TJL-3W
Copyright 2006 Troels Gravesen

CROSSOVER     CABINET     MEASUREMENTS     CROSSOVER KIT    


Download specs:
SEAS W18E001 + SEAS W12CY001 + HIQUPHON OWI

I do think I've been writing a lot about the TJL 3-way. The experiences gained from the Ekta and Zahra made me even more anxious to get started. If the latter two constructions hadn't come along, I would have done the TJL3W a long time ago. But events took their own course and I have been pleased with the results of the former constructions and the TJL3W was very much next.

The TJL2W is my favourite among a number of 2-way floorstanders and you always ask how to make a good thing better. The TJL shares the limitations of most other 6" + tweeter two-ways: Lack of dispersion in the upper midrange/lower treble due to the bass driver having to handle everything up into the treble area despite the low point of crossover around 2500 Hz. When a single driver has to pump the deep bass and at the same time handle the delicate upper midrange and lower treble there are limitations to the loudness we can expect from such a design. Suitable for most, but it can't play excessively loud. However, the TJL3W wasn't made to play loud, rather to play low. Below some thoughts about crossing over in the "critical" midrange and some thoughts on having a uniform power response from our speakers.

Download TJL3W files:
http://www.troelsgravesen.dk/diy_loudspeaker/TJL3W.zip
The zip file contains all large graphics and pics.

(back to top)

The paradigm:

Don't cross over between drivers in the critical midrange!

Just to recapitulate: lower midrange = 160-320 Hz, middle midrange = 320-640 Hz, upper midrange = 640-1280 Hz. Above 1280 Hz we're getting into treble territory and no wonder we find 6-7" drivers having trouble here. People are always surprised when I play a 315 Hz warble tone from a test CD. Is it really this high?!

So, even 3-4" drivers can cope with a 300 Hz point of crossover and we do see middomes taken this low. Does it sound natural? I case you have an electronic crossover, preferably 24 dB/octave, try make a 320-1280 Hz midband and play voices on a 4", a 5" and a 6" driver. Add a coil to the drivers in order to make the response reasonably linear. All drivers on suitable baffles. Try adjust to the same loudness from the drivers and listen to voices. Quite some difference. Remarkably different. I've always found 6" drivers to give the best lower midrange. It adds a low-end weight to vocals that smaller drivers cannot do. However, depending on the woofer below and the chosen slope of the crossover you can make a perfect blend of sounds. If you can get away with a 2nd order filter the results may be very good indeed. With higher order filters you start depending very much on the actual performance of the driver itself and there's no "help from your friends" here. I've heard excellent vocal performance from a 6" midrange driver taken down to 150 Hz with - presumably - 4th order filters (Von Schweikert VR4 mkIII).

In large speaker systems we often see 6" or 8" drivers used for lower midrange and not seldom a small 4" driver helping in the upper midrange, e.g. Krell Resolution 1. This is an interesting speaker as it uses a "low-cost", 4" Peerless pp midrange. This is a dedicated upper midrange driver with flat polypropylene surround making a low-sloped blend with the SEAS W22 around 450 Hz. Read here.


Measurements from
Stereophile: Blue is Peerless M122 mid-driver.
(Very nice and smooth slopes, by the way.)

Now, few manufacturers go 4-way due to complexity and overall costs, but no wonder this approach make it possible to use a driver with a diameter best suited for the particular frequency range it has too handle.
The compromise - and we're not talking 2-ways - is to find a 5" driver that will handle anything from 300 Hz to 3 kHz or even up 4-4.5 kHz and there are a few around. The best example is the AudioTechnology
4H52 Flexunits or C-Quenze used in the Stradivari and other constructions. A 4H52 C-Quenze really is a broadbanded driver doing well up to 4 kHz. If it had to go higher it would need a smaller voice coil like seen in the larger Dynaudio speakers with dedicated midrange drivers.

I have once tried making a 2-way from my 4H52 Flex-Unit with a point of crossover (4th order) at 4.5 kHz and I wasn't impressed. But making a point of crossover in an area where the drivers have an easy one-octave surplus capability is luxury and we may get the best of both worlds.

Don't be afraid to cross over in the "critical" midrange!

There are - to my mind - great benefits to be gained from this approach, but make sure you have good control over amplitude and phase when you do so. It's not particularly difficult to cross over at 800 Hz. The wavelength is 43 cm here and when you model the bass and mid in LspCAD, the distance from drivers to listening position is much less critical compared to the upper mid-treble situation.

Dispersion characteristics of loudspeakers.

The dispersion characteristics of loudspeakers is a controversial subject and an enormous amount of written material is published on the web and even trying to summarise the various opinions regarding this aspect of speaker performance is risky.

Read an introduction here from Dick Olsher: http://www.blackdahlia.com/samadhi/cbae.htm:
Although audio magazines continue to make a living off on-axis frequency response measurements, it is actually well established that speakers that measure identically on-axis in an anechoic chamber sound much different in a  typical domestic environment. These differences are due to how these speakers radiate sound off-axis as a function of frequency. Even at a listening  distance of only 8-feet, the soundfield at the listening seat is already  more than 50% room reflections in the bass and midrange. Speakers whose power response is deficient in the midrange will color the soundfield in the same  direction. For example, a two-way design with an 8-inch woofer crossing  over to a 1-inch dome at 3 kHz, may measure fine on-axis (anechoically), but show a 10 dB dip in the midrange off axis. Such a speaker will sound recessed  and lifeless in a real-world room.
More here at:
http://mixguides.com/studiomonitors/products_features/audio-listening-space/

The problem relates to having and wide and uniform versus a narrow and - uneven power response. Both the choice of drivers and the filters used will have a major influence on what is achieved. What can be seen from this construction is a uniform horizontal and vertical sound dispersion and to get my feet back on the ground, this has one distinct feature: My writing desk is close to the door into our living room where the speakers are located, giving me rich opportunity to the "listening from another room" situation. Actually I sit here an awful lot writing these web pages! With the TJL3W I can clearly follow what is going on even at low levels. It's even better than the Acapella SE. The "flat" ribbons have a nice horizontal dispersion, but are usually poor in the vertical plane (The ATD or NDRL ribbons are exceptions due to their special construction). As the Acapella SE is more or less optimised for on-axis listening, the combined upper frequency power response is probably significantly lower compared to the TJL3W. This does not say that the Acap. SE is a crappy speaker, but was balanced to produce the flattest response on-axis.
A "domed"
Acapella SEas has been made for those who may want to experiment with this. It makes quite a difference and there's a lesson to be learned from this exercise too.

Now, with a speaker having a wide and uniform sound distribution I can more fully enjoy the music off-axis, even if it has to pass a door opening - well, I'm only metre away from that door opening - where a speaker with a narrow and focused radiation will force me to pay more attention to details at low levels - and even sometimes draw me into the sofa in the living room for some "on-axis" action - and eventually some power napping, which is good.
Before we start building a loudspeaker we should really analyse our listening room to get an idea of whether this room is suitable at all for music reproduction. At low frequencies we don't have much choice and we're stuck with the room modes (resonances) and we can go Linkwitz's route and use a dipole bass in order to try overcoming this problem. Going higher in frequency we start having desirable or undesirable reflections supporting or distracting the sound images we project.
"Early reflections ruin perspective" we read all the time. The Stradivari has nothing but early reflections, however renders good reviews and is praised for what it does. How does this correlate? Not even "reflections" derived from diffraction seems to disturb our peace. We praise speakers for their 3-dimensionality despite having loads of diffraction whether we round our cabinet edges or not.

My point is that the Ekta, Zahra and TJL3W due to the combination of drivers having suitable diameters for the frequency ranges they have to handle and from the use of higher order filters produce a more even power response and are less fuzzy when it comes to placement. Speakers that have energized my listening room better than more directional constructions. Speakers that makes me enjoy the music also at low levels, that can play louder due to lack of intermodulation distortion from overlapping driver's response*, that can play female vocals much louder due to lack of beaming - and maybe distortion - in the 1-3 kHz range. Essentially I get speakers that are "easy to listen to".... Read here: http://www.troelsgravesen.dk/HES.htm
I have an old pair of HECO P3000 speakers. 3-way (6"+3"+1") bookshelf speakers from the Seventies. 3" midrange driver. There are a lot of things these speakers are not good at, but... they're easy to listen to and essentially does what a speaker is supposed to do. Provide enjoyment and relaxation rather than adding to the daily struggle for survival.

*: By this I mean not trying to combine a 15-20 grams midbass cone to a 300 milligrams tweeter diaphragm. 5 grams appear to be much more suitable. No matter what kind of crossover we use in order to produce the flattest possible amplitude and phase performance, we cannot do anything about the fact that where the drivers have to reproduce the same frequencies, a tweeter dome of 300 milligrams will rise much, much faster compared to a 10-15 grams midbass cone. And I don't care if the whole construction can reproduce square waves or not. I don't listen to square waves. I don't think it's an important parameter. Linear distortion and dispersion appear to be more important.

So, back to the TJL3W:

Should you consider building a speaker like this, keep in mind that this is a relatively small speaker of 24 litres net. volume and there are limitations to what a 126 cm^2 bass cone can do. In this construction you get a decent deep bass (F3 = 40 Hz) and I render the construction well suited for rooms up to 25-35 m^2.
The relatively low sensitivity of 84-85 dB is what we can expect from a 6" driver with a 15 grams membrane weight and a small magnet. The advantage of this was deep bass from a small cabinet with a high WAF. All about compromises. Very much in line with the 2.5 clone but with much better sound.

I had the cabs made in June 05 and we're now in January 2006. I routed the test front panels to be attached by screws, as I wanted to experiment with both the W11 and W12 middrivers and the HiQUPHON OWI and the Fountek JP3 tweeters. The panels seen above are routed for the OWI and this will be the initial set-up. For those who may want to upgrade their TJL, I'll try the JP3 ribbon later.
The frequency response and impedance data files for all the drivers were made a long time ago and from time to time I've made a few simulations to explore the possible crossovers needed to make the three drivers gel into a - hopefully - coherent soundstage.

Having the TJL3Ws running for the first time I used these CDs:
1. Marie Boine: "Unfolding", Verve 533889-2,
2. Siri's Svale Band: "necessarily so...", Sonor Records, SONCD 2005, and
3. Maddy Prior and The Girls: "Bib and Tuck", Park Records, PRKCD61.
Thanks again to Finn Dahle in Norway for sending med the two first CDs:
2_95_finndahle.htm. I would probably never have found these magnificent recordings had it not been for you.
Marie Boine sings the most exciting songs I've heard in a long time and she does it in her native tongue coming from Samiland (Lappland). The recording was done in a studio a little North from where I live, PUK-studio, Denmark. And it's a great recording. Play track 6 and the TJL3Ws will shake your guts and tell you that these speakers will give you a decent sense of what's below 40Hz. Marie Boine's vocal is recorded a little too close and dry to my taste, but the TJL3Ws handle it all very well.

Siri's Svale Band has been my favourite recoding for a long time for demonstrating loudspeakers. People may like or dislike jazz music, but this recording will leave no one unaffected. Recorded in Trondhjem, Norway. Excellent! The TJL3W will tell you a lot of what's going on in the studio. If your speaker has a good level of transparency, this recording will make you want to take the few step into the studio and watch what's going on.

Maddy Prior: I had this CD from my friend Darryl and it's a killer! Three female vocals producing narrow peaks of 10-15 dB in the 700-1000 Hz range. Very few speakers can handle these vocals without severe distortion. The Zahra and the Ekta did these recordings very well and thanks to W12 upper midrange driver, the TJL3W is not left behind. The Zahra may beat the TJL3W by a small margin due to higher sensitivity (larger membrane area) but the TJL3W is more neutral due to the magnesium drivers.
Well, the very first music I fed the TJL3Ws were some vinyl stuff, a 45 rpm audiophile recording, Kei Akagi: New Smiles and Travelled Miles. This LP comes with a bonus 45 rpm LP. Try this on a good turntable and who cares about CD when this is possible? Well, back to the present and ...

Crossover:

(back to top)


Do I hear something about a lot of components here? Well, we're dealing with metal cone drivers and we need to completely eliminate the cone break-ups. Basically the crossover between bass and mid is close to a third order on the acoustic side. The bass needs a 4th order electric filter to render this and the benefit of this is we don't need any notch filter for this driver.

The middriver need a 3rd order HP section and a 4th order LP section. The tweeter section is a bit tricky and this is basically a 4th order with some minor equalisation at the inlet.
The placement of the attenuation resistor after the 13.3 uF cap. in the mid section doesn't leave much room for adjustment here. However, you can reduce value to 3R3 without problems to increase midband level and to reduce midband level you may ad 47R across mid terminals and 1R0 in series before the 47R.
Basically I don't think either the midband or the tweeter will need further adjustments.
The notch filter in the mid section takes care of a small bump around 2 kHz. I don't think you should leave it out.

A high order crossover may seem rather complex, but breaking it up into the various sections, it becomes quite manageable. The bass section is placed at the bottom of the cabinet and consists of only 2 coils, 3 caps and a resistor. I'll make separate boards for the mid and tweeter sections to maintain overview.

And by the way: I've recently auditioned the Von Schweikerts VR4 mkIII speakers and if you still think 4th order filters will kill the music, go listen to these speakers. If you think paper cone (Aerogel) midrange drivers lack homogeneity and transparency, go listen to the VR4 mkIII. If you think the Vifa XT25TG tweeter is crap, go listen to the VR4 mkIII. It's not the end of the world, but a great lesson in doing things clever. Forget about "Global Axis Integration Networks" and cascaded 1st order filters - whatever that is. Common marketing rubbish. Schweikert uses coils and caps like the rest of us, so if you use higher order crossovers, do it right. Well done!

The Kit

(back to top)


The list of crossover components for one of the options. Download all below.

The two coils in the bass section are important. Used cored and very low-ohm coils < 0.2 ohms DC resistance. We need all the sensitivity we can get from the W18. The 4.7 mH coil for the LCR filter can be a high-ohm cored coil made from thin wire as the LCR carries a 15 ohms resistor anyway. Don't use a monster coil for this small circuit.

Complete TJL3W kit including drivers,
crossover components, vents, terminals, wires, etc. from:
Jantzen Audio

Download Kit Salen Presentation here.

Not included: HIQUPHON OWI, cabinets and damping materials.
HIQUPHON OWI tweeter available here. Mail info@hiquphon.dk


TJL3W cabinet

(back to top)

The question on placement of drivers is one I often receive from builders, like: "Can you explain the reason for placing some of the drivers off-centre on the front baffle? Are there any guidelines here or should I just place them so they look good?"

There's a short and a long explanation. The long one is found in this link: http://www.speakerdesign.net/understand.html

The short one is that the baffle provides an acoustic support for the soundwaves generated by the drivers. When the soundwave reaches an edge, it "falls down" creating irregularities in the frequency response of the driver. This is compensated in the crossover to render a flat response. By off-centering the drivers the intrinsic response of the drivers is often smoothed and this makes it possible to make a less complicated crossover.

So the drivers have to placed exactly as shown in the drawings, otherwise the crossover will not work as intended.
The front panel dimensions, width and length, has to be exactly as on the drawing to render the frequency response shown.
And don't forget to chamfer the driver holes, in particular for the midrange driver to allow free air ventilation from the rear of the driver. Obviously the tweeter don't need this.
The amount of energy released from the rear of the driver is exactly the same as from the front.   
More links to understand why:

http://sound.westhost.com/bafflestep.htm#bafflestepresponse
http://www.t-linespeakers.org/tech/bafflestep/index.htm

Also check this file for a transmission line solution: TJL3W built by Ingvar/Sweden 


The net volume of the mid cabinet is 2.15 litres, a little more than needed,
but I wanted to experiment with various volumes and damping materials.


Routing the front panel for drivers. Use a fresh, sharp router bit when doing solid wood.


Test front panel with drivers. To the right the SEAS W11 paper cone. Not a bad option
for a 3-way - and easy when it comes to making crossovers. The W11 will not be used here.


Damping materials used for TJL3W. The black wavy polyester foam is used
only right behind the bass driver. Middle: MDM3. Right: 10-12 mm sheep's wool felt.


Damping material behind bass driver.


Gluing front panel and cab ready for oil.


Measurements

(back to top)

The starting point, frequency response and impedance of drivers:

Impedance of drivers in cabinet. The chosen vent (Monacor BR60-TR) works spot on in this cabinet. Vent tuning is approx. 37 Hz.
Middriver's basic resonance is at 90 Hz. No reason to do anything about this as the middriver is only going to work down to 700-800 Hz.
HIQUPHON QWI resonance is around 900 Hz as declared. Smooth and symmetrical. How does
Oscar Wroending make these tweeters? No problem with these tweeters.

SPL response of individual drivers. Red = W18E001, green = W12CY001, blue = OWI . This tweeter is a darling, isn't it. Flat response, etc. Does everything right - and it sounds great too!
The W18 break-up is two octaves above point of crossover.

The high-Q peaks of the magnesium drivers are hard to overlook. Adding the MLS signal to the drivers during testing, the sound is terrible. It shreds your ears. In particular the W18 having the peak at exactly 5 kHz. The W12 doesn't sound nearly as bad having the peaks from 10 kHz and above. I took special precautions in the two-way TJL to ensure the W18's 5 kHz would be gone. Steep filters and a notch filter as well. But once these peaks are gone, there's a calm, "dark" and uncoloured sound left. "Dark"? The magnesium drivers have a quietness contrary to most other cone materials and for lack of better words, I call it "darkness". Reading a review recently on a high-end speaker using the Accuton drivers, it appeared that the reviewer almost thought they were too un-coloured. We're so used to the mixed blessings of cones breaking up in all sorts of ways that once we're left with a speaker made from drivers with significant reduced levels of cone colouration, we get a little uneasy from not hearing what we use to do.

The finished speakers:


SPL response of the finished speaker. An overall smooth response with the usual minor tilt downwards towards the highs rendering a smooth presentation of overtones. The HIQUPHON OWI indeed is a fine tweeter.
Blue is minimum phase.


Power response:

I'm going to spend some more time than usual on the dispersion characteristics of this speaker and show three measurements on horizontal and vertical dispersion to display the overall power response.
Obviously a speaker should have a reasonably flat response measured on-axis as this is where we tend to place ourselves when doing some "serious" listening. However, off-axis response may be just as important and here we go:

Vertical dispersion:
What I have done here is to record the vertical SPL at one metre distance varying the height of the microphone in 10 cm increments, starting at between-tweeter-and-mid height. This will correspond to a listening height of +30 cm to +60 cm above normal listening height at 3 metres distance. Downwards from 0 cm to -90 cm. Well, few people may listen from lying under the sofa, but just to show the extreme.

Red = -90 cm, blue = -60 cm, green = -30, yellow = 0 cm, purple = +30 cm and black = +60 cm.
I don't think I have to comment the results. An even vertical dispersion is the most difficult thing to do and only if you are seriously "grounded" will you have a -5dB response around 3 kHz.


Horizontal dispersion:

Horizontal dispersion is measured at 1 metre distance and in 10 deg. increments.
The TJL3W is a non-symmetrical speaker as the mid and tweeter drivers are offset, thus giving different horizontal dispersion depending on from "which side" you are listening.
The TJL3W should be set up with the mid and tweeter towards each other as seen on the drawing. The measurements shown here were done on the right speaker turning the speaker to the left (L) and right (R).

Left speaker turned right (outwards):
Purple = 40 deg. The rest: Well, take your pick!

Left speaker turned left (inwards):
Purple = 40 deg. The rest: It doesn't matter.
What can be seen is that due to lack of symmetry in drivers' placement on front panel, the response is different depending on listening angle.

What happens if we invert polarity of the midrange?
In a two-way system you can always find a spot where the phase tracking is perfect and make a nice and deep suck-out. At least it has to be specified where the reading was taken from and those previously shown have been at listening height = usually at tweeter height or a little below. I have also warned you not to put too much into this, as it's so easy to manipulate to impressive performance. But if conditions are specified it may tell that something appear to have been done right.
A 3-way is more nerve wrecking and reversing the polarity of the middriver must be the litmus test for a three-way system when you still place the microphone at tweeter height.
I have to confess I hadn't done this test before the crossover was finished because even it hadn't been as nice as seen above, I wouldn't have changed the crossover anyway. Now, this all looks very nice but what happens at 2 metre distance? Next graph below.

Two metres distance is a bit tricky because my workshop isn't very big and the readings are only valid down to 500 Hz.
Red = tweeter level, green = mic lifted 15 cm and blue = mic lowered 15 cm.
So, despite different listening level, we have a decent phase tracking between drivers.


Same exercise at 1 metre distance.
Blue = correct polarity of mid.
Red = reverse mid polarity at tweeter height.
Green = reverse mid polarity with mic lifted 10 cm.
Yellow = reverse mid polarity with mic lowered 10 cm.
Purple = reverse mid polarity with mic lowered 20 cm.


More SPLs:


Red = same SPL as seen above: On-axis, 1 metre at tweeter-mid height. Not valid to lower than 350 Hz due to gating.
Blue = nearfield reading of bass driver showing decent response down
to 60 Hz and minimum at 35 Hz where the vent will take over and produce the lowest octave.
Green = vent (response not corrected for actual out-put).


Step-response showing tweeter connected with negative polarity
and the mid and bass with positive polarity.


Impedance profile. As always an interesting measurement giving you an
idea of how tough the speaker will be on your power amp. Impedance is not below 5 ohms in the critical bass area and an overall moderate load to your amplifier. My 50 wpc valve power amp has no objections running this speaker loud.


Construction of crossover

Bass section:


TJL3W bass section crossover layout.

Space is a serious concern in this construction due to a limiting 24 litre cabinet volume and the bass crossover section is placed at the bottom panel. The two coils were connected in series and moved around while measuring the inductance and the shown placement provided minimum interaction. Placing the coils in parallel as far apart as possible did not work despite some 15 cm distance.
The coils used here provide a combined series resistance of only 0.33 ohm.
Don't be afraid to use an electrolytic capacitor here in the bass section. These ELKOs from Intertechnik works great and are often seen in high-end German speakers.
Use brass or stainless steel screws in order not to change the inductance of the coils.


Midrange section:


TJL3W mid section crossover layout.


Pretty close to the layout above.


Tweeter section:


TJL3W tweeter section crossover layout.


Slightly different from the layout above.

Go to TJL3W - Fountek JP3 version.