I share Dr. Geddes opinion on this.
I am just a hobby DIY'er, but have worked with something like 10 different waveguides the past 5 years.
I don't belive that normalized plots are the best way to display directivity. They simply show the response in comparison to some axis (often on-axis). It doesnt show the true energy that is on that axis.
I have owned and used 4 different SEOS waveguides. In many of them sonograms look wery good to the eye when normalized. The -6db angle is often very consistent with the on-axis response. This is ofcourse not a bad thing. But I have with several of them experienced that there is much more variation between the response inside the 0-30 deg window. And this is what JBL defines as the "listening window".
Like them I have found that the average of the listening window has been a much more consistent with what I subjectively hear. Voicing a speaker a flat on axis is simply has lnot led to a good result on several ocasions. Also with the SEOS WG's.
The normalized plots can be very misleading. Say you have a small -2db hole on the "on-axis" frequency plot and small +2db at the -6db angle at the same frequency. The normalized plot will have you believe that there is a large excess of energy (+4db) at the edge of the pattern at this frequency. That is not the case.
In my opinion you look for a large and smooth listening window (Direct sound) and you want a flattish smooth DI in the 800-8khz region. This will often produce a flat controlled in room response without aditional EQ. This is excactly what JBL do and Floyd Toole has shown how close the predicted steady state response in the listening position will be when you used this method. Again a normalized response can fool you in this regard, because you dont see the true energy output, you just see the output compared to the on-axis.
The hole point of measuring is trying to measure in a way that has relevance to what we hear.
I have not heard a WG with SQ bettering the M2. Ofc this is just my subjective opinion. Many of the problems in the posted plot can be there due to normalizing, and from my experience therefor not necesarely problems. I also find the plot unlikely to be correct in the lower frequencies. I have not seen any speaker of this size that has such narrow directivity down to 400hz.
Apologies for some typing errors here and there, I'm a bit dyslectic. Hope the essence comes through.
I am just a hobby DIY'er, but have worked with something like 10 different waveguides the past 5 years.
I don't belive that normalized plots are the best way to display directivity. They simply show the response in comparison to some axis (often on-axis). It doesnt show the true energy that is on that axis.
I have owned and used 4 different SEOS waveguides. In many of them sonograms look wery good to the eye when normalized. The -6db angle is often very consistent with the on-axis response. This is ofcourse not a bad thing. But I have with several of them experienced that there is much more variation between the response inside the 0-30 deg window. And this is what JBL defines as the "listening window".
Like them I have found that the average of the listening window has been a much more consistent with what I subjectively hear. Voicing a speaker a flat on axis is simply has lnot led to a good result on several ocasions. Also with the SEOS WG's.
The normalized plots can be very misleading. Say you have a small -2db hole on the "on-axis" frequency plot and small +2db at the -6db angle at the same frequency. The normalized plot will have you believe that there is a large excess of energy (+4db) at the edge of the pattern at this frequency. That is not the case.
In my opinion you look for a large and smooth listening window (Direct sound) and you want a flattish smooth DI in the 800-8khz region. This will often produce a flat controlled in room response without aditional EQ. This is excactly what JBL do and Floyd Toole has shown how close the predicted steady state response in the listening position will be when you used this method. Again a normalized response can fool you in this regard, because you dont see the true energy output, you just see the output compared to the on-axis.
The hole point of measuring is trying to measure in a way that has relevance to what we hear.
I have not heard a WG with SQ bettering the M2. Ofc this is just my subjective opinion. Many of the problems in the posted plot can be there due to normalizing, and from my experience therefor not necesarely problems. I also find the plot unlikely to be correct in the lower frequencies. I have not seen any speaker of this size that has such narrow directivity down to 400hz.
Apologies for some typing errors here and there, I'm a bit dyslectic. Hope the essence comes through.
Last edited:
This is your choice.
I think we have different assumptions on how a speaker should be heard. Maybe there is a misunderstanding.
I always listen under 0° and I always equalize 0° to a flat frequency response, because I want correct direct sound. This is a basic setup for every "usual" loudspeaker and it this assumption is wide spread in the industry. When you equalize a hole on 0° you raise the power of higher angles, too. So it is indeed equivalent to a widening.
So you are correct in a non-equalized setup. And I'm correct in a equalized setup. But regarding directvity only (which equals to an equalized setup), you are wrong.
@FrederikC
Please read my answer above, since you share the same assumption as Earl.
FredrikC - I don't think many people understand this issue - that the power of an axial point is way below that of an off axis point. It goes as sin( theta). Consider an extremely small wide hole of some -12 dB depth. One could not even detect this hole in the power response, but on a normalized plot this hole would become immense, way out of proportion to its importance.
Then there is the choice of axis. Why would I ever publish data that is normalized to an axis that I do not optimize for? If one want to see "flat and well controlled directivity then look at an S-15 at 20 degrees off axis - the intended listening axis. It is exemplary at this angle, but people keep looking only at the erroneous axial response.
No waveguide is perfectly "constant directivity", not mine not anyone's. What should be rated is the smoothness of the DI (Directivity Index) along the listening axis and over what bandwidth is it nearly flat, and how high is it over this bandwidth. That is what is perceptually correct. The rest is just arguments over unimportant aspects.
Then there is the choice of axis. Why would I ever publish data that is normalized to an axis that I do not optimize for? If one want to see "flat and well controlled directivity then look at an S-15 at 20 degrees off axis - the intended listening axis. It is exemplary at this angle, but people keep looking only at the erroneous axial response.
No waveguide is perfectly "constant directivity", not mine not anyone's. What should be rated is the smoothness of the DI (Directivity Index) along the listening axis and over what bandwidth is it nearly flat, and how high is it over this bandwidth. That is what is perceptually correct. The rest is just arguments over unimportant aspects.
FolGoTT - you continue to ignore my point that the axial response is not the direct response when my speakers are properly setup (design intent.) This is not some arbitrary thing - it is a precisely defined and a logically derived approach to maximizing the sweat spot in a listening position, as well as minimizing the very early reflections found in a small room. both highly important aspects of loudspeaker design. If you refuse to accept my recommended setup then we really do not have anything to talk about because we are not using the same premise.
Loudspeaker should not have horizontal symmetric responses. The easiest way to achieve this is simply by pointing them inward, (outward works too, but with other ensuing problems like increased very early reflections.) One could make a system that "looked like" it was pointed straight ahead, but in fact had a non-symmetric polar response. That's just a lot more difficult and expensive and ends up just doing the same thing as turning the speakers.
It simply makes no sense to arbitrarily choose the "normal" vector of a loudspeaker as the reference.
The direct response needs to be near-flat and smooth, but the power response and hence the room response needs that as well. On needs to find that axis along which these two requirements are best achieved. It will seldom be the normal. Designing specifically for a non-normal axis is, I agree, a radical departure from the "norm" (pun intended) and it requires a rethinking of many of the "normal" way things are measured and displayed, but I'm perfectly OK with that. You aren't. I am not going back to the "old" ways, maybe you should look at the newer ideas and adjust your position accordingly.
Loudspeaker should not have horizontal symmetric responses. The easiest way to achieve this is simply by pointing them inward, (outward works too, but with other ensuing problems like increased very early reflections.) One could make a system that "looked like" it was pointed straight ahead, but in fact had a non-symmetric polar response. That's just a lot more difficult and expensive and ends up just doing the same thing as turning the speakers.
It simply makes no sense to arbitrarily choose the "normal" vector of a loudspeaker as the reference.
The direct response needs to be near-flat and smooth, but the power response and hence the room response needs that as well. On needs to find that axis along which these two requirements are best achieved. It will seldom be the normal. Designing specifically for a non-normal axis is, I agree, a radical departure from the "norm" (pun intended) and it requires a rethinking of many of the "normal" way things are measured and displayed, but I'm perfectly OK with that. You aren't. I am not going back to the "old" ways, maybe you should look at the newer ideas and adjust your position accordingly.
Last edited:
I guess one of the first problems in these types of discussions is the terminology and how it is used. Constant directivity is used by many to mean different things and has been one of those marketing terms since I remember seeing the Altec MantaRay, the JBL Bi-Radial, EV's horns and many other horn designs that I would have to say were far from a true constant directivity. I would guess technically speaking the only horn shape that could have a constant directivity would be an infinite conic section. We don't use infinite length horns and so this becomes only a matter of semantics.
On the JBL front you have to realize that change is part of the structure of the Harman Corporate marketing plan. Even if they made a perfect horn shape that did everything you could desire that couldn't stay that way forever. Change is demanded to follow their marketing based structure, they have to change products for no other reason than to have something new to sell, a new product that is desired by the public.
The best car I have ever driven and owned was a 1974 Mercedes 450SEL. It had a better ride than any other car at the time, an engine that just would run and last forever, mine went almost 500K miles with no major issues and it was comfortable and easy to drive, even at very high speeds. But you can't sell the same thing forever, you have to change the model to bring in new buyers. Believe me the newer S class Mercedes while being technically superior are not as well constructed, the mechanical issues are much greater. Does that mean that Mercedes should go back to building that previous car, no, but that doesn't mean change is improvement, that is unless you consider all the extra complexity and electronics absolutely necessary to have a great car. With today's cars if a mechanic can not plug in a scanner and see a problem in the codes they can spend months and never solve a problem hidden in all those miles of wires. A simple dirty ground on a single ground point can turn an expensive car into a pain in the rear and a very unenjoyable one at that. Been there done that, had to find the issue myself after every major electronic sensor was replaced. Just a dirty ground sending a modulated signal the computer never senses. No check engine light, no codes to follow, something so simple yet so hard to find given the time constraints given to the technicians, just throw in new parts and hope for the best.
So my conclusion is that in the discussion of constant directivity we have to realize it truly is a marketing terminology, not a factual reality. You will never get away from the room device interface, there are no perfect solutions.
On the JBL front you have to realize that change is part of the structure of the Harman Corporate marketing plan. Even if they made a perfect horn shape that did everything you could desire that couldn't stay that way forever. Change is demanded to follow their marketing based structure, they have to change products for no other reason than to have something new to sell, a new product that is desired by the public.
The best car I have ever driven and owned was a 1974 Mercedes 450SEL. It had a better ride than any other car at the time, an engine that just would run and last forever, mine went almost 500K miles with no major issues and it was comfortable and easy to drive, even at very high speeds. But you can't sell the same thing forever, you have to change the model to bring in new buyers. Believe me the newer S class Mercedes while being technically superior are not as well constructed, the mechanical issues are much greater. Does that mean that Mercedes should go back to building that previous car, no, but that doesn't mean change is improvement, that is unless you consider all the extra complexity and electronics absolutely necessary to have a great car. With today's cars if a mechanic can not plug in a scanner and see a problem in the codes they can spend months and never solve a problem hidden in all those miles of wires. A simple dirty ground on a single ground point can turn an expensive car into a pain in the rear and a very unenjoyable one at that. Been there done that, had to find the issue myself after every major electronic sensor was replaced. Just a dirty ground sending a modulated signal the computer never senses. No check engine light, no codes to follow, something so simple yet so hard to find given the time constraints given to the technicians, just throw in new parts and hope for the best.
So my conclusion is that in the discussion of constant directivity we have to realize it truly is a marketing terminology, not a factual reality. You will never get away from the room device interface, there are no perfect solutions.
I know your approach and I don't ignore it. But you ignore the usual standards when publishing measurements.
You design for 30°. So why you don't publish a diagram which is normalized to 30°?
Why?
Is this your personal opinion or are there objective facts which proves this thesis.
I agree completely.
Please show me examples. It should strongly depend on room dimensions, speakers position and listening position. There is no standard listeing environment.
And maybe you should arrive in the present where waveguides don't have holes on-axis.
Hi gedlee,
If I have understood You, driver&horn combo such as JBL 2446&2311, or 2421&2312 would be an example of deep hole in-axis response,as either of them are very small mouth, round mouth (about 4-inch diameter), with quite narrow horn wall off-axis angle (may be about 20deg at the mouth), but, honestly to say with 2308 lenses, they behave 'almost' flat from 700Hz up to over 10kHz .
Is such in-axis FR hole influenced by interference in-axis sound wave and from the mouth edge 'diffracted' sound wave, as it is 'delayed' due to the different way length?
regards
ivica
FoLLgoTT - considering Earl has a pretty flat response on his design axis (ok not ruler flat) isn't that fairly close to normalizing to the design axis? The contour plots you posted look to me like they are more "trad" horn designs which in my experience seem to be more consistent off axis....they just progressively narrow the pattern.
I'd be interested in knowing how the SEOS waveguides achieve such nice polars in the horizontal. The vertical isn't that great imo. One thing I've noticed in Earl's larger waveguides, the waveguides I've built, and other commercial offerings is a bit of excess off-axis energy in the roughly 6k-8khz region. Particularly around 30deg off axis. What I've noticed is that the level at this angle seems to rise and meet the level at 0deg. In my waveguides the 30deg measurement actually exceeds the 0deg in level by about 1dB at 7khz. As was said earlier if you look closely at some SEOS15 measurements the curves from 0-20deg also tend to get "mixed up" above 5khz. What causes this, and is it the fact that the SEOS and similar waveguides have such a high aspect ratio that the effect isn't as dramatic?
I'd be interested in knowing how the SEOS waveguides achieve such nice polars in the horizontal. The vertical isn't that great imo. One thing I've noticed in Earl's larger waveguides, the waveguides I've built, and other commercial offerings is a bit of excess off-axis energy in the roughly 6k-8khz region. Particularly around 30deg off axis. What I've noticed is that the level at this angle seems to rise and meet the level at 0deg. In my waveguides the 30deg measurement actually exceeds the 0deg in level by about 1dB at 7khz. As was said earlier if you look closely at some SEOS15 measurements the curves from 0-20deg also tend to get "mixed up" above 5khz. What causes this, and is it the fact that the SEOS and similar waveguides have such a high aspect ratio that the effect isn't as dramatic?
Yes, you are right. I missed that fact.
This is pretty easy. You have to optimize it by hand until directivity looks good enough. For simple designs I use AxiDriver which supports only circular waveguides/horns. For more complex designs 3D CAD and another BEM simulator is necessary.
Here are two examples. One is pretty conical with rounded edges and one is optimized in a half hour. Of course with more time the contour will get even better, but it shall show how much influence the contour has on directvitiy.
Conical:
Roughly optimized by hand:
Attachments
Member
Joined 2003
I don't know exactly. But the model is not exact in this frequency range. AxiDriver simulates a ideal piston driver. The reality is different. So I wouldn't optimize too much above 10 kHz. With a real driver it can behave differently.
I 3D-printed a prototype horn a few months ago and the simulationw as very acourate. But it also differed in the highest highs.
To alleviate the "sweet spot" situation the response should increase as the angle moves away from the listening axis towards the opposite speaker and do the opposite in the other direction. This helps to trade-off the image shift (per Blauert.) This is well known and even used in the past by JBL and Advent. I am shocked that you didn't know this.
And Nate Hanson is correct. The NS-15 is darn near flat at 20-30 degrees so it is in essence normalized. That I don't use "standard approaches" is, to me, a compliment. I do very little that is conventional and I will not apologize for that fact.
And maybe you should arrive in the present where waveguides don't have holes on-axis.
Again, you are stuck in what you consider "on-axis" - open up your mind!
I have used those old JBl devices and they were pretty bad overall. I sought to get all their advantages without there flaws. I did my MS thesis on slant plate lenses and published an AES paper on them - so I know them well.
Yes the normal-axis hole is caused by interference from the mouth edge diffraction, and it is delayed by the extra path length but also reversed in phase because of the diffraction - so the net result can get quite complicated.
Correct - and non-normal response control is more important than the normal response, so your simulation examples are not very meaningful. Until you build a real device and show it in all its detail glory - ala a PolarMap - we are not talking about comparative things.
I have and continue to offer the implementation of any device in my PolarMap database thus allowing everyone to do an apples-to-apples comparison. Not doing so just gives us all a reason to discount the claims.
Member
Joined 2003
I know about the frequency dependand image shift. And there is even software which tries to compensate that.
But on the other hand this special listening angle does not represents the typical situation when the music gets mixed. And you never know if the person who mixed the content may have considered this phenomenon and compensate for it.
I respect your approach. But I don't see the "must" in it.
Hi gedlee,
so, about 2311/2312, are there FR holes as You have expected, and on what frequencies, as I have not seen them, or they are not too deep
As I have calculated, concerning mentioned horns, the difference in the path length (without 2308 lenses) is about 1.2mm, that would correspond to the frequency much more over 20kHz, looking , looking from about 1m distance from the horn mouth.
regards
ivica
Last edited:
It is meaningful if you know about the differences to reality.
I'm tired in posting pictures and proofs and from you comes nothing. Here is the last one.
3D printed device:
Simulation:
Measurement:
Note: the simulation is on an infinite baffle. The measurement was taken without a baffle.
Attachments
I was involved in the SEOS development and can say that the secret to its horizontal pattern is largely.. luck! SEOS was largely designed by committee by diyers on the avs and audiokarma sites. The largest non active technical contributions are from the papers written by Earl, actually. As Earl says, its vertical and particularly its diagonal performance isn't as impressive (how much that matters comparatively is up for debate).