The Usual Suspects

Parts Express #297-464  approx. $8  Textile dome despite the shiny appearance

These were purchased a couple of years ago just for their response and cost. No consideration was given to harmonic distortion as this is one of the characteristics that differentiates the sonic quality of a loudspeaker. Most manufacturers don't publish this data for obvious reasons.  One is that the distortion produced by any loudspeaker can be several orders of magnitude higher than that of the amplifier.  See fig.1

I wasn't looking for anything esoteric, else a buyout would probably have not been considered. They actually proved to be more than worth the $5 each as they tested my mettle.

This project isn't something one would consider as it involves characteristics of loudspeakers not known by the average hobbyist.  In fact, considering the work involved, it wouldn't be something anyone would consider to be a practical venture.

Looking at the mfr's data sheets (HiVi T20), the unit is flat within about 3dB from 1500hz to 22000hz. This, of course, ignores THD which is probably why the response is stated as being from 4khz

This graph shows the response to be flat within 3dB from about 2400hz before THD begins to rise.

This graph shows that at 1khz THD≈31%  Above 3khz, THD≈2.5%  The mfr's stated response from 4khz could pushed to 2400hz and still have less than 2.5%THD. However, this would reduce the power handling capacity, especially with a first order high pass filter.

This response was measured at 1w1m with a 2khz first order filter.

Response of the 5" at 1w1m and high pass filtered at 150hz. Here we have a peak around 900hz, a full octave dip between 1khz and 2khz and a nasty rise to about 8khz. This poor little fella sounded awful

The initial 3-way with the Wharfedale W12RS and the HiVi T20 tweeter. Crossover frequencies are at 400hz and 2700hz, (later changed to 2300hz) second order with exception of the tweeter which is first order high pass at 7900hz. This frequency spread was necessary to flatten the response above around 2khz. The initial design crossing both the midrange and tweeter at the same frequency all failed. Several crossover points were tried. This is attributed to the rising high frequency response of the midrange.  See fig.4

The peak at 1khz and the dip around 1300hz were of concern, especially the latter which will be discussed later. The reason for this latter concern is because it seemed to defy all initial attempts to remove it. The rise in THD around that frequency was also of concern. Suffice it to say at the moment that the two are related.

The bass reflex is raised to get the midrange and tweeter midway between the floor and ceiling which is optimum for a gated response.  Bass response was not considered as the room takes effect below about 300hz. This also adversely affects a gated response, limiting its low frequency to about 300hz.

A closeup of the midrange unit. The dome and cone may look odd and that will be discussed later.

Just a peek at two of the several  worksheets

The final crossover 5 with the notch filters, N3 and N4. Note the woofer's polarity reversal.

The tainted paper colour is due to its age, around 40 years, dating to the eighties.

Now, we come to the solution of the aforementioned response peaks and dips. Note that two of these curves have no such peaks and dips and that has nothing to do with the gating process which removes room reflections by some 95% above 300hz.

The RED is midrange by itself without filters. The green is that of the midrange and tweeter only, with crossover filters, hence the drop below 600hz. The woofer was disconnected. The black curve is the whole system with crossover #5. the slight shift around 5khz is attributed to high frequency output from the woofer despite its second order 300hz low pass filter and possibly overall system impedance. Keep in mind that the three sections of the crossover do affect each other, however subtly.

OK, so where went the peaks and dips?  Refer back to photo 7. the dustcap has been coated with RTV, an uncured rubber resembling toothpaste with an odour of vinegar. This cured some of the nasty high frequency stringency. Notice the narrow rubber strips of speaker gasket stuck to the cone circumference and the annulus.

These were added for the following reasons. The one attached to the cone was intended to dampen the cone. It was spiraled from the apex to the circumference with one revolution. It helped a little but the dip at 1300hz (fig 3) was still present, along with its THD component. It is known that frequency wavelengths and medium dimensions will interact. So, a similarity was investigated. It just so happens that the wavelength of 1300hz is 10.47" and the circumference of the cone at the annulus is 10.4".  What is happening is that the annulus is resonating at 1300hz and the transverse wave along the cone is being reflected back into the cone, causing the dip in response. Dampening this resonance in the annulus cured the problem

1w1m Xover5 GREY N3 BLK N3 N4

N3 affects the 5khz to 7khz peak    See fig 3

The upper grey curve is with the N3 notch applied. It tames the peak around 5khz to 7khz.  see fig.3

The lower black curve is with both notch filters applied. The resistors in these notches are variable, hence their effect is adjustable. This isn't recommended for actual use as these potentiometers are rated at 5 watts and being inserted at the beginning of the crossover, will take the full current of the system.  they created slight problems if inserted into each section of the crossover for the following reasons.

First, they'll affect the section impedance.

Second, they will only affect the frequency in that section. Since the crossover filters have a slope, 5khz is present in the midrange also, however suppressed by its 2300hz low pass filter section.

1w1m Xover5 N3 N4

The response of the system. This is the same curve as the black one in fig.7. The lower curve is THD. Note that the large distortion hump between 1khz and 2khz is gone. see fig.3

Ignore the response below 400hz as this is room dependent.