Copyright 2022 © Troels Gravesen

   Go to on this page:

This is my 6th construction featuring Ellipticor drivers (not all published) and I shan't praise the Ellipticor drivers this time. Please read files on Ellipticor-1 and Ellipticor-4 if you want to know why I'm so enthusiastic about these drivers. With the introduction of the 8" version of Ellipticor, new opportunities were emerging for creating larger speaker systems with significant power handling and still - overall simplicity. Even more simple than I had imagined. These drivers have made me do stuff I've never done before - at least not with good results.
The special thing about this Ellipticor-3 is that we do not have a high-pass filter for the two midrange drivers as is standard for larger speaker systems and the two drivers are working in a 2½-way fashion, where the 21WE adds to the low-end of the 18WE. And due to inherent smooth response, we only have a single coil to each of the two drivers, creating target roll-off. I've never before had drivers that could do this. We are dealing with 1st order filters.
The Ellipticor-3 has been a playground for numerous crossovers and initially I meant to publish two, but one of them continuously  proved the other inferior, so only one will be presented - and it's a rather unusual one. I had a good friend doing a blind test, where I played two pieces of classical music on two different crossovers creating the same frequency response within ½ dB. He didn't know what he was listening to and he immediately picked the simplest crossover, which was my own favourite too - which he obviously didn't know in advance. "This just sounds more natural!".
I guess it was Sonus Faber that many years ago launched a smaller two-way without any tweeter series capacitors: "Sine Cap". This takes a high efficiency tweeter as the attenuation resistor combined with a parallel coil goes directly to ground. With the D3404 it can be done without hitting rock bottom on the impedance curve.
The basic crossovers topology was - as always - done by simulation, but the final tuning was done by ear, regardless of computer simulation - and there was a lot of fine-tuning! In particular L3 was critical in getting midrange the balance right. It seems the simpler the crossover, the more fine-tuning it takes.
Having the 8" helping the 6" midrange driver at the low end allows a very low point of crossover, thus all of the middle/upper bass and midrange is handled by these two drivers. Having the 6" driver in a fairly small closed box helps keeping control of cone movement. Below ~80 Hz we have help from a brand-new and hefty ScanSpeak high-efficiency, 96 dB, 13"/32W bass driver giving the oomph we like from the lower octaves. This driver is driven from a 500 watt plate-amp/DSP module.
If the 21WE driver is operated in a vented box, the three upper drivers alone would make quite a 3-way with substantial bass. No plans for launch of such a thing, so please do not ask. The 21WE performance is depending on the baffle below the driver, so it's not as simple as cutting off the bass cabinet and place the upper small pyramid on a stand. And it would be tuned for a overall lower sensitivity too.

Should you need to increase power handling of the system, you may add a simple 1st order filter to your power amplifier handling the MMT- section:
This has the benefit of reducing the input to the 21WE and 18WE drivers by some 6 dB @ 40 Hz depending on your preference. In all simplicity you add a small capacitor in the signal line to your power amplifier. The size of the capacitor depends on the input impedance of your power amplifier and your preferred cut-off frequency. I suggest -3dB @ 80 Hz. This doesn't impact the fullness of upper-bass/lower-mid but it will increase power handling of your MMT-section, give cleaner midrange at high levels and reduced back-EMF to your power amp. Win-win!

Four-driver speaker from 13" bass driver, 8"+6" mid-bass + 34 mm dome tweeter. Active Hypex module for bass driver.
Dimensions: 40 x 53 x 115 cm, WxDxH.
System sensitivity: 90 dB/2.8V/1 meter.
Impedance: 4-8 Ohms for MMT section.
Point of crossover: 80 and 2700 Hz.
Power requirement: 20+ watts/channel. And you need a line signal for running the Hypex module.
Power handling: 200 watts. Please also read:, and remember any burned driver is a misused driver.

Useful links (Please follow all links before e-mailing!):

Click images to view large




Click images to view large.
Here the 8 Ohm versions of 18WE and 21WE are used.

Download specs:

D3404/552000    21WE/8542-T00     18WE/8542-T00    32W/4878-T11

Prototype 32W/4878-T11 woofers. Brand-new woofers from SS. The final version will have a rubber cover for the magnet. Not shown here.
Click images above to view large.

The final 32W/4878-T11.



The crossover, thanks to an inherent smooth response of the 18WE and 21WE drivers, features a simplistic approach with only a single coil to the drivers, which produce 1st order roll-off of the drivers - and essentially we have a 2½-way set-up. The tweeter section features a series resistor and a coil. That's all.
Should you be fed up with capacitors and which one is best, you here have the opportunity to evaluate your system without worrying about any sonic degradation from crossover capacitors. Three coils, two resistors, that's all.
Now, to pull off this unusual feature, we need a high-efficiency tweeter that needs quite some attenuation as the series resistor (R1) combined with the coil (L1) goes directly to ground. The minimum impedance reaches 2.9 Ohms and my EAR 961 tube power amp doesn't mind at all - and this even from the 8 Ohms taps. For once the 4 Ohms taps goes well too, although at a slight decrease in level.
R1 resistor directly connected to ground via L1 can draw some current, hence here made up from 3 x 10 watt resistors as seen on lay-out. This to have them stay cool. Initially I use 2 x 10 watts and they were never more than lukewarm at very loud levels, but better stay on the safe side.
The two midrange drivers, having a combined membrane area of 366 cm^2 equal to a large 10" driver, really don't do much to produce significant sound levels. And the Ellipticor drivers have an unusual dynamic headroom and can be run hard before breaking up.

Now, this all looks very nice and simple but hardcore 1st order aficionados may object to the tweeter being connected with inverted polarity. What does the step response look like? We may produce a flat on-axis frequency response from drivers being out of phase but is this as good as can be? What does a square wave pulse look like? Etc., etc. 1st order filters are tricky and well-known examples can show elaborate crossovers that may shock people who think a 1st order filter is a just cap and a coil and nothing more.

There are very few true 1st order filter speakers in this world. Some may claim they do 1. order filters, but it usually comes with a big grain of salt. The filter may be 1st order on the electrical side, but even so, the roll-off can be 2nd, 3rd or even 4th order from very simple filters - due to the drivers' inherent roll-off. What matters is what happens on the acoustic side. And it all depends, because how far beyond the point of crossover must a driver go 1. order before we can call it a true 1st order filter? There are no rules for this, not even any general practice. Vandersteen is probably the company that does best in this area. My best bid on this issue is 2 octaves above and below point of crossover. Based on measurements in Stereophile, Vandersteen doesn't seem to always meet that target.
Here the 18WE driver follows 1st order roll-off up to around 6 kHz, then a dip around 8-9 kHz and it stops around 15 kHz but never exceeds target profile. The tweeter is pretty much 1st order all the way from 15 kHz down to 800 Hz, where is starts rolling off 2nd order. This is pretty darn good!

To produce a time-aligned, phase-correct system we could turn the phase of the tweeter by adding an all-pass filter plus an impedance flattening circuit as well, so the tweeter crossover all of a sudden looks like above to the right. Quite elaborate compared to a single resistor and a coil. Well, there's only one way to tell - to do it. Thus, the all-pass filter was introduced and my trusted ears were invited again for an audition: Blind test +/- all-pass section and impedance correcting circuit. Frequency response <0.5 dB between crossovers. Some may argue that even 0.5 dB difference between two identical crossovers/drivers are audible, but this is the best we can do and the trained ear may be able to focus on other things than amplitude.
So done, the all-pass version had thumbs down. This despite premium components and careful fine-tuning. The simple "sine-cap" version had better transparency and generally more life-like presentation. Another thing is that when we use an all-pass filter we introduce a delay, thus a negative response on the step response. We can't use all-pass filters in a true 1. order crossover if we want the step response right.

Hypex FUSION plate-amp/DSP
Since launch the FA252 has been replaced by FA501


Using the Hypex FUSION  FA501 for high-level input operation

An easier way of powering the bass driver is by using the FA501, which takes a high level input, i.e. connect to speaker cables and adjust level with potentiometer on Hypex faceplate.

Here's the simple setup without the high-pass filter before the power amplifier.
This give the full bass for 21WE and 18WE as these do not have a passive high-pass filter.
This may work OK for most applications, but to get the full potential, this is what is needed:

Here we have a high-pass filter inserted before the power amplifier, making a 1st order roll-off @ 100 Hz. Here you must know the input impedance of your power amplifier, in my case with the EAR-861, 10 kOhm, thus ~220 nF is needed for the high-pass filter. As I run balanced operation I needed four 200 nF caps, here Miflex capacitors. These caps obviously must be premium quality, thus copper foil caps.
Now, what the Hypex now sees from taking a high-level input, is the blue graph and this must be adjusted in the Hypex program to initially make a flat response before we add our low-pass filter to the software. Software for both options will be provided.

Download software here for both scenarios.

- or from here.

Scenario 1: No high-pass filter before power amp
Scenario 2: High-pass filter before power amp.

Do not use Ellipticor-3_FA501_HP-power-amp file if you do not have the the high-pass filter
-or you may damage the bass driver.

Open the zip file and extract the files to your FilterData folder in your MyDocuments folder.

Go to FA501 page for further instructions.


Strictly calculated the cabinets are 199 mm wide at top and 399 depth without the front panel. I reality you aim at 200 mm width at top and 400 mm depth. This makes the angle a tiny bit less than 5 deg. tilt of both sides and front.
The material used here was 27 mm Baltic birch, which in fact was only 26 mm, just showing these things have tolerances. The panel between bass and mid-bass drivers was 26 mm and for bracing and 18WE cabinet I used 20 mm BB. The rear panel of the Hypex compartment was 15 mm BB.
You can make it all from e.g. 22-25 mm MDF or similar. Do not use standard plywood as this is much too light-weight and resonant for speaker enclosures.
Faceting the top front panel was done on my table saw set to 30 deg. angle. The table slider guide was set to around 2 deg., but must be tried out to make it right. It can de done by hand like seen here:
Obviously you can make the front panel in one piece if preferred. I needed flexibility for experimention, hence the upper part attached by screws.

Do not change the front panel dimensions or drivers' placement or you'll need a new crossover - and I can't help.

Click images to view large


Workshop images

Side panel fronts were cut free hand and trimmed with the router.

Cut-offs from side panels were glued with fillet to make rear panels.
Gluing sides, rear and top panels in one operation.

Routing for the Hypex module.

Mounting panel for bass driver.

Preparing for driver front panel and routing for ports. Holes are Ø96 and Ø149 mm.
The port for the 21WE is placed close to the bracing to allow space for crossover on rear panel.

Lining internal panels with felt. Also felt on 32W front panel.
No felt on rear panel behind 21WE driver.
Make room for terminals and wire from Hypex module. Red circle.
The 18WE cavity is filled with 125 gram wool.
Drill holes for wires before adding felt. From terminals through 21WE bottom panel and from 21WE cabinet to 18WE cavity.
Seal wires holes like seen here. Make sure it's absolutely air tight.

Making the front top panels, here attached by screws, as I may want to try out other things. Normally I would use glue.
Faceting the top front panel was done on my table saw set to 30 deg. angle. The table slider guide was set to around 2 deg., but must be tried out to make it right. It can de done by hand like seen here:

Ports in place and feet nuts fastened with araldite.

Hypex wires sealed from bass cabinet.

Cut two pieces of 30 x 50 cm acoustilux, roll and stuff the rear compartment of the 21WE driver.
Weigh out 125 grams wool for 18WE compartment.

21W cabinet: Place a piece of 30 x 50 cm acoustilux op top of the crossover on rear panel.
A piece of 50 x 35 cm acoustilux is placed at the bottom of the cabinet and up on the sides as seen on image below.

Left: Place two pieces of 30 x 50 cm around the port at the top of the bass cabinet. Right: First time set-up of speakers.

All wires ready for drivers.
Cut piece of 10 x 50 cm acoustilux, roll gently and stuff the 21WE port. Aperiodic tuning.

Installing the 32W bass drivers.

Making the alu grilles for the bass cab.
See ATS4-HE for details on the alu grilles:

Mount crossover on rear panel below port.

Fold a piece of 25 x 50 cm acoustilux to cover the crossover.
30 x 50 cm acoustilux is placed at bottom and sides of 21WE cabinet - as seen on image above.
Roll a piece of 10 x 50 cm acoustilux and stuff the port. Aperiodic tuning.




SPL of drivers on front panel. Green = 18WE, red = D3404 and orange = 21WE.
As can be seen the intrinsic response of the bass drivers are as flat as can be for the 1st order crossover to these drivers.

Let's have a look at the 21WE alone. Here merged with near-field response @ 300 Hz.
This driver helps the 18WE maintain a solid response below 100 Hz, which was the target: Having the 21WE + 18WE handle all of the middle-bass, upper-bass and midrange.

Red: Impedance of 21WE+ 18WE + D3404 driven from crossover. Green: Electrical phase.
This proves an unusual flat profile, an easy load despite minimum impedance of 2.9 Ohms.

SPL of 18WE and D3404 driven from crossover.
The 21WE driver will lift the 200-500 Hz range up to around 91 dB, which is the area that I would consider relevant to measuring overall system sensitivity. Don't for a second think this speaker is treble shy, despite the sloped response profile. It isn't.

Above the response of 18WE and D3404 driven from crossover and summed response (red).

To the left the 18WE response vs. target profile. Right: Same for tweeter.

Horizontal dispersion @ 0, 10, 20 and 30 deg. off-axis. 21WE not included in these measurements.


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

Download Kit Sale Presentations:

All technical questions to

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


Board is 225 x 130 mm. Cut corners 5 deg. to half width.

Place crossover on rear panel below port in 21WE compartment.


Pay notice to tweeter connected with inverted polarity.

Click images to view large.
All R1 resistors not shown. See layout.


Bringing the Ellipticor-3 into our living room for the first time and putting on the above 45 rpm vinyl (Vincent Belanger), something happened that rarely happens. The dynamics, the see-through, the visceral presence of the cello there in our living room. No speaker entering our living room has ever had this coherent presentation from the lower register to the very top. This kind put the whole thing into a dilemma, because it clearly demonstrates how higher order filters add an edginess to every note played and how transparency is reduced. But not all driver can handle 1st order filters and we compromise. And the "compromise" can sound very well; very, very well indeed, but not exactly the same way as drivers freed from time delay or phase distortion. Only first order filter are free of time delays and phase distortion.
Equally exciting was it to put on the re-mastered "White Album". Not all mixed to my liking, but the level of detail is something if you were young in the Sixties, where my first vinyl purchase was Sgt. Pepper, at the horrific sum of 42.50 Danish Kroner, which is equivalent to around 53 EUR in today's money. New LPs were expensive!

We cannot tell what a speaker sounds like, but clearly this is a large full-range system with a significant capability in the lower registers. Next we notice significant sensitivity. This speaker doesn't need a lot of power to run the top system. My 32 wpc tube amp does very well and I never noticed any sign of distortion playing seriously loud..
I recently went to a hifi show in Copenhagen, and no different from any other show, a lot of pling-plong electronic music was played that doesn't tell a thing about amps, speakers, cables, or anything else in the chain. Electronic music can sound like anything and we have no reference for what it should be like. Acoustic music recorded in real world rooms with not too much artificial processing is what counts when it comes to evaluating systems and admittedly there were a few playing decent acoustic music.

Response from visitors were unanimously this: "Natural". Rather than creating a bubble of sound around the speaker, it's "see through", or rather "hear through". Depth and three-dimensionality are the key words in describing the sound of this loudspeaker. Conventional crossovers may be smooth and easy on the ear, but compared to 1st order filters, they always have an edginess in their transient response, which may be perceived as positive - to a certain degree - but is mostly related to the rapid phase shift generated by the crossover - and driver. So, why aren't all my crossovers 1st order? Simply because most drivers do not allow 1st order filters due to irregular frequency response and it would take all kinds of correction circuits to reach target roll-off. All speakers are compromises.
More often than usual I have gone to the speakers to see if I had the speaker wires connected correctly. On very good recordings you come aware of phase issues and mixing with regard to placement of microphones in the studio/hall where the recording was made. You clearly hear what instruments are mixed in pure mono, being in one channel or spread across channels.

If you don't like the front panel screws, it looks like this, thanks to photoshop:


Ellipticor-3 at a local hifi show, 2020.
Driven by double Gamut amplifiers and with my high-pass filter in front of the amp running the MMT section.