Copyright 2017 © Troels Gravesen

   Go to on this page:

Visitor quote: "I thought I'd be hearing just another "six + one", the world's most common and boring speakers. Hmm...These speakers makes you forget about the 100 litre bass cabs missing".

Good fortune brought me in touch with the producer of the ribbon tweeter seen below and after thorough testing, I set up a two-way from a new AudioTechnology 18H52 driver fitted with a new titanium voice coil former, thus two new drivers.
I've used ribbon tweeters before but never felt I had to go back, but this ribbon changed my feelings. Being a fairly large ribbon, it displays an impressive frequency response and very low distortion all the way down to 1 kHz, allowing a wider range of crossover options than usual, thus a 2nd order filter at 2 kHz is possible.

The "thing" about this ribbon tweeter is that the gaps between the ribbon and poles of magnets are sealed with a U-shaped mylar suspension, thus reducing air turbulence, increasing sensitivity and reducing distortion. And it works! This ribbon tweeter delivers some of the best highs I've had, regardless of brand, principle and price. The naturalness of cymbals, violins, whatever is astonishing. Put the best money can buy in front of it - and it delivers.
Now, no tweeter can do it alone. We need a seriously good mating partner, thus -

Per Skaaning at AudioTechnology made me a pair of 18H52 mid-woofers featuring titanium voice coil formers. The result is a driver with an exceptional linear response, in fact ruler flat up to 5 kHz with a small notch at 6 kHz after which the response rolls off smoothly. Easy on crossover - if baffle allows. The actual baffle is what can ruin a nice infinite baffle response, thus several baffles and driver positions were tried before I settled on what can be seen below. This will always be a compromise because we have constraints on baffle design. We could probably come up with an even better front panel - and nobody would build it. And just having the best intrinsic frequency response doesn't necessarily guarantee better sound. All depends!
For the ATS4-HE I used a passive radiator with the midrange. Works very well, and here we take full advantage of the PR units rather than the usual port. Admitted, this PR unit is many times the cost of a plastic port and should you want to cut cost, you can use a 68 mm port instead. Cut it to 18 cm length. Using the PR unit delivers a solid, potent bass for the size of the midbass. Don't expect earthquake scenarios from a 6" driver, although bass does remarkably well. This driver is fast, clean and has dynamic headroom in abundance to put it short.

I may have failed in staying cool and calm in describing the drivers presented here, but so be it. Yes, I'm freaking nuts about this small speaker, having given me a new dimension to my record collection and pointing to future projects. This is all I ask. Having built more than 200 speakers over the years, I need drivers that give me new experiences and point to new directions.
Visitors can't believe the bass these speakers can deliver, the clarity of the midrange - even at high SPLs - and the uncoloured treble. We listen to dubious tracks we usually skip, because they simply don't sound bad any more. We discover details previously hidden and musical layers being freed of congestion. Clarity!


2-way speaker from  6" midbass + ribbon tweeter.
System sensitivity: 87 dB/2.8V/1 meter.
Impedance: 8 Ohms, minimum 5 Ohms at 150 Hz.
Point of crossover: 2.3 kHz.
Power requirement: 20+ watts/channel.
Power handling: 150 watts, but please also read this:, and remember any burned driver is a misused driver.
ATiRi: AudioTechnology-Titanium-Ribbon


Click images to view large

Data sheets: AudioTechnology 18H52-17-06-SDT, PR23 and Viawave GRT-145-4

Test of GRT-145-4


The crossover follows an LR2 topology with a point of crossover around 2200 Hz. Having a tweeter sensitivity around 98 dB obviously takes some attenuation to align the tweeter with the midbass around 87 dB, thus R1-4.
For the midbass R7/L3/C3 dampens the 6 kHz peak. During crossover development this peak was unattended as it is so low it doesn't compromise treble performance, but it turned out that the LCR circuit improved phase tracking in all of the crossover region and significantly improved sound, so it stays in place.


Click image to view large

Ad squares and rectangles to your brace and distribute holes like seen here.

Click image to view large

The tweeter front panel must have the dimensions shown on drawing to the left. What's behind the tweeter is fully up to you.
To the right a suggestion to add more visual weight to the tweeter housing. Tweeter side panels must be chamfered 45° not to impact tweeter frequency response. The tweeter front panel must be 12 mm behind mid-bass front panel.

As always: Any change to front panel design and drivers' placement and you need a new crossover and I can't help. Please read carefully before asking:


Everything starts with building prototypes and testing different crossovers under normal living room conditions.
From simulation it's easy to set up different crossovers that basically deliver the same tonal balance, yet sound quite different.
In addition to this, four different amplifiers, solid state as well as tube amps, are used for digital and vinyl playback.
The speakers here run very well from my 32 wpc EAR-861 power amplifier.

Workshop images

What a thrill making small cabs! No huge, chunky bass cabs to move around...

Front panel made from 30 mm laminated BB with mahogany fillets. Takes some time, but I love the look of it.

PR unit seen from inside and rear.

Routing for the 18H driver. DO NOT forget to chamfer driver holes! This driver needs air to breathe.
With laminated front panels like these I add one coat of lacquer before routing to reduce edge ripping.
Start with 0.1-0.2 mm at a time and move router slowly!

Bitumen pads in place. 0.5 sqm supplied with kit.

Felt in place. For once, also cover front panel where possible.

Gluing front panels.

Ready for lacquer and first coat.

Parts for tweeter housing ready. To the right gluing tweeter cabs. Takes quite a few clamps!

Place crossover at bottom of cabinet and fold a piece of acoustilux to cover.

Damping behind 18H52 driver. One piece of acoustilux is covering the bracing frame and up on the top of cabinet.
Cabinet here seen from top.


A few comments on MEASUREMENTS before you start interpreting all the readings below.
First of all, if we think measurements will tell us how a speaker sounds, we're wrong. The perception of sound is way too subjective to be reflected in any measurements we can perform. A loudspeaker system is meant to give us a satisfying idea of an acoustic event and for some people a pair of 5 USD ear-plugs are enough, others spend 200 kUSD on a truly full-range pair of speakers - and the latter may not be happier than the former.
Measurements may give us an idea of tonal balance of a system, i.e. too much or too little energy in certain areas, although dispersion characteristics play a vital role here. A two-way 7+1 and a three-way 7+4+1 may display similar horizontal dispersion, yet sound very different. Measurements may tell us about bass extension if far-field measurements are merged with near-field measurements. In addition to this, ports may contribute to bass extension. Most of we diy'ers do not have access to an anechoic room for full-range measurements from 20-20000 Hz.  
What cannot be seen is what kind of bass performance we get in a given room. Bass performance is highly dependent on in-room placement of your speaker and the same speaker can be boomy in one place and lean in another. Actual SPL level at 1 meter distance and 2.8V input is useful for en estimate of system sensitivity and combined with the impedance profile may give an idea of how powerful an amplifier is needed to drive the speaker to adequate levels.
What measurements do not tell is the very sound of the speaker unless displaying serious linear distortion. The level of transparency, the ability to resolve micro-details, the "speed" of the bass, etc., cannot be derived from these data. Distortion measurements rarely tell much unless seriously bad, and most modern drivers display low distortion within their specified operating range. 
Many people put way too much into these graphs and my comments here are only meant as warning against over-interpretation. There are more to good sound than what can be extracted from a few graphs. Every graph needs interpretation in terms of what it means sonically and how it impacts our choice of mating drivers, cabinet and crossover design.
What measurements certainly do not tell is the sonic signature of the speaker, because speaker cones made from polypropylene, aluminum, Kevlar, paper, glass fiber, carbon fiber, magnesium, ceramics or even diamonds all have their way of adding spices to the stew. Nor do measurements tell what impact the quality of the crossover components add to the sound, from state of the art components to the cheapest of coils and caps, they all measure the same if values are correct, yet sound very different.


Ribbon GRT-145: Distortion measured at 0.25 m distance @ input equivalent to ~92 dB/1 meter.
Green = 2nd harm., blue = 3rd harm.

GRT-145 horizontal dispersion at 0, 10, 20, 30 and 40 deg. off-axis. Very good indeed.


GRT-145: More interestingly is the vertical dispersion, usually the weak point of ribbons, but here we see a very good result.
Vertical dispersion at 0, 10, 20, 30 and 40 deg. off axis.


18H52-17-06-SDT frequency response on 65 x 100 cm baffle (green) and actual speaker baffle (red).
On an infinite baffle we have a sensitivity around 90 dB/2.8V/1 meter. On a narrower baffle we obviously have the inevitable baffle step loss, thus around 87 dB sensitivity i all of the midrange.
What pleases me is the extreme linearity in the crossover region, here all the way up to 5 kHz.

Frenquency response of finished speaker normalised for 2.8V/1 meter. Sensitivity around 87 dB.
Reading merged @ 200 Hz with near-field response summed with response of PR23 unit.


Impedance of final system with 22 gram added PR23. Minimum impedance = 5 Ohm.


Horizontal dispersion at 0, 10, 20, 30 and 40 deg. off-axis.



All kit and component prices may be subject to change and are always to be confirmed by Jantzen Audio Denmark.

Download Complete Kit Sale Presentations (pdf file):

All technical questions to

All questions regarding purchase of kits, please mail Jantzen Audio at





ATiRi wiring


ScanSpeak D2908/714000 Beryllium dome version
We may call this the ATiBe version

ScanSpeak D2908/714000 dome.
Place the tweeter housing 18 mm from mid-bass front panel.

Should have fallen in love with Be domes - and I know a lot of people think it's the holy grail - then this version. I have a lot of questions regarding replacement of tweeters with Be domes, and normally not possible without changes to the crossover. Implementing the ScanSpeak D2908/714000 was easy as I already had a housing suitable for such application. The ATS4-HE dome cabinet fits well on the top of the ATiRi cabinet, so just follow the design and you're safe. For any of the tweeters you can obviously make the cabinet completely to your liking as long as you stick to the exact front panel dimensions and tweeter's placement. It can be as simple as my test panels and it will work as intended.

Use the tweeter cabinet from ATS4-HE.

Crossover as simple as the ribbon version, only here we can use a smaller C1 due to higher impedance.

Crossover layout is the same as for the ribbon version, only leave out R4.

Wiring is the same as for ATiRi


All kit and component prices may be subject to change and are always to be confirmed by Jantzen Audio Denmark.

Download Complete Kit Sale Presentations (pdf file): Delivery mid Feb 18

All technical questions to

All questions regarding purchase of kits, please mail Jantzen Audio at