Category: Altec 1005B – Multicellular Horn Replica

Initial Prototype:

From the previous post (here), we discovered potentially problematic resonances at the Mouth Section (due to the low wall stiffness and large span.) We identified, through simulation, the first three vibration modes to be quite low and will probably require treatment.

Several methods of treatment were discussed:
1. Thicker Walls: increases wall stiffness
2. Damping Material: absorbs vibrational energy
3. Metal Stiffener: increases wall stiffness (where thicker walls are not possible)

An initial prototype, with the simplest geometry, was designed and printed. This is the baseline design and had thin walls, no damping, and no metal stiffener.

Test Fixture:

A test fixture was designed and built to facilitate testing. The compression driver for this test is the wide band AXI2050 (capable of reaching 500Hz).

An adapter transitions the CD exit to the rectangular geometry of the Mouth Section. The planar expansion area is shown below. Unfortunately, the throat adapter could not continue the exponential geometry of the Mouth Section. A slight linear reduction was needed to match.

Damping Experiment:

Felt damping material was taped onto the mouth, near the mouth, and near the throat. Sweeps were taken to gather initial data.

Test Setup:

The horn was placed vertically with the compression driver resting on the floor. Fixture Base was not used. Microphone was placed directly in front and centered on the mouth.

Care was taken to swap out each configuration without moving the compression driver or microphone. Capture settings and gains were kept identical for all measurements.

Frequency: 300 – 15000Hz (no smoothing)
Gating: 300ms

Raw Measurements:

In the FRQ (frequency response), we can see response variations mostly below 1500Hz. There are several resonance dips especially for the “As Printed” config. Above 1500Hz, response variation between configs is relatively small at 4000, 5100, and 11500Hz.

All configs show a big dip at 2700Hz for all configs and this appears to be from mouth diffraction.

The phase graph is useful for picking out resonances, appearing as dips / spikes surrounding the problem frequency. The pattern is similar to the above FRQ.

Measurements (300-1500Hz):

Below, the results have been truncated down to 1500Hz to reveal more details. Configs are individually plotted with the baseline as reference. Three resonance dips are clearly visible and with various degree of improved based on damping.

Baseline: Three dips observed at 540, 760, and 1000Hz which seem to correspond to the modal simulation.

Mouth Damping: Adding felt around the mouth reduced all three dips. But there is still something going on around 540 and 1000Hz.

Near Mouth Damping: Moving the felt down smoothed all three dips even better.

Near Throat Damping: 540Hz dip is back while the other two are gone. This aligns well with the modal simulation as the mouth opening is allowed to resonate again.

Modal Simulation / Correlation:

The three resonances above (540, 760, and 1000Hz) don’t really line up with any of the simulation results on the previous page. This is to be expected because the physical model is very different from the simulation geometries. More material is added at various places and each side of the horn is actually slightly different.

I decided to rerun the modal simulation again with the actual horn geometry. I had to split up the horn into four triangular quadrants (Top, Side, Side, Bottom) as the mesh was too big otherwise.

For each quadrant, the splitting flanges and mouth flange were fixed to eliminate torsional and bending modes. Simulation results are shown below grouped by modes. Each quadrant showed slightly different resonant frequencies and a simple average was taken to arrive at a single value. This, surprisingly, actually got within 50Hz of the physical values! I would call this data “usable” for now and have some confidence simulation can be used in design.

Study Method / Setup:

We will examine the behavior of a single cell in the 1005B and, in fact, only one side wall of a single cell. The assumption is that all cells and all cell walls are identical.

The image below shows a typical “wedge” for this study. Three ends are fixed as shown. A static deflection simulation will be performed by applying pressure onto the entire interior wall. A modal simulation will be performed with the same constraints minus pressure.

Note: This assumes the Mouth Opening is unconstrained by adjacent cells and should represent the worst case free deflection.

Assumptions:

The following properties are assumed for the simulation. Only elastic properties will be used to due low deflection.

The original 1005B is assumed to have cell wall thickness of 20ga (~1.0mm). This is the best guess I could find online.

For configurations with stiffeners, the bond is assumed to be ideal line-to-line contact with no glue thickness.

Design Configurations:

Many brainstorming concepts are shown below, ranging from the simplest case to rather elaborate designs. Hollow internal features and various stiffener configurations were explored.

Static Deflection Results:

ANSYS was used to evaluate the design. The mesh was generated using the automatic tool (manual refinement was not done.) As visible below, the software defaulted to a mix of quad and triangular meshes.

Pressure of 1Pa was applied to the internal surface (equivalent to 93dB SPL). This should represent a home listening level. As reference, the Altec 288-16K compression driver is capable of producing 110dB SPL @ 1W (or 6Pa @ 1W) and has a maximum input limit of 15W!

The resultant deflection from the 1Pa pressure is minuscule and well within the linear region. As an example, the Steel – Baseline config shows a maximum deflection of 4.4×10^-8m or 44 nanometer. We can expect the displacement to increase for higher operational pressure but it should still be within the linear region.

The weakest point is at the mouth opening for most configs. This is expected due to the large span. PETG configs are much weaker than Steel but adding cross section helps greatly. PETG configs with metal stiffeners are able to further reduce maximum deflection by an order of magnitude.

Modal Simulation Results:

The modal simulation output and first four deflected shapes for the Steel – Baseline config is shown below. The first mode, lowest frequency, is located at the mouth and higher modes involve material farther away.

Interestingly, the third mode is transverse across the mouth. I’m curious if this mode can actually be excited by an acoustic pressure wave since the air displacement volume cancels out. Many such transverse modes exist in the simulation.

Conclusion:

In summary, eleven different configurations were tested for both static and modal behavior. The configs started as constant wall thickness cases to establish baseline and iterated into more elaborate designs. Majority of the effort was made to minimized static deflection in the mouth region.

Note: many material assumptions and mesh simplifications were made to simplify the simulation. The results should only be used for comparative study and have not been correlated to physical behavior.

We are able to infer the following static deflection behavior:

• For PETG only configs: adding material at the mouth is able reduce max deflection. Some configs are able to match max. deflection of the baseline.
• For PETG + stiffener configs: max. deflection can be further decreased 10x of the baseline. Is this actually beneficial for acoustic performance though?
• Deflection Zone: max. deflections typically occur at the mouth where the span is largest.

We are able to infer the following modal behavior:

• For thin wall and PETG only configs: low resonant frequencies are observed. Many have the 1st Mode below the 500Hz intended cutoff. Maybe this means they won’t be excited? But for many, the following three modes fall between 500 – 1000Hz and can definitely be excited. It will be important to find damping material that can be effective at these frequencies.
• For thick wall and stiffener configs: First resonance is pushed much higher to the 500-1000Hz region and following modes are way above 1000Hz. This is beneficial as higher frequencies as easier to absorb with thinner damping material.
• Modal Zone: All first four modes occur near the mouth.

Next Steps:

The above result confirms my initial suspicion that the mouth section can indeed be problematic from a rigidity and resonance perspective. There are several configs that look hopefully (especially the stiffener configs.) It is optimistic to see that adding local cross section in PETG is able to match the mouth stiffness of the baseline.

The next step is to start building some of these configs.

Original Manufacturing Method:

The original multicellular horns were manufactured using sheet metal brazing. Individual cells were first assembled consisting of the four sides. Two of the sides have stamped flanges to allow the braze to fill. This is visible in the images below. The four sides were then placed into a fixture and carefully hand brazed by skilled technicians.

Damping material was then applied over the majority of the cell. The ten cells (ie: for building a 1005B) were then fixtured into the correct position and brazed into the final assembly. I assume there is leeway to “massage” the length and shape of each cell to nicely fit each other. A sheet metal cap is then added to the mouth region which stiffens the edges, fills the in-between gaps, and provides a nice cosmetic finish.

On the throat end, cell-to-cell regions were brazed to seal the gaps. A cap (appearing to be brass casting) is then brazed around the terminate to allow attachment of the throat adapter. The multicellular is now essentially complete. This is all very labor intensive work! It requires many tools to correctly manufacture and assemble the components. To achieve this quality is quite incredible and I wonder if it is still a feasible product today.

Subcomponent Division:

While it is possible to completely replicate the original method, it is the 21st century, and we now have access to more tools than ever. The goal of this project, as previously stated, is to explore 3D printing as the primary manufacturing process. It will allow prototyping and manufacturing of near-net-shape components requiring minimal labor.

The first step is to decide how to split the 1005B into smaller chunks for manufacturing. The entire horn is much too big to fit onto a standard 3D printer. In this case, the size limit is a Bambu Lab X1C with a moderate print volume of 256 x 256 x 256 mm (~10.25 x 10.25 x 10.25 in).

I decided to attempt splitting each cell into three pieces (mouth, mid, and throat section). The mouth and mid section should ideally be the same for all ten locations.

The throat section should ideally be combined to capture the detailed geometry and thin walls where sound is split. The region is of particular interest because it handles the transition from plane wave to “spherical” wave. It will be interesting to optimize this transition and be able to do so without rebuilding the whole prototype.

The throat adapter is for expanding the CD (compression driver) exit to the correct rectangular shape. Building this as a separate piece will allow experimenting with different expansion rates and also CD’s with different exit diameters.

Assembly Method:

The horn should be easy to assemble using reversible fasteners. The preference is to use machine screws with threaded inserts. This will allow the horn to be dis-assembled without accumulating wear on the plastic.

Each section needs to have a tight fit to ensure acoustic performance. Gaskets will be used at joints to ensure seal.

---

---

---

---

Wall Thickness / Damping / Acoustic Performance:

At this point, the biggest question is the acoustic performance of the 3D printed parts. There is an almost infinite combinations of wall thickness / infill / damping / stiffener to experiment with. Intuitive, the mouth section appears to be the most challenging as it interacts with the lowest frequencies and has the thinnest walls with largest span. Wall resonance here seems likely.

The other sections are physically smaller, don’t have such space constraints, and can be easily made stiffer. They are, however, subject to higher pressure from the compression driver. This may become an issue at high levels but shouldn’t be for home use (maybe I need to clarify my design requirement). I’m also curious if the lower elastic modulus of plastic will regardlessly color the sound (compared to a metal surface).

Next Steps:

It is perhaps worth evaluating a smaller section of the horn instead of building the entire 1005B at once. The mouth section will be investigated first for wall resonances and design alternatives. It will be a good read on feasibility of 3D printed horns.

Starting Point / Available Information:

We will make the following assumptions: 
- Cells follow pure exponential expansion to provide even loading and sound distribution. 
- Cells are square and identical to each other and also identical between 1005B / 1505B. 

This should appear obvious by examining the various resources online.

The official dimensions from the Altec datasheet are shown below. We can deduce which dimension corresponds to which axis by comparing the differences between the 1005B and 1505B.

The width is 30in corresponding to the common 5-cell direction of the horn. The depth is 13in / 18.5in corresponding to the 2-cell and 3-cell direction. The height is the 17.25in / 16.75in dimension which also includes the brazed on adapter plate.

At The Mouth:

Markus Klug (who builds beautiful handcrafted wooden replicas) shared the 1505B cell mouth dimension as ~16cm (~6.3in). From the datasheet above, we can also deduce each cell is ~6.5in in height.

We observe that Altec first assembles the cells into “columns” with the minimum gaps across the horizontal joint lines (red lines, below). Each column is then joined the adjacent columns. This creates a wedge shaped region (black lines, below) which is filled in. The size of the wedge is defined by the outer two contact points.

At The Throat:

The cells are joined in a “pseudo-spherical” manner and brazed to get everything sealed up. The adapter plate is planar and seems to assume a planar wavefront from the compression driver.

We can approximate the cell throat dimension by examining the throat adapter. The 1005B throat adapter is Altec PN 30210 which has external dimensions of “125 x 120 x 110mm (4.9 x 4.7 x 4.3in)” (according to various online sources.)

Using a front-projection image, the throat opening width can be calculated to be around 80mm (3.15in). This gives roughly ~16mm for throat opening size of each cell (ignoring off-axis projection).

Virtual Pivot:

Using a topdown image (below) and drawing a centerline through each cell, a virtual pivot can be observed beyond the length of the horn. Note: the datasheet dimensions do not include the throat adapter.

In Summary:

1005B Skeleton Sketch:

The info above is sufficient for creating a skeleton sketch of the horn in 3D. This represents the placement and stacking of each individual cells within the 1005B. After a lot of tweaking, the coverage angle is able to closely match the original datasheet. The cell-to-cell mouth spacing was slightly increased to allow thicker walls. This is important since the intentional is to design for 3D printing.

Cell Design:

The specific exponential flare rate can then be found using the above mouth, throat, and length dimensions.

S = S₀e^mx

Then the side wall profile can be calculated, plotted in X/Y coordinates, and imported into CAD. Using a higher point density will improve curvature quality near the mouth.

Initial Hornresp Simulation:

The first sanity check is to simulate a simple exponential horn with Hornresp.

Three sets of S1/S2 surface area values (representing a single cell, 1005B, and 1505B) are plotted below for acoustic impedance and power.

---

Single Cell Results:

---

10x Cell Results:

---

15x Cell Results:

---

All above configurations have the same F12 cutoff frequency at 286.5Hz since they have the same flare rate. The advertised operational cutoff frequency of 500Hz is ~1.75x the F12 which respects the typical horn design rule-of-thumb.

The response ripple is typical of a finite horn termination and is more severe on the single cell. By itself, the single cell is not recommended to be used down to 500Hz. However, this is where the multi-cell benefit comes in: The combined surface area of all the cells makes for a much smoother response . The response ripple decreases significantly and is more suitable down to 500Hz. “Strength in numbers” is applicable here.

The results here make sense and aligns with the original Altec datasheet. I’m sure those fine Altec engineers knew exactly what they were doing. But it is amazing that they were able to get here with rudimental equipments over 75 years ago.

Note: Many simplifications were made with this simulation. The driver surface area (S_d) was assumed to be equal to the horn throat area (S₁) .This is not realistic but is simpler than creating throat transition geometries which will definitely affect the result. The acoustic power noticeably increases with the larger S_d. Driver parameters were also simplified to avoid high frequency roll off.

Patent Excerpts:

We can refer back to the original patent by E.C. Wente for more insight. The entire four page patent is worth a read and contains great details. An excerpt of the patent is included below:

Note: The horn described by E.C. Wente converts a planar wavefront to spherical by having curved cell walls. This is not the case with the Altec design which does not really treat this issue at all. It will be interesting to see the performance difference between these two designs.

Additional Reading:

Acoustic Devices
(US Patent#1992268) by E.C. Wente

Quarter Wavelength Loudspeaker Design
by Martin J. King

Introduction to Horn Theory
By Bjørn Kolbrek

Inspiration:

I have always admired the Altec “Voice of the Theater” since first becoming interested in horn speakers. The design is simple, practical, imposing, and almost intentionally brutalist. It clearly shows its intent: to deliver a wall of sound to your face.

The image above shows the Altec A5 with two separate speaker unit. The top half is the high frequency unit (1005B multicellular horn) and the bottom half is the low frequency unit.

Buying and owning one seemed highly impractical.
Where would I even put something this big? Wait, how much does it weigh? 
Hmm, maybe I can buy the top half and build a smaller bass horn? 
I mean, the top half doesn’t look that complicated. Maybe, I can just build one?
Ya! Let’s just build one!

Rational reasoning falls apart when one tries looking for a deeper meaning in these projects. I guess this will still be a good engineering exercise to learn more about classic exponential expansion and get into the really nitty-gritty of horn building.

Why Altec 1005B?

The 1005B was chosen because it was the most “moderately” sized of the Altec Multi-cell catalog. The naming convention is as follows:

The 500Hz cut-off horns are actually smaller than the 300Hz series due to the shorter wavelength at cut-off. Examining the brochure above, all 500Hz single cells “appear” to be the same dimensions. This doesn’t appear to apply to the 300Hz series. Especially not for the 203B which is significantly larger than the other two models.

One of my primary limitation is the size of my 3D printer which is the Bambu Lab X1C. Rough calculations show that a single 300Hz cell mouth won’t fit on the print bed. The 500Hz cell will fit but need to be divided length-wise to realize the full horn. Given that all 500Hz cells appear to the be same size, I was hopefully that I could adapt the design to the other configurations later if desired.

Why 3D Print?

I wanted to use 3D printing because I wanted a design that is easy to iterate and easy to build without metalworking tools. 3D printing also offers great flexibility in geometries / materials that are difficult to manufacture otherwise. The design could also be manufactured on-demand without significant investment in tooling and setup.

The foreseeable challenge with 3D printing is matching the acoustic performance of the original metal horn. Plastic is inherently not as stiff and 3D printing often uses partial infill which may be more transmissive. Damping / resonance performance will be critical in the thin regions. Reliability is also a concern with exposure to sunlight / temperature in hot environments. Some plastics (PLA) may not be suitable for outdoor use.

Goals:

• Replicate the Altec 1005B Multicellular Horn
• Match (or exceed) acoustic performance of original metal horn
• Use 3D printing as primary manufacturing method
• Easy to assembly without specialized tools, glue, or welding

Existing Work / References:

I cannot claim to be the first to attempt this. There is a great thread “The Construction of a Multicell Horn” on diyaudio.com with contributions from various member building similar designs with wood and 3D printing.

I also cannot claim to be an expert in acoustic sciences as some of the math far exceed my level. I am currently studying “High Quality Horn Loudspeaker Systems” which has been an incredible reference for all things horn related.

The original patent for the multicellular horn by E.C. Wente is also worth a read at “Acoustic Devices” (US Patent#1992268)