(spoken Demokritty)

I guess this will be my last experiment with Demokrit for the time being.

The initial thought was to replace the reflector by a tweeter to find out about the sonic differences if the diffraction at the cone is avoided. The used tweeter is a Vifa OX20SC00-04, one of the smallest available. This makes it possible that the sound waves can travel around it to maintain widest possible dispersion.

As can be seen from the initial measurements, though, the speaker will now exhibit some amount of forward bias. I also expect, for what it's worth, that the vertical polar response is now wider than with the reflector.
I have saved myself some time for now so, I only show the responses under significant angles (0°, 45°, 90° and 180°). As always, the ripple between 200Hz and 1KHz are due to a floor reflection.

The response from the back at 180° actually shows more output than to the sides because now the magnet cup of the tweeter acts like a diffraction device. It is the same phenomenon that can be observed with the reflector version when measuring straight from the top of the speaker only to a much smaller degree.

From a mono listening test with this reasonably flat response it is obvious that Demokrit-T has become a very different speaker now compared to the reflector version only because of this "little" additional directionality. Hence, the equalization will be somewhat different as well. The cross-over point might also change long term.
The looks: Well...I needed an easy and quick retrofit so that I can compare both versions without building another set of pipes. The mounting is stretched copper lead wire, which is used as signal conductor at the same time.

Is this yet another Pluto inspired design ? No !
Although they share driver arrangement and the goal to minimize diffraction, the directional behavior is very different. If I would talk of similarities, then Demokrit-T is something like the so-called ZDL speaker (Zero Diffraction Loudspeaker) started by Joachim Gerhard.


Update 18-Sep-2012:

Stereo listening tests and the equalization process have started now. Before that, I had to modify the tweeter mounting because the tweeter was too far in front of the pipe and thus creating phase issues in the cross-over region, which could not be fixed by any amount of time delay (visible in the pictures above). The tweeter is now further back, right above the edge of the pipe and the cross over region looks much smoother.

The equalization process has started with this flat response, which already contains some response shaping. The blue trace is again the response at 90°.
Also with the tweeter and this flat response the sound is too bright but it is much less wrong compared to the reflector version with a flat response. That is mainly due to the decreased vertical output (the 9BN is band pass filtered now) and the beaming of the tweeter.

It is most amazing how this amount of directivity changes the auditory scene. The directionality is now approx. in the middle between Pluto and Demokrit at 10KHz. Only this experience was worth the work. The stage is now less wide and the side walls are noticeably a bit less illuminated. The distance of the AS has grown from a front row seat to a small amount of rows back in the audience.
The phantom image has changed as well of course. But more on that later.

Update 23-Sep-2012:

The equalization that is being used for the first long term test is shown at the right hand side. From now on, changes are typically within tenths of a dB.

As can be seen, also with this speaker a non-flat on axis frequency response is required for a balanced sound in my live sounding room. Compared to Demokrit, the sag between 4KHz and 14KHz has to be flatter now, which perfectly correlates with the increased directivity of the tweeter and hence the less diffusivity of the sound field in that frequency range.

The peak centered around 2.5KHz again performs the magician trick:
It sharpens and emphasizes the phantom image nicely and naturally. Without it, the image is more blurred, muddier, duller and the features of the center image are pulled towards the pipes. The peak is absolutely required for such speakers.

So far, Demokrit-T sounds a bit smoother and cleaner with more focus in the center image. The stage has a pinch less depth but the highs are more shiny and have a nice brilliance. Another change that was apparent immediately is the reduced IACC in the highs.

Update 07-Jan-2013:

Highs are somehow attractive or even addictive at first with their clarity and details. But I had to cut them back meanwhile in order to achieve a more neutral representation. Also, the approximation to the loudness curve was not good enough and had to be improved in the 4KHz...5KHz range because there was a small amount shrieking where it does not belong (visible in the graph above). As already learned previously with Demokrit, this makes an adaption in the 5KHz...14KHz range mandatory as the two regions interact perceptually. The result of the work is shown to the left.

This is yet another proof for the absolute applicability of the equal loudness function for speakers that create a more diffuse sound field in a small room. Without this target function the equalization process would have been much more difficult.

Meanwhile, I find it rather amazing that the folks, who deal with PA systems know about these neuraglic frequency bands whereas the home speaker builders seem to more or less neglect this correlation.

As it stands today and as already indicated previously, the reflectors have severe long term potential for the basement...

Update 29-Jan-2013:

Meanwhile, I have convinced myself once more that a monotonically falling FR as equalization method for an omni does not do “the same things” like the EQ used above. I varied the magnitude of the shelving filter (100Hz...20KHz applied to the flat response as of 18-Sep-2012) from -2dB to -5dB (and even beyond) in 0.1 dB steps. But no matter what...

It is simply not applicable !

It does not create any natural balance at all and ranges from dull to shrieking otherwise. Therefore, I saved myself the measurement of the real FR.

Update 23-Feb-2013:

The picture to the right now shows the accurate but equalized to flat response of the Vifa OX20 tweeter when rotated around its center (unlike in the first three pictures above).


  • Ungated responses
  • Measured indoors
  • Excitation: periodic noise
  • Mic. distance 30cm
  • 92cm above ground
  • LR8 HP @ 1KHz to block low frequencies


From this data the Directivity Index of Demokrit-T has been approximated. The graph is representative for the entire speaker as it is perfectly omni below 2KHz.
The DI is reasonably uniform and ends below 6dB due to the wide dispersion of the tweeter. Other tweeters reach 10dB or even more especially when used in a waveguide.

I am not generating the sonogram, waterfall or polar plots until known calculation inaccuracies concerning such diagrams in Arta have been addressed. They are rather useless to look at until then.

The red graph at the right hand side again shows the response from the back of the tweeter, which is partially higher in magnitude because of diffraction.

Last updated 03-Feb-2019