Correcting inaccuracies and misconceptions
"This month's issue of "Test and Measurement World" had a write-up on Harman Audio's methodology of applying their own unique measurement criteria to build the "best sounding inexpensive loudspeakers."
”Harman Audio, for those of you who didn't know, currently owns JBL, Infinity, and of course Harman Audio speakers.”
Harman International owns the following speaker brands: JBL, Infinity, Harman-Kardon (mostly used for multimedia/computer speakers) and Revel. Engineering, design and testing for these brands takes place at Harman’s Northridge, CA facility.
Aside from how successfully they're balancing their findings with cost saving techniques, they had some interesting notions:
1) They downplay the importance of on-axis frequency response flatness as measured on a DB meter (industry standard for response curve and rating)
Harman does not “downplay” the importance of on-axis frequency response per say. More correctly stated, through years of research, qualified by rigorous double blind testing Harman has determined (and published papers in the Journal of the Audio Engineering Society) that on-axis frequency response is but one of several important and measurable parameters that define and identify a speaker’s performance envelope.
“DB” meters are NOT the “industry standard for response curve and rating”. It is generally agreed world-wide that a reflection-free, free-air frequency response taken by recognized loudspeaker testing programs such as LMS, Sys-ED, Clio or MLLSA, etc. using a properly calibrated microphone is the accepted method for accurate measurement. Such programs can be accurate to within 2Hz if so desired but the measurement takes a bit more time. Lacking that, a 4 pi anechoic chamber, preferably accurate down to 20Hz, is utilized. Smaller chambers such as Harman’s can be calibrated to the 0.5dB accuracy as noted in the article.
”2) Their designs attempt to compensate for early reflections caused by potentially poor placements in people's homes (they gave out sets of speakers to 15 people and then went to their homes to measure)”
Again, improperly stated. Harman’s designs do NOT “attempt to compensate” for early reflections. Harman’s engineers design to a fairly flat on-axis frequency response with a similar off- axis response. The difference being that as the frequency increases the off-axis response is expected to decrease smoothly, with no apparent dips showing, as the speaker system transitions through the crossover points.
I will detail, where, within the “flat frequency response” dictum Harman engineers often deviate by model and by price point, so as to achieve the best balance of measured
And psychoacoustically flat response in a future article. There is quite a bit a wiggle room within a ±2dB range if you know what you’re doing and you know your competition.
Early reflections from a “flat on-axis & polar response” speaker system will have all frequencies reflected evenly if no absorption is applied at the point of the first axial reflection. So the system will act more like an ideal “acoustic mirror”. This effect is often appealing to those who like to think that their speakers can “image way beyond their enclosures” or produce “bigger than natural” sound.
“3) They conduct blind A/B/C/D testing of their speakers on people who:
a) Have no prior critical listening skill or bias
b) Have gone through their special "training" on how they should be listening”
Correctly stated, Sean Olive based his now famous paper
“Differences in Performance and Preference of Trained versus Untrained Listeners in Loudspeaker Tests: A Case Study” on 3), a) and b) above.
”4) Any audible difference or quality in speakers is measurable objectively”
I was one of Harman’s trained listeners. I averaged about 3 test sessions per week. I have heard (and believe) that the conclusions of Sean’s paper are entirely accurate. If you are a member of the AES you can obtain the full text of Sean’s paper through
www.aes.com
“For those of us who arrange our listening rooms to properly accomodate our speakers away from back and side walls and furniture, it seems a bad thing to have the speakers designed to avoid early reflection distortion. Wouldn't that potentially lead to poor off-axis response, narrowing the sweet spot?”
An improper assertion > “designed to avoid early reflection distortion”. A flat polar response will yield, at the ears, a time delayed but spectrally balanced image, seeming larger, but NOT distorted by narrower-band frequency aberrations. This is, by definition, excellent polar frequency response and it yields the widest possible sweet spot.
”As for using specially trained laymen as listeners, that one truly depends upon target audience. If you're trying to cater to the masses, you'll use a cross section of the masses to make your speakers match their preference. I can't fault them for that, though it's not likely to produce an accurate speaker as it will a "catchy" speaker that people might be initially impressed with pumping out Pop music at Best Buy.”
Read Sean’s paper. To quote “
The loudspeaker preferences of trained listeners were generally the same as those measured using a group of nominally untrained listeners composed of audio retailers, marketing and sales people, audio reviewers and college students…..The most preferred loudspeakers had the smoothest, flattest, and most extended frequency responses maintained uniformly off axis.”
“Personally, I'd want a group of classical or jazz fans & musicians to sit in a small theatre listening to a live acoustic performance, then go directly next door to a listening room where a recording of the same group is being played back on several different speakers in a blind test.”
Again, Sean’s paper showed that both trained and untrained listeners would come to the same conclusions regarding response. The only difference was that trained listeners would be able to come to those same conclusions 3 to 27 times sooner.
”Is such fidelity measurable? Or would the qualities that separate a good speaker from a bad one too inconsistent and, well, subjectively human in nature?
Yes, such “fidelity” (truth) has been proven, through hundreds of listeners, both trained and untrained, while listening double-blind, to be definable (see above).
Well, inexpensive within the Infinity brand name means that they start at $150/pr for bookshelfs and $200/pr for floor standing. Having also listened to the $1,200/pr floor standing versions at a store, I'd say the same designers were at work on those as well because they had similar sound "qualities"...
There are six full time Systems Engineers at Harman. And yes the same guy will usually design an entire line.
“The magazine's editor's note in the front of the magazine shows his lack of audio awareness as well as his stout arrogance with the line:
"Unlike preferences for wine, loudspeaker preferences aren't a matter of personal taste."”
If you had had the opportunity to attend Sean’s October presentation to Los Angeles Section of the AES you would understand the genesis of this statement. Sean starts his presentation by demonstrating that there can be no perfect wine because different wines have different and more prized taste qualities such as bouquet, aroma, body etc. (And there are many, many varieties with many, many anticipated "best of" qualities.) He shows a slide of a wine qualification “taste wheel”. Yes, this is an actual double-wheel, made of cardstock, with a pin in the center which has the different adjectives usually ascribed to specific varieties of wine, arrayed around the wheel. This makes the judging of “best” in any varietals category a matter of evaluating the proportions of an adjective which a particular wine may possess.
He invites those who disagree to go to www.tmworld.com/audio to respond. I think I'll drop in and say "hi."
Yes, say “hi” and ask the writer if he too had heard Sean’s presentation before writing his article.
“I lost faith and interest in Infinity speakers when Arnie Nudell sold the company.”
So you would prefer that Infinity speakers still be designed with dips at specific frequencies going down to 1 ohm? That doesn’t happen anymore. No speaker is Ever designed which dips below 4 ohms. For any of the Harman brands. Department policy.