I think this is why B&W speakers have sold so well at their price points. A lot of people are much more into what "sounds good" than what is neutral to recordings. Among my friends who are buyers of premium audio systems, I can't think of one who is concerned about neutrality first.
This doesn't make much sense. We know that what's neutral
is what sounds good, that's the whole point of all these blind tests. Unless you mean specifically what might catch your attention for a few minutes in a showroom, in which case
maybe there's an argument for elevated treble and bass.
My understanding is that B&W
used to make more neutral-measuring speakers but have recently abandoned that. That said, their recent formation Duo wasn't too bad, on axis at least.
Hi,
Do they exist? speakers not flat on-axis (with a treble peak for instance) that sound good and hi-fi nevertheless?
Should their positionning be different from the flat on-axis ones? (different toe-in for instance)
I'm after examples and explanating why they can sound good.
Cheers.
It's certainly possible, but I also think this depends on exactly what degree of "not-flat" you mean. As others have mentioned, many speakers balance a slightly faulty on axis with the off-axis sound. Many designers will also prioritize slightly off axis response over on-axis, because many/most consumers don't toe in their speakers for aesthetic reasons. both the KEF R3 and LS50W measure slightly better at 15ish degrees off axis than right at 0 for instance.
On a more personal level, the JBL L100 I recently measured was, subjectively, one of the most enjoyable speakers I've heard, even though the measurements don't seem particularly fantastic.
Dispersion looks pretty nice and even, but, but the on-axis (and LW) shows some clear, broad peaks and dips. Moreover, the 2K peak and 3.5k scoop remained at almost every angle I measured. It's not
awful, but not great for a $4,000 speaker. I'd even worried I screwed up my measurements, but
Hifi-news and
Audio.com.pl measured a very similar response.
(Note that these measurements are for the L100 with their default knob setting. It did look to me like turning up the treble and mids knobs on this speaker actually provided a more neutral response and is what I preferred, but this is what JBL claims should measure flattest).
This puzzled me for a few reasons. One, I hadn't seen any recent Harman speakers show such seemingly middling measurements. Even though the L100 classic was a remake of an old speaker, it struck me as odd. 2) I enjoyed this speaker tremendously, and I've always had a preference for neutral-measuring speakers (even back in my purely subjective days). I'd not tested a speaker where my subjective impressions seemed so distinct from the measurements. 3) I didn't find this peak particularly audible, even when specifically looking for it, and peaks are typically more audible than dips.
Now, this is
purely conjecture and I could be totally wrong and misunderstanding the science, but my theory was that JBL aimed to balance out the frequency response effects of interaural crosstalk to some degree with this frequency response.
If you're not familiar, interaural crosstalk causes a significant dip at approximately 1.8k Hz when listening to stereo speakers in an equilateral triangle setup, (moving higher the closer the speakers are). This is a big reason why listening to a center speaker sounds so much clearer for dialogue than a pair of stereo speakers - the 2k region is important for intelligibility.
See this figure(7.2) from Toole's book for a summary:
While it seems we may adapt to this dip to some degree and it becomes less of an issue with wide dispersion speakers in a typically reflective room, it does still seem to be present to some degree. Note then, how the L100's response is curiously quite close to the inverse of the artifacts described in 7.2e. It's possible that's a coincidence, but it seems suspiciously close to someone trying to equalize that 2K dip and 3-5k hump.
Normally one doesn't want to correct much for this dip because as soon as you move much out of the sweet spot, the peak
would become noticeable. But most audiophiles
do sit right in the sweet spot without moving much. It seems to me trying to correct for this would make some sense on a speaker made for audiophiles.
You might then wonder why JBL/harman and others wouldn't do this for all their speakers. Well, most of the harman speaker lines also offer matching center speakers. Correcting for IAC would throw things off if you ever bought a center, so it's probably not worth considering we adapt to it to some extent. The L100s do
not have a matching center. It will pretty much only ever be used in a stereo setup by audiophiles, who will likely plant their butts right in the sweet spot when listening "seriously" and not care enough when they're not. That's certainly the case for me.
Of course, without any blind comparisons, it's very well possible this is a coincidence and the speakers work well with my particular room, or I'm just rationalizing how much I liked the speakers. But nonetheless, I
did like the speakers, and I thought it might be an interesting example of how one might voice speakers if you have a particular use case in mind. It compelled me to try equalizing for this dip on my other speakers by adding one or 2 db of energy around 2k (as I have an Atmos setup but don't use a center) and I find I prefer the sound this way.
