• Item: 3-way high efficiency horn speaker system v.2  
  • Type: Fostex FW405N + JBL 2450J-Be + Arai A320FL + Fostex T925A  
  • Posted: 2 February 2017  
  • Builder: marco_gea  
  • Country: United Kingdom  
  • Comments: 5  

The Beryllium project

WOOFER: Fostex FW405N

This is a modern woofer of recent design, with a composite cellulose+carbon fibre cone and a newly developed aluminum die-cast frame based on FEA.

Magnet is ferrite, and sensitivity is medium-high at ~93 dB/W(m).

The features that attracted me were:

(i) Medium Efficiency Bandwidth Product (EBP) = Fs/Qes = 56.

This allows the use of an overdamped bass-reflex box characterized by an almost 12dB/oct initial roll-off and associated low Group Delay, while still obtaining an F-3 of approximately 40 Hz in anechoic conditions (which leads to a nice almost flat response in room, given the typical Room Gain over the last octave):

Vb = 2 * VAS * Qt^2 = 180 L
Fb = 0.39 * Fs/Qt = 23 Hz

(ii) Very low mechanical damping = Fs/Qms = 4.0.

This, in my experience, is a recipe for excellent low-level detail retrieval. In other words, the woofer sounds good and does not mask or smear micro-details also at very low SPLs.

By way of comparison, this is roughly the same mechanical damping as that of the Great Plains Audio 416-8B (Fs/Qms = 3.4) and of the TAD TL-1601a (Fs/Qms = 4.1), while it is HALF of that of the more PA-oriented JBL 2226H (Fs/Qms = 8.0).

(iii) Extremely low distortion.
The levels of 2nd and 3rd harmonic distortion are amongst the lowest ever measured by the German magazine "Hobby-Hi-Fi": < 0.1% at 90dB(1m) all the way down to 20Hz.

COMPRESSION DRIVER: JBL 2450J + Truextent BeX4016 beryllium diaphragm.

(https://www.jblpro.com/pages/components/cmp_drvs.htm ;

This is an 'old style' 2" driver with a very powerful Neodymium magnet.
Its internal structure is still very similar the original JBL 375 driver designed in 1954, including an internal conical throat characterized by a slow flare rate (the 375 model was itself a slightly updated version of the Western Electric 594A of 1935!).

However, what sets the 2450 driver apart from its predecessors is an original and very advanced "Coherent Wave" concentric slit phase plug which results in a perfectly in-phase wavefront at the throat all the way up to 20 kHz.

Interestingly, this driver was originally developed (in the 1990s) to accommodate a Beryllium diaphragm, as JBL was considering going head to head with TAD (who had been using Be diaphragms in their compression drivers since 1978). In fact, since the more common Aluminium and Titanium diaphragms are in breakup mode over most of their bandwidth, they cannot fully benefit from the "Coherent Wave" design. Instead, a Be diaphragm, being pistonic for most of its bandwidth, would theoretically benefit the most. However, for economic reasons JBL then decided to forego the Be diaphragm and use a Ti one instead.

In recent years, however, compatible Be diaphragms have been made available by Materion, thus realising the full potential of this interesting design for the first time.

HORN: Yuichi Arai A320FL (also sometimes referred to as A290FL)


This is a refined take on the classic radial horn, originally introduced in the Japanese audio magazine "MJ" in 1995.

It differs from many "classic" radial horns (such as, e.g., the Altec 311, 511 and 811) in several important ways:

(i) Its expansion is Hypex (hyperbolic-exponential) instead of simply Exponential, resulting in better acoustic loading of the driver at the lower end of the pass-band. Specifically, its cut-off frequency is 320Hz and its T (hypex factor) is 0.6.

(ii) It is carved out of a block of dense high-grade plywood (in my case, baltic birch, with a density = 0.7 kg/dm3). This gives it excellent self-damping and prevents the notorious 'ringing' of the classic metal horns.

(iii) Its throat adapter operates an extremely smooth transition from the driver's internal exit angle and 2" (actually, 49.3mm) circular throat to the horn's 50x60mm rectangular throat.

Mr. Arai developed a number of similar horns designs which feature internal dividers ("fins" or "vanes") in order to achieve an even more controlled directivity in the upper-most octave (>10 kHz). However, such "fins" also cause internal reflections and diffraction, and I therefore decided to go for the "fin-less" (FL) version instead.

I had my horns manufactured by ALG Audio designs in France

Super-TWEETER: Fostex T925A


A classic 'bullet' tweeter, with an aluminium ring diaphragm and large AlNiCo magnet.

+ P30 stand


A nice and heavy tweeter stand made of solid brass.


For the Woofer-Mid crossover, I got inspiration from the design developed by Shozo Kinoshita for Pioneer's TAD/Exclusive monitors

(e.g., http://audio-database.com/PIONEER-EXCLUSIVE/speaker/model2402-e.html),

and later reprised in his own Rey Audio monitors


which is based on an asymmetrical 6th-order electrical low pass / 2nd order electrical high pass, with both drivers down -6dB at Fx.

In my case, the chosen crossover frequency is 600 Hz, so as to stay approximately one octave above the horn's cut-off.

When both the Woofer and Mid are connected with positive polarity and the Mid is physically set back so as to create a suitable offset between the two acoustic centres, this crossover results in the emissions of both drivers to be in phase over a relatively wide frequency range around the crossover frequency. As further proof of the excellent phase match, when reversing the polarity of the Mid, a deep notch at Fx is observed.

This type of crossover is handy in two important ways:

(i) the front-to-back offset allows the convenient positioning of the horn-loaded mid atop the Woofer box without requiring any form of delay (be it analogue or digital);

(ii) the 6th-order low pass effectively does away with any unwanted resonances at the top of the woofer's operating range.

For the second crossover between Mid and super-Tweeter, I opted for a simple symmetrical 2nd-order "Constant Power" design at 6kHz, with both drivers down -3dB at Fx.

Inevitably, given the short wavelengths at play and the physical separation between the acoustic centres, the super-Tweeter is not time-aligned. However,

(i) this was not deemed important at such high frequencies, where the human auditory system's sensitivity to phase is reduced;

(ii) the resulting comb filtering between the Mid and Tweeter outputs falls within the Equivalent Rectangular Bandwith (ERB), and is thus essentially inaudible.

Additionally, having an intentionally misaligned constant power crossover at such high frequencies results in a 1/6th octave-smoothed summed response that is essentially invariant over a +/- 30 degree listening angle, since any additional misalignment introduced by moving the head laterally is swamped by the >360 degree original misalignment.

I realize this design choice may be somewhat controversial, but IMO while it may "look" bad on paper, it actually sounds better than just going for a 2 way system and making do with the compression driver's own response over the top octave.

From a practical point of view:

- The woofer box was constructed out of 25mm-thick baltic birch plywood, with ample internal bracing and wool + polyurethane foam damping, plus an additional square sub-baffle to further mass-load and stiffen the area around the Woofer hole.

- The mid-high section is decoupled from the woofer box by vibration-dissipating sorbothane sheets.

- The crossovers were built using high quality parts (Mundorf coils and Jantzen MKP capacitors), and housed in two external wooden boxes.

The woofer low-pass section uses a balanced topology to reduce crosstalk with the mid and tweeter sections.

Both the compression driver and the super-tweeter are attenuated using variable units by Fostex


respectively the transformer-based R-100T for the compression driver and the R-80B potentiometer for the super-Tweeter.


  • St├ęphane
  • 11 April 2017 at 05:44AM
  • Shozo Kino****a did not take up the idea of filtering, he has the designer at Pioneer / Tad
  • marco_gea
  • 24 April 2017 at 09:56AM
  • Crossover description re-worded to avoid misunderstandings (thanks St├ęphane!)
  • marco_gea
  • 7 May 2019 at 02:14PM
  • May 2019: edited with new system configuration
  • dkalsi
  • 20 June 2019 at 06:54PM
  • Do you have any frequency response measurements from your build?
  • hallcon83
  • 24 January 2020 at 01:49AM
  • Very nice horn system! And thank you for the detailed discription, you did a great job explaining how and why of your design.I especially liked your bass Ideas.

    Best Regards,

Leave a Comment