The Preference for Direct Radiators

Mitch asked a question in another thread which I think is the most interesting debate in audio right now:

In a nutshell, if two loudspeakers measure the same, why do people have a subjective preference for direct radiators?

Here's some background on this question, from my own perspective:

Seventeen years ago, I worked for one of those failing dot com companies. We didn't have any business, so I would show up to work and study patents all day long. It was a way to kill time and appear to be busy. Just sit at my desk staring at a technical doc for eight hours that looked like engineering work. That's when I started learning about horns, Tom Danley's stuff, Keele's stuff, etc.

I've built a ton of speakers over the years, nearly all with a horn or waveguide, but there's a consistent problem with nearly every one:

They can sound great with great recordings, but with bad recordings they sound mono.

Here's an example of what I mean:

IMG_0661.JPG


A few years back, I rented some Danley SH50s. They have a fairly flat frequency response and they're really well behaved in the phase domain.

vandersteen-3A_3.jpg


My reference speakers are Vandersteen. They have a fairly flat frequency response and they're really well behaved in the phase domain.

The Danley speakers cost about 20X as much as the Vandersteens. You can find the latter on Craigslist for $300, easily.

On an excellent recording, the SH50s were breathtaking. For instance, you can turn the lights off and put on a track with a really great soundstage, and the SH50 is one of those speakers where the image is almost holographic. Where the center is so solid it sounds like you have a center channel, and you can almost get a sense of "front to back" depth. The dynamics are far beyond anything you would ever need in a home.

The Vandersteen can't reach those highs, the stage is never as well defined and the dynamic limits are nowhere near as high.

But here's the crazy part - on most recordings, the Vandersteen's image sounds bigger than the SH50.

This is particularly noticeable if you listen to crummy recordings. I basically listen to podcasts, EDM, some 80s music, and some punk rock. I don't listen to jazz, or orchestral, or classical.

My 'hunch' is that the much wider directivity of the Vandersteens is 'lighting up' the room in a way that the SH50s can't. Basically the Vandy's can't extract all the information in a really good recording the way that the SH50s can, but the Vandy's are also 'glossing over' the bad qualities of many recordings.

I've built a lot of Unity horns for my car, and ran into a similar issue:

There were some recordings that sounded bigger than the car itself, but there were also many recordings where it sounded like I had a mono speaker sitting in the center of my dash. It's like buying a 2160P HDTV, then realizing that a lot of your favorite movies were filmed with a potato.
 
Here's some data, to support my argument that people prefer direct radiators:

1220682-revel-salon-2.jpg

1) Possibly the best example is this: at AVS Forum, two speakers which both share nearly identical response were evaluated. Despite one having more headroom, the direct radiator was preferred. The speakers were Revel Salon versus JBL M2, both Harman products and designed with the same target criteria : Speaker Shootout - two of the most accurate and well reviewed speakers ever made - AVS Forum | Home Theater Discussions And Reviews

An externally hosted image should be here but it was not working when we last tested it.

2) I did my own dumb version of the above test. At the CES this year, I listened to the JBL 4367 back-to-back with a number of Revel speakers, including the Salon. My conclusion was that the Revel speakers have dynamics that are surprisingly competitive with the 4367, while creating a soundstage that was bigger and more diffuse. Similar to the SH50 versus Vandersteen thing, the direct radiator speaker was not as 'pinpoint' but the additional reflections in the room created a euphonic and pleasant ambience. If cost was no factor, I might consider the JBL, but the inexpensive Revel is a third of the price.


DSC03693 by Gary Eickmeier, on Flickr
3) Eight years ago, the Linkwitz speakers were beaten by Behringer. I wound up buying a set, and relegated my Vandersteens to the den. Possibly the weirdest thing about this thread is that a DIY clone of a Bose(!) speaker beat them both! Having a speaker that radiates most of it's sound into the back wall is taking this "reflections" thing to the extreme, no doubt. But the results don't lie - it won. The designer of the speaker has been arguing for this type of design for well over 25 years, his posts predate the world wide web itself. (Gary was active on Usenet back before Mosaic was A Thing.)

620046d1496829365-great-balls-prestige-img_20170607_104843.jpg

4) Bushmeister did a showdown between his direct radiators and his Synergy Horns, and the former won. Note that Bushmeister and Revel are both using a giant roundover on their cabinets - I think this plays a big role here, I'll get to that... Comparison of 'Xbush Sphere synergy horn', versus 'Balls of Prestige'.
 
I think that most of it is room/speaker synergy.

In my experience:

-larger rooms "like" wider directivity, much diffused sound that makes the ambient look bigger
-small rooms work best with narrower directivity because most of the reflections are early and are detrimental to sound.

But this is the problem by definition. Narrower directivity (90-60 deg and lower) is only doable by horns and OB. OB demands very specific positioning in rooms (almost impossible in small rooms) and horns tend to be very large to control directivity through the midrange so they don't fit in small rooms also.

Comparing some (small by size) horn-loaded loudspeakers and direct radiators in a small room (25sq m or less) could be interesting and i think it would yield different results.

Say Usher BE718 and something like JBL LSR708i
 
Last edited:
If memory serves me:

1) Earl Geddes was using JBL speakers for a while

2) He designed the Summas to address deficiencies in the JBLs

3) I bought a set (I'm certain of that part lol)

4) During an ABX comparison between the JBLs, the Summas, and a set of Gradient Revolutions, the JBLs came in last, the Summas second, and the Gradients were first.

I can't find the original article. Here's a pic of the comparison, the Gradients are the small black speakers:

geddes2d.jpg


image.php

Are these the same Gradients? Not sure.

Earl discusses directivity index, and further improvements to the Summa:

YouTube
 
I think that most of it is room/speaker synergy.

In my experience:

-larger rooms "like" wider directivity, much diffused sound that makes the ambient look bigger
-small rooms work best with narrower directivity because most of the reflections are early and are detrimental to sound.

But this is the problem by definition. Narrower directivity (90-60 deg and lower) is only doable by horns and OB. OB demands very specific positioning in rooms (almost impossible in small rooms) and horns tend to be very large to control directivity through the midrange so they don't fit in small rooms also.

Comparing some (small by size) horn-loaded loudspeakers and direct radiators in a small room (25sq m or less) could be interesting and i think it would yield different results.

Say Usher BE718 and something like JBL LSR708i

That makes a lot of sense:

Back when I was a bachelor, I tried my Summas all over the house. Everywhere from the living room to the TV room to the main floor. At one point I had a crazy looking set up, where these giant loudspeakers were about three feet from my head.

An externally hosted image should be here but it was not working when we last tested it.

Similar to this. Looks silly, images great.

When I put the speakers in the main room, with a much longer distance from listener to speaker, the treble would sound "smoother" but the imaging would suffer.

Obviously, all of this is subjective, but to me the Vandersteens sound more "euphonic" when listened to at a distance. There's additional ambience that seems to be generated by the much wider directivity.
 
After listening to some other peoples systems I was findind I preferred the higher frequencies from theirs to my synergy inspired waveguide spkrs. The image solidity and clarity from mine was much better but lacked the shimmer and (I think) the effect of sound arriving from other directions (side and behind) that happens with actual live music. I recently added rear "ambience tweeters" aimed to reflect off the walls and ceiling behind the main speakers and fed from a ~20msec delay. That preserves the solid image while livening up the overall sound, I think that brings the sound liveness up with those other systems (have to go hear those again to be sure)
 
It would be an interesting test to copy the directivity of a waveguide speaker system using direct radiators to see if they become subjectively very similar at that point. If someone wanted to pick something (A summa?) and commit to building a variant, I could help in the design process of that variant.
 
Through the years Ive noticed that wide dispersion seems to always have better WAF. Sense they can actually hear I conclude the obvious ;)

My only experience with wide and narrow dispersion, yes I am recklessly assuming here a similarity between direct and guided, is with large single panel planers vs multi driver ribbon /planer and ribbon / piston systems. IMO the wide dispersion stuff always sounds, well, more "pleasant", less irritating in any major way. The narrower dispersion stuff can often sound more present, more dynamic, and image solidity can be outstanding, BUT over time I find I miss the sound of wider dispersion.
 
My 'hunch' is that the much wider directivity of the Vandersteens is 'lighting up' the room in a way that the SH50s can't. Basically the Vandy's can't extract all the information in a really good recording the way that the SH50s can, but the Vandy's are also 'glossing over' the bad qualities of many recordings.
I'd still rather have my Unitys and I listen to a vast range of genres except rap and some EDM.
 
It would be an interesting test to copy the directivity of a waveguide speaker system using direct radiators to see if they become subjectively very similar at that point. If someone wanted to pick something (A summa?) and commit to building a variant, I could help in the design process of that variant.

As I understand it, Harman has copied the power response and used it over and over and over again.

So it gives us some interesting data; if the power response of their speakers is nearly always the same, then why do the speakers sound different?

Some data:

prelude%20pfr_small.jpg


912IPFRfig2.jpg


Infinity PFR, designed by Andrew Jones, anechoic response on tweeter axis at 50", averaged across 30° horizontal window (1996)

Revel-Salon2-front-1024x680.jpg


708Revfig04.jpg


Revel Ultima Salon2, designed by Kevin Voecks, anechoic response without grille on listening axis at 50", averaged across 30° horizontal window (2008)

An externally hosted image should be here but it was not working when we last tested it.


510JBLfig4.jpg


JBL 1400 Array, designed by Greg Timbers, anechoic response on tweeter axis at 50", averaged across 30° horizontal window

DSC05175.JPG

g7yzUbpIPu2pA4EJ_LLsBRFuf-vX4xM82yw_qCMncgbfx56rA2C1SoCLxIU6SC87uOMgdz0AYON4njZsyn2e6ju48acGvtx8VFsjIUBl-J8kcJzv2rTtdBDdLxUrxl4v23BVObNs=s800

JBL LSR305 polars from JBL
JBL_LSR_305_Random.jpg

JBL LSR305, frequency response measurement from noaudiophile.com


If you look at the JBL, Infinity and Revel measurements, going back to the 90s, a clear pattern emerges:

1) directivity control on the tweeter

2) flat frequency response

3) The flagship models often have an exaggerated bass

The performance of the $129 JBL LSR 305 is particularly astonishing; above 300hz, it's measured performance isn't much different than speakers that cost almost 100 times as much! (WTF)
 
Last edited:
As I understand it, Harman has copied the power response and used it over and over and over again...
1) directivity control on the tweeter
2) flat frequency response
3) The flagship models often have an exaggerated bass
(WTF)

Patrick, I don't think that's quite right.

I have attached the image from page 399 of the previous edition of Floyd Toole's Sound Reproduction. I'm virtually positive that Speaker A is the Revel Salon 2 and Speaker B is the JBL Array 1400. The bottom dotted curves show the power response. I believe that it would be accurate to say that Harman seems to aim for a similar listening window curve for their speakers. Stereophile's measurements usually include a sentence like "the broad rise in output in the upper bass will be mainly due to the nearfield measurement technique" (from Salon 2).
 

Attachments

  • Page 399.png
    Page 399.png
    38.6 KB · Views: 1,987
I think there is way too much going on, in terms of room interaction complications, to try to make a case for direct radiator preference.

A 'for instance'.... before I'd try to compare two different radiation patterns, I would want a room with an even rt60 throughout the spectrum.
And I'd want a fairly symmetrical reflection/absorption pattern spread evenly around the room.
Otherwise I'm just hearing what given room anomalies suit which particular speaker....

And even with a "perfect room" for evaluation, I still think preference would probably come down to 'what song sounds best on what system'. Horses for courses.

A Danley technique I always find fascinating...listening to a system through headphones, from a measurement mic set up at listening position...
It really helps hear what the room is doing, and the interplay with radiation patterns.
Try it with stereo, and then each side mono....often big differences..
 
Patrick, I don't think that's quite right.

I have attached the image from page 399 of the previous edition of Floyd Toole's Sound Reproduction. I'm virtually positive that Speaker A is the Revel Salon 2 and Speaker B is the JBL Array 1400. The bottom dotted curves show the power response. I believe that it would be accurate to say that Harman seems to aim for a similar listening window curve for their speakers. Stereophile's measurements usually include a sentence like "the broad rise in output in the upper bass will be mainly due to the nearfield measurement technique" (from Salon 2).

Interesting, thank you!

According to Stereophile:

"When JBL's chief systems designer, Greg Timbers—see —applied vertical horns to the design of speakers for the home, he found that they increased the three-dimensionality of the speakers' soundstaging. However, as Japanese audio critics dislike the appearance of vertical horns, the horns were arrayed horizontally in the flagship K2 models. But, as JBL's Synthesis line was designed for the US market, Timbers was free to stand the midrange horn up again: it stands atop the 1400 Array BG, though the tweeter is still loaded by a horizontal horn."

attachment.php


You can see the effect in the measurement you posted; there's a 'blip' in the directivity index at 8khz.

If I recall correctly, the midrange compression driver in the JBL is around 2" in diameter. So that 'blip' happens as the midrange compression driver loses pattern control due to the width of the compression driver throat.


I have a hunch that the loss of "air" isn't just due to the directivity index, but may also be related to the width of the cabinet. For instance, I listened to the Summas back-to-back with the Danley SH50s, and though their directivity index differs, their baffle width is similar. The two speakers sound completely different, so clearly the crossover and cabinet construction is playing a part here too. And to clarify that comment - the "air" of the two speakers is similar, but the soundstage sounds different, and percussion sounds different. IMHO, percussion is a litmus test for phase coherent speakers. My Yamahas use FIR filters to fix the phase, and percussion is noticeably clear on them.

It's too bad I never listened to Sheldon's Unities back-to-back with my Summas; IIRC, both use the same compression driver and a similar woofer, so it might be a closer comparison than the SH50 vs Summa.

An externally hosted image should be here but it was not working when we last tested it.


Obviously, narrow cabinets have long been known to image well.
 
Last edited:
I have read Toole and Linkwitz and lately the white paper on The Beo 90. I built some nearly omni directional speakers, some narrow baffle and wide baffle speakers for comparison. All to conclude that wide dispersion pattern speakers, or those tending towards omni pattern, bounce lots of sound off your room walls and those reflections give the listener lots of information to really precisely locate the speakers in the room. So it creates a new performance into the room. Really dead, no reverb, dual mono stuff like a few early Beatles, sounds like the performer is sitting there where the speaker is. If there is recorded ambient information, it gets mixed with and confused with the room ambient information. The Beo paper says the image exists almost on a line between the speakers for omnis. Narrow controlled dispersion speakers, however the radiation pattern is achieved, minimize room interaction and sort of open a window into the original performance space, as the recorded ambient sound can dominate. So the performance seems to be back through the window. So depending on what you are listening to, it would be great to have both wide and narrow available.
 
I long ago came to the conclusion that sofit mounting or in wall speakers are the way forward if you can have them.

In room,..... I strongly agree...and it seems many studios do too.

best sound of all though, is outdoors...
And another Danley technique, for finding best sound there....
generation-loss recordings ....highly recommended...
see how many recordings the speaker can survive outdoors,
then move it inside ;)