Creating a Soundstage with Waveguides and Psychoacoustics

[Patrick Bateman]

In this thread I am going to demonstrate how to create a convincing soundstage using waveguides and psychoacoustics. This thread will cover the construction of a loudspeaker in a car. However the experts on this subject rarely participate in car audio forums, so I'm going to delve into the subject here. (Hopefully the moderators don't mind.)

Here's some background on the project.

Some people may have seen my Unity horn projects, which were also in the car. Recently I attempted to take some of the "unity goodness" and scale it down. Unfortunately, it was not physically possible to do this, and I scrapped the project.

This new project will be similar to the last one. I'll be building an affordable, relatively accessible loudspeaker using waveguides.

This will not be as outrageous as my Unity horns, but it will also be a heck of a lot easier to build (hopefully!)

A number of people expressed interest in creating an excellent soundstage, but weren't willing to put Unities in the car.

So if you fall into that camp, this thread is for you. In addition, if you just want to learn some psychoacoustic "shortcuts" to create a good image, keep reading...

[Patrick Bateman]

For a little background on my previous projects, which are all related, see these URLs:

Latest & Greatest :

http://www.diymobileaudio.com/forum/...oundstage.html

The Unity from 2006:

http://audiogroupforum.com/csforum/s...ad.php?t=62789

A lot of good discussion on Unities and waveguides:

Another Unity Horn

The most recent attempt at an "accessible" alternative:
Not Another Unity Horn

[Patrick Bateman]

Like a lot of my car projects, this one is heavily based on the waveguide work done by USD audio. In a typical USD audio setup, there are horn-loaded compression drivers under the dash, midbasses up front or in the quarter panels, and subs in the trunk.

IMHO, the main reason to use horn-loaded compression drivers in the car is that it gives you enormous dynamics and it "equalizes" the path lengths.

For instance, if you put speakers in the door of your car, the pathlength difference between the left and the right will be a couple feet. With a waveguide under the dash the pathlength difference can be as little as 4" or so. That's the trick - it creates a center image by pushing the speakers very VERY far back.

Here's a picture from USD's web site of the mold for the horn in what is likely the best-known installation in the history of car stereo competition:



This project will use a similar configuration. Instead of an uber-expensive compression driver, I'm going to find out if we can get comparable results using more practical components.

[Patrick Bateman]

When people use waveguides in their car, they often have two common complaints. The first is that the soundstage is too low. The second is that the soundstage is too narrow.

There's a few things that can be done to fix this:

• Move the waveguides onto the dash, a la my Unity horn project
• Mount the waveguides under the dash, but point them up. This is met with limited success; it's still easy to tell the stage is low.
• Modify the frequency response to 'fool' the listener. Also an improvement, but with limited success.

In this thread I'll explore an alternative. We'll leverage recent pyschoacoustic discoveries to seperate the frequency range.

In other words, we'll use two waveguides, physically seperated, with a waveguide for the midrange and a waveguide for the treble.

It's not as elegant as the Unity horn, but a heck of a lot easier to build (and hide.)

[Patrick Bateman]

According to "Spatial Hearing"*, The opinion of the overwhelming majority of authors on this subject is that interaural time differences are the most important attributes of the ear input signals relating to the formation of lateral displacements of auditory events.

This single fact is why equalizing the path lengths is so effective in creating a phantom center channel in the car.

While this is well known, and frequently exploited in numerous winning designs, many find that the treble is lacking. The soundstage is too low, or the high frequencies suffer from audible colorations.

In the past few years there's been significant research into these hearing mechanisms, and mounting evidence that interaural time delays are virtually irrelevant above 1600hz.

For example, on page 172 of Spatial Hearing we find that "for signals with components above 1.6khz, the intensity image may dominate even when there are also low frequency components." (italics mine)

This aspect of hearing is why the left and right channel must match above 1600hz. Even a small difference in intensity will disturb the soundstage.

Yet ironically, it is the range about 1600hz that's prone to reflections in an automotive environment.

So we will get that range above the dash, where reflections are minimized, while minimizing pathlength differences via waveguides under the dash.

* Spatial Hearing, By Jens Blauert, John S. Allen

[Patrick Bateman]

I've finished the plans for a midrange horn which will go under the dash. It was a lot more work than a tweeter horn, because it's a challenge to get flat response with such an under-sized horn.

In my previous designs I've used the dash to extend the curve, but since we're going UNDER the dash, we don't have that luxury. The dash WILL help extend the low end a bit, but nowhere near as much as when you couple with the windshield and the dash.

The number one goal of this design is to push the midranges back as far as possible, to equalize the left and the right pathlength. Based on that goal, I managed to cram a 500hz horn into a depth of just 6.75"!

I literally spent hours juggling the numbers, this is one of the trickiest horns I've ever made. When it's finished it will look a lot like a USD waveguide, but with a midrange where the compression driver normally goes. Here's a pic of mine:



Here's an illustration of the horn:

[Patrick Bateman]

Here is a "printable" version of the horn plans from above.

Basically print it out on your printer, transfer the shape to wood, and make sawdust.

To make it fit in 8" x 10", I had to cut out some parts, but it should be obvious if you look at the previous pic.

The height of the horn is 3.5" internally, except at the mouth. At the mouth it flares to 16" wide x 4.5" tall. The dimensions of each segment are printed in the illustration.

I will post horn response simulations shortly, and pics of the actual horn.

[trusound]

most current horn guys... well those who go about using current best practices in a car.. well tell you to get them as low and and wide under the dash as you can... to use the dash as an extension of the guide and to get the motor as wide as you can... most strive to get the motors to tuck into the kick panel cavity

[Patrick Bateman]

Originally posted by trusound
most current horn guys... well those who go about using current best practices in a car.. well tell you to get them as low and and wide under the dash as you can... to use the dash as an extension of the guide and to get the motor as wide as you can... most strive to get the motors to tuck into the kick panel cavity

Hopefully we can best the "state of the art."

Pushing the horns back increases the path length. That increases stage depth, and makes the phantom center "more solid."

Some things that we'll do here to improve upon the state of the art include:

• We'll use Geddes-style reticulated foam to reduce higher order modes. This is important in a waveguide, but even more so in a horn. A waveguide reduces HOMs by carefully reducing diffraction. Our horns, due to their contour, will have higher diffraction than a waveguide. Therefore, the foam is particularly important.
• Due to reflections and diffraction, it isn't possible to make a horn sound like the sound source is at the throat. Typically the sound appears to emanate at a point near the mouth. But the aforementioned technology should improve the impulse response to a point that the soundstage is much deeper than we're accustomed to in a car.
• Because this is a midrange horn, not a compression driver, we should play an octave lower than what is possible with even the best compression drivers. Of course we're sacrificing extension to get there.

[Patrick Bateman]

OK, don't bother making the waveguide I posted a couple days ago, it doesn't fit.

Basically it gets in the way of the clutch pedal. It might work OK if you drive an automatic.

So I went back to the drawing board, and came up with a 70.56 x 18.92 degree conical waveguide. I didn't even bother with a CAD drawing, just created a mold and glassed it.



From left to right, there's a tractrix horn from my 2003 project, the new waveguide described above, a USD audio waveguide with a Radian 450PB compression driver, a mold for the new waveguide, the horn that I published two days ago, and a hunk of foam which was whittled down to create the waveguide mold.



On the right is the new waveguide mold, in the middle is the USD waveguide.



The new mold looks a lot like this: