December 2011

Heil AMT1 And The Heil Horn
Article By Jeff Poth

Difficulty Level

  In my enthusiasm for hi-fi, I have acquired an eclectic mix of toys that find themselves waiting for a project. Few of them get the attention they deserve, but one particularly unique toy is the Heil AMT1. This is a legendary speaker driver that was used in some speaker systems, mostly in the 70s and early 80s. Its unique topology comes from the mind of a brilliant man, Dr. Oskar Heil (1908 to 1994). He came up with 2 particularly interesting inventions, you may have heard of the first- the Field Effect Transistor, or FET. Yep, like the MOSFET you see touted in transistor amplification. Whether he actually "invented" this or not could be debated, as there were several engineers working on similar patents, but he certainly was a major driving force behind the design and the ability to utilize this in electronics.

Here we are concerned with his other invention, AMT, or Air Motion Transformer. The AMT1 is the "biggun", though as the patent protection recently expired there are now a plethora of AMTs on the market, coming from Elac, Adam, Eton, Mundorf, and even Beyma (who produces loudspeaker drivers for the Pro sound market as opposed to hi-fi). Heil had several versions of this driver, but as I said the AMT1 is the biggun, the one that started it all, and as such is the most interesting both historically and from a hi-fi perspective.

The mode of operation of a Heil AMT is unique amongst loudspeakers, in that it's not a "in-out" transducer. Rather, the pleated diaphragm is compressed and air is forced out of the space in between folds like shooting a seed out from between fingers by pinching. The fact that it's folded the way it is and the way the air moves makes the name "Air Motion Transformer" very accurate. The diaphragm in an AMT1 is described as being approximately equal in area to an 8" cone driver- but is a very tiny strip.  This allows the AMT1 to operate both lower, and at a much higher level, than an equivalently sized cone, dome, or ribbon, while maintaining the relatively broad horizontal dispersion. Some Heil speakers crossed the AMT1 as low as 500 Hz, though that seems to be pushing it, to my way of thinking (and is attested to by the high failure rate of AMTs used thusly).

At this point in our article, the Heil AMT1 seems to be nothing more than a curiosity, but when you hear one…. you understand. The AMT is a superb way of reproducing sound, and all the moreso because it's a dipole. There's no box coloration, just a very clean, open, dynamic sounding treble. I cannot overstate the value of a large tweeter. Small domes just don't have the juice to do full-scale sound. They have to move too much and dissipate too much heat. The AMT, on the other hand, has its voicecoil (something of a misnomer here, it's a ribbon attached to the plastic diaphragm) in free air, with a huge amount of surface area both for heat dissipation and for sound reproduction. Simply stated, used within their limits, thermal handling or power compression is not an issue with an AMT. So, what does one do with an impressive driver like this?

Are you really surprised? The AMT assembly is already waveguiding to some extent via the 90 degree walls. There's another interesting horn that has a 90 degree straight profile, though conical (as covered in this article). Since we already have a defined throat section and horn profile, we simply need to extend the walls and smoothly terminate the horn to hopefully achieve two goals- boosting lower frequencies, and a reduction in diffraction. By use of the waveguide, the expectation is that there will be some efficiency jump in the 800 Hz to 2 kHz range. A reduction in diffraction (note how the 45 degree straight termination of the original assembly becomes a large, smooth, rounded profile) and get overall smoother frequency response, particularly off-axis. The Heil as-is has somewhat rough polar behavior, which is likely due to interference patterns from the diffraction of the edge. By smoothing that edge, the source of the interference pattern is eliminated and that should give us a more consistent on-and-off axis response. This is crucial since reflected sound plays a big role in what we hear. The OS Waveguide article goes into a fair amount more detail.

Construction
The impetus for this assembly was the "because I can" factor. I'd always pictured a front/rear horn for the AMT1 in my mind's eye, but it wasn't until I had a compound miter saw that the angles were easy for me. As you can see in the below picture, we were starting with rounded pieces (

This "Edge Round" is designed to act as a large round corner on box constructions, but is useful to this application too. One could do jigs or manage to cut it some other way, but the miter saw made it easy to cut the angles on the top section in the MDF edge rounds. Even so, they still required some cut to fit work to allow the vertical section of the horn (the V-shaped section on top). You can see how a flat panel fits via the Rush LP below. Additionally, the edge-round needed to cease to be round. More on this later.

A bendable side panel will also be required. I did this by kerf cutting, which is cutting grooves most but not all the way through a panel, so that you're only bending a very thin layer. The kerf cuts were performed with a table saw but can be achieved with a straight edge and a circular saw, a radial armsaw (best!) or by using bendable ply, or curving thin layers of materials. If you can't figure out how to do the rectangular 0.5" thick side pieces.... You shouldn't be using anything with spinning blades.

Once the pieces are cut, you must glue them up. The side panel is glued in mine with a simple pressure fit. While not ideal, with appropriate care you can get it firmly attached. Use too much glue and carefully balance it in place.  Same goes for attaching the second piece of the edgeround.

Once you have the side panels and roundovers assembled, you have to flex the kerfcut piece and glue it into place. I suggest gently clamping it into shape before applying glue, then liberally apply glue to both the inset edge on the edge round and the edge of the kerfcut piece. Gently place as though it were a completed assembly, and tighten the clamp, adjusting to get the most surface area in contact. This is not ideal, again, but with appropriate care it works fine. You're working with a flexible section of wood so be careful you aren't overclamping. When everything's glued up you're going to be coating all the weak sections with bondo anyway so it doesn't need to be perfect, just firmly attached.  After attaching one side, you'll use a bunch of wood glue (enough to grab but not so much as to leak out everywhere) and rub it into the slits of the kerf cuts before affixing the next side. You'll then glue and clamp the second edge of this piece as you did the first.

Once all this is done you have 2 side sections. This is the first configuration I tested, leaving the top section off. Between the dipole operation and the off-axis cancellation of the 4" tall source, there is some significant directionality already in place vertically, which reduces the need for a top section. Sometime later, I'll test the top section more thoroughly. For now, I'll give you a pic of the 4 sections with some primer on them.

I Was Wrong
As it turns out, the construction method used and described above was far from ideal for the horn. If I had it to do over, I'd use bendable ply or pre-kerfed stock, and use a single piece per side. This would be faster and easier than what I did. Important to note that with my construction, the initial assembly doesn't show one important feature- the front sections have been flattened to maximize the size of the conical (straight-walled) horn section. A different assembly would have eliminated the need to sand down the rounded parts, and minimized the need for as much filling.

As such, I'll leave the figgerin' to you, my reader, if you should desire to build such a thing. The key components are the extension of, and smooth transition from the 90 degree flare of the original assembly. The extension reinforces lower frequencies by preventing dipole cancellation, and the smooth termination prevents diffraction related bugaboos.

By profiling it this way, I get a less flat response than stock, and there are 2 notable features that need correction with my waveguided amt1. The first is a 5 kHz peak, which is present in the stock AMT1. This gets a notch filter. The second is that the bottom end has now been boosted significantly, and needs to be smoothed back down. Why boost it just to suppress it? Smoother response with much better off-axis performance, and much less power input and diaphragm motion required for a given SPL. It's important to note here that the rougher response is only in measuring a driver without a crossover. There are no added high-q peaks and no rough horizontal off axis misbehaviors (there ARE some stock).

Shaping
From a crossover perspective, the AMT1 is a pleasure to work with. The reason is simple- the impedance is ruler-flat. No meaningful Fs peak, no inductive rise, you might as well be working with a resistor. So below are two graphs indicating the shapes and the behavior of the corrective measures. These are mocked up so don't mind the lack of resolution. The behavior of the tweeter is very smooth, and would look much better with a more resolute graph. It is from 1 kHz to 20 kHz (+/-1dB ) with filters in place, very good performance by any measure. More importantly, it's consistent in this regard. Constant directivity? Naw. It's too complex for that, especially given that the diaphragm is tall. But it's very close, within its coverage window. Certainly close enough that there's a very wide range of listening positions possible in the horizontal plane. Vertically… well, you better have it close to ear height.

Sound
As my wonderful readers already know, I'm a little bit hands-off on the sonic descriptions. There's a good reason for this- it's boring reading to hear someone wax poetic about their latest and greatest. The advantages to the Heil, it is clean, unfettered, dynamic response are retained, with even better dynamic performance in the mids due to the reduced demands, and more consistent imaging as one moves around. It's very open and airy, due to the good high frequency extension and dipole behavior. I can only speculate that it'll improve further as I move further down the dipole road.