As you know I've been involved in the directivity issue since the 4430 days and my opinions on the importance of constant directivity have vacilated over time.
Its good to point out that we are really talkng about smooth directivity or directivity that is constant over a partial range. Conventional speakers do not have uniform directivty over 10 Octaves.
We also would be better off talking about polar uniformity. There are a number of units out there with constant directivity and very nonuniform polars. If they beam in one plane they broaden in the other. I think "constant directivity" was a catch phrase for marketing and it generally emphasizes the wrong aspect. Good drivers can have polar uniformity or at least slowly changing polars, and that is a more rigorous requirment than the 1 dimensional directivity curve.
When we talk about directivity, more than anything, we are just setting the relationship between the speaker's power response and its axial response. If you know any 2 curves you also know the third: axial response curve - directivity index curve = power response curve. This gets you directly to the fundamental question of loudspeaker design: what response do we optimize? Anechoic response? Power response? Room response?
Opinions alone are not the answer here, we need to rely on listening tests, many of which have been done. I like to look at the early tests of Toole where he rank ordered several dozen speakers and then compared all the measurable parameters of them. My conclusion is that there is a strong correlation between axial response and listener preference and a very poor correlation with power response.
This alone makes you wonder if directivity matters at all?
Lipshitz and Vanderkooy did an interesting study where they independantly adjusted axial response and power response. The could make power response flat, axial response flat, power response roll down hill with flat axial. Axial rise with flat power. The only natural combination was with flat axial response and a falling power response (rising directivity). Flat axial with flat power response was too bright. Flat power response achieved by rising axial response was way too bright.
They even found that holes in power response were fairly innocuous.
If you look at the Soren Bech studies then he goes into the audibility of reflections, primarily based on their direction of arrival. He finds most reflections are fairly "invisible" at least at the levels you might expect to naturally occur in a home listening room. One exception was the floor bounce that could give problems up into the midrange and the rear wall bounce. The reduced side wall energy that a "more directivity" speaker would have didn't seem to buy you much. (If the reflections are inaudible, then lowering them isn't beneficial.)
When you talk to audiophiles you run into the attitude that "I can hear everything" and "everything is important". Based, I think, on this there is a growing consensous that directivity is important and maybe part of the magic between the great and the merely good.
I'm not so sure.
I do like having smooth polars and a response that doesn't vary with listener position. The ability to have several listeners on a couch and the same initial arrival response for all. I also know that people will look at room curves and want to equalize them. This is a big can of worms that gets you somewhere between axial response and power response. Having a smooth directivity function may help in that regards: it won't drag response away from the axial or anechoic response to a totally different power response curve.
Smooth polar curves are also generally an indicator of low reflections. Thats a good thing too. But is there a requirement of constant directivity for a speaker to sound good? Can anybody give evidence?
David S.
Its good to point out that we are really talkng about smooth directivity or directivity that is constant over a partial range. Conventional speakers do not have uniform directivty over 10 Octaves.
We also would be better off talking about polar uniformity. There are a number of units out there with constant directivity and very nonuniform polars. If they beam in one plane they broaden in the other. I think "constant directivity" was a catch phrase for marketing and it generally emphasizes the wrong aspect. Good drivers can have polar uniformity or at least slowly changing polars, and that is a more rigorous requirment than the 1 dimensional directivity curve.
When we talk about directivity, more than anything, we are just setting the relationship between the speaker's power response and its axial response. If you know any 2 curves you also know the third: axial response curve - directivity index curve = power response curve. This gets you directly to the fundamental question of loudspeaker design: what response do we optimize? Anechoic response? Power response? Room response?
Opinions alone are not the answer here, we need to rely on listening tests, many of which have been done. I like to look at the early tests of Toole where he rank ordered several dozen speakers and then compared all the measurable parameters of them. My conclusion is that there is a strong correlation between axial response and listener preference and a very poor correlation with power response.
This alone makes you wonder if directivity matters at all?
Lipshitz and Vanderkooy did an interesting study where they independantly adjusted axial response and power response. The could make power response flat, axial response flat, power response roll down hill with flat axial. Axial rise with flat power. The only natural combination was with flat axial response and a falling power response (rising directivity). Flat axial with flat power response was too bright. Flat power response achieved by rising axial response was way too bright.
They even found that holes in power response were fairly innocuous.
If you look at the Soren Bech studies then he goes into the audibility of reflections, primarily based on their direction of arrival. He finds most reflections are fairly "invisible" at least at the levels you might expect to naturally occur in a home listening room. One exception was the floor bounce that could give problems up into the midrange and the rear wall bounce. The reduced side wall energy that a "more directivity" speaker would have didn't seem to buy you much. (If the reflections are inaudible, then lowering them isn't beneficial.)
When you talk to audiophiles you run into the attitude that "I can hear everything" and "everything is important". Based, I think, on this there is a growing consensous that directivity is important and maybe part of the magic between the great and the merely good.
I'm not so sure.
I do like having smooth polars and a response that doesn't vary with listener position. The ability to have several listeners on a couch and the same initial arrival response for all. I also know that people will look at room curves and want to equalize them. This is a big can of worms that gets you somewhere between axial response and power response. Having a smooth directivity function may help in that regards: it won't drag response away from the axial or anechoic response to a totally different power response curve.
Smooth polar curves are also generally an indicator of low reflections. Thats a good thing too. But is there a requirement of constant directivity for a speaker to sound good? Can anybody give evidence?
David S.
Did they also test flat axial and flat power response equalized to different house curves? Now this would have been interesting.
Reflections clearly aren't "invisible" at any reasonable level. I suppose he must be talking about some particular effect of reflections? It's awfully easy to disprove the notion if "hearing" a reflection can involve any difference it causes. A room with one glass side wall and all the other walls covered in 10cm of insulation, for example, or a mirror placed next to one speaker.
"Equalized to a house curve" of course wouldn't be flat. They did't try to control the room curve but to turn the speaker into a system with independently variable power response and axial response to see if any particular combination was better or worse than others.
The experiment was to recreate a typical living room system within an anechoic chamber. A large number of speakers surounding the test subjects were set to mimic the arrival direction, strength and response of all the initial room reflections. Based on that setup they could vary the strength of any particular reflection. At some level every reflection would be audible as either a directional shift, a response change or an echo. These levels of audibility were compared to the real levels found in a typical room.
For most of the reflections the real life levels were found to be lower than the level of detection. In other words you could turn off a number of simulated reflections and not hear a difference. The exceptions were the floor bounce in front of the speaker and the rear wall behind.
How would higher directivity help us? If we found that early side wall bounces were typically audible then a more directional speaker might just get those bounces to fall below the threshold of detection. This was not found to be the case.
If higher directivity isn't helpful with regard to particular reflections, and late power response spectrum shape is a low level effect (relative to the early sound) then how important is speaker directivity?
David S.
When considering the original question posed by Wayne, I think it only natural to consider the room and application BEFORE generalizing what performance parameters to prioritize. CD may not take top priority in a large space or well damped/treated....or in the nearfield. I know my desktop mains have a significant discontinuity here at and below 2khz but I never miss it at such close proximity.
Would be to your ears.
In my mind if we leave out the physical impact the room imposes on our hearing we are only half way there. Not to say there is anything wrong with the body of research. At first glace would appear to agree with my experience. Would also appear to be a good jumping off point to further refine our theory(s).
Should we start a checklist of sorts? A "Start here, follow these rules and it'll be magical, lists. LOL
1. Proper wavefront generation
2. Proper wavefront propagation
3. linear phase crossover (active or passive) +/-1 or more octaves
4. smooth natural on and off axis response spectraly balanced
etc.
not necessarily in that order or inclusive of those mentioned.
Thanks for chiming in, Dave. I see you as one of the grandfathers of the matched-directivity two-way design approach, so your input here means a lot.
Mayhem, I generally agree with you about needing to define the room before talking about what kinds of directivity matter. I am most interested in small rooms, those typically found in homes. Large rooms are easy to make sound good, just like outdoors. But indoors, we have all these tricky boundary reflections to deal with.
I wouldn't say my opinions on the importance of constant directivity have changed, but I would say that my position compared to other audiophiles has changed. Like I said in the first post in this thread, not so long ago, directivity wasn't even on the list of concerns for any audiophile I knew of. In that environment, I probably came across as being obsessed with directivity.
But I wasn't really - I chose not to use any of the horns with sharp edges in them because they sounded spitty to me. Instead, I always used horns that traded a little bit of directivity for smoother response. The radials I chose had pretty constant directivity in the horizontal, but they had collapsing directivity in the vertical. And I still think that's fine.
Now that waveguides have become the rage, this makes me a little bit on the side of saying be careful not to trade everything for polars. You can find a constant directivity horn from the 1970s that has really constant beamwidth, but there are a whole lot of anomalies in the trade. So my position now - while not having changed - still prefers uniform directivity but not at the cost of sound quality in other respects.
As for uniform directivity loudspeakers, my favorite design type has always been constant directivity cornerhorns. But they're somewhat uncommon, really, because most rooms aren't well suited for that kind of setup. So when I saw your 4430 design, I immediately embraced that approach, the matched-directivity two-way design.
In hindsight, I think the reason the constant directivity cornerhorn sounds so good to me is it eliminates any self-interference from reflections off the (front) wall that would normally be behind the speakers. I must admit to thinking it helps that it eliminates self-interference from the ipsilateral side wall too. Maybe not - Maybe it's just the removal of the front wall reflection that's most important. But I would think that if the distance to the ipsilateral wall is in a certain range, it could be as big a problem as the front wall. My guess is if it's really far, it doesn't matter. Or if it's really close (like flush, within 1/4λ), then it also doesn't matter. But in between, I would think the ipsilateral wall could make problems something like the front wall does.
Back to the matched-directivity two-way design approach, I should clarify something, briefly. I've had a bad habit of using the phrases "DI-matched" and "directivity-matched" interchangeably. When I say either one, what I mean is matched-directivity, not matched-index. Specifically, I'm talking about matched in the horizontal. I tend to like an approach of matching the horizontals using the crossover point and the horn's horizontal wall angle and setting the vertical pattern using the crossover to place the vertical null angles just outside the vertical wall angle.
Extra points if the horn's vertical beam is controlled down to the crossover point, but this is rare, in practice. Seems like the verticals are usually about equal to the horizontals at the crossover point - starting to do the pattern flip thing - and the nulls cut into that. Larger horns would push the vertical control point lower in frequency, but would also increase the distance between adjacent drivers. So like all other things, the size becomes a balance of competing priorities.
As you might guess from my infatuation with the constant directivity cornerhorn approach, I definitely agree that the bounce off the wall behind the speakers is the most objectionable, and that floor bounce might also be. This midrange anomaly is fully audible, and room treatments are ineffective. A soffit mounted speaker with baffle flush to the wall prevents it, as does the constant directivity cornerhorn configuration. But short of that, I kind of like using helper woofers or flanking subs to smooth those notches, sort of like a truncated array. Where self-interference from one source creates a deep notch, another source in another position fills it in. Not as good as flush soffit mounting or cornerhorns, but definitely worthwhile for stand-mounted speakers as are commonly employed.
Mayhem, I generally agree with you about needing to define the room before talking about what kinds of directivity matter. I am most interested in small rooms, those typically found in homes. Large rooms are easy to make sound good, just like outdoors. But indoors, we have all these tricky boundary reflections to deal with.
I wouldn't say my opinions on the importance of constant directivity have changed, but I would say that my position compared to other audiophiles has changed. Like I said in the first post in this thread, not so long ago, directivity wasn't even on the list of concerns for any audiophile I knew of. In that environment, I probably came across as being obsessed with directivity.
But I wasn't really - I chose not to use any of the horns with sharp edges in them because they sounded spitty to me. Instead, I always used horns that traded a little bit of directivity for smoother response. The radials I chose had pretty constant directivity in the horizontal, but they had collapsing directivity in the vertical. And I still think that's fine.
Now that waveguides have become the rage, this makes me a little bit on the side of saying be careful not to trade everything for polars. You can find a constant directivity horn from the 1970s that has really constant beamwidth, but there are a whole lot of anomalies in the trade. So my position now - while not having changed - still prefers uniform directivity but not at the cost of sound quality in other respects.
As for uniform directivity loudspeakers, my favorite design type has always been constant directivity cornerhorns. But they're somewhat uncommon, really, because most rooms aren't well suited for that kind of setup. So when I saw your 4430 design, I immediately embraced that approach, the matched-directivity two-way design.
In hindsight, I think the reason the constant directivity cornerhorn sounds so good to me is it eliminates any self-interference from reflections off the (front) wall that would normally be behind the speakers. I must admit to thinking it helps that it eliminates self-interference from the ipsilateral side wall too. Maybe not - Maybe it's just the removal of the front wall reflection that's most important. But I would think that if the distance to the ipsilateral wall is in a certain range, it could be as big a problem as the front wall. My guess is if it's really far, it doesn't matter. Or if it's really close (like flush, within 1/4λ), then it also doesn't matter. But in between, I would think the ipsilateral wall could make problems something like the front wall does.
Back to the matched-directivity two-way design approach, I should clarify something, briefly. I've had a bad habit of using the phrases "DI-matched" and "directivity-matched" interchangeably. When I say either one, what I mean is matched-directivity, not matched-index. Specifically, I'm talking about matched in the horizontal. I tend to like an approach of matching the horizontals using the crossover point and the horn's horizontal wall angle and setting the vertical pattern using the crossover to place the vertical null angles just outside the vertical wall angle.
Extra points if the horn's vertical beam is controlled down to the crossover point, but this is rare, in practice. Seems like the verticals are usually about equal to the horizontals at the crossover point - starting to do the pattern flip thing - and the nulls cut into that. Larger horns would push the vertical control point lower in frequency, but would also increase the distance between adjacent drivers. So like all other things, the size becomes a balance of competing priorities.
As you might guess from my infatuation with the constant directivity cornerhorn approach, I definitely agree that the bounce off the wall behind the speakers is the most objectionable, and that floor bounce might also be. This midrange anomaly is fully audible, and room treatments are ineffective. A soffit mounted speaker with baffle flush to the wall prevents it, as does the constant directivity cornerhorn configuration. But short of that, I kind of like using helper woofers or flanking subs to smooth those notches, sort of like a truncated array. Where self-interference from one source creates a deep notch, another source in another position fills it in. Not as good as flush soffit mounting or cornerhorns, but definitely worthwhile for stand-mounted speakers as are commonly employed.
I would be shocked if floor bounce is among the most objectionable, frankly. Not saying it isn't, necessarily, just saying I would be a bit shocked. I know in cases where you end up with an especially bad lower midrange / midbass cancellation, you can hear it pretty well and it's not good, but it also seems less bad than it measures, and (at home) it's often in the range where vertical room modes are also getting in on the fun so that makes amateur analysis kinda difficult.
Anyway, my point is our brains grow up hearing everything bouncing off the ground/floor all the time, and we evolved with a ground too (and probably to at least some extent, a floor). It's the most normal reflection. Or, maybe it would be more accurate to say it's the expected (by the brain) reflection?
Anyway, my point is our brains grow up hearing everything bouncing off the ground/floor all the time, and we evolved with a ground too (and probably to at least some extent, a floor). It's the most normal reflection. Or, maybe it would be more accurate to say it's the expected (by the brain) reflection?
I think the self-interference notch from the wall behind the speakers is probably the worst offender in the lower midrange. The floor bounce notch can appear on the same frequency range, depending on speaker height and distance to the listeners though, and so mitigation of both can be achieved by the same mechanism, in many cases. And I agree with you that the vertical modes tend to start in this same low-midrange band, so all can be dealt with using helper woofers or flanking subs to good effect.
Another thing I hate is HF ceiling reflections. I call that ceiling slap because that's exactly what it sounds like. It makes a pinging sound, sort of like flutter echo, but it happens even in some rooms that are pretty well carpeted. It seems worse in rooms with gabled ceilings, probably because they reflect more into the center of the room. Speakers with tall vertical radiation sound really cluttered in rooms like that, sort of like a tinny radio in the bathroom. Speakers with narrow vertical HF beamwidth really help. In my experience, the worst part seems to center around 4kHz, so as long as the radiation pattern above ~2kHz prevents much energy at the angle of reflection back to the listeners, that particular problem is greatly reduced. It isn't a hard goal to achieve, but speakers that provide too tall vertical beamwidth in some rooms sound bad to me.
Thanks for the clarification. Seems to conflict with my experience... more than a little bit. How large of a room were they emulating? Did they try different spectral content for the reflections for determining the level of detection? Should I just go look this up somewhere?
Http://www.aes.org/e-lib/browse.cfm?elib=7673
He basically did two generations of the test. Initial tests used a simple cardioid model for the speaker and its off axis curves. A later test used actual measured curves for the response of each reflection's radiation angle. i.e., if a side wall reflection was caused by a ray 30 degrees off axis to the loudspeaker, the actual 30 degree response was used in the simulation.
He did find that where the more realistic model had higher directivity the threshold of detection for those reflections was higher (reflections less audible). Since the level of detection was already above the natural level it was of no consequence.
As reflection levels were varied, test subjects noticed differences in source direction, loudness and timbre (frequency response). As reflection levels were reduced, timbre changes were the last to disappear. Tests were done with speech and pink noise. Pink noise was more revealing.
The natural level for most reflections in his typical room were 10 to 15 dB below the direct sound. This was due to the off axis level of the speaker, the increased air travel distance and the loss at each boundary. Noise was more revealing than speech as a test signal, but still had detection levels 3 to 9 dB higher than the natural level, except for the first front floor bounce.
David S
I've heard this as a comb filter effect. Move nearer and farther from the speaker and the pitch changes. You can also hear significant treble bounce off of most thin carpet.
David S
You are still jumping to the unproven conclusion.
You will not measure its effect because it is a perceptual issue. You have to do what Soren Bech did, find a way to vary the off axis part only and then search for thresholds of detection or just noticable difference levels.
David S
I was most disappointed when I discovered this when with speakers toed toward the opposite walls at hotter than the speaker's listening axis, 2" of rockwool 3'x3' made only a slightly noticeable difference.
When you ask the question 'how important is speaker directivity?', I'm sure it has driven many to make inadvertent improvements including better baffling, keeping sound away from room clutter, even the change to compression tweeters.
It wouldn't be the first time a mass misconception with illogical reasoning has led to better speakers.
Lipshitz's conclusions:
"1) Flat power produced at the expense of non-flat axial response
produces a colored overbright and definitely 'wrong' sound in the
case of loudspeaker #1. The effect is worse in the smaller
listening room.
2) Flat power together with flat axial response is still colored and
overbright in the case of loudspeaker #1, but less so than in (1)
above. In the case of loudspeaker #2 no midrange coloration
results, only excess brightness.
3) The midrange coloration in the case of Experiment 3 appears to be
caused by the unbalanced reverberant power put into the room due to
the sway-back nature of EQ3. In principle, this should have exactly
compensated for the complementary power curve of the
forward-pointing 'monopole'. Clearly the spatial distribution of
the reverberant power must be important, and hence presumably the
loudspeaker's directivity index curve does affect the perceived
coloration.
4) If the directivity index curve is monotonic the impression of flat
power is purely one of excess brightness and some delocalization.
5) In all cases, flat power into the room appears undesirable. As
previously argued, rolled-off highs on axis are also undesirable,
and so we are led to conclude that for domestic use a loudspeaker
should have a smooth monotonically increasing directivity index as a
function of frequency. Precisely what the shape and ultimate rise
should be remain to be investigated. Probably the best that one can
do is to find a range of acceptable contours. Personal preferences
and listening room acoustics as well as loudspeaker positioning will
also affect the decision.
6) A flat power 'dipole' is less objectionable than a flat-power
'monopole'.
7) Adding spectrally neutral reverberant power to the room produces a
somewhat bright sound.
8) It follows that the frequency balance of the reverberant sound is
perceptually significant.
9) These effects are greater in smaller, acoustically more reflective
rooms.
10) A small power dip in the crossover region is relatively innocuous,
whereas a peak is audibly undesirable. Although our conclusion
referes to reverberant power peaks and dips, it is consistent with
Bücklein [7].
11) Consequently, the perceived tone color is a product of both the
direct spectral balance and that of the early reflections, and not
of the direct sound or of the reverberant sound alone.
12) Unbalancing both the direct and reverberant spectrum produces
colorations broadly similar to, but more pronounced than, spectrally
unbalancing only the reverberant sound.
13) A more directional loudspeaker is less affected by the room.
14) The greater the delocalization the less pronounced is the effect on
the perceived tone quality." (AES Conv. Paper 2301)
Unproven? Pretty much the same others have found.
If we can correlate reflection patterns to perception, we can "measure" perception. That's what psychoacoustics is all about.
I think it is somehow unproven, because what is the optimal power response, for example, for the speaker depends mostly on how the music recording was created and which kind of power response the original monitor speakers once had.
If they had falling power response then sure the same recording sounds bright with constant power response speakers.
- Elias
I have messed around a dipole speaker project with numerous measurements. For a sanity check I measured also my 2-way BR speaker, a variation of MarkK's ER18DXT. I can verify that in my conventional living room with RT around 0.3 - 0.5, The dipole spekers produce significantly more accurate and pleasant sound. Reflections, seen as spectral decay up to 300ms are markedly less with the dipoles (one speaker's measurements shown here.
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.