Originally Posted by neutralguy
Back to science?
Distortion is probably underrated in this thread. Most of this thread's focus has been on and off axis frequency response and resonances, and rightly so, because most speakers don't even get this right. However, as mentioned in this thread, that's only a good starting point for a good sounding speaker. Bad performance in these measurements will be indicative of a bad speaker, but among speakers that get this right, there are still sound quality differences that are not explained by the frequency response charts.
Case in point is the M2 vs Salon2 shootout between two speakers with excellent spinorama measurements. The Salon2 destroyed the M2 by 73% average A/B preference and 8.40 to 7.18 average preference scores, even though the M2's measurements look at least as good as the Salon2's, if not better. Spinoramas did not explain differences among these two high achievers.
I get the point you're trying to make, but when comparing the M2 and Salon2, there's enough information in the spinorama alone to understand why they sound different. These two speakers are like an apple and an orange. Both have "good" frequency response measurements, but you're comparing a controlled directivity speaker designed for very high output with a conventional home hi-fi speaker.
The M2 becomes more directional than the Salon2 beginning above 100 Hz and remains more directional all the way up to 10 KHz. In a typical non-treated or minimally-treated room, the perceived tonal balance is determined more by the power response than the on-axis frequency response. Compared to the Salon2, the power response of the M2 has a steeper downward slope through most of the midrange. Also, the M2 transitions from increasing directivity to constant directivity around its crossover point of 800 Hz, producing a discontinuity in the DI (and change of slope in the power response) right in the midrange. In comparison, the DI of the Salon2 doesn't change much at the woofer to midrange crossover point around 500 Hz. It follows a fairly constant slope representing a gently increasing directivity up to its mid/tweet crossover. I suspect that if I listened to them back to back, I would find the Salon2 to be subjectively more "neutral" in the critical midrange. Also, the Salon2 has an off-axis dip where the mid crosses to the tweet, centered around 2.2-2.3k, which is subjectively favored by many listeners.
Of course, the measurements among these speakers are not completely identical, so perhaps there are still signals that can be extracted in their minor variations from one another that could further explain differences. Barring that, we have to look at aspects of speaker performance that are not captured by these plots. Frequency response plots capture only linear behavior, so nonlinear distortions such as THD and IMD are not reflected. When the speaker's frequency response is broken, distortion is probably the lesser offender, but among speakers that have good frequency response, I'm not convinced that distortion can be ignored. Previous studies have shown high tolerance thresholds due to musical masking. However, that does not mean distortion is solved. Two reasons I can think of:
1. THD is not a good measurement of distortion that correlates with human perception. More recently work using more complex stimulus than single tones and more complex analyses have yielded metrics of nonlinear distortion that better correlate with perception. There's the perceptual model based metrics PEAQ and Tan et al 2004
2. My own hypothesis is that we also have to consider off axis distribution of distortion, just as we look at off axis frequency response. Most previous experiments have experimented with perceptual correlations by adding distortion into the signal. This means the distortion takes on the same dispersion as the rest of the content when played by a loudspeaker. Masking makes sense in this case since all we're doing is introducing harmonics and other distortions that are similar to those that occur naturally in the recorded music anyway. However, for loudspeakers, if the source of distortion is from a vibrating side panel, the distortion may be strongest at 90 degrees off axis. That's a sound field that cannot be duplicated by any distortions of the source signal. My guess is off-axis distortion is what causes some loudspeakers not to "disappear" and to sound "small", even when they have excellent on and off axis frequency response measurements.
One example is a speaker I've heard that had near perfect spinorama measurements but sounded chesty in the lower midrange, resulting in a somewhat small and colored sound. Standing next to the speaker and hearing the sound emanating from the flimsy side panel made me think that was a likely cause.
Distortion from enclosure resonances is a concern, but it's a concern whose magnitude isn't well characterized IMO. John Atkinson of Stereophile likes to measure them with an accelerometer, but that doesn't tell you whether a vibrating cabinet panel produces enough acoustic output to be audible.
But I appreciate the fact that he measures the port. Sometimes, the port noise and resonances are only 10-20 dB below the direct output!
Here's an example: https://www.stereophile.com/content/...r-measurements
Even though the port is facing the rear, that energy is going to reflect off the wall and reach your ears. Some of the resonant energy is also leaking out the front through the driver, which (I think) explains the jaggies that Soundstage measured under anechoic conditions in that same frequency range:
I also think that explains why JA couldn't get rid of the midrange congestion he noticed by loading the cabinet. He was incorrectly blaming it on a cabinet resonance when the real problem is the port.
I think distortion from the drivers comes into play mainly at higher levels, e.g. reference level HT, outdoor, or sound reinforcement applications. I've measured harmonic and IM distortion of some of the "good" hi-fi loudspeakers that I've owned, and at comfortable home listening levels (75-85 dB SPL at the listening position) the distortion products are usually quite low. One of my first experiments with it is described in another forum: https://db.audioasylum.com/mhtml/m.h...2Bmeasurements
I've since measured others, and achieved similar results, with distortion products down in the -50 to -60 dB range at the listening levels I prefer.
You can also tell a lot about how a speaker will work and sound by looking at the impedance plots. That can be a topic for another day. But in general, I agree with you that spinorama doesn't tell the whole story.