Aleksandar Radisavljevic on RAAL speaker

[mige0]

Thread opened to provide a platform about RAAL speakers - in continuation on the postings to be found here:

On the directivity of dipole tweeters

Originally Posted by RAAL

Hi guys,

a friend told me about this directivity discussion, so I thought I drop in a few facts for your understanding of what influences the directivity. I hope you won't mind as much as I did your conclusion.

First, in horizontal plane, it's not the width of the baffle that determines it. Its the width in the diaphragm, which in this case is just 15mm, which gives us about 120 deg on 20k.

Second, in vertical plane, even though the diaphragm is a lot longer than the wavelength of 20k, the use of acoustical lens, in form of those two foam pads on the face plate, gives us about 20 deg of vertical coverage without any loss highs compared to mids and no loss of flat tone balance.

With proper baffling, the directivity pattern is almost perfect, rounded, figure-of-eight pattern in a horizontal plane, as the ribbon sprays wider that the open baffle arrangement permits, so the baffle rules that plane, governing the figure-of-eight pattern.
In vertical plane, the foam pads rule the situation, and that behaves exactly the same as closed 140-15D.
Measurements attached, but no polar plots, sorry. Too tedious to prepare for publishing but I'm sure you'll figure out the point from these plots.

I also need to say this: thinking that my designs neglect directivity and that my customers are happy with it, couldn't be further from the truth. The whole point of my designs and of my customers desires, revolves around just the opposite. That is why I felt compelled to explain you gents what's this all about.

On that note, model DIPOLE 140-15D can not be used for stacking into line array, it was designed that way, so the accuracy of a near-point source would not be impaired, as well as symmetry of the front-back radiation, which in turn, allows different usage of acoustic lens front and back, so user could fit the directivity to room acoustics in terms of back wall reflections and stereo depth imaging.

As for the prices, I know, they're expensive, but if it was possible to make them cheaper, we'd do it. If any loudspeaker company out there could make a cheaper magnetically shielded, fully front-back symmetrical, open ribbon with this directivity, efficiency and reliability, they would, trust me, and they would have a market for it. The point is that they can't speed it up more that we did. To retain the quality I aim for, it takes a company of five highly skilled people and a few outside machining services to be able to build 50 of those a month, making them only. Once you put things on paper, add shipping and distribution, that what you end up with for a retail price. I can't fight the way the world of economy works, and nobody can. At least we have a choice, and my choice is this: Hopefully, with ever-growing support of our customers that graciously buy our stuff, RAAL will eventually grow into a large company that will improve the manufacturing efficiency and bring out good new products for affordable money. Right now, we are more of an artisan's shop that does things the way any artisan's shop does.

And so on and so forth...I could talk about decisions I needed to make all day long and positively bore you to death!

I wish you a very happy New Year!
I'm off now to "enjoy" some excessive drinking with friends that won't take no for an answer!
Cheers!

Michael

[mige0]

Originally Posted by RAAL

Hi Michael, John, everyone!

Measurements are done in IEC standard baffle for 8" driver, only the ribbon is flush mounted. Drawing attached for the people that aren't familiar with it.
Smoothing is 1/6th oct, brickwall. In any case, CSD plots can't be smoothed out, so they provide clues to what the manufacturer did to make other plots look nice. It will show if it was tampered with too much.

John, I couldn't agree more!
However...You know it better than I do, that polar plot will look as good as it's measurement baffle shape is, as well as driver's position on it.
We also know, that most people while browsing for a driver to buy, just have a glance at the plots and go on. What happens if they spot a big dip at 1k5? Would they contribute it to a measurement baffle size or to the driver? Will they read the fine print and a lenghty explanation why it is there? What will the polar plot look like when that edge diffraction dip start shifting around with measurement angle and even worse, multiply itself by the number of baffle edges as soon as you move the mike away from the dead center?
Naturally, I would like to present the driver at it's best, but in essence, you are actually suggesting that I should design the baffle shape solution, just like you had to do, to yield the best results with your new speaker. But then, would that shape fit the midrange that someone wants to use? Would it fit their taste in design?
Considering the above, and that I would like to make sure that the driver polar plots will not suffer from the baffle shape, the thing I should do is to measure the polar plots in IEC baffle, but then, they wouldn't really look like anything dipole, or at least, like anything that the user would get in their all-too-different baffles than than mine. Would that make me a liar? Would that situation of users getting different measurements by default than what I published, help in increasing or in decreasing the customer satisfaction in the future? Would RAAL drivers remain respectable in that case?

Whatever the thing may be, whether I'm right or wrong, my decision is that upon request, people that can read that data, will get the data acquired with IEC baffle. The success of the completed speaker itself, rests is upon them, so here it is, PDF attached.
The actual driver that has been measured is regular, closed 140-15D. I hope you will trust me on this, but the DIPOLE 140-15D spits out the same graphs in that same baffle.

I hope this helps and I hope it sheds some light on why driver manufacturers aren't jumping ahead and show all kinds of data by default. Too many misconceptions out there and the ones that possess them are the quickest to judge and tell about it.

Cheers!

[mige0]

Originally Posted by RAAL

Rudolf, you're reading my mind. The design of the driver was made with having in mind the wish to use it as bare-naked as possible ( I'll tell my opinion of that a bit later). That throws us back to the question of what is the best baffle shape for linear sources like this one, and I, too, have an opinion of that. So...

In the attached pictures you can see what the driver looks like from behind.

The pear-shaped structure is the iron yoke that closes the magnetic flux. It was designed in FEM simulation and completely closes the flux with a minimal amount of iron. The iron thickness is 16mm. To that, add 2 + 2 mm of front and back plate.

The front plate is just a mounting flange and it also can be made having the same shape, or completely removed. If removed, the mounting will be a problem, but as it has always been, we make custom front plates without surcharge. If clients start asking frequently for that, I even might decide to redesign it and have it eventually offered as such.

Since this driver is not designed to be used lower than 1.6k, and the mean width of this "pear" allows for about 1k5 cut-off, the driver can be used without this rectangular front plate, without impeding it's low-end response.

That leaves us with this pear shaped yoke, which then plays the role of acoustic baffle, which I believe, is the best possible shape for linear radiators like ribbons. Also, this shape falls in line with this tendency to have double-circle figure-of-eight response extended as much as possible toward highs, but with maintaining the proper low-end (this is a contradictory demand and I'll turn to that later) and without creating too much dips and peaks in the response. It should be a tiny little bit better than naked Neo 3PDR .
Speaking in general now: I mean, c'mon, equidistant edge all around the active surface, are you kidding me?! Show me the naked Neo 3PDR polar plot, horizontal and vertical, please. Better yet, call BG and ask them to send you one. Gee...

later:
Now, to be honest, I never thought that developments of OB design will go towards pushing the ideal, double-circle figure-of-eight all the way up. By definition, you can't have it! Not even with naked Neo3PDR above 2k5. Well, you may get something of a sort, if the active surface size itself is large enough to start beaming, narrowing the kidney, but if it does, it will be frequency dependent and then what? Lobeing and further narrowing toward highs. Double-circle figure of eight is acheivable ONLY below the cut-off frequency of the baffle and that's that.
In short, you can't have double-circle dipole response if your baffle is wider than half of the wavelength at the frequency you are looking at.

Personally, even if it could be done the proper way, I consider it unnecessary, as RT60 times in any room can only have severe nonlinearity and prolongation below 300Hz, and in my book, figure-of-eight (OB woofers) serves as a problem solver in untreated, boomy-bass, rooms by decreasing the excitation of the room modes. I still don't see why it has to be pushed all the way up, as it will only decrease RT60 up there, too, maintaining the RT60 unbalance with low frequencies, solving nothing that OB was originally used for.

To me, whether you have a normal, wide baffle double-kidney or an ideal, double-circle, response of an OB speaker above 3k, isn't something that would make much of a difference, unless you live in one of those ultra-modern, designer apartments that I call "fish tanks" because of the way they sound, because idiots designed them, along with putting the damn kitchen wide open, right in the middle, so you could listen your cutlery rattle with bass and your microwave humming along behind you while you listen to your music waiting on your dinner. If you do live in one of those, my recommendation would be to turn to headphones and forget the high-end speakers based audio altogether, OB or not, and spend more time out of the fish tank.

In any case, you will still have a dipole response all the way up with a null on the side in wide OB's. The main thing is to have that back radiation that reflects from the back wall and gives excellent spatial cues for us to enjoy. A very good reading is Linkwitz's website, and I warmly recommend to anyone that hasn't read it, to do so. A couple of years ago, I've listened to his speaker in his home and from that experience, I can't see what's wrong with having wide baffles. It was THE best stereo imaging I've ever heard in my entire life, and luckily, managed to recreate it or exceed it in my listening room.

Anyway, that was my best shot in making a symmetrical true ribbon tweeter with the best vertical response of all ribbons of the same size out there, with 1k5 cut-off baffle with pretty good acoustical shape, with custom mounting flange when needed. I'm also preparing a small dipole driver, a version of 70-20XR that will be in free sale, not just to manufacturers, which my be better if small pear-shaped baffle ribbon would be better for those design requests mentioned earlier.

Now, as I have a principle of not to use DIYAUDIO forums for business, and things I've already said can be regarded as promotion, I won't be posting any more of anything that has to do with my products. I have my own website for that and any info about new products that I will regard as relevant, will be posted there in due time.

So, when I can join you guys, we can talk principles of operation and such stuff like I tried to do in this post and rattle your cage a little bit

Cheers!

[mige0]

Originally Posted by RAAL

Thank you, gents, I appreciate your welcome. I'll try to be useful, even though I'm coming from omnidirectional with OB woofers camp.

Of course, biased we all are, but we shouldn't forget that we're dealing with objectively measurable stuff and objectiveness will have to unleash us from of our biases once we accept the physical truth and compare it to definitions we claim to follow.

So let's talk principles and definitions we so easily throw around.

For example, cuibono, you do have a fairly nice result, but just at a first glance. It isn't anything near of what it should be.

Let's analyze those results in depth, and someone tell me if I'm wrong. I may be and/or the measurements may be wrong.

Definition: If I remember correctly, almost an ideal dipole response must have about -6 db at 45 degrees, -12 dB at 60 degrees and at least -24 at 90 degrees at all frequencies compared to dead on-axis. Correct me if this is untrue, I'm rusty in math department.

Measured result: That "almost ideal" you mentioned should read "+6 dB to the goal", which is double the energy. Doubling things, or adding 100%, is not a near miss. This happens between the crossover frequency to 6k, above which, things get rough.
Ideal dipole response does not account for such lobeing that manages to make 3dB more on 30 and 45 deg and +1dB at 60 deg. than on axis, turning a response with octave wide -6dB dip on axis into a response with 3dB peak at 60 degrees. That is about 12 to 15 dB more than the goal. That is quadrupling things, which again, can't be called a near miss.

Consequence:
The first reflection off the side wall, which is between 30 to 60 degrees, depending on the room size and speakers-listener layout within it, will on average between 6k to 10k, receive the same energy as your listening position. If anything gives spatial cues, that is this frequency band and if anything, dipole shouldn't do that.

Verdict:
Considering that below 6-7k, dipole is not nearly as fully functional, we can call it a moderate success to the goal that has been set. What we have here is a double-kidney equally obtainable with a wide baffle.

Considering that, centered at 8k, we have a cloverleaf pattern of two lobes front and two back, and depression of 3dB on axis, and that on average it has equal energy dispersion up to 45 degrees, falling off only 11 to 12 dB on average at 90 degrees, we can call it a failure.
9k is a very interesting point, as we have equal energy distribution with only 8-9dB drop at 90 degrees. I also suspect that there is a 45-ish degrees phase-shift break there that allows the 90 deg. 9k peak not to be canceled properly with back radiation. It may be due to slots resonance and/or spacing. Examination of anything that may fall into 38, 19 or 9.5 mm with driver's construction will provide a cause to problems between 6 to 10k.
If it is just edge diffraction related, it is correctable with a baffle/frame of enough width to allow for transposing the dip below the crossover frequency and framing the driver in such a way that the face plate becomes flush with frame front and back surface. If the baffle/frame does not have equidistant edges to the active driver surface (slots area), it may be possible to use the lesser width than what the crossover frequency suggests.

Considering that above 13k we have just simple beaming, but working exactly the dipole below baffle cut-off should, we can consider it a full success.

As far as the room interaction is concerned, all of the above is easily corrected with absorbers on side walls at the point of first reflection. If you don't already have them, that shouldn't be hard to do, so try it out and see what you should be aiming for. Let us know is it better. Follow the speaker-listener-room layout that Linkwitz suggested.

As for the audibility of -6dB at 8k, you certainly hear it, just don't know how to pinpoint it's attributes. A very good exercise is to use headphones and try tampering with any EQ from any media player on your computer.

One of these days, I'll make measurements with this dipole tweeter of mine without front plate and we'll see what is it like. I suspect it will turn out to be allota work to try to correct for anomalies if they appear so I'll give report about it and the correcting methods. Right now I'm too busy, but I'll do it, eventually.

Cheers,

[mige0]

Originally Posted by RAAL

Whoa! I'm off by a laaarge margin! Then, very impressive result you have there!

I officially declare the most part of my previous post as **, coming from ignorance and I apologize for being an ****.

Since those are the numbers (thank you), it means that 67 mm wide "baffle" of Neo3PDR will extend the dipole behavior all the way up to at least 6k cleanly.
Now I'm even more interested in the nature of that 9k anomaly.
I strongly suspect the 9k anomaly has mostly to do with Neo3 being bare naked. Can you frame it up somehow, keeping the same overall size, but getting rid of that edge-step and see if it helps? What is that soft, oily, pliable putty that always stays soft and is easily shaped called in English?

I wonder how far it will go, given the size, if the anomaly wasn't there and it makes me wonder how far the dipole ribbon will go, given it's size and pear shape. I thought it won't go far up, but it just might go up enough...I'll do some measurements these days.

You have an interesting observation about 1 to 4k. Most "empty" rooms sound harsh and annoying right there, so having dipole response in that region will certainly help.

Haven't read Toole's book, aside from some papers published in AES Journal, but I know that Linkwitz agrees with side reflections issue.
I tend to agree with that, but only up to an extent, as they tend to smear the focus in the middle of stereo image. I don't know if Toole and Linkwitz specifically mention that, but from my experience, I need to emphasize the importance of having speakers very, very accurately symmetrically positioned, with less than 1cm error between left and right channel. Whatever the distance from side walls, it must be right-left equal, as well as from the back wall. Equally important is the symmetry of furnishing the room. Cabinets, sofas, chairs, tables, anything, wherever it is, especially in region from the wall behind the speakers all the way to the listening position must be symmetrical. Also, move away the table in front of the listener when listening, if it's wide enough to reflect the direct sound towards the listener. Push it all the way to between the speakers, if you can't put it behind you. If you don't, it will force the image down and smear the focus again.

The side wall distance error will easily be heard like sibilants that aren't having the same position in stereo image like the bulk of the vocal. For an inch of positioning error, if the tweeter is about 2 - 2.5 feet from the side wall, the offset is about 4-5 inches in stereo image and it's very annoying.

That is why I consider this 8-9k anomaly very dangerous as it will probably emphasize the positioning error as sibilants are falling right in it.
Omnis are particularly sensitive to symmetry errors in layout, whatever the side wall distance may be, even 6 feet, and you have more of an omni thing going on in sibilants than anything else. (Most people never heard omni speaker accurately positioned in a symmetrically furnished room and attribute the focus smearing to "nature of the omnis". Earl Geddes once dropped in my room at CES with large omnis and didn't like the focus, of course. Neither have I, but he was kinda' deaf to my explanation that because we weren't allowed to move this darn TV cabinet, we had to move both speakers towards right, as well as the listening position, and all symmetry was lost. It still sounded better than having the TV cabinet almost in front of the left speaker. I'll never exhibit anything again at stupid Venetian, BTW).

Dangerous game this is. Many manufacturers when exhibit at shows, place speakers very poorly and furnish the rooms disregarding symmetry. Actually, accurately laid out monopoles will whoop the **** of inaccurately laid out dipoles any day, and you can hear it going on all the time at audio shows.
I know audiophiles spending their entire audiophile "career" listening to poorly laid out rooms and speakers within, spending megabucks on new gear, hoping they'll get lucky, but for some incomprehensive reason to me, fail to listen when you tell them they should play with positioning.
I can't even begin to explain how important the room layout is and how often it's the more decisive factor than the speaker itself.

[mige0]

Originally Posted by RAAL

Hi, twest820!

Right. Cuibono reminded me of it, too, thanks to both.

Even if I was close to the real-life figures, he was referring to ideal dipole, and because of that my discussion is still grossly off target.

As for Clio, if I were you, I wouldn't think twice about it. Maybe it's just me, but I can't use Clio because of it's handling that annoys the hell out of me. I'm a big fan of Praxis, though.

Wow, thank you for the info! 8k5 !!!! Crazy!
That's about 3.3 times it's cut-off frequency!

So, do you have a dip at 8k?
If not because it's EQ'd flat, what does it say off-axis?

[mige0]

Originally Posted by RAAL

Hi John,

You're a highly regarded loudspeaker designer and your work is always thorough, it brought so much useful data to wide audience and it was always discovering the good features of affordable components, giving valuable advice to DIY community.
Surely, you must know what would you like to have, regarding dipole tweeters format. You can't just sit back and let me guess what it is. If you're not happy with current offerings, then please define what optimal is. You may be used to manufacturers that don't listen, but I do listen and try to do my best to serve the needs.
Because of that, I have question to ask you, after I point out one thing.

To the difficulties you noted, I would also add the sensitivity issue, which is disregarded here.
Let's not mix apples and oranges, as 90dB/Watt allows a speaker designer a completely different vantage point than 95dB/Watt.

If your choice of amplifier is, say, 25 Watt Firstwatt J2 or any other 15-25 Watt beauty out there, problems start to jump out of the closet. You'd be aware of the potential, but you wouldn't be able to use it with 90 dB/Watt speaker, and that is very frustrating. You'd want at least 95 dB/Watt speaker to match.

For example, closed 140-15 has 95 dB/Watt (not 2.83V), or more, depending on the foam pads positioning. The dipole version, due to increased air loading mass, has about 1.3 dB less with same response curve.
Experience shows that for a good tone balance of the loudspeaker, they both require about 1.5dB more efficient midrange due to dynamic compression issues on the midrange side and due to wider dispersion on the ribbon side.

The problem is that there aren't much mids out there that are churning out 96.5 or 95 dB/Watt flat responses and the ones that do that, are not small filler drivers, but 8" exotics, like Supravox 215 variety, Seas exotic 8, and such. No planars in the game, either.

The mid will start beaming fairly soon and you can't use a filler driver, because there isn't any variety of 4-5" 95 dB's to choose from.

Just 3dB, let alone 5dB in increased sensitivity demand by the customer, changes the game completely and 50 bucks is not considered a chip any more, due to exponential increase in difficulty of designing and building such a thing. 90 dB/ Watt allows much more liberty and it's easy compared to 95dB/Watt.

So, my question is, if you were to design a dipole 95-96 dB/Watt speaker, having in mind that you'll have to use an 8" midrange, crossing is as low as possible, what would be the tweeter format that you'd like to use?

I think I nailed it pretty close, for what it s intended for, but I'd like to know what would you do, if it ain't a secret.

I'm asking because I may want to provide certain features to the baffle so everybody would be happy.

One possible way to do this is to increase the iron yoke size, but drill a large number of 1/2" to 3/4" holes through it. That way, you can choose which holes to plug and which to leave open, creating different patterns that just might fit any dipole taste, without increasing efficiency.

[mige0]

Originally Posted by RAAL

Since I share this view to the last point, that's why I can't really figure out what would be the benefit of having equal dispersion everywhere, as both rooms and our psyche doesn't accept it equally.
As I pointed out before, all rooms have problems in bass, and some rooms have problems in presence region (centered about 2k5). As cuibono pointed out he prefers dipole at least up to 4k, which falls in line with mentioned room problem.

I just want to add that I stand behind the room treatment and prefer omni down to 200 Hz in such case. Rooms must be treated, or they will ruin a lot of our efforts. However, once appropriately treated according to their size and volume, the results are nothing short of astonishing and the sound is in a different class altogether. Nothing like it, if you want to hear the actual size and feel of the instruments, and horns can't touch it, with their "in your face" presentation. The way I like to put it, they have a big sound, but not big instruments.

One of the people from the whole of audio industry that I admire the most is Amar Bose. He went through the trouble of designing excellent 5" whizzer-less FR drivers and stuffed 9 of them in a clever box, with one shooting forward.
That speaker, with it's dedicated EQ box, kicks *** big time in recreating the proper size of the orchestra and all that it contains.

For me, a wide baffle, big midrange (even 12"), OB is the next best thing in untreated rooms.

[mige0]

Originally Posted by RAAL

I second that about Rudolf's work! The same goes to twest820!

I could, but you should know that a good cone driver does things better up to 1k5-2k than a ribbon ever could dream of. That is why electrostats and big planars sound wimpy with drums, percussion and piano.
In midrange, parachute effect, or large air loading, needs to be controlled with greater force per surface area than any ribbon, stat or planar can give.
Trust me, you're much better off with cones in midrange.
If you want to get truly remarkable results in midrange, use 3" FR driver in clusters of 2x4, 3x3 or 6 pieces in hexa and one in the middle for total of 7. Do not use line sources, they don't add together well, as top driver doesn't know what bottom driver does and vice versa, meaning that they don't share the load and sound equally wimpy just like one feeble 3" driver, only you can cross it a bit lower. Sharing the air loading is the key, a such 3" FR drivers clusters are the way to go. They get so much authority and slam that it;s not even funny. The main advantage to a comparable surface area of just one cone driver, is that 3" FR will have first cone breakup problems even as high as 11k, so you'll end up having a piston of a large area all the way up.
Choose only ones without Farraday rings and such tricks. Look for saturated pole plate and central pole piece. BTW, very few good driver use a Farraday ring as it wastes the signal current on showing you a nice impedance and they need 150 Watts to get dynamics and start sounding lively. Only a few bright examples exist, made by people that understand the drawback of it.

[mige0]

Originally Posted by RAAL

Oh, yes, they certainly are in driving mids and highs! Passive XO between mid and high doesn't hurt it, if designed nicely. It goes without saying that EQ headroom for the low 15" or two, will a big solid state or a D-class.
I assume OB should use EQ, and I assume you need a separate amp for the job that wouldn't be 25 Watts.

As for the number and kind of mids I like to use, please read my previous post.