Beyond on axis response - how to properly consider off-axis contribution?

[fjhuerta]

So, it's late, I'm at work waiting for a backup to finish, and I got time to think. Lots of it.

I have two projects - one of them is basically finished (a simple 2 way with 4th order acoustical x-overs, nothing special), and a little 2 way I built using scrap parts I ended up with from project #1.

The little one (let's call it "Tails", since that's the name my son uses for it - it's a yellow cabinet...) is actually nothing special, either. It uses a spare set of C-CAM gold anodized (I believe) Monitor Audio 25mm tweeters. They really suck, now that I think about it. Its Fs is 1.6 KHz, the breakup is at 20 KHz or so, and the way it was used originally, it crossed over at 3 KHz (!!!) to an 6.5" metal cone mid (think about it). More details here: http://www.stereophile.com/content/m...r-measurements I used to have the RS8 speakers. Fun, party-like sound - but tiring in the long run. But I digress.

The woofer is the little Dayton ND105-4. They are really tiny things. So that got me thinking - I'd like to try as close to a source point as possible. I know - I should have gone all out and used an SB29, crossed over at 1.2~1.3 KHz, 6th order, and do it right, but I already had the MA tweeter. So I decided on a 2.8 KHz, 6th order, L-R crossover for Tails. The spacing is as tight as I could make it. The woofer and tweeter are pretty close (I'll have to upload pics later).

So, here's the real thing I've been wondering.

1) If my design is aimed at a flat on-axis response, I get a lovely flare at the point where the woofer starts becoming directional and the tweeter is still non-directional - ie, at 2.8 KHz. And the flare is sort of strong at 30°. Maybe not that bad, but I've found I really dislike any emphasis around that region.



2) If I take into consideration off axis response at 15° and 30° while designing the crossover, I end up with a slightly de-emphasized response on-axis from 2.5 to around 5 KHz on-axis, but off axis response looks far better at 30° (there's still a bit of a flare, though), and 15° is basically flat.

0°


15°


30°


So, here's what I'm thinking... please correct me if I'm wrong here.

1) Nobody ever listens to a speaker exactly on-axis, so there's no harm in deviating a bit from flat response there - what's perfect is not actually the best.

2) Stereophile does the whole +-30° average measurements to show FR because of this exact same thing - on axis response is not an accurate representation of a speaker behavior when taken alone.

3) Floyd Toole's school of thought (flat on axis response, smooth power response, etc.) can be achieved by carefully weighing the pros and cons of a perfectly flat on axis response vs. the actual, measured off axis response. If he had settled for a single thing (ie, flat on axis response), speaker design would be so much easier...

In short, I think I'll build the crossover with the smoothest off axis curves and the decent on axis response... any thoughts will be welcome!

[r-carpenter]

this is a very very very and I mean extremely broad and interesting subject for discussion. Harman has probably been researching it for the last 30 or 40 years.
http://www.amazon.com/Sound-Reproduc...5220284&sr=1-4

Where did I read that they are actually using additional loudspeakers in the anechoic chamber to model first and second reflections and it's interpretations.

The book is very helpful.

[fjhuerta]

this is a very very very and I mean extremely broad and interesting subject for discussion. Harman has probably been researching it for the last 30 or 40 years.
http://www.amazon.com/Sound-Reproduc...5220284&sr=1-4

Where did I read that they are actually using additional loudspeakers in the anechoic chamber to model first and second reflections and it's interpretations.

The book is very helpful.

Yeah, I've already read it. It radically changed my mind about reflections and power response. A must buy.

[Pete Schumacher ®]

This is why a waveguide may be preferred over a flat baffle, because you can have smooth/flat both on axis, and off. Below is a 1KHz LR4 XO with a 8.5" waveguide loaded dome. With your crossover frequency, a 5 or 6" guide crossing to a smaller woofer, say 5 or 6", would help keep the off axis (power response) more uniform.

With many tweeter flanges exceeding 4", it seems odd that almost no manufacturers include a guide, which would have the above stated benefits, as well as helping to align acoustic centers.

20, 45, 60 degrees, on axis is about 1dB higher than the red line.

[ErinH]

Keep in mind, I don't build speakers. I test raw drivers and love it. I put car audio systems together and tune like crazy but I've got extensive DSP and a stupid environment. Home audio just isn't my "thing" like it is here for so many. building speakers is different than measuring a single driver or actively tuning a stereo. That said, I love audio in general so strive to understand it in all avenues as the laws of physics doesn't change from one environment to the other; just how those laws are affected by the environment. So, I spend a lot of time reading (toole's book is friggin' awesome, but have you read Geddes' book on home theater design?... if not, it's a must read IMO).

Regarding #1, I don't agree. I think a lot of people actually do line their speakers up on-axis. It means you have a sweet spot and the surrounding areas will suffer, but some do this anyway. So, I wouldn't say no one listens to speakers on axis.

I like what you said in your last sentence and personally agree:
Go for a smooth off-axis response while not striving for a perfectly flat on-axis response. But, that should have some sort of note in it stating that an on-axis response with +/- xdB deviations is out of tolerance. I start to look at things like placement, yes, but also how you hear; namely ITD and ILD. With the research I've found, it seems 0.5dB is within the threshold of audibility when ILD was tested. So, that lends itself to some consideration on just how much deviation you can accept. That said, what I'm getting at is my goal would be for smooth when it makes sense. I'd rather take an overall decrease in SPL toward the high end than one with a more 'flat' response with audible bumps in response here and there. Of course, maybe it's easy to add a notch filter in where it's not easier to increase response via a shelf filter of sorts?

Then there's the room. Stupid f*cking room!!!!
Until you get the speaker in the room, you won't know the effect of the room itself on the response but, I wouldn't expect much trouble in the upper end response (ie: 5khz+) as much as I would lower midrange and especially mid/subbass. The goal should be to push the first reflection as far out as possible. In car audio there's this idea that you should sync the two; I've never found this to work as easily as keeping the driver further away from a reflecting surface. I've achieved better (and more dynamic, I might add) stage depth, and width by moving my tweeters from right next to the glass to approximately 2" away. Comb filtering sucks. I believe (don't quote me on this), Geddes' book says for spaciousness you want a high reverb to directivity ratio and I think the value for this was given as 10-20ms of delay in reflected sound in the lateral plane.


One thing I don't see much of when speaker designs are evaluated is off-axis data for the vertical axis. I realize that under a given frequency related to effective diameter, a driver is pistonic and therefore beaming has not yet occured so it should be that when measuring a speaker, a 30* measurement in one location would be the same as the other. But, when you introduce a speaker system in to the fold, does this still shake out the same? Your location changes the way the response is delivered; floor bounce, ceiling bounce, wall bounce... all that come in to play. so, I would think a speaker system could very well have differing results in the horizontal plane vs. vertical plane namely due to floor bounce given it's typically the first reflection point.



Sorry for the ramblings.

[GranteedEV]

1) Nobody ever listens to a speaker exactly on-axis, so there's no harm in deviating a bit from flat response there - what's perfect is not actually the best.

Yes, you should measure the listening window response. I think Toole defined it as 15 deg horizontal both ways and 10 deg vertical both ways but I could be wrong.

3) Floyd Toole's school of thought (flat on axis response, smooth power response, etc.) can be achieved by carefully weighing the pros and cons of a perfectly flat on axis response vs. the actual, measured off axis response. If he had settled for a single thing (ie, flat on axis response), speaker design would be so much easier..

As pete mentioned, matching directivity index is the key. to do so you normally need a waveguide unless it's an acoustically small OB speaker like the Nao Note (which technically... still uses a waveguide).

If you look at all the speakers that seem to follow the Toole School (...) of (.....COOL?) design, they all employ tweeter waveguides. The Revel Salons, The JBL LSR 6332s, the KEF Blade, TAD Reference, Genelec 8260, Mark K's ER18DXT, Pete's MiniByz. Geddes is on the same page although he prefers to employ a much narrower directivity index for imaging/clarity over spaciousness/diffuseness (although he makes an effort to liven up the room as much as possible to counter the effect, and crossfires the speakers). The other key is to use tweeters that can take a beating at the bottom of their passband. If they can't cross low enough (and by low enough, i really mean the response as far out as 60deg off axis has to match the tweeter) then things get tricky.

Using a narrow mid is a solution - Infinity used to have a very cool mid that they used on the Cascade speakers, but they used it as a midwoofer which throws away its wicked headroom instead of high passing it. Dennis Murphy has used the 4.5" accuton + RAAL 70-20XR and the thin Neo8 mids with the 70-10D - so getting a reasonably matching DI is possible without a waveguide in a 3-way speaker. For a 2-way though... you really want to cross lower, and use a waveguide.

So to address your question - driver selection is really important. If you're not planning the full speaker design out with power response in mind, then you're just trying to fit a square peg (smooth power response) in a round hole (2-way with 6.5" woofer playing up to 3khz w/ flush mount tweeter).

So my advice is, have fun with this but if you really want to try out a speaker that has good listening window and horizontal off-axis response, start from scratch and plan it carefully

As for vertical off axis response, only coaxials are really acceptable in that realm. I recommend absorbing 1st vertical reflections as it's pretty offensive VERY early reflections without localization cues. I'm not sure MTMs are really a solution to vertical reflections either. One one hand, they're probably "better than nothing" as they operate mostly in the speech bands where the energy might be most offensive, but on the other hand they kind of take that tweeter flare to the next level on the vertical axis. I think i'm pretty much sure that absorbing early vertical reflections, diffusing and delaying reflections behind you and behind the speakers, and reflecting or diffusing sound from the sides, is how i'd deal with sound in a room.

[Deward Hastings]

25mm tweeters. They really suck, now that I think about it. . . I know - I should have gone all out and used [any decent tweeter] but . . .

You really did answer your own question, you know. There are plenty of inexpensive small-flange tweeters that would work well with the ND105 (and not need a 6th order crossover, either). Make the baffle small enough and the whole thing becomes effectively omni up to 3-4kHz at least, and the "problem" goes away.

But the real take-away is that in almost all listening environments there is more energy in the "reflected" sound than "on-axis", so that's what you actually hear, and make your frequency response judgments about. It's not just about getting "flat" on *any* one axis, be it 0 or 30 or 60, it's about getting the overall power response of the speaker flat. It's especially nice if you can get power response and listening axis response *both* flat (or, actually, drooping a bit on the high end) . . . that's why omnis and dipoles sound "right" (if the room's response isn't so far off as to spoil them).

ErinH is correct about the general neglect of the vertical pattern, too . . . and it's especially critical since in most rooms the ceiling bounce is the first (and loudest) reflection that we hear . . .

[johnnyrichards]

Hell, even the dirt-cheap Infinity Primus series uses a waveguide loaded tweeter. Of course, they don't load the system up like the spendy Salon series BUT: they (the Harmon engineers responsible for this series) spend a ****-ton more time worrying about how it works beyond a simple simulation using manufacturer supplied FR data

I love this hobby to no end, but there are times when I realize that even the best of us are playing catchup with some engineers at a commercial outfit

[GranteedEV]

Hell, even the dirt-cheap Infinity Primus series uses a waveguide loaded tweeter. Of course, they don't load the system up like the spendy Salon series BUT: they (the Harmon engineers responsible for this series) spend a ****-ton more time worrying about how it works beyond a simple simulation using manufacturer supplied FR data

Same with those andrew jones Pioneer cheapo speakers. The use of the waveguide is probably the dominant factor in their performance. Andrew Jones sure does love his grills though!

[fjhuerta]

This is why a waveguide may be preferred over a flat baffle, because you can have smooth/flat both on axis, and off. Below is a 1KHz LR4 XO with a 8.5" waveguide loaded dome. With your crossover frequency, a 5 or 6" guide crossing to a smaller woofer, say 5 or 6", would help keep the off axis (power response) more uniform.
[/IMG]

When I read the commentary about how Harman, etc. have so many tools and knowledge that are beyond our grasp / means, I always think about your MiniByzy. Harman would be proud to have someone like you on their ranks.

Yeah, you are right. It took me years to finally understand the relationship between power response, perceived frequency response, reflections, on-axis response, etc. I think I'm getting to fully get the big picture. And I agree - why aren't other companies building waveguide tweeters? As of now, it's a pain to get a proper waveguide working for someone with no CNC routers. I wonder if the SEAS DXT-18 would work properly on such an implementation with, say, a Dayton RS150... that's as close as I could get to a proper waveguide right now.

That said, what I'm getting at is my goal would be for smooth when it makes sense. I'd rather take an overall decrease in SPL toward the high end than one with a more 'flat' response with audible bumps in response here and there.

When I first got my Behringer Ultracurve, and dialed in a completely flat response, I always wondered where my bass went... and I tried to convince myself all those "reviewers" who said "veils had been lifted" when EQ'ng for flat were right. They weren't...

So my advice is, have fun with this but if you really want to try out a speaker that has good listening window and horizontal off-axis response, start from scratch and plan it carefully

That was sort of my intention in the first place. I thought a 3 5/8" woofer would mate pretty well with the MA tweeter. I wasn't aware of how high they needed to be crossed over. If I did it all over again, I'd definitely use a much beefier tweeter. And yeah, my next project will definitely explore the smoothest off axis response I can get. And once again, I'm thinking about using the SEAS tweeter. I've always found that I really dislike upper midrange flare, and I always double checked my designs, to see if I had any issues there. I didn't - why should I? I was only looking on axis! Now I suppose the flare I dislike comes from the off axis tweeter response at the bottom of the passband. I could never get over how much I noticed this while listening to the Nat-P design I built.. even though I measured them and they were pretty flat off axis.

You really did answer your own question, you know. There are plenty of inexpensive small-flange tweeters that would work well with the ND105 (and not need a 6th order crossover, either). Make the baffle small enough and the whole thing becomes effectively omni up to 3-4kHz at least, and the "problem" goes away.

Incredible what kind of stuff is used on $1.5K commercial speakers! Yeah, the tweeter was a disappointment. I still think I can make everything sound right. I was actually thinking about your comments when I decided to build this speaker... omni as high as I could make it. Next time, for sure!

[Deward Hastings]

omni as high as I could make it. Next time, for sure!

. . . we have had this conversation before, haven't we . . .

I've got a pair of ND105 and a pair of D26NC55 sitting on the shelf that I've been meaning to mount in some sort of "blade" . . . the D26 should be able to cross low enough to "blend" smoothly. Got a sweet old Yamaha M35 4-channel amp to drive them, too . . . just have to solder up a crossover board (or use a miniDSP). Yet another project for the New Year (as if I don't have enough already).

[Face]

Same with those andrew jones Pioneer cheapo speakers. The use of the waveguide is probably the dominant factor in their performance. Andrew Jones sure does love his grills though!

I have the larger pair of his bookshelves in my bedroom setup. They're a little bass shy and don't sound great when used for nearfield listening, but otherwise sound really nice. Judging by your photograph, it appears I'm not the only one who had a difficult time photographing them.

Speaking of Andrew Jones: http://www.stereophile.com/content/t...r1-loudspeaker

[GranteedEV]

very nice polars. if stereophile didnt normalize to the zero degree response they would look even nicer above 8khz.

[fjhuerta]

. . . we have had this conversation before, haven't we . . .

I've got a pair of ND105 and a pair of D26NC55 sitting on the shelf that I've been meaning to mount in some sort of "blade" . . . the D26 should be able to cross low enough to "blend" smoothly. Got a sweet old Yamaha M35 4-channel amp to drive them, too . . . just have to solder up a crossover board (or use a miniDSP). Yet another project for the New Year (as if I don't have enough already).

What's the idea behind the blade? Sounds interesting...

[Deward Hastings]

What's the idea behind the blade?

That the "baffle step" be moved *up* as high as possible. Think PLUTO with the mid-woofer mounted on the side of the pipe rather than the top, but the enclosure made deeper (blade like) to reduce wraparound issues. Don't know just how deep yet . . .

I've already done a PLUTO variant with the woofer set at 30 degrees (but still at the end of the pipe) . . . works fine. The "blade" would just bring it further around and down the face a bit. I don't believe it would work with the "full size" woofer, but the ND105 is significantly smaller. There are a lot of unanswered questions, like where best to "hinge" the downward slope of the power response (complicated by trying to do it with radiation pattern rather than equalization). PLUTO does the "hinge" with the beaming of the NSW2 . . . somewhere around 3-4 kHz. . . . I'd like to move that down an octave. A baffle no wider than the ND105, with the tweeter close coupled on the same baffle, should do that. It's more an experiment than a fully thought out design . . .

[mzisserson]

From my design experiences and preferences from others Listening to many of each side by side over the years, for the home environment a traditional speaker which is properly designed to account for all of the normal driver directivity patters yielding smooth and extended power responce with very good off axis performance will always be preferred over a speaker employing controlled directivity.

You will never be able to truly eliminate the room. It is far better to work with it than run from it, and smooth power responce allows any given speaker to blend much better in multiple environments. I will say for the few controlled direcitivity speakers that I have liked, when you are between them they truly can sound wonderful. IMHO, save it for pro audio where it is needed. The 40's are over and drivers are efficient enough not not require horn loading.

I am unsure anyone else is in this boat as I am, but directly relating to off axis performance, the room counts, weather we like it or not.

The speakers I have preferred over the years all exhibit these same attributes:

1) No driver ever becomes directional employing small midranges and small dome tweeters.

2) Are 3 +way

3) If any waveguide is employed, it is extremely mild acting more as baffle control than directional control.

4) Have very smooth and very gentle roll off off axis, and measure in-room (any room) very well.

5) All use appropriately sized drivers with appropriate crossivers. i.e. no 8" 2-ways, or 10" with a horn from 500Hz and up etc. All are well integrated such as 10" with a 5.5" extended range mid, to a 1" tweeter.

If controlled directivity were so important in the home environment "high end" mfg's would have switched to them years ago. I know there is a great deal of venom for commercial speakers out there, but from knowing and having some deep discussions with some acoustic designers and engineers out there the common theme is always power response dominates over directivity in small (less than concert hall) listening environments.

FWIW, IMHO, I may be the only one in this boat but, but ima' stickin' to it as the feedback I have received continues to validate these design philosophies.

I would love if anyone close by has a pair of waveguided speakers would lend them for a day and we can do some side by side spatially avarage measurements of them aganst a speaker that is designed specifically for smooth off axis response and power response.

[Deward Hastings]

preferred over a speaker employing controlled directivity.

It would help if you were a little more . . . precise . . . in your use of terms . . . full range omni, full range dipole, and full range 2pi (wall mount) speakers are all "controlled directivity" (so is a speaker that starts omni and smoothly tapers it's pattern as frequency rises) . . . they are just "controlling" for a different (and typically more uniform) radiation pattern/power response than horns/waveguides.

The "difficulty" with horns/waveguides is wavelength . . . the necessary size typically (and detrimentally) compromises other things, not always with a net gain in performance. The two areas where "box" speakers typically fail the worst are baffle step (any speaker that has it, or "corrects" for it with equalization in the crossover, is wrong), and beaming/blooming at the M/T transition (which "waveguides" aim to correct, but if they are big enough then c to c spacing gets too large and there is a loss of control on the vertical axis).

power response dominates over directivity

Yes . . . but that is not to say that flat response on the listening axis is not important too. Getting flat (or smoothly declining) power response while maintaining flat-on-axis response means controlling the polar substantially beyond what waveguides on a tweeter are able to do. 30 degree (or even 60 degree) horizontal polars (popular with the waveguide proponents) simply do not disclose what the speaker is doing in the room. For that you need 90 degree and 120 and 150 degree horizontal, and a full set of vertical as well. Or direct measurement in a room with a known absorption profile.

[mzisserson]

It would help if you were a little more . . . precise . . . in your use of terms . . . full range omni, full range dipole, and full range 2pi (wall mount) speakers are all "controlled directivity" (so is a speaker that starts omni and smoothly tapers it's pattern as frequency rises) . . . they are just "controlling" for a different (and typically more uniform) radiation pattern/power response than horns/waveguides.

The "difficulty" with horns/waveguides is wavelength . . . the necessary size typically (and detrimentally) compromises other things, not always with a net gain in performance. The two areas where "box" speakers typically fail the worst are baffle step (any speaker that has it, or "corrects" for it with equalization in the crossover, is wrong), and beaming/blooming at the M/T transition (which "waveguides" aim to correct, but if they are big enough then c to c spacing gets too large and there is a loss of control on the vertical axis).

Yes . . . but that is not to say that flat response on the listening axis is not important too. Getting flat (or smoothly declining) power response while maintaining flat-on-axis response means controlling the polar substantially beyond what waveguides on a tweeter are able to do. 30 degree (or even 60 degree) horizontal polars (popular with the waveguide proponents) simply do not disclose what the speaker is doing in the room. For that you need 90 degree and 120 and 150 degree horizontal, and a full set of vertical as well. Or direct measurement in a room with a known absorption profile.

To the bolded point I agree.

A speaker with controlled driectivity will always measure well. You are focusing the sound directly onto the mic (which mics have no correlation to our ears and how they work, but that's another story), you can get BOTH flat on axis fq performance and good power response from a standard, baffle mounted tweeter/mid combo.

BSC and diffraction are a larger issue in a traditional design I agree, but again good design and a 1/2" chamfer or roundover can solve these issues without really creating more.

Notice you never see any spatially averaged measurements of speakers employing waveguides? You are always giving up to gain. In-room and for most pratical applications, speakers with too much directivity in home use end up with poor performance off axis. If you are really looking for sweet spot listening this is fine, but for the avagrage room and even for most hifi geeks, it just does not pan out well. It is an overcomplexity that can be done without if design details such as driver selection and crossover points (passive or active) are properly wrapped up. Also, if there is "beaming/blooming" m/t transition, then there is a design issue, not a speaker one.

In fact, controlled directivity pundits are so worried about in-room performance I asked one to borrow a design of theirs to measure against a similar size speakers using the design convention of smooth power response and I was told to build my own since they did not want me measuring theirs. Obviously there is something to hide.

I would like to reiterate that I am not sating waveguides are a bad thing as I have heard plenty of good speakers which employ them. I am saying they are not the only, and far from the "best" (which as we know best is a subjective term) option for small rooms. (By small I mean living-room sized where most listen from 6-15 feet (give or take) from the loudspeaker) There are always exceptions.

[Deward Hastings]

A speaker with controlled driectivity will always measure well.

Not hardly. The Geddes horns, for example, look quite . . . bad . . . on the forward axis. You have to measure (and listen) off axis to get them to measure (and sound) "good". Their power response is another thing altogether, and "fails" at baffle step. Earl's answer is (at least for his bigger boxes) that the baffle step is low enough that it doesnt matter, it's masked by the room.

There is, um, room for disagreement.

[GranteedEV]

In-room and for most pratical applications, speakers with too much directivity in home use end up with poor performance off axis. If you are really looking for sweet spot listening

Er...that's what cross firing is for... in fact the sweet spot is wider because there's less influence from the NEAR side wall and more influence from the FAR side speaker (you get more on-axis) - in turn balancing things out than a wide directivity speaker facing straight out like most people's setups.

I mean, controlled pattern wide directivity speakers can still have a good sweet spot, but your argument doesn't run.

and no one's using 60 deg horns here. 90 - 100 degree pattern control is plenty wide.