink noise complicates matters.

As does wide band pink noise. With low-frequency components present, there’s lots of level bouncing around, and when played back in a room, the effects of standing waves are to add a frequency-by-frequency ±15 dB variation to the level typically. One way to solve this problem is to band limit the noise to above the Schroeder frequency; that room-volume dependent frequency where not much else above it in the soundfield can be called diffuse. For small rooms, 500 Hz will do.

At the high-frequency end of the range, there’s lots of trouble, too. Half-inch measurement microphones exhibit an 8 dB difference between direct sound and diffuse field sound at 20 kHz, and 6 dB at 10 kHz. These are big differences that will show up in an overall measurement, depending on whether you are direct- or diffuse-field dominated, and which way you aim the mic. Also, for interchangeability with the largest range of differing systems where “house curves” or the X curve of motion picture sound is in effect starting their rolloff at 2 kHz, rejecting frequencies above 2 kHz leads to maximum utility.

So pink noise, band limited to 500 Hz to 2 kHz is the preferred source to set acoustic levels. It is broad enough so that single-frequency standing waves don’t make the measurement inaccurate, and narrow enough that low- and high-frequency acoustical and sound system effects are minimized.

Using this two-octave band noise also solves another problem. The standard of the motion-picture industry for level setting is the ubiquitous Radio Shack sound level meter. Everybody should have one. Get the cheaper analog meter that has better resolution than the more expensive digital one. I have calibrated more than 100 of these meters over a period of years, and find them out of the factory to mostly be within 1 dB, pretty good for such a cheap device. However, they do vary one to the next in their C weighting filter, so that broadband noise may show a different level meter-to-meter when they are calibrated on a midrange tone (or noise).

Where do you get this narrow band noise? Well you could make it with pink noise and two steep filters, and then calibrate it for level using a good AC voltmeter (there’s differences in methods here, including bandwidth of the measurement (made moot by the narrow band noise) and type of detector, rms or average-responding rms calibrated, a 1 dB difference). Now you’ve got to get that level to studio bus level, say +4 dBu, rms, (note probably +10 dB peaks; if using SMPTE –20 dBFS reference then to –10 dBFS) out of the console, for each channel in turn. Then you want to set that to a standard, by setting the room equalizer or power amp gain so that the noise measures the following on a C weighted slow reading sound level meter. Among them are:

Television mixing...............................78 dBC slow
Film sound, small room....................83 dBC slow
Film sound, large room.....................85 dBC slow
Music mixing Varies, try the range above

How do I set the subwoofer’s LFE level?
Now there’s that pesky subwoofer level to set. There are two things to do: get the splice right with the main channels if you are using bass management, and get the LFE level right. If you can filter your noise to below 120 Hz, play it at the same electrical level and measure it with a C weighted slow reading sound level meter. It should read +4 dB compared to the table above.

“Where did +4 come from, isn’t it +10?” you ask. Yes both are true! How come? Because the +10 dB specification is for “in band gain,” boosting just the range below 120 Hz, not measuring the full frequency range. How is it that this can work and our band limited 500 Hz to 2 kHz noise work, too? Because in filtering the wideband noise to two octaves for this noise we reduced its level (by stripping off the energy outside the band), but then by definition we “turned it up” to get “back” to –20 dBFSrms. So these numbers are correct.

Of course, a better way to align the subwoofer level is to get a spectral level match between mains and subwoofer of the bass managed system with a spectrum analyzer. This is what we always do. Then, when an extra 10 dB gain is put in the LFE path, it will read 10 dB greater in those bands that are in the range, and measure to +4 dB compared to wide band noise.

A lot of this work has been done for you in our Multichannel Test Tape, available from Martinsound as CheckMAX (www.martinsound.com). It is aimed at the monitor system, is in the DTRS (DA-88) format, and includes the narrow band noise, subwoofer level setting tracks, and much more (noise floor testing, headroom testing, phantom image testing). While this is yet another shameless plug in the pages of Surround Professional for products I’ve been involved in, it really is the tool that is needed to set up multichannel sound systems. What can I say? Among the users are George Massenburg and Bob Ludwig, who have told me they use it frequently.