Your browser is out-of-date!

Update your browser to view this website correctly. Update my browser now

×

Real Coverage, Part 2

A follow-up to the March article, this month's discussion includes sound levels, power compression and how to figure out real SPL capability.

Real Coverage, Part 2

May 1, 2001 12:00 PM,
Rick Kamlet

We need to consider the real, long-term sound pressure levelcapability of today’s systems to make sure they’re not beingoperated at levels where the components are stressed and don’tsound good.

While acoustic design programs do a good job of predicting howthe system will perform in a given room, the system designer mustdetermine what the goal is for each application.

THE WORLD OF BUSINESS MUSIC IS CHANGING, AS PART 1 of thisarticle noted in the March 2001 issue of S&VC.Customers are more sophisticated about music quality than in thepast. There is a growing recognition on the part of businesses thattop-quality sound is an integral part of the customer’s experience.The demands on businesses to create effective experiences for theirpatrons are higher than ever.

Toward an understanding of how to create a premium businessmusic system, Part 1 outlined some questions to ask the customerbefore starting the sound-system design. We discussed objectivesfor power amplifiers and some accessories, and we delved into thecoverage of ceiling speakers, noting that the polar coveragespecification can’t be used for laying out ceiling speakers; you’vegot to look at the real coverage as projected onto the listeningplane. Part 1 ended with a discussion of the need for speakers thatprovide even coverage of the listening area. It doesn’t work wellto use speakers with different coverage angles at everyfrequency.

This month our discussion expands, starting with sound levelsand how to figure out the real SPL capability of a system, plus howto avoid a common computational mistake with 70-volt speakers, howto consider power compression, and some technical considerations.In future articles we’ll get into laying out the speakers anddiscussing today’s trend toward incorporating subwoofers intobusiness systems. So let’s get to the real coverage!

WHAT SPL IS NOT

SPL is not a substitute for fidelity! Have you everbeen to one of the many locations where management seems to thinkit’s meeting today’s need for higher quality music simply byputting in a louder sound system? Either your thoughts get drownedout in the barrage of noise, or the speakers sound awful becausethey’re turned up beyond their capacity. SPL cannot make up forlack of fidelity. However, we do need to consider the real,long-term SPL capability of today’s systems to make sure they’renot being operated at levels where the components are stressed anddon’t sound good.

SOUND PRESSURE LEVEL

Maximum SPL Capability. In the simplest theoreticalterms, loudspeakers produce sound levels according to threefactors: 1) the speaker’s power handling capability (driven with anindustry-standard pink noise signal); 2) the speaker’s sensitivity(the sound level produced with 1 watt of power); and 3) thedistance between the listening plane and the speakers.

Power handling and sensitivity are generally provided on thespeaker’s spec sheet. Consider the distance from the speakers to bethe ceiling height minus the listening-plane height (sometimescalled the ear height). It’s common to assume that ear height for aseated audience is around 3.5 feet above the floor, and ear heightfor a standing audience is about 5 feet above the floor, onaverage. For every doubling of distance beyond 1 meter, subtract 6dB.

The formula for computing maximum SPL capability for a pinknoise signal is:

M=S + 10(logP) -20(logD)

where M is maximum average SPL for pink noise, S is SPL fromsensitivity, P is pink noise power handling and D is distance inmeters

However, this resulting SPL figure, even at the standard 1-meterdistance, is not a measure of the typical music or speech that canbe expected from the system. Pink noise is different from music andspeech. A downward adjustment of at least 4 dB must be made to themaximum pink noise computation to reflect the maximum average SPLcapability for music or speech.

Why does a 4dB adjustment need to be made? Pink noise has apeak-to-average ratio of 6 dB. In other words, the average signalis only 6 dB lower than the peak signal. Music or speech, on theother hand, has a peak-to-average ratio of at least 10 dB (if nothigher). If we keep the peak capability at the same point as whereit has been tested with the pink noise, then average music orspeech is 10 dB below that (or 4 dB below the pink noise’s averagelevel). Take the pink noise rating and lower it by at least 4 dB toget the maximum SPL capability for music or speech. Voila!

Notice that this adjustment factor is based on an assumed 10dBpeak-to-average ratio. I use 4 dB as the highest average music orspeech will be able to attain. You may choose to make an evenbigger adjustment, maybe -6 dB or -10 dB, corresponding topeak-to-average ratios of 12 dB or 16 dB.

Adjusting our formula, the maximum SPL of music or speech of asingle loudspeaker is:

M=S + 10(logP) -20(logD) – 4

where M is maximum average SPL for music and speech, S isSPL from sensitivity, P is pink noise power handling and D isdistance in meters

SPL Capability of Multiple vs. Single Speakers. Oneother thing to consider is that the SPL capability of amultispeaker system is higher than that of a single speaker becausemultiple speakers overlap to a degree that depends on their layoutand spacing.

SPL of 70V/100V vs. Low-Impedance Speakers. Speakerswith 70/100 volts produce lower maximum output SPL than do lowimpedance (4-, 8- or 16-ohm) speakers. This is because, first,70/100-volt speakers are typically tapped at lower power levelsthan the low-impedance speaker’s maximum capabilities. Second,there is always some insertion loss through the speakertransformer.

Possibly the biggest factor, though frequently overlooked, isthat the driving amplifiers clip at 70 volts (on a sinewave). The highest the audio signal can get before clipping is 70volts (or 100 volts outside the United States). The average formusic or speech is at least 10 dB below that. So, if you’re tappedat 15 watts, you’re really only averaging somewhere around 1.5watts of sound out of that speaker, and that’s with the peakshitting clipping!

If you compute SPL based on 70 volts going into the speaker,your estimate is going to be at least 10 dB higher than the averagemusic or speech that can actually come out of that speaker!

Power Compression. Speakers compress the sound at highlevels. As their voicecoil temperatures rise, the impedance of thevoicecoil goes up, resulting in less draw of audio power from thesame voltage drive signal). This is called “powercompression.” We have left power compression out of theformulas because speakers differ in this regard, and because thedegree of compression is highly dependent on operational factorssuch as the peak-to-average ratio of the signal source. Powercompression is also affected by the degree to which the speaker isbeing run below its maximum capability. You can generally assumearound 2 or 3 dB of compression for quality speakers (such as thosewith large diameter voicecoils) with typical music or speech; andassume as much as 5 or 6 dB of compression with inexpensivespeakers (especially those with small diameter voicecoils and/orlow power ratings), operating with compressed music sources.

Now that we’ve talked about the need to understand the realcoverage of the speaker (in Part 1) and how to figure out how loudthe system can get, we can apply this coverage information tolaying out systems.

SETTING SPL TARGET GOALS

While the JBL Distributed System Design software, acousticdesign programs (like EASE), or computerized programs from othermanufacturers do a good job of predicting how the system willperform in a given room, the system designer must determine whatthe goal is for each application. The following is intended to helpset an SPL goal for the system design.

For systems requiring paging intelligibility, here are somegeneral guidelines to use as starting points in your systemdesigns. The SPL levels we’re considering here correspond to themaximum average SPL for music and speech (which, you’ll recall, isat least 4 dB less than the maximum average SPL for pinknoise).

Factors to consider include the ambient noise, loudspeakerdistance from listeners, loudspeaker overlap, the type of programmaterial and fidelity expectations of the listener.

Determining SPL Above Ambient. Economy background musicinstallations require an average music level capability of at least5 dB above the ambient noise. For good paging intelligibility, thesystem needs to be 10 dB higher than the ambient noise. Levels of15 to 20 dB above ambient yields excellent intelligibility. (SeeTable 1.)

Finding the Ambient Sound Level. If the installation isin a location that is already in use, use an SPL meter set for slowresponse to measure the A-weighted ambient sound at the listener’sear position. Take measurements during the noisiest time and makesure the HVAC (air handling) system is operating during the test.If the installation will be in a new facility, try measuring theambient sound level in a similar type of venue.

EQUALIZING CEILING SPEAKERS

Setting EQ for ceiling speakers can be different than with soundreinforcement speakers. Incorrect measurement techniques willresult in the system not sounding very good, and it would be ashame to design and install a great system and then misadjust it soit sounds like an amateur job.

An example of poor measuring technique is positioning themeasurement mic in the overlap region between adjacent speakers.This can lead to faulty measurements. Microphones tend to showsignal additions and cancellations at various frequencies that onlyoccur in that one ¼-inch space and are notrepresentative of the listening space as a whole.

In addition, more than a few installers have tried to equalizefor floor reflections, which sometimes results in large boosts andcuts in adjacent EQ filter bands. You can not equalize outfloor reflections (or any other reflections). But for good, solidmeasurements, here are some suggestions:

Mic Placement Within the Speaker Coverage Pattern.Place the microphone either on-axis or up to 20° off-axis. Tryto stay within the coverage pattern of a single speaker. Whenequalizing on-axis, you equalize the direct sound in the mid tohigh frequencies, whereas at low frequencies, you’re still takinginto account the low-frequency summation of adjacent speakers.

Mic Height. While it is best to place the microphone atthe typical listening height for the application, the measurementscan be contaminated by floor reflections that can artificially addor subtract to various frequencies as displayed on your testequipment. For example, a measurement taken at a 4-foot (1.2-meter)height may show dips at odd multiples of 80 Hz (240 Hz, 400 Hz, 560Hz, 720 Hz, 880 Hz, etc.) and peaks at even multiples of 80 Hz (160Hz, 320 Hz, 480 Hz, 640 Hz, etc.). Depending on the resolution andbandwidth characteristics of your measuring device, these can showup as various boosts and dips in your measurement bands. Thesereflections are not equalizable, and trying to equalize them outcan result in a very bad sounding system. It is best to eliminatefloor reflections from your measurement. How do you do that?

Eliminating Floor Reflections by Mic Positioning. Toeliminate the floor reflection from your measurements, use themicrophone in pressure zone mic mode by placing the mic on a hardsurface on the ground or by setting it on a large piece of plywoodat ear height. For this type of measurement, the microphone istypically laid on its side. If you’re using plywood, place the micslightly off center on the plywood to minimize complications fromthe addition of symmetrical diffraction effects from the edges ofthe plywood plane. Placement about 4 to 6 inches away from thecenter point, toward one of the corners, is appropriate.

To maximize the high frequency accuracy, make sure that themicrophone diaphragm is as close to the plywood plane (or floor)surface as possible. If the natural contour of the mic case makesthe element sit off the surface, it is beneficial to angle the caseso the mic diaphragm is within ¼ inch (6 mm) of the woodsurface, without actually touching the wood (or floor).

You may get some strange looks by placing your microphone on thefloor, but it’s the best way to get an accurate measurement.Anyway, funny looks during the testing phase sure beat funny looksafter the system is up and running.

Microphone Type. If possible, use aninstrumentation-grade microphone. To get the most accuratemeasurement in PZM mode, a small-diaphragm mic with small housingis best because it allows the mic diaphragm to get as close aspossible to the plywood (or floor), minimizing any interferencebetween direct waves and those reflected back onto the diaphragmfrom the plywood itself and thereby minimizing false information.Your curve is only going to be as good as the measurement micyou’re using, so if at all possible, use an accurate, top-qualityinstrumentation mic.

EQUALIZATION

It is often best to adopt the strategy of only cutting the EQfilters, not boosting them. Peaks in the frequency response can bebothersome and can be equalized out. Dips in the frequency responsetend not to be as audible. Because they are frequently caused bytime-related phenomena, they are often not equalizable. So it’susually best not to bother with equalizing out dips. As mentionedbefore, a graphic EQ with only a few bands is probably not of muchuse in EQing a system. In order to remove the peaks, you end uptaking out way too much good sound content.

Note that as with all equalization, you must be wary of largeboosts and cuts in adjacent filter bands. Use equalizationgently.

Parametric EQs are often more useful than graphic EQs becauseyou can really pinpoint the frequency required and narrow thefilter enough to take out only as much as needed without taking outtoo much content. Unfortunately, only a few business musiccontrollers come with parametric EQs, although more digital controlsystems are being used, and many of these are built withparametric-EQ capability.

Do not try to equalize reflections because they change withevery mic position in the room. If you’re not sure what’s real,move the mic to a number of locations. It is also useful to averagea number of locations.

I like to set the EQ for a gentle roll-off of high frequencies.A roll-off of somewhere around 3 dB, starting at 2 kHz per octave(4 kHz being down 3 dB) often sounds better than setting itflat.

For bass frequencies, you may have to experiment with whatsounds best for that application. You might need to set the bassfrequencies (below 100 Hz) up as much as 10 dB higher than the midsand highs in order to balance the sound. Be careful not to boostbass frequencies that the speake can’t handle. Ported speakerscan’t handle much power below their tuning frequency, so find outfrom the manufacturer what frequency the loudspeaker is tuned at,and make sure not to send it frequencies much lower than what itwas built to handle.

One factor affecting how much higher to set the bass is how loudthe system will be played. Unfortunately, music itself goes up anddown in level, so the balance is going to be wrong at some musiclevels. Unless you’ve got a dynamic system that automaticallyadjusts for this (such as the JBL Soundzone’s Autowarmth function),you’re probably going to need to just split the difference and setthe bass volume somewhere between the loudest and the softestmusical volumes.

SUMMING UP FOR NOW

Now you know some key steps in putting together a top-qualitybusiness music system. So, let me summarize a few points:

  • Lay out the system based on the real coverage as projected ontothe listening plane — don’t base it on the polar coveragespecification from the spec sheet.
  • Help customers avoid the assumption that a loud sound system isa good sound system. Design a system that has the real SPLcapability that’s going to be required, without distorting orbecoming harsh.
  • Don’t assume that the computed pink noise capability of a systemis what they’re going to be able to get with music or speech.
  • Guard against the common computational mistake of assuming thatyou’ll get the SPL you computed at 70 volts from a 70 volt system.Decide the speaker density you need based on the requirements ofthe application.
  • After the installation, be sure to EQ the system with the rightmicrophone setup, avoiding some of the common EQ mistakes mentionedabove.

In upcoming issues, we will delve into the specifics of speakerlayout patterns and spacing, and how this affects the coverage ofthe space and the SPL capability of the system. We will alsoexplore the emerging trend to incorporate subwoofers in businessmusic systems, the advantages and drawbacks of various crossovertopologies, target SPLs and placement aspects.

To be continued…

Rick Kamlet is the senior director of installed sound at JBLProfessional. He would like to acknowledge some of the people whocontributed concepts and/or research used in this series ofarticles, including John Eargle, Joe Etrick and KurtGraffy.

SVConline

Lost the magazine? You can read Part 1 of “RealCoverage” in the March 2001 issue of S&VC onlineat www.svconline.com

Designing for Speech Intelligibility

Choose the Appropriate Speaker for Output Capability— Make sure the speaker you choose has adequate powerhandling and sensitivity to produce the sound levels required. Forgood paging, the speaker is typically required to sustain averagespeech levels that are at least 10 dB higher than the ambient noiselevel.

Cover Mid and High Frequencies — Choose a speakerwith the appropriate coverage pattern for your application. Forintelligibility, select a speaker that has especially even coveragein the 1 to 6kHz range.

Power the Speaker Adequately — Make sure thespeakers are driven with enough amplifier power to sustain theexpected sound levels. It doesn’t help to have speakers with enoughoutput capability if you don’t provide enough amplifier power todrive them to that capability. Clipping the power amplifier addsconsiderable distortion, degrades intelligibility, and results inunacceptable sound quality. Clipping is especially hard on thehigh-frequency speaker components in a system and can cause damageor failure.

Featured Articles

Close