Loudspeakers: When Is Good Enough, Enough? Part 2
In truth, it is very difficult to know for sure how a speaker by itself really sounds unless you can listen to it in an almost perfectly neutral room. Any listening situation short of that perfectly neutral room and you have no idea how much of what you're hearing is room and how much is the speaker.
At my last gig with Harman Consumer Group I designed the just-released Infinity Beta Series speakers. As a part of every speaker's design process, each Beta system (except subwoofers) had to go through the well-known Multi-channel Listening Lab set up by Dr. Floyd Toole and Sean Olive. This neutral listening environment has a "shuffler" mechanism in the floor which can perform a double-blind test with four speakers at a time. Each speaker can be shuffled to the same listening location while still behind a black, acoustically transparent curtain.
The room is approximately 30 feet wide by ~25 feet deep. The ceiling is varies around 8 feet high in the listening area and has a combination of absorptive tiles and 2'x 2' foam RPG primitive root diffusors placed at strategic locations to randomly scatter reflections. You sit in the same fixed-position listening chair for each test sequence. The chair is located off the left-right centerline of the room and about two-thirds of the way into the room. There is a very shear black veil across the halfway point of the room. It is so sheer in fact that all of the set-up lights have to be turned off by the lab tech before a listener can enter. Otherwise one could easily see the speakers under test. The shuffler mechanism places the speaker under test directly in front of the listener, behind the shear curtain and about 3 meters away. Before each test series all speaker system levels are level-matched as closely as possible.
This speaker-in-a-neutral-room facility is set up to play a single Harman speaker and two or three other competitive speakers which Harman purchases, usually from a retailer. There are four specific, short vocal cuts, which are always looped. These are played back randomly for each of the 8 sequences that comprise a test. That is, in each single test the speaker that was "A" in the last test might still be "A" or it could be "B" or "C" or "D". The selections are- a very poor quality recording of a guy who sounds exactly like Jackson Browne with Little Feat, a James Taylor, a Tracy Chapman and a Jennifer Warnes cut.
Before they enter the MLL, all speakers to be tested, competitors included, have a series of tests run on them in Harman's large 4pi chamber. The most revealing test we used was called the "spin". The spin would rotate a speaker every 10 degrees for a complete rotation (360 degrees). This was done in both the horizontal and vertical planes. There was then a custom program, which would assign a weighting factor to each of the different angles. What we got outputted onto a single graph sheet was a set of six curves that gave us on-axis, listening window, early reflections and sound power curves plus two directivity indices. All in all these six curves paint an exceptionally accurate prediction of how a loudspeaker will sound in a "neutral" room.
During the time I was at Harman I participated as one of the twenty or so regulars who had been trained to listen critically in the MLL. I've got to tell you, it was very tiring doing this two or three times a week. Especially if you had to listen to several sets of really inexpensive home-theater-in-a-box systems!
The problem with inexpensive HTIB systems is that, as a category, they are sold for their small size and cosmetics. Thus, when held up to the standard of double-blind testing which helps to determine overall faithfulness of sound reproduction via controlled measurements, it is the HTIB systems which tend to stray furthest from the flat frequency and power response ideal. So, listening double-blind, what you end up listening for is the satellite system whose deviation from flat response tends to fall in the most psycho-acoustically "enjoyable" frequency areas. This can become a judgment call among the various listeners.
Another caveat in testing at this extreme low end is that only a single subwoofer is used to A-B-C-D each satellite. Matching each satellite with its own companion subwoofer would insert yet another highly variable parameter into the test, making choosing a "winner" almost impossible. Since I will be evaluating loudspeakers for Audioholics I will attempt to set up a listening room and test methodology, which hopefully can be repeated by an Audioholic in an accommodating dealer's listening room. Audioholics President Gene DellaSala tells me I'll have the use of the Rives Audio Pro Test Kit in a couple of months so I'll be able to verify what I'm hearing.
I may also try to get other Audioholics listening impressions. It will be interesting to see what "sighted" listening and A-B-only tests will reveal. I do know that Sean Olive's* research indicates the discrimination and reliability of trained listeners is between 3-27 times better than that of untrained listeners. Fortunately, loudspeaker preferences between trained and untrained listeners are the same. Trained listeners are just a little more critical and fussy.
Lessons learned from Harman's optimized double-blind listening. Or, trying to make sighted listening comparisons as fair as possible.
To minimize first order reflections from side walls, listen to a single speaker set just off the centerline of a room about one-third of the distance from the room's back wall. You may want to try adding a second speaker for comparison purposes. But you'll have to swap speaker positions and run through all listening tests again just to make sure you're not hearing position-sensitive anomalies. First order reflections can come from nearby sidewalls, the floor or ceiling. The sidewalls are usually the biggest offenders because most direct radiating loudspeakers have a wide horizontal radiation pattern and a more narrow vertical dispersion pattern. Note: There are a couple exceptions to the wide horizontal radiation rule. One is a typical two-way center channel system, Another are dynamic hybrid systems which might have a woofer or woofer and midrange firing forward and a tweeter firing upward into some form of dispersion device. Last exceptions are dipole radiators which require exacting room placement to sound their best. With a dipole radiator you must use your rooms surfaces to get the response that sounds good to you. Yes, I have heard dipoles (like the Quads) that are truly wonderful. It's just that trying to test dipoles versus direct radiators is an exercise in futility. They're two very different animals.
Pending the (hopefully positive) review by gds: install inline an Automated Controlled Environments, inc. Subwoofer Optimizer System unit. In subwoofers and full-range loudspeakers the SOS should be able to resolve and smooth out the single high Q bass peak that is present in about 90% of enclosed or partially-enclosed room systems.
Custom record a stereo DVD/CD into mono. Use 20 second snippets of no more than four music selections. Preferably pick music containing female and/or male vocals and fairly intricate arrangements. The recordings' quality should vary from the best you've heard to a least one which is poorly recorded. A great speaker can pull more out of a poor recording than you think possible.
If you know an accommodating dealer "calibrate" your ears to all the minute details of your four chosen recordings when listening back over a truly great speaker, like a Revel Salon. If possible, see if the dealer is game for letting you perform steps 1 and 2 so you can try to take the room out of the equation. Listening to single speaker like the Salon is instructive and enjoyable. Its sound power curve declines very smoothly as frequency increases; your brain relaxes as it detects the smoothness, the correctness, the "sound spreading out from the box" that reviewers often describe so enthusiastically.
Yes, from a single, very high quality direct-radiating design, the soundstage width and depth within a mono'ed recording can be heard. The degree of spread-versus-focus, of the vocalist's voice, versus processing that was used in recording the voice, can be easily delineated with a speaker like the Revel Salon.
Once you know intimately every detail on your four sound clips it's time to pick three or four likely speaker candidates that fit your budget/WAF. Here, as before, you'll still want to listen to no more than two (single) contenders in the middle of the room at a time. Determine which of the two single speakers you like best when playing your same four selections. Then, just to make sure, swap their exact positions and listen again. The winner stays. Bring up the next contender.
The grilles: decide whether you will listen with grilles on or grilles off. Finished grille assemblies are often not available during a speaker's design process. So the performance of many speaker systems are qualified by the system designer using the raw prototype cabinet only. Many otherwise excellently performing systems have their polar response curves adversely affected when the completed grille assembly comes in at the very end of the system design process. By then it's too late to make changes to either the speaker system or the grille.
When you get down to the final two single candidates try varying the volume levels between the lowest to the highest levels you'll listen. With low listening levels the ears' lower sensitivity to high and low notes comes into play. So it may be difficult to make any absolute judgments as to a system's sound quality here. However, a trained ear can monitor, to a limited degree, the speaker system's self-noise. The self-noise of a system is akin to signal-to-noise ratio for electronics. For example, when we're first trying out a new woofer driver we'll usually hold or suspend the raw driver in air and run it up and down the frequency scale. We'll listen to how much noise is generated from the spider, the surround, etc. Well equipped labs like Harman's have a Klippel system which can map a raw driver's noise characteristics at all drive levels. A driver with a noisy component part adds to the S/N of your entire system. Getting back to low level system listening; in ported systems especially there is a low drive level at which the woofer and the port start working together efficiently. A well designed port, usually flared at both ends, will couple with its driver/cabinet in a smooth transition, with no hint of turbulent air noise.