An Improved Paraline.

I've built a bunch of Paralines*, but I wasn't satisfied with the sound. A lot of dips in the response, and a character that was generally harsh.

I believe the problem is that the sound radiated from the diaphragm hits the bends in the Paraline and is reflected back to the throat. The reflected wave is a higher order mode. HOMs make a horn sound harsh.

So...

I wanted to use some of the same tricks that we use with waveguides to reduce higher order modes. I wanted to use gentle curves to reduce reflections back into the throat.

Screen%252520Shot%2525202016-01-16%252520at%25252012.13.18%252520PM.png


The first thing I do is draw that familiar 'eye' shape. The point of this shape is to take a spherical wavefront and flatten it into a ribbon shaped wavefront...

* For more info on how Paralines work, read my threads here:

http://www.diyaudio.com/forums/multi-way/133745-i-dont-understand.html

and here: http://www.diyaudio.com/forums/multi-way/217298-square-pegs.html

and here: http://www.diyaudio.com/forums/multi-way/225832-stargate.html

Those are in chronological order, but the real 'eureka' moment was the second thread. Danley has comments in all three, which are valuable insight into the devices.
 
audio-rendering-from-sound-diffusion-to-sound-projection-27-638.jpg

The L'Acoustic VDOSC was patented in the 80s iirc; you can see it's shape in the top left. Geometrically, it's perfect, but I wonder if those abrupt transitions will create higher order modes.

psa1_components.jpg

Danley Paralines are similar, but with some new tricks thrown in, such as the ability to generate various wavefront shapes, not just 'ribbon' shapes.

Screen%252520Shot%2525202016-01-16%252520at%25252012.29.58%252520PM.png

My device is using a diffraction reducing curve, similar to what JBL and Genelec use. Basically the way that this curve works is that the exit angle is 90 degrees, but it slooooooooowly expands to 180 degrees at the mouth. The Geddes speakers work in a somewhat similar fashion, but the curve on the Geddes speakers changes more abruptly; the Genelec in particular is verrrrrrry slow.

Genelec-G-Five-tweeter.jpg

Here's a pic of the Genelec waveguide, you can see the curve is the same as what I've drawn.
 
Screen%252520Shot%2525202016-01-16%252520at%2525201.04.24%252520PM.png


I took some liberties with the Paraline idea here.

In a Danley Paraline, the waveguide is an 'eye' shape, and that eye shape is a very specific shape. The distance from the center of the eye to the top of the eye is EXACTLY half the distance from the center of the eye to the left of the eye.

paraline.jpg

Here's one of the illustrations from the Danley patent.

So...

Why did I use a rectangular shape?

I did this for a few reasons:

1) A plain ol' rectangle satisfies the requirement that the distance from the throat of the waveguide to the top of the waveguide is half of the distance from the throat to the left of the waveguide.
2) While it's true that the distance is too long on the obliques, I don't really care about the obliques. I'm just trying to get the correct horizontal coverage. I don't care about oblique coverage.
3) By far the main reason that I did it is because the area of a rectangle is about 27% higher than the area of an 'eye' shaped Paraline. I believe that additional 27% should smooth out the overall response. (It is my belief that the very small volume of a Paraline is one of the reasons that the response is ragged. You can model this in Hornresp; very small horns have a lot of ripple. An additional 27% of volume should smooth out the ripple.)
4) The last reason that I used a rectangular shape is because I intend to curve the wavefront so that the listening axis is 30 degrees off the center. IE, if you're in front of this waveguide, you're 30 degrees off axis to the listening axis. And if you're 30 degrees off axis, you're on the listening axis of this waveguide. Hope that makes sense. The idea here is that you could put this speaker in front of you and it will 'throw' the sound to a spot that's NEXT to you.
 
Hi Patrick:
Have you gotten around to 3D printing a paraline yet?

Yeah, it was one of the first things I printed with my 3D printer.

At the time, I wasn't aware that 3D prints are porous.

So all of my early 3D prints are basically worthless, because they don't work until you seal them up. That's why you'll notice all my new ones are slathered in glue, I do that to seal them.
 
Screen%252520Shot%2525202016-01-16%252520at%2525201.32.04%252520PM.png


Screen%252520Shot%2525202016-01-16%252520at%2525201.32.34%252520PM.png


In these pics, I've added the center piece. This center piece is functionally equivalent to the center piece in a paraline. But a Paraline is two dimension and this is three dimensional. The V-DOSC is also three dimensional.

Screen%252520Shot%2525202016-01-16%252520at%2525201.32.49%252520PM.png

Here's a pic of the center piece with the waveguide that surrounds it hidden from view. Tolerances here are insanely tight; that gap is just five millimeters(!) Here's where the 3D printer comes in handy...

cdx1-1425%2520and%2520paraline%2520phase%2520plug.png

Here's an illustration of a sideview of a Paraline that I drew up for the 'Square Pegs' thread a few years back. What I'm doing today is the same idea, but it's three dimensional, so that the bends here are no longer 180 degrees, they're just ninety. Slower transitions means less diffraction.
 
  • Like
Reactions: 1 user
Founder of XSA-Labs
Joined 2012
Paid Member
At the time, I wasn't aware that 3D prints are porous.

Patrick Bateman, not sure what printer software you are using but on Simplify 3d, select in the options tab to apply something like 3 solid layers on surface boundaries to make airtight. You may also want a manual over-ride on the cooling fan or to slow down the filament laydown speed and use a bit less cooling to let the filament fuse together longer before hardening. If you do 1 layer it will probably leak. You can tell it's airtight when the surface filament melts together to form a smooth glossy surface. There is a lot of little little adjustments in 3d printing that makes these things go right and a lot of it is on the job training. Calling the tech support at the printer software company is often useful too. You don't need glue to make airtight.
 
Screen%252520Shot%2525202016-01-16%252520at%2525203.55.12%252520PM.png


Screen%252520Shot%2525202016-01-16%252520at%2525203.55.41%252520PM.png


One 'funky' thing I'm doing with this waveguide is that it's designed to bend the wavefront. If you're immediately in front of it, the sound is 'thrown' fifteen degrees off axis. I'm doing this by tweaking the pathlengths in the waveguide, to create a curved wavefront. So the device itself has a flat face, but the wavefront that comes out of it is shifted fifteen degrees off axis.

The pics above show the 'normal' curve on the left, and the 'tweaked' curve on the right.
 
Screen%252520Shot%2525202016-01-16%252520at%2525204.33.52%252520PM.png


Screen%252520Shot%2525202016-01-16%252520at%2525204.33.06%252520PM.png


We're pretty close to finished here. I wanted to obliterate any rough edges, I wanted the waveguide to look like a bar of soap. I *think* those 180 degree bends in the Paraline and the sharp edges of the VDOSC are creating diffraction.

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

Here's another pic of the VDOSC for comparison

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

While unintentional, it's starting to look like the speakers from Funktion One
 
Wow, this is possibly the worst performing waveguide I've ever made:

whisper10.jpg


It fails on so many levels:

1) there's a huge high Q peak in the midrange. Likely caused by horn gain, but normally horn gain is much broader. I'm guessing it's narrow here because the expansion rate is so high.

2) The device doesn't do what it's supposed to; it was supposed to curve the wavefront so that it's louder OFF axis than ON. But it doesn't do that.

3) at 3000hz you see a combination of peaks ON axis and nulls OFF axis. I'd speculate that what's happening here is that my attempts to alter the pathlength caused it to sum ON axis and create a null OFF axis.

Well that was a waste of about twelve hours of work...
 
Wow, this is possibly the worst performing waveguide I've ever made:

1) there's a huge high Q peak in the midrange. Likely caused by horn gain, but normally horn gain is much broader. I'm guessing it's narrow here because the expansion rate is so high.

2) The device doesn't do what it's supposed to; it was supposed to curve the wavefront so that it's louder OFF axis than ON. But it doesn't do that.

3) at 3000hz you see a combination of peaks ON axis and nulls OFF axis. I'd speculate that what's happening here is that my attempts to alter the pathlength caused it to sum ON axis and create a null OFF axis.

Well that was a waste of about twelve hours of work...
"Patrick",

Glad you posted the results, they were about what I had predicted in my head while reading the first posts.
The problems stem from the bifurcated (quadfurcated?) paths exits being much more than 1/4 wavelength apart, this causes the massive comb filtering. The large lower peak results from the horns combining constructively where the wavelength is longer, but the short expansion does not load very low, so drops off rapidly.

Anyway, the horn looks cool..

If it makes you feel any better, I have recently put about four times the time you did on your horn testing various combinations of multi-cell "full range driver" horns that resulted in similar response problems as yours.

Art
 
Last edited:
BMW-i3-production-car-dash.jpg


The thing that I'm trying to do is create a device which will 'throw' sound forty five degrees off axis. This is for a car.

The obvious solution is to build a waveguide and simply tilt it 45 degrees. The problem wit that is that it gets big in a hurry. To control directivity down to 1000hz you need something that's 13.5" wide.

The reason that I want a flat front is that I want to put it under the dash or above the dash.

So, ideally I'd want something that's shaped like a horizontal ribbon, but which emits a curved wavefront. I could do it with DSP, but then you're looking at something like sixteen channels of amplification and DSP for a stereo pair of speakers. (Because the individual drivers would have to be delayed to create a curved wavefront.)
 
Founder of XSA-Labs
Joined 2012
Paid Member
Wow, this is possibly the worst performing waveguide I've ever made:

whisper10.jpg


It fails on so many levels:

1) there's a huge high Q peak in the midrange. Likely caused by horn gain, but normally horn gain is much broader. I'm guessing it's narrow here because the expansion rate is so high.

2) The device doesn't do what it's supposed to; it was supposed to curve the wavefront so that it's louder OFF axis than ON. But it doesn't do that.

3) at 3000hz you see a combination of peaks ON axis and nulls OFF axis. I'd speculate that what's happening here is that my attempts to alter the pathlength caused it to sum ON axis and create a null OFF axis.

Well that was a waste of about twelve hours of work...

Patrick Bateman - thanks for showing your dirty laundry. Takes guts to show big fails. But if there is any consolation I, shared my Paraline measurement in your square pegs thread a while back and I think mine were even worse.

I think Weltersys has some great observations of the cause of your peaks and dips. Can we see a photo of the printed WG and what driver you are using and how it's mounted? I think the main dip at 4.7k you have may be just a reflection cancellation from the driver to horn mouth impedance mismatch or it is the throat to driver cancellation dip, in either case you have something that is spaced 1.43in apart that is causing this. Is the horn axial length 1.43 in? Or throat diameter or radial distance from throat to first main turn at centerbody?

I have no idea what the overall scale of this WG is.

Thanks
 
Last edited:
Hello everyone,

I'm resurrecting an old thread but I need to understand one thing about the Paraline.

The "oval" cat's eye shape of the Paraline has the purpose of giving the "correct" or rather intended directivity of the system out of the slot.

But what would happen if we gave it a circular shape ?

Would it expand the vertical directivity ? To the expense of the horizontal one ?

Please, enlighten me !
 
For the most part, you can get whatever vertical beamwidth you'd like from a Paraline.

The "eye" shape yields a vertical beamwidth that's basically zero degrees.

A round Paraline would yield a vertical beamwidth that's negative. It would behave like a parabolic reflector.

Basically, the "eye" shape gives you a flat wavefront. If you "squeeze" the sides you'll get a wider vertical beamwidth. If you expand the sides you'll make the wavefront converge.
 
In a different thread, Patrick said that the aperature shape of a Paraline could shape the dispertion to almost any direction wanted. This one, I think:


Square Pegs


He explains it better in this thread.

One thing to keep in mind is that the directivity control is dependent on the size of the device. IE, if the exit of your Paraline measures 7" in height by 1" in width, it can control directivity down to 1928Hz on the vertical axis and 13500Hz on the horizontal axis. There's no getting around physics.