Pierre Aubert runs an online repository of over 500 loudspeaker measurements gleaned from other online measurement sites, volunteers, and even the manufacturers themselves. It can be found here:
www.spinorama.org The quality of the raw data is flagged, with loudspeakers often scoring zero for insufficient data rather than poor performance. While less ideal than using something like N/A (not available), this does avoid the garbage in = garbage out problem that plagues the research fields of nutritional sciences, psychology and pharmacology.
The measurements themselves are used to calculate the Olive scores of each loudspeaker. Higher scores have frequency responses that are considered to sound better than those with lower scores. There’s threshold of 1.0 where, below this number, two loudspeakers are within a statistical marginal error and can be treated as essentially sounding equally good under double blind listening conditions.
Why should a consumer bother with audio measurements? To avoid buyers remorse. Good sound is to bad sound (the absence of bad sound) as hot is to cold (the absence of heat) and bright is to dark (the absence of light). You don’t add good sound, you subtract bad sound. There are only a handful of ways of achieving neutral sound, and countless inventive ways of screwing up audio playback. Your goal should be to avoid products that add coloration to all recordings.
General trends for the spreadsheet warrior
Most loudspeakers on the market are terrible, with scores between -6.0 and + 2.0. This is where you find the Bose, Sonos and IKEA lifestyle speakers. I don’t think it’s a stretch to infer that most of us consider a loudspeaker an Olive score > 3.0 to be high fidelity. 6.0 appears to be the threshold where major defects in the listening window and sound power directivity index curves stop being obvious to the untrained eye (spikes, slope changes, and deviations from straight).
Audiophiles are prone to saying that upgrade XYZ blows away ABC, particularly when they’re placebos like digital interconnects or audiophile grade network switches. But if you’re willing to accept that an Olive score of 1.0 is the margin of error before there’s a noticeable improvement, then how many levels of high fidelity are there? The Olive score implies a firm answer: six. Why? Because those are the gradients between 3 and 9 (perfection). You can take a guess as to where I place most consumer electronics and wiring driven within their intended performance envelope to my objectivist ears.
Almost all loudspeakers, even the ones with bass extension (-3 dB) of <25 Hz, have scores that improve with a perfect subwoofer, usually by 2.0 or more. What’s a perfect subwoofer? I don’t know, but it’s probably not cheap, and you might need 2 or 4 of them. Go scratch that itch!
The scale tops out at around 9 with a perfect subwoofer. That represents the state of the art, and those able to assemble such a system ought to quit and take up another hobby to expend their bottomless neuroticism. Remember that 1.0 is the margin of error? If you can find a loudspeaker that’s 8.1 or higher with perfect subwoofer, you’ve touched the sun. Over 60 loudspeakers meet that metric. Icarus would be proud.
Price is not a predictor of absolute sound quality, only relative performance within the brand. Only a handful of full model range brands are consistently good: Genelec, Neumann, Magico, and Perlisten. All of the Harman brands have inconsistent scoring, despite being leaders in applied loudspeaker research at the time. Some brands such as KEF and Ascend Audio appear to have a sharp increase in quality between their older and newer products.
Smart speakers are a category to watch. The Apple HomePod, Google Nest Audio, Sonos Move, and Sonos Roam score better than most hifi brands, albeit with limited SPL output.
Concert loudspeakers
If you filter by “Tour Sound”, “Column” or “Cinema”, you will find this category of cinema, night club, and concert hall loudspeakers. Public spaces have performance demands and construction materials that are inappropriate for home loudspeakers (thermal capacity, extreme ambient temperatures, weatherproofing, drop-proof construction, quick release scaffold mounting, directivity control to minimize echos, line arrays, etc.)
Unfortunately, it seems to be at the expense of good measurements. Populated by brands such as Alcon Audio, Axiom Pro Audio (no relationship to Axiom Audio), Danley Sound Labs, Electro-Voice, Fulcrum Acoustic, JBL, JTR Noesis, Kling & Freitag, and Turbosound, as a general rule, they measure poorly. It’s almost as if we get better audio at home than at a live concert.
Much of the $$$ high end audio industry such as Burmeister, Fyne Audio, Goldmund, Klipschorn, and MBL resemble their industrial counterparts. I hate to think how they would measure.
Open questions about the future
There’s a few open questions that objectivists would like answered.
1. How do current sound bars measure? We know rtings use a proprietary measurement score on their website, but can we calibrate their scoring with the Olive score for a basis of comparison?
2. How much fidelity headroom is there for sound bars before their form factor and placement run into ceiling?
3. How do all-in-one wireless loudspeakers (such as a Sonos Five) measure, with and without active equalization?
4. Can wireless loudspeaker manufacturers manage latency without forcing the consumer to be locked into an ecosystem? I can see a wireless product category where a pre-pro or AVR is connected to a loudspeaker’s 7.4.4 multichannel wireless box via speaker-level/RCA/XLR; once invented, you can lock a customer into their loudspeaker family.
Shut up and tell me what to buy
The cat’s out of the bag. With measurements now readily available to the general consumer, there’s no longer a reason to mistakenly own a dud. Unless you have specific usability requirements (wireless, integration with Roon, etc.), you ought to be able to assemble an audio system that can provide decades of listening pleasure.
While there’s no guarantees, you’d do fine if you restricted your search to companies that:
a) Avoids known tricks of marketing (breakthrough engineering, celebrity endorsements (including designers), promise of exclusivity)
b) Use certain keywords in their marketing, such as “NRC”, “family of curves”, “Klippel Near Field Scanner”, or “anechoic chamber”.
c) Publish their listening window and sound power directivity curve on their website, and not just an on-axis frequency response curve.
