This is a study into rather nerdy high-efficiency speakers using extended range drivers, allowing point of crossover to be taken above 7-8 kHz, thus a single driver covering the entire range of basic notes plus most harmonics.
Nerdy? Well, this is not a speaker for everyone - I think. It's a speaker for those who want to explore the world of low-wattage, single-ended triodes - and not least - the delicate presentation of the very best recordings from either CD or vinyl. It's a discriminating speaker and if you have less than the best source material, this speaker will tell you. Those addicted to heavy metal or contemporary type of music having a constant "noise" level over the entire frequency band, please look elsewhere. I like music having breaks, music with low level passages, etc. How can we build up expectation and excitement if we don't have breaks? Anyway....

Driven by the best of SET amps it can bring you close to the music like only good fullrange drivers or electrostatic speakers can do - but there's a price to pay: The paper cone of the 215RTF64 is extremely thin and careful damping behind the driver is necessary to avoid rear reflections and there's a limit to loud we can play before the membrane starts breaking up and smears details. That's not to say this speaker cannot play loud - my goodness it can - but the limitations are audible. At average listening levels it has given me some very good moments with my vinyls. It's a great pleasure to hear e.g. flute and cymbals not being a blend of sounds from two different drivers, no matter how well they may integrate.
How about beaming? Hmm... I had low expectations here. That is, I thought there would be some serious hot-spot. There is a hot spot, but not as bad as I had anticipated from the large cone driver. I guess the waveguided tweeter blending into upper treble is responsible for this not being as big a problem as expected.

The 215RTF64 is a large 8" driver having a membrane area of 265 cm^2 and efficiency is around 96 dB. In commercial terms this means 98-99 dB from the usual overrating of products. In the treble range it actually does deliver some 100 dB/2.8V, but this won't help if the upper bass and lower mid does not follow - and it doesn't.

TS data on 215RTF64 driver.

Frequency bands

If we study music recordings with regard to treble content, not much happens (in terms of acoustic energy) above 5 kHz. What's above 5 kHz provides the airiness of the presentation. Sorry to reiterate this definition of frequency ranges, but many people regard treble being something far higher in terms of cycles per second. This actually means that in many 2-way systems having a point of crossover at 2.5-3 kHz, the midbass is actually playing a significant part of the treble content. No wonder it sometimes leaves a lot to be desired. But a very light-weight paper cone can actually play treble very well - if well constructed.

Ingredients:

Left: Supravox 215RTF-64. Middle LE26 with waveguide. Right: Supravox 215GMF.
Click to go to Supravox website.
Check here for LE26 details.
Click here for cabinet details.

DTQWT test cabs fitted with Supravox/JBL drivers. Right: Series connected 215GMFs on rear of cabinet.

Crossover v1, "single-cap" version

First set-up is a single cap situation. Well, we need a low-pass filter for the bass drivers. I guess even the single-cap people can live with that....
Having no equalisation on the main front driver takes some very good driver providing a reasonable flat response in all of the midrange and treble area. The Supravox 215RTF-64 is such a driver. Compared to the terrible Lowther drivers, this Supravox driver is a dream. I never tried the 215RFT64 on a rather narrow front baffle before, and I was pleased to see a flat response from 300 Hz to 9 kHz, actually +/- 2.5 dB. Excellent!
The waveguided LE26 is a quite capable driver down to 1.5 kHz and from a single 4.7 uF cap we have high response down to 1.3 kHz and phase-tracking is not good at all.

The set-up here sounds better than I would like to admit, most likely because the elevated 1-7 kHz region is reduced some 2-4 dB by the tweeter out-phasing the response from the 215RTF64. Not an ideal situation, but lower/middle is now well balanced to the important middle/upper midband. However, on vocals and some instruments it can be heard that something is seriously wrong due to poor phase tracking.
Reversing tweeter polarity will increase treble region some 3-5 dB - and it sounds really bad I can tell.

Overall this set-up gives promise of better things to come. Having no point of crossover up to 8 kHz is special. We get very close to the music and as said, beaming from the 215RTF64 is not as bad as I had feared. I suspect the waveguided tweeter plays a role here as it did with the JA8008/TW034/DTQWT construction.

One reason for good mid and treble sound can be seen above. Blue = frequency response of 215RTF64 on DTQWT front panel. Red = minimum phase. Now, what's important here is that we have no excessive phase shifts in the 200 Hz - 8 kHz range. Take a look at the Lowther EX3 where we have close to 180 deg. phase shift at 3.5 kHz, close to where the ear is most sensitive. This can be heard and be even painful from certain instruments, like the obo.
You may think the bump at 1800 Hz is really bad. It isn't. Nothing severe is going on here in terms of sudden major large phase shifts and from even mild equalisation as can be seen below, it even starts looking good on paper. This, by the way, tells how bad frequency response graphs are in terms of telling how a speaker actually sounds. The phase graph may actually be more interesting.

Left: Response of drivers without crossover. Right: Red = summed response. Blue = 215RTF64, green = LE26wg.
As predicted by LspCAD, the 1500-5000 Hz range is reduced by tweeter being out of phase with 215RTF64.

Crossover v2

Left: Response from V2 crossover. Right: V2 (blue) and V1 (red) for comparison. V1 not normalised for 2.8 V input.

With the V2 crossover we have overcome some weird sounding problems with the tweeter mingling with upper mid and treble and being out of phase with the 215RTF64. Basically the 215RTF is tuned for a sensitivity of 97-98 dB/2.8 volts and the treble range almost flat. All this can be heard.
The two bass drivers really can't make up to a tuning of 98 dB and the overall presentation is rather slim and forward. Overall it sounds good, but it's very selective on recordings. Solo, due and trio instrumental stuff goes very well and you're close to the music as ever, but more complex orchestral works incl. vocals become just too massive and forward. The 215RTF needs some baffle step compensation to take the voicing down some 2 dB - well in accordance with the calculated efficiency of the driver.

Crossover v3

To flatten the response of 215RTF driver we need a series coil by-passed by a resistor. This will reduce overall response without sacrifying upper extension as can be seen above. The value (mH) of the coil determines where the tilt should set in and 0.33 mH, 0.47 and 0.56 mH was tried in the simulation program. Finally I decided on 0.56 mH bypassed by 22 or 33 ohms. This will reduce the 800-8000 Hz range to some 95 dB/2.8 volts. The treble unit had a 3rd order filter doing very much the same as the LCR circuit in V1, but with better phase tracking.

Left: Response of drivers driven from crossover and summed response (blue).
Right: Summed response at ½ meter distance normalised for 1 meter/2.8 volts. Blue = minimum phase.

Left: Distortion measured at 0.5 meter distance. Input equivalent to ~90 dB @ 1 meter distance.
Right: Same as left, but input raised to resemble 96 dB @ 1 meter distance.
Red = response from driver. Blue = 2nd., green = 3rd., yellow = 4th., purple = 5th.

I thought it might be interesting to see distortion from this set-up and indeed it was. I had never expected such low values in the treble range. Compared to the Lowther EX3, this driver is doing extremely well.

Cumulative spectral decay. Left: 20 dB level, right, 40 dB level (not really relevant).
I've seen much worse than this from conventional 2-way systems.

Left: Impulse response. Right: Step response.
The step response looks almost ideal. Tweeter is a tiny bit ahead of the 215RTF,
but this is very much how a time-coherent system should look like. Perfect!
Microphone position is exactly between the two drivers.

I've nicked this from Stereophile with John Atkinson's comments.
The 215RTF has a faster decay compared to the graph shown.

Above Schroeder PLot and ETC plot displaying low stored energy.

Above horizontal dispersion. I didn't have to high hopes on these measurements but was positively surprised.
Left: 0-10-20-30 and 40 deg. (red/blue/green/purple/black). Right: Only 0-10-20 deg. shown.
This is not bad at all, considering the large size cone of the 215RTF driver.

Vertical dispersion at 1 meter distance with microphone placed at various heights, i.e. from lower edge of 215RTF to upper edge of waveguide.
Now, crossing over at 8 kHz is calling for trouble unless we have a dual-concentric driver, and this set-up is no exception.
However, what we see here is hardly what we hear. Our two ears are not compressed into a few square millimeter sized microphone capsule excluding all reflected acoustic energy and fortunately our brain does not work as the CLIO system - although it might be convenient some times :-)

So, what about the sound from the V3? Well, taking the mid and treble range down to what the bass drivers can manage helped a lot. Upper bass, lower mid now seems well balanced and vocals get the fullness we are looking for. The rather extreme forwardness of v1 and v2 is gone and we can manage a wider range of musical sources. It still very much sounds like a fullrange driver - for better or for worse - but none of the intimacy with the recording seems lost. After some listening the upper treble seemed a little high pitched and I was kind of aware I had a tweeter in the system. If we can "hear" the tweeter - if you get my meaning - it's playing too loud. We should hear "treble", not "tweeter", thus a mkII version of the v3 crossover, where R2021 is raised to 33 ohms and the tweeter series resistor is slightly increased and the first cab in the 3rd order filter is lowered to 1.5 uF. The difference is shown in the simulation below and the idea came from simply disconnecting the tweeter. We might expect the overall sound to collapse from this - but it doesn't! It actually sounded better with the tweeter off, although the airiness of the presentation was pretty much gone. This just shows the 215RTF is really handling a significant part of the presentation.
(Maybe those expensive French PHY-PH drivers with the tiny piezo-"thing" mounted concentrically are really worth trying some day).

V3, mkII

The mkII turned out to just a bit too laid back and simply removing R1011 did the job. At the same time I changed all caps from standard propylenes to Superior Z-caps. I hadn't expected much from the caps at 8 kHz - but I was surprised. Clean and sparkling treble from this minor tweak.

V3, mkIII

Next I tried reducing L3011 from 18 mH to 8.2 mH to gain some more weight in lower octaves. Very little difference it seems. I don't want this speaker to be a bipole (yes, bi-pole, not dipole), e.g. look here for an example of what a bipole is: http://diyaudioprojects.com/Speakers/FE127E_SB/FE127E_SB.htm
In theory we can avoid bafflestep loss from a bipole, very interesting actually, but some other time.

I guess the 215RTF64 drivers are breaking in further during this study. They haven't been used since my 2007-open-baffle experiments and they seem smoother than ever.

Now, this all look pretty cool, but how about the sound?
Well, no matter what we do, we can't do much about the beaming from the large extended range driver that has to cover all up to 8 kHz:
1. If I was willing to sacrifice the vocal recordings that simply do not sound good on this system, I'd give it a green light.
2. For the nerdy "single-cap" people - a full recommendation.

For making a pair of speakers I suggest the V3, mkIII crossovers and these components:
2 x LE26, search eBay. Necessary mods can be found here.
2 x waveguides from here.
2 x Supravox 215RFT64 drivers from here.
4 x Eminence Deltalite II, 2510 bass drivers, lowest price found here.
Cabinet details here.

That's it - and please do not ask for other tweeter options, it won't be given.

I do not recommend the 215GMF bass drivers. They are good, but too expensive for this application, around 200 EUR each. My estimate is 800-900 EUR for all drivers + crossover components + cabinet materials.

V4, final experiment

Left: SPL from drivers driven from crossover. Red = summed response.
Right: Summed response (red) and minimum phase (blue).

Left: THD.
Right: System impedance. Pretty low at 7 kHz but no problem for even NFB SET amps.

Left: CSD, 20 dB. Right: CSD, 40 dB.

Remembering my OB7 experiments, the final experiment here was to take the point of crossover down to 4-4.5 kHz. You can read my comments to this experiments at the bottom of the OB7 page. Only comment this time: My vocal recordings returned to the turntable and into the CD drawer.....

Left: THD, V4, shown again for comparing to V3 (right).
Taking the LE26 down to 4.5 kHz introduces higher THD in the 5-10 kHz range.