The time has come for the ears to take over. Listening to the speakers, the drivers sound fine crossed at 1,500Hz. My greatest concern, the tweeter, did not exhibit the difficulty it was going through earlier when it was crossed at 1,250Hz.

As for the woofer, I am encouraged that all the hashes at the upper end are well suppressed. That "brittleness" that's associated to cone breakup is totally absent. Sonically, the crossover point is seamless. Overlapping of drivers is inaudible, resulting in a distinct Midrange transparency.

With the cross point established, the next step is to refine the sound. Even with 4th Order Linkwitz Riley networks, the music is still very "Raw". Much more work needs to be done to bring out the best from these Vifas.

Voicing a Speaker

This is where it makes or breaks the speakers. Over a period of a few weeks, the irregularities of the system were largely ironed out. It was accomplished mainly by ear, using LMS only to verify the outcome. The most difficult component was the PL27 tweeter. About 90% of the work done was in fact on the PL27. Let's have a look at what was involved.


In Figure 31, the top graph is the response of the PL27 without any network.

The lower graph is with the crossover, L-Pad and a Notch Filter centered at 700Hz to null the Free Air Resonance Impedance Peak of the PL27.

Note that apart from the reduced SPL, the response is largely unchanged from 20kHz to 3kHz. But from 3kHz to 1.8kHz, a large chunk is missing.

This was eventually traced to the Notch Filter. Apparently, the Notch Filter, being so close to the crossover point has an adverse effect on that region.

(Figure 31) 4th Order Linkwitz Riley High Pass


Removing the Anti-Resonant Notch Filter restored the natural response as can be seen in the solid curve in Figure 32.

However, the tweeter is far from finished. The peaks and valleys span almost a 10dB window. On playback, sibilance and shrillness in violins were clearly discernible.

A Notch Filter (L3, C3, R1) eventually tamed the peak at about 4kHz (Figure 33).

A tweeter conjugate filter for impedance compensation that was installed earlier, had to be removed and in its' place is a shelving filter (R2, L4) instead. This prevented the higher frequencies from climbing and allowed the shelving filter to level off the frequencies above 10kHz.

In the final crossover, the irregularities have been reduced to about a 5dB window. This is a massive improvement from its' initial 10dB.

This may not seem much on paper but on listening, the improvements are remarkable. Personally, I would have preferred to have the frequency response flat to within + / - 1dB but with this tweeter, this is about the practical limit.

(Figure 32) Final Frequency Response of PL27TG35

(Figure 33) Final High Pass Crossover Network for PL27TG35


The PL18 woofer fortunately, is less problematic compared to the PL27 tweeter. With the crossover point well away from the upper end, whatever nonsense from 3kHz upwards is well suppressed (Figure 34).

However, it presented an entirely different set of problems. Unlike the tweeter, where LMS can be used to verify the equalization circuits, the PL18 had to be tuned strictly by ear alone. I did not have the luxury of LMS assisting me with measurements.

For LMS to measure down to 100Hz, the Meter On Time will have to be 10mSec. The space constraints of my lab will compromise a Gated Measurement with a 10mSec window. The only practical solution is to do a Ground Plane measurement. This of course will have to be done out of the lab, which is impractical when one is in the process of voicing a speaker.

Fortunately, over the years working with this woofer, I have acquired an intimate feel of its behavior. Using my collection of CDs, it was obvious that the frequencies between approximately 200Hz to 600Hz could do with some cut. Due to this emphasis, vocals pitch sounded like it has shifted slightly lower. A very broad notch filter consisting of R5, L5, C4 in Figure 35, restored the vocals pitch. Cleared up the "veil" in the voices too.

In contrast to the tweeter, the conjugate network (R6, C7) is a necessity. On removing this network, I was greeted with a pronounced peak right at the shoulder of the rolloff.

(Figure 34) Final Frequency Response of PL18WO09

(Figure 35) Final Low Pass Network of PL18WO09


60 Downes Street | Calais | ME 04619 | USA