TJL-2W W18E001/JP3/NeoCD3  or  W18E001/T25C003
Copyright 2005 © Troels Gravesen

CROSSOVER    CABINET    MEASUREMENTS    CROSSOVER KIT     T25CF001 version

Digitally manipulated picture in middle. Drivers in yew veneered TJL cabinet.

The following project describes a 2-way floorstander of modest dimensions: 19 x 98 x 26 cm (W x H x D). As a follow-up to the 2.5 clone this project is targeting a better midrange, better treble, bass comparable to the 2.5 clone and a more moderate sized cabinet. Well, can all these requirements be accomplished in a smaller size construction? And from what drivers?

From the Point75 and Acapella projects the use of the SEAS W18 drivers seemed an obvious choice for a 2-way floorstander. And from the bass performance of the Point75 a smaller cabinet might provide similar or better qualities compared to the 2.5 clone. But you can never tell the bass quality until you've tried. What I also wanted to try in this construction was the use of a passive radiator and SEAS provides suitable PRs for 6" drivers. This PR proved to be a major disappointment, but more on this later.

Getting more bass out of a SEAS W18 design, the magnet obviously has to be smaller - or the cone must be heavier - compared to the W18EX001 driver used in the Acapella, hence the W18E001 was the one to look for. This driver has TS data suggesting a 24-litre cabinet to be suitable for some decent bass and a port/PR tuning of 34-38 Hz should provide a f3 at 40 Hz. The JP3 ribbon tweeter is described in the Acapella file and needs no further introduction here. This is a fine tweeter at a reasonable price. You may substitute the Fountek JP3 by the Aurum Cantus G2Si ribbon without changes to the crossover. I think the JP3 has a better faceplate compared to the AC G2Si despite being made from plastic. The JP3 plastic faceplate follows the curvature of the horn better and makes a good termination.

Cabinet construction:
BACK TO INDEX

Click image to download cabinet drawing, zip file, 34 KB

It's a bit difficult to describe the overall damping of the cabs, but all internal panels are added 5 mm bitumen pads and a layer of 10 mm polyester foam. For further damping I use a mixture of loose polyester and sheep's wool (33/66) MDM3; available from Monacor. I guess Acoustilux will make a good substitute for this material. At bottom, top and right behind bass driver several layers of this material is placed. For further damping I would add small rolls of the same material in the section between the vent and the drivers and in the section below the vent, but be sure the damping material does not prevent free airflow between vent and driver. Experiment with different amounts of the loose damping material. It will have an impact on the overall sound of the loudspeaker and generally you shouldn't stuff the cabinet too heavily.

Fountek JP driver cut-out.

Test cabs made from Baltic birch. Front panel from 19 mm MDF.

The prototype cabinets were made from 20 mm Baltic birch with 19 mm MDF front panel. Internal panels were added 4 mm self-adhesive bitumen pads (can be found in car accessory stores). All internal panels were damped with 10 mm polyester foam and behind the W18 driver 2 sheets of 4 x 15 x 20 cm MDM3 (Monacor) were placed. At the top and bottom of the cabinet more of this damping material was placed to reduce standing waves between top and bottom.

Chamfering driver holes:

Much to often heavy front panels are produced without chamfering like seen on the picture to the right. This seriously prevents the air from having free access to the enclosure behind the driver. Remember, the amount of energy released from the rear of the driver membrane is exactly the same as from the front. Don't forget to chamfer the driver holes 45 deg. as seen on the picture to the left.

Measurements:
BACK TO INDEX

Left: TJL impedance curves: As can be seen the port tuning is around 36 Hz. Vent 72 (ID) x 200 mm. Overall impedance never falls below 7 ohm.
Right: TJL frequency response, 1 meter distance, tweeter height.

Left: TJL BM and T response. Point of crossover = 2600 Hz. Right: Blue = frequency response with reverse polarity of drivers. Red = frequency response with same polarity of drivers.

Left: TJL, cumulative spectral decay. Right: TJL, step response.

Crossover
BACK TO INDEX

Sound of the TJL?

Initially I listened to the version with a passive radiator:
Having been listening to semi-dipoles for almost a year it took me some time to adjust to the sound of a conventional construction again. The upper midrange and treble was superb, but the bass was not as deep as I remember from the 2.5 clones. And something was seriously wrong about the lower midrange. Transients appeared dull and constrained. Vocals had a peculiar sound as not being properly "liberated" from the construction. After listening for two weeks the time was ripe to introduce a vented version and the PR was substituted with a 72 (ID) x 200 mm vent placed where the PR had been. Then all of a sudden these W18 drivers came to life.
The bass turned out a lot deeper and transients improved considerably. And best of all, vocals regained naturalness not heard before from this construction. Very strange indeed, quite a large number of - mostly high-end - loudspeakers utilises passive radiators with reported good results. Not here, this passive radiator simply killed the liveliness of the speaker and did serious damage to the naturalness of vocals. After this I'd be more than hesitant to again trying a passive radiator*. I have to add that several tunings were tested with the PR adding more and more weight to the membrane. None of these tunings appeared to increase the overall performance.
After finding the subjectively optimal tuning of the vent - 35 Hz - the bass appeared to match the 2.5 clone being less boomy, thus improving the impact on the lower midrange. The midrange is clear and transparent throwing a wide soundstage.

These magnesium drivers are revealing and improper recordings will be presented as such. The construction does not favour any particular type of music. Very often we read that a construction excels from simple acoustic music, jazz trios, etc., where large orchestral works fall apart. Not here. Of coarse there's a limit to the loudness due the size of the driver membrane, but it can play louder than most other 6" drivers I have tried. Very low distortion that is. I found the sound the best with some 70-90 cm from the front of the cabinets to the wall behind the speakers. For good reasons I haven't mentioned the treble from the ribbons. This you will find in the Acapella files. The JP3s in this set-up holds all the qualities from the Acapella. Given the simplicity of the construction and the modest size, I think the TJL handles all musical genres remarkably well.

Download cabinet drawing here.

Read about Kevin's TJLs here.

*: 13-02-2006: The poor results from the passive radiator set-up may have been caused by the radiator simply being too far away from the bass driver.

The NeoCD3 version
BACK TO INDEX

20-03-2006:
I've had quite a number of requests on a NeoCD3 version of the TJL, as some would like to use this less fragile ribbon to replace the Fountek JP3. The NeoCD3 has a plastic laminated ribbon, which does not mean you can remove dust from your drivers with a vacuum cleaner! But the plastic laminated ribbon has a higher flexibility compared to pure aluminium ribbon and will take some more air turbulence compared to the JP3 ribbon.
Please do not ask which version is the best, because I won't answer. I would probably not be able to detect either ribbon in a blind test.

It's been a long time since I've heard the TJL and comparing the TJL 2-way to the TJL 3-way was obvious. With the experiences gained from the Ekta and Zahra and carried through to the TJL3W from adding a small driver to handle the upper midrange/lower treble, the TJL doesn't have the dispersion in this area and will off-axis sound more closed and with a lack of presence as often heard from similar sized 2-ways. On-axis, this is less noticeable, but I would have liked you all to have the possibility of compared a speaker made from the same basic drivers, having similar sensitivity and amplitude profiles, but where one is a 2-way and the other a 3-way. The sound is very different. No surprise, the TJL3W is a better speaker, but it obviously comes at a cost. Should you decide to engage in a TJL project, I seriously recommend taking the 3-way into consideration.

The NeoCD3 and the JP3 are quite alike in their frequency response, but they are not compatible and the crossover will have to be modified to accommodate the NeoCD3 ribbon.

Polarity of NeoCD3: The NeoCD3 has it's positive terminal to the left when you view the driver from the rear and the terminals are up. Remember, the NeoCD3 - like the JP3 - has to be connected with inverted polarity.

Fountek NeoCD3 ribbon tweeter.

So, the TJL was set up in the test cabinets again and this time I took the help of LspCAD to do most of the work. Actually the new measurements were so good I didn't have to change any component values to reach the target frequency and phase response. Great!

The lowpass section stays the same, but tweeter section has new component values.

Measurements:

Left: SPL response from 1 metre distance, tweeter height. Compare to simulation later. Right: SPL response from 1 metre distance, tweeter height. What is apparent, is a minor depression around 400-500 Hz. Otherwise a smooth response close to the famous BBC dip (2 dB decline from 100 Hz to 10 kHz).

Go to top

Revised TJL-Classic

The "problem" with re-visiting a construction you made a couple of years ago is that what you have leaned since then makes you reconsider the whole project. What are the strengths and weaknesses of your construction? A few things here:
Strengths: Great, deep bass for the size, un-coloured sound from using magnesium cone drivers and ribbon tweeters, high level of transparency.
Weaknesses: Relatively low sensitivity, reduced dispersion in the crossover region, lack of presence compared to TJL3W. Will it be possible to do something about the listed weaknesses? Let's take a closer look at the drivers.

Some basics: The TJL uses a bass driver with a relatively small magnet (same situation as the 18W/8535 driver in the 2.5 Clone). This means relatively low sensitivity, but we get deep bass from a small cabinet - and the W18E001 really goes deep. It also means the frequency response of the driver is relatively flat, where a driver with a large magnet often has a steep rising response towards higher frequences. It also means the baffle step compensation has to be relatively small for the W18E001. To equalise the response of a driver on a narrow baffle we add a large coil to 1) provide baffle step compensation and 2) tilt the rising frequency response towards higher frequences. The result of this can produce several things: Quite often drivers have a bump around 800-1200 Hz due to baffle step loss and relatively high magnet strength (there are more to this, but these are two of the main factors). If we want a flat response we can do two things: 1) We can use a large coil to basically align the 1 kHz response with the lower midrange and 2) we can equalise less (use smaller inductor) and remove the bump at 800-1200 Hz by adding a notch filter for the same region. The benefit from the latter option is that we sometimes can avoid a dip around 400-600 Hz, which may be the result from simply using a large coil at the intro of the crossover to render a flat response at 1 kHz. The first option is to some extent the situation for the TJL.

With LspCAD at hand I decided to wipe the table clean and start all over again. From the merged nearfield and farfield frequency response profiles of the W18E001 driver we can see a phenomenal bass extension from a 6" driver. I have 10" drivers that will start to roll off from 80-90 Hz, but this little fellow goes down to 60-70 Hz before descending. The price for this - as said - is sensitivity. We're talking 83-84 dB/2.8V in the upper bass/lower midrange.

The TJL has a minor depression around 400-500 Hz (upper graph) to render a sonically optimal level at 1 kHz, so what if we tried to balance the 100-500 Hz region first, and then equalised the 1 kHz region with a notch filter = lower graph. What seems to be the result of this is 1) a flatter overall response, and 2) increased sensitivity in the midrange and treble. Well, we cannot increase system sensitivity from doing so. What the W18E001 can manage in the low bass is what determines the basic sensitivity of the construction. When we increase the response in the midrange and treble we sacrifice relative bass level, not extension but level. But the bass level of the TJL is so good that it may be worth sacrificing 1-1½ dB to get an apparently more sensitive speaker. How this will sound is a good question.

Next thing: Dispersion in the crossover region. Almost any 6-7" driver will have poor dispersion characteristics around 2-3 kHz where the point of crossover often is placed. Ribbon tweeters have (mostly) good horizontal dispersion and poor vertical dispersion, so the TJL is likely to have a reduced power response in the crossover region (read about the Acapella SEas experiments). It's a strange thing, because we can measure a 6" driver to perform well up to 3 kHz, but try a 4" driver handling the same upper midrange/lower treble region and you will hear a world of difference. To improve the power response in the crossover region we need a dome rather than a ribbon and thus the SEAS T25C-001 tweeter (Acapella SEas) was brought in again and inserted in the TJL cabinet and new measurements were performed with subsequent LspCAD simulations. To further improve power response, the point of crossover was taken down to approx. 2 kHz. Ideally I don't think a 6-7" driver should be taken much higher than 1.5 kHz, read SP38 file.

I have to say that what is depicted here is highly dependent on proper splicing of the two SPL graphs of the W18E001 response. The near-field and far-field response has to be valid at 350 Hz for both readings. The far-field reading was an average of 5 measurements at ½ metre distance to have a lower reliable limit of approx 250 Hz and the near-field measurement was an average of 5 measurements taken at 1 mm (one millimetre) distance to the membrane and with a 64 millisecond window to render a lower limit of 15 Hz. Repeated measurements gave then same results.
So here you have two new versions of the TJL loudspeaker:

TJL-NeoCD3 revised.

Fig. 1. TJLNeoCD versions. Upper = TJLNeoCD3, lower = new simulation.

This version has not been tested in hardwiring, so no measurements.
However, the low pass section has been tested in the T25 version and performs spot on,
so to deviate from normal* practice, I don't think you run a big risk of doing this version.

The SEAS T25CF001 version (my favourite)
BACK TO INDEX

T25C003 shown here. Use T25CF001

The Crossover

Crossover simulation.

The Kit

Complete kit incl drivers and crossovers available from Jantzen Audio, mail at: contact@jantzen-audio.com

TJL-T25 version in test cabinet.
Fortunately the T25 has the same 110 mm faceplate as the Fountek ribbons.

Measurements TJL-T25

Left: SPL from individual drivers without crossover attached. As can be seen, the response from the W18 is reasonably linear up to 400-500 Hz, then climbing some 5-6 dB up to 1 kHz. Last but not least, the horrific cone break-ups peaking at 5 kHz that has to be carefully suppressed by the notch filter. The T25 has a minor bump at 1700 Hz, but nothing that causes trouble in making the crossover work.

Summed response (red) with near-field bass response from 350 Hz. Now this is some smooth linear response and the sonic impact from this is not to be ignored. Blue = minimum phase.

Left: SPL from drivers with crossover attached. As can be seen the cone break-up from the W18 is thoroughly suppressed. Point of crossover is around 2300 Hz. (sorry for merging the tweeter also! Doesn't do much below 600 Hz)

Right: Impedance profile of TJL-T25. This is a relatively easy load on your amplifier.

The T25 Sound

So, what about the sound from this T25 version? Well, most of the new design criteria seem to be accomplished here. Right, the basic sensitivity of the construction hasn't changed - and yet the midrange/treble has been lifted some 1-1½ dB compared to the TJL-"Classic". And you really do not feel any lack of bass compared to the "Classic" despite the "brighter" voicing of the speaker. The upper midrange/lower treble is quite different from the "Classic" due to an improved dispersion in the crossover region.

To cut a long story short you get a speaker with added presence from the T25 version. And you get an "easier-to-listen-to" speaker due to enhanced dispersion. You also get a more forward presentation and this may suit some, others may prefer the more laid-back voicing of the TJL-Classic.

To summarise the options:
1. The TJL Classic with the JP3 ribbon.
2. The TJL Classic with the NeoCD3 ribbon.
3. Try the revised TJL-NeoCD3 version (not tested in hardwiring) and report back what you think.
4. The SEAS T25CF001 dome version.

Comments from Andrej/Slovenia on the TJL

Hello Troels.
This weekend I took some time and did some serious listening test with TJL_JP3 speakers. My comparison speakers were Proac 2.5 (8513 version) with your latest crossover update. My test equipment was NAD 541i (completely modified CD player),  Pass BZLS preamplifier and 120 watts AKSA power amplifier. Interconnect were Van den Hull D102, speaker cables Kimber 4 VS.

Here we go... This speaker is totally opposite of so called "lushness" sound. It is very detailed and can play VERY loud without noticable distorsion considering only 7 inch driver. My Proacs cannot even approach this volume level without severe distorsion. Imaging is very good. Due to more laidback voicing laying more behind speakers. Transparency is exceptional. This speaker with Fountek treble really excels where gitars and percussion plays major role. Just listen to Nils Lofgren's "Keith don`t go"...

It is really a great design (thanks of course to your great knowledge about speaker design!), but with my equipment it is not the best partner. I borrowed my friends good LP turntable and things got better. When I add a valve power amplifier things get more rounded and smoother with more enjoyable sound , but due low wattage, output dynamics suffer.

I chose this design because I really wanted to know what a good design with magnesium cones would sound like. I know now why Joseph audio loudspeakers  are each time represented with big Manley valve power amps.  Cause valves are great match to magnesium low distorsion cones. That's are my experiences from today's listening session. But with my equipment and CD collection it's just not the best match.

Because I've got nice TJL cabinets done, I decided to keep these boxes and choose between CNO or 8545 + 9500/9700 projects. I am targeting to some kind more forgiving sound compared to the Excel units. Some of my CDs sound great, but most of them sound less enjoyable with TJL JP3. I just wish to have more warm and smoother sound, like old Sonus faber designs. The Excel magnesium designs  have great transparency, but I miss that some kind of "lushness" which give voices and instruments more natural and less edgy sound.

I could say goodbye to some deep bass and trasparency from TJL JP3 speaker to get little bit more sensitivity speaker with more relaxed sound. I've listened to Sonus Faber Electa Amator speakers which use 8545 driver paired to 19 mm tweeter from ScanSpeak. To my ears it sound marvellous...But I wish I could also hear the Seas 7 inch Nextel coated driver. Is there any major differences between these heavy coated units? If you got good a LP turntable in your system or valve preamp or valve power amplifier, than this system could sound magnificient. If you will pair it with all clean sounding solid state electronics than you might get a too transparent sound. Not agressive, just very revealing. Regards, Andrej

Andrej, thanks for your comments to "the clone" and the TJLs. System matching is a never ending story, but in comment to your mail, I can't help mentioning that I a few days ago received the latest recording done by John Atkinson at Stereophile: Attention Screen/Live at Merkin Hall. This is what my wife calls stress-jazz, but I like it and the recording is excellent. Ordering CDs from the US I also included all the Stereophile test CDs and here we find recordings so good, that I started wondering if my LPs could ever deliver something similar. Magnificent recordings! I was in the process of fine-tuning a small mini-monitor in replacement of the W11, a W12 Nextel construction: http://www.troelsgravesen.dk/W12.htm, and I couldn't help thinking that some of our speakers are really better than we think they are because most CDs really sound crappy, when all of a sudden you're confronted with some high-quality recordings.
Regarding the TJL, I don't think we can find drivers being more natural than the SEAS magnesium drivers except for ceramic cone drivers. If the sound is edgy, we may have to look at the whole system and try to identify bottlenecks. To my experience good valve amps don't smooth things; on the contrary, they rather display the shortcomings of a lot of transistor gear. I recently had someone bringing in a huge NAD integrated amp, NAD M3 I think it was. Very well built and very impressive features, but the sound was dull and life-less. It seriously lacked dynamics and transparency.

TJL2W-T25, built by Max/Taiwan
TJL2W, built by Henrik/Norway
TJL2W, built by Kevin/US, the most thorough report ever received.
TJL2W, built by Andy, US

BACK TO INDEX