What matters most in speaker design?

It's been reported that John Dunlavy thought step response was what mattered most in speaker design.

>11-16-11: Cdc
>I originally thought flat frequency response was a primary goal. And it is, IF damage done to achieve this is minimal.

>BUT having listened to single driver speakers with no x-over and then adding different tone controls, [DEQ24/96, Parasound Z-pre, 3-band tone control] to flatten the response,

1) Unless you were doing some form of reflection free measurement (anechoic chamber, speakers buried in the ground and measured in half space, or a gated measurement of some sort most likely using Maximum Length Sequences and software like ARTA) you were adjusting for flat power response which you don't want instead of flat on-axis response with smooth off-axis curves which you do.

2) Unless you were using a calibrated measurement microphone you weren't adjusting for flat response there either.

3) Toole and Olive have found that listeners prefer flat on-axis response with smooth off-axis curves regardless of nationality and preferred musical genres. Olive turned that research into a speaker ranking formula which correlates well with subjective listening tests.

>11-16-11: Drew_eckhardt
>1) Unless you were doing some form of reflection free measurement (anechoic chamber, speakers buried in the ground and measured in half space, or a gated measurement of some sort most likely using Maximum Length Sequences and software like ARTA) you were adjusting for flat power response which you don't want instead of flat on-axis response with smooth off-axis curves which you do.

Even then it's not enough. It's possible to have an on-axis dip with a power-response peak where it's better to cut power response at the expense of degraded on-axis response.

Drew, That is interesting. I thought flat power response was the goal:
Earle Geddes - waveguide loudspeakers

I was using a calibrated mic and measuring the F-R 3 feet from the driver. This is very different from what I measured where I sit. Anechoic is nice for theoretical design but in-room (like JA's Stereophile measurements) can be a whole other animal. So what is correct in the reality of the listening room?

You can analyze it to death and if you want to be picky enough - angels dance on the head of a pin -, never come to the "ultimate" solution. This is a legitimate endeavor but I just want something useful for my 'umble setup.

I am kicking around Seas W15LY001

As you can see, at 30 degrees off-axis, it's a pretty flat response. Just add a supertweeter + L-pad above 8kHz.

11-21-11: Cdc
>Drew, That is interesting. I thought flat power response was the goal: Earle Geddes - waveguide loudspeakers

Earl's speakers don't have flat power response. They're omnidirectional at low frequencies, narrow gradually as the mid-bass becomes acoustically large with declining power response, have fairly constant directivity and power response for around a decade, and then narrow with decreasing power response. With flat power response they'd sound unnaturally bright. He has nomographs on his web-site plus a Windows program which can display them for various speaker data.

If you haven't yet you need to read _Sound Reproduction: The Acoustics and Psychoacoustics of Loudspeakers and Rooms_ by Floyd Toole and peruse Sean Olive's blog (they're both members of the Harman International research group with a financial interest in figuring out what people like in speakers) which get you to flat on the listening axis (Earl does that) with smooth off-axis curves decreasing monotonically with frequency (Earl does that too).

Most speakers lack smooth monotonic off-axis response and therefore don't sound natural. Conventional cone and dome two-ways have directivity mismatches between mid-woofer and tweeter. Panel speakers are acoustically large at high frequencies which produces a multi-lobed mess. Large full-range drivers are ill-mannered.

The ones which get that right sound surprisingly similar within their passbands and output level limits in spite of radically different designs. Examples include wave guide tweeters crossed to large mid-woofers where directivity matches, some coaxials, the Plutos where the drivers are acoustically small with perhaps 1/4 wave length between acoustic centers at the cross-over point, the RAAL Requisite Eternities with uniform horizontal off-axis response and mid-range array that should have vertical polar response which matches well to the ribbon tweeters, and small full-range drivers.

>I was using a calibrated mic and measuring the F-R 3 feet from the driver.

That doesn't correlate well with perception without gated measurements (like MLS signals via ARTA or MLSSA) because about half the power is coming from the reverberant field without consideration for the time domain although our brains seem to perceive timbre based on the spectra from the original sound and what they identify as reflections.

>Anechoic is nice for theoretical design but in-room (like JA's Stereophile measurements) can be a whole other animal. So what is correct in the reality of the listening room?

Flat on-axis with smooth monotonically decreasing off-axis curves where there's some latitude to balance too much output in one area with less elsewhere. That makes reflections which sound like the source probable regardless of what angle they're coming off the speaker and what they bounce off of.

I'd speculate it comes from evolution - we needed to identify predator + prey animals and their locations in spite of reflections off objects and the ground plane where outdoor environments are more absorbtive and diffusive at higher frequencies.

Siegfried Linkwitz put a shallow (about 2dB/decade) shelving low-pass filter on current Orion incarnations which sounds less bright and more natural, with an offered explanation being that the slope closely compliments HRTF differences between a real source straight ahead and sounds arriving at angles from a pair of speakers although polar response broadens going from the mid-range to tweeters and compensation for that is a plausible explanation consistent with John Krevosky's experiments (he built Orions with the original transfer function and perceptually matched better behaved speakers with acoustically small dipole tweeters to them with a high frequency boost).

At Burning Amp 2010 he said that his Plutos were no longer flat on-axis either although I didn't inquire for details or measure mine.

>I am kicking around Seas W15LY001 As you can see, at 30 degrees off-axis, it's a pretty flat response. Just add a supertweeter + L-pad above 8kHz.

Behavior at 30 degrees isn't enough. For example, in one of my listening rooms I looked at the first reflection angles as a thought exercise and noted that the side-wall reflection came from the sound 37 degrees off-axis which became 67 degrees off-axis with toe-in.