Click for REVISION ADDED March 13, 2022

This project started a few months ago, Sept. '21 when a member of an audio forum contacted me about repairing a Super 3 tweeter that had the braided wires ripped from the cone as well as the terminal strip being removed from the basket spoke.  For future reference to this fella here, his name is Jeff.

Anyway, he sent the tweeter to me which I managed to repair.  It was tested for lengthy periods above 3khz at levels it wouldn't experience if one valued their ears.  This was done to verify the structural integrity of the repair. Initially, the test failed as the solder joint of the braid to the aluminum voice coil wire opened.  This was not a surprise to me.  Aluminum wire solder requires about 30% more heat that lead/tin alloy solder and I was over cautious so as not to burn the cone.

Subsequent to this repair, a conversation started regarding the pair of W70 enclosures he recently purchased.  His main concern was the brightness of the upper mid-range.  Having the three drivers here in my collection along with a 3 ft^3 enclosure, duplication the W70 was a piece of cake, which I enjoyed with coffee.

Initially, a band reject filter was employed to lower the peak in the 10 inch unit.  This worked with the 10 inch in and by itself but failed when coupled with the tweeter in the crossover.  After some testing, two notch filters were employed as the width of the band was too wide to be handled easily by one filter.  This gave a better result than the single filter but was still not enough; I was trying for a response flat within about 7dB.  Further thought brought forth the idea of placing the higher band filter into the tweeter circuit but this was also for naught; the tweeter response had no peak in its lower passband.  It was then that the peak was being exacerbated by the effect of interaction of the upper mid and lower tweet radiation patterns, so the filters were placed at the input to the system.  This would reject that band from all speakers and theoretically null the peak.  It worked.  Now, an aural test had to be done so the system was moved into my living room.  It initially sounded ok until female vocals were played, like those of the country music genre where it failed dismally, especially at levels around 90dB and worse above that.  Back to the drawing board.

All practical further attempts failed.  By practical, I mean keeping the cost down and also that I was under the impression that this fella wanted to keep the system as original as possible.  During this experimentation, a second order low pass filter was employed in the upper mid-range. The mid-range polarity was reversed as per the norm for second order filters but this produced a serious null in the upper bass.  See, the 10 inch mid-range isn't high pass filtered. The lack of a high pass filter around 500hz in the original design will create a serious null below the crossover point due to the 180 degree opposite phase relationship between the woofer and mid-range if the polarity between them is reversed.  This second order filter used was only for the low pass upper section of the mid-range. So, the mid-range was changed to normal polarity and the polarity of the tweeter was reversed.  This reduced the mid to tweet peak and removed the null between the 12" and the 10".

The lengthy testing at high levels proved the ability of the tweeter level control to hold that power, despite their being rated at 5 watts.  However, that rating is for DC.  If the peak AC power is assumed to be that of the DC level, then the AC power transfer is 0.707 times less, the rms AC power.  Testing with a scope showed a power transfer of about 15 watts on most of the peaks and usually in the low end.  This is about 10 watts rms.  Despite that this is twice the potentiometer's rating, it's also not DC but AC and fluctuating with frequency, so the power is only momentarily applied. Again, an all afternoon jam session produced no detectable increase in the potentiometer's temperature.

It was then that I was notified by Jeff that an alternative mid-range unit be used.  So, I searched for a reasonably priced unit with a good response to above 5 khz and Lady Luck was with me.  The Visaton BG17 fit that requirement so two were purchased.  They were measured for a sealed cabinet size to about 300hz and such cabinet was made. The original crossover design, such as that is, was employed with disappointing results. So, a new crossover was designed.

This new crossover uses a first order low pass filter for the woofer.  The mid-range uses a first order high pass at 300hz and a second order low pass at 3000hz.  The Super 3 performs much better between 3khz and 5 khz than the Visaton.  The Super 3 is first order high pass filtered at 3khz, close to the designs in Jeff's units.

Temporarily, I suggested his reversing the tweeter polarity and changing the 3.5uf filter capacitor to 1uf.  This will begin a roll-off of the tweeter around 9khz.  This will result in a reduction of the peak between 1.5khz and 7khz.  As of this writing, a reply from him is still pending.  He, as of this writing, has no idea what I've achieved with the Visation and a new crossover.  These erzatz W70's sound very close to my Wharfedale corner 3-ways, albeit a little louder in the mid and tweet due to the W70 facing the listener and my corner units have the mid and tweet upward firing.  (see fig.7 way down under)  However, the new crossover employs a level pad in the mid-range as well as the tweeter and a little juggling can remedy that difference.  The corner Wharfies are used as a reference as no one as yet has criticized their sound, despite the speakers used being 65 years old.

A scan of an original W70D brochure.  Note, it's a 4 way and a sealed enclosure,   It's called an acoustic suspension system but it really isn't.  An acoustic suspension system is completely sealed whereas this uses a passive radiator which, in this case, is opposite to usual passive radiator design where the passive radiator usually has an effective piston area of at least 50% higher than the powered unit. based on that, the 12 inch woofer would require a 15 inch passive radiator.

In this case, if the 12" were to be driven to it's Xmax of, say, 4mm, the poor 8" unit will try to move 11mm.  While it could be designed to do that, it wouldn't do any good for it's upper register as modulation distortion would go high.  Again, this is not to say that such a condition can't actually sound good.

It is said that beauty lies in the eyes of the beholder.  With sound, one can say that a pleasing sound lies in the ears of the beholder.

The innards of another W70 found on the net.  Not Jeff's but extremely similar.  This has an alnico magnet woofer and uses compressed wool dampening.

Compare this to photo 3.

Photo taken by Jeff; the innards of his W70.

The chamber for the 10" had to be made with a cylinder because there wasn't enough clearance for a square chamber without making it with 5 sides to clear the tweeter. See photo 6.

The cylinder is stuffed with 130g pure washed but otherwise untreated Australian wool.

One of the early adjustable depth notch filters. This was using the same crossover as in Jeff's unit with exception of the 3.5uf capacitor for the tweeter being changed to 1uf, raising that crossover point to about 9khz.  The 10" unit operates full range.

The tape is used to hold the components while moving the enclosure to the living room.

In the living room

The Visaton BG17.  Very good midrange speaker with a response of 140hz to 8khz within 8dB.  Available at Parts Express   Here's the data sheet

Installing the baffle and at my age, that wasn't easy.

Under test in it's chamber, stuffed with 130g of wool.  The chamber is screwed to the inner face of the front baffle.  the speaker is front loaded.  This should eliminate modifying the square hole for the original 10".

Under test in the music room

Under test in the living room but here it gets the acid test, listening to various types of music and different levels and comparing to the left corner Wharfedale 3-way.

The design involving the Visaton mid-range is worthy of further description.  After much listening to various music at various levels, another slight change to the crossover was implemented.

The 10" is chambered here; the system is otherwise original with the low pass filter in the 12" woofer circuit and the 3.5uf high pass filter in the tweeter circuit..  The RED curve is the system measured at 1w1m and gated to reduce room reflections.  The sweep is from 585hz and up as the LMS will lose the difference between the direct signal from the speakers and the first reflection from the floor or ceiling. The BLACK curve is near field and amplitude adjusted to coincide with the red curve at 585hz.  That sweep isn't gated as being near field, that isn't necessary.  It runs from 10hz to 585hz.

There were 3 near field measurements made with the same input signal level. The mic was positioned 1/4" from the diaphragms and 1/4" from the end of the vent. 

The BLACK curve is that of the 12", the 10" and the VENT run independently.  The three are summed and SPL level corrected.  As can be seen, there is strong bass output between 50hz and 350hz centered around 80hz.  This produces and drummy effect for lack of a better word.  It can ruin the sound of a male voice and produce a harsh bass similar to that of an early juke box dating to the fifties and sixties. I realize this description may be meaningless to anyone younger than 55.  I'm 79 and remember that sound well, so much so that several years ago, I found on ebay a 15" that came from, yup, an early jukebox.  The idea was to build a thin walled open back enclosure to recreate that memorable sound.  The drawings are complete but the construction has yet to begin. Nostalgia and procrastination.

The nasty peak at 5500hz is cause of that which Jeff described as exceptionally bright. Removing it became more of a challenge than expected and is briefly described in the introduction.

Near field curves of the 10" unit under the same conditions as stated in fig.1 above with the following exception. The BLACK curve is that of the open back (unchambered) 10" unit and the GREEN is that of the chambered 10" unit.  There are two reasons for the difference, at least two of which I can think.

One.  The small sealed enclosure highly damps the cone motion.  This is typical of sealed box woofers which roll off very fast below 45hz.  The advantage is a tighter bass; a kick drum will sound more like the real thing.

Two. The open back 10" is also being acoustically/mechanically driven by the 12" unit, acting like a passive radiator. The damping effect of this type of passive radiator isn't very good due to the light mass and highly compliant suspension of the 10" unit. Being used full range, adding mass to adjust it isn't an option as that would deteriorate the high frequency response. 

While this may appear to be a serious drop in bass response, curves can be deceiving.  These are near field responses and as a result, are not affected by the room.  Also, the system with the isolated ten can tolerate a bass boost if desired without sounding boomy.

The final crossover.  A larger copy can be obtained by clicking on the figure.

The woofer is first order low pass filtered at 325hz. This figure was obtained by determining the smoothest response in the area without resorting to a second order high pass filter for the mid-range. This eliminates L13.

As a result, the mid-range high pass section was raised to about 825hz, first order (C6=24uf). The low pass upper section of the mid-range is second order at about 3khz with a slight unorthodox method, that being the value of C13.  The result if this will be seen and explained in the next figure. L13 (1mh) is the other component of this section.

Of the two 50W potentiometers (pads/pots), the one in the tweeter circuit is necessary but the one in the mid-range circuit is optional, actually not needed. 

Highly modified W70, new xover and Visaton 6"

The result of changing the value of C13.  The GREY curve is that of C13=3.5uf and the BLACK curve is that of C13=6.8uf.  The effect is mostly in the decade from 200hz to 2khz and is of the order of 1dB.  While 1dB isn't very much, it becomes apparent at 90dB levels with certain music, one of which I noticed was female country. At moderate levels around 83dB, the effect is hardly noticed even when switching the caps during a song. 

or, to be more specific, band reject filter

Highly modified W70, new xover and Visaton 6"

RED is without the notch; BLACK is with the notch.  This leveled the response of this 500hz to 2khz band from flat within 11dB to flat within 7dB.

GREY=Original W70 design; BLACK=highly modified W70, new crossover and Visaton 6"

 These curves were run in my living room at a distance from mic to speakers of about 7 feet. (Photo 13)

Modified means replacing the mid-range and adding a crossover. (fig 3)

The tweeter level control will lower the rise above 7khz but has no effect on the nasty peak between about 2khz and 5.khz.

Highly modded W70, new xover and Visaton 6"

Like I said earlier, somewhere, these modified W70's sound very much like the Briggs' 3 way albeit louder due to the mid-range and tweeter in the Briggs facing up.

The Briggs' bass seems a little stronger but consider that the Briggs is tucked into a corner and also has a 15" woofer, the W15FS in 6 ft^3. Actually, the W15 is  more of a 13" to 14" woofer if one measures the actual piston diameter.

Yellow is the W70 and black is the left corner Briggs.

Using the same 12" woofer as used in the W70 and also one of the woofers used in the W90, essentially a W12RS, an analysis of its behaviour in both size cabinets can be studied.

The cabinet used in this analysis is 3 ft^3, shown in photo 15 below (left).  Photo 16 (right) below shows the section added to increase the volume to 5 ft^3, being clamped on with 4 clamps.  The pair are sealed with a foam strip around the perimeter edge. BassBox Pro v6 was used to juggle some of the possibilities.

Photo 15

Photo 16

What was done to get around this and look at the speakers as a single unit, the pair were wired in parallel and the TS parameters were derived from that.  The method used with LMS was the delta mass method, where impedance is measured twice; once with a known mass was stuck to the cone and once without the added mass.  The trick was to find out how much of an effect these masses would have on the result as two masses would be needed, one for each speaker.  Typically, 30g will do it but where to divide that for each speaker.  Arbitrarily, 10g was added to the 10" and 21g added to the 12".  Curiosity got the better part of me so the test was repeated with 5g on the 10" and 21g on the 12".  This changed the resonant peaks of each speaker but the TS parameters changed very little.  A few changed a little but not enough to have an adverse effect on an optimum enclosure.  The Q's and Fs(Fo)are identical.  Plugging both sets of data into BassBox Pro produced the same result.

It would have been preferred to use the delta compliance method where the speaker impedance is measured on and off a sealed box of known volume.  However, the cloth annulus of the ten incher prevented using that method.

I remember a time when measuring a W15FS using the compliance method produced weird results.  Wondering when went wrong, it suddenly donned on me.  Dummkopf.  Foam annulus.

Below are the results of obtaining the TS parameters using two distributions of added mass.

The above tables show the combined figures for the speaker pair.  For example, the diaphragm area for the pair is 122.72 in^2.  This is the sum of the ten, 44.18 in^2 and the twelve, 78.54 in^2.  The program calls for the dc resistance, Revc which is the product over the sum of the two dcr's.  The Revc of the ten is 7.1W and that of the twelve is 10.3W. The result is 4.2W. The resonant peaks are surprisingly the same, probably due to the open air fs of the ten and the twelve being within 2W.(fig.8 left).  Figure 9 (right) shows the two weighted impedances, one curve is with 5g and 26g on the ten and twelve, respectively and the other is the reverse.

Figure 8                                                                                              Figure 9

The next step was to plug these figures into BassBoxPro.  Now, the two speakers had to be considered as one unit, hence the Revc of 4.2W and the diaphragm area, Sd of 122.72 in^2B  Box volumes of 3 ft^3 were used to comply with the W70 internal volume.  Keep in mind that BassBox Pro sees the ten and twelve as one slightly larger unit.  To be specific, the effective piston area of the two is equivalent to a 16" speaker.

Here's the results from BBP.

Three designs, the first two being the original design with it's 2" diameter, 8" long vent.  Design 3 is the W12RS by itself in a 3 ft^3 cabinet, tuned for optimum bass output.  More on this and associated caveats later.

The colours here correspond to the following three sets of curves below.

NOTE.  5184 in^3 = 3 ft^3.

Here, for the single 12" can be seen the extra 3dB to 5dB boost in the bass below 60hz. (yellow) and the removal of the 2dB bump between 80 hz and 150hz 

This 2dB bumb, in and by itself is mostly harmless until one boosts the bass to bring out the bass below about 50hz.  The problem is that the bump will also be boosted and may have an obnoxious sound, depending on the music, the room and personal preference.

The excursion is held below 4mm with power levels up to 50 watts.  Here's one of the caveats. The yellow trace will exceed Xmax at 50 watts below 35hz.  That hasn't been a problem with some of my listening levels approaching 100dB.  This speaker in this enclosure can output 98dB at 30hz and 101dB at 35hz with 50 watts. See fig 13.  Wanting more would be tantamount to madness. 

Based on that, a true wattmeter would most likely show that at that ear shattering playing level, the system may have been on the threshold of exceeding Xmax with less power delivered than expected.  At 35hz, the system impedance for the single 12" is about 10 ohms.

The smaller diameter and longer vent holds the diaphragm excursion well at the cost of bass output.  Designing a speaker system,, like most other things is a compriomise.

For more on my determination of Xmax, see this page.

Custom amplitude response showing output vs power input, while staying within Xmax

The system impedances as calculated by BBP.  The yellow trace is that of the W12RS alone; the red and orange are those of the W12RS in parallel with the W10CSB, hence the lower impedance.

It should be noted that the 10 and 12 are not really in parallel.  The 12 has a low pass filter in series with it and the 10 is actually in parallel with the whole mid-section network. (see fig 3)  In the original system, the ten is connected directly to the input.

These traces match that of the actual measured system impedance. (fig 15)  The left bump is that of the vent resonance.  The right bump is that of the driver resonance.  The null between them is the box resonance, Fb.

The green bump on the right is caused by the 3/4 inch length of the 3" vent.  Nominal impedance is 5.5W around 140hz.

Now, let's look at some near field responses.

All are near field with the original crossover design and the ten is open back.  Later, we'll look at a gated frequency response, followed by my living room response and the ten isolated with the tweeter polarity reversed..  That will be followed by replacing the ten with the Visaton 6" and a new crossover, as shown in figure 3.

Figs 16, 17, 18 and 19 are near field responses of the 10", the 12", the vent and the summation of those three.  There's no point in doing into a dissertation as to which curve is which in the first four as they are described in the fifth, fig.20.  The idea was to get an idea of the bass output under quasi-anechoic conditions.  The alternative was to take the speaker system outdoors along with the electronic paraphernalia.  There, the only non-anechoic deviation would have been the bounce from the ground which can usually be seen in the graph.  The back yard is 55' by 85' and empty.  The closest reflection other than the ground would be the block fence and the house wall, both some 27 feet distant, a wavelength of 42hz  Considering that the block wall is 6' high and the house wall is 9' high, the reflection from either would be less than that from the ground.

Ten or more years ago, dragging a 75 pound speaker system outdoors was a lot easier.

A larger view can be obtained by clicking on the piccy

3ft^3  3"x8" vent

3 ft^3  2"x8" vent

5 ft^3  3"x8" vent

5 ft^3  3"x3/4" vent

Speaker and vent summations for clarity

The red curve is the one I initially used since they abound here.  Later, I made a 2" vent.  See photos 17 and 18 below fig 21.

RED  3ft^3  3"x8" vent  fig 16

It should be noted here that the 3" vent was used because they are plentiful here.  The biggest difference with this vent is that the speaker can be much easier overdriven.  As long as one is aware of this, it wouldn't otherwise be a problem.

BLK  3 ft^3  2"x8" vent  fig 17

GREY  5 ft^3  3"x8" vent  fig 18

LT BLUE  5 ft^3  3"x3/4" vent  fig 19

Judging from these curves, it seems as if Wharfedale did their work well.  The black curve is the 3 ft^3 design with then 2" by 8"  vent, as per Jeff's enclosures.  The grey and light blue are the best I could get with a 5 ft^3 enclosure, including watching diaphragm excursion limits, negating the need for the extra 2 cubic feet.

So, there ya have it, Folks. The 3 ft^3 is optimum.

These two responses were run in my living room.  The bass was a little hot but turning the bass control to 9 o'clock rendered them satisfactory, most of the time. I knew they could sound better, especially for one who likes some music loud, close to 90dB.

When this thing hits 88dB, you're gonna hear some serious schitt. (Variation from Back To The Future, pt.1) I couldn't resist.

RED: original design, 10" open back and 3.5uf tweeter high pass and tweeter normal polarity

BLACK:  10" open back and 1uf tweeter high pass and tweeter polarity reversed.  The tweeter trimpot is set at 3 o'clock to lower the response above 7khz by about 2dB.  The 1uf high pass filter allows the tweeter to work above about 9khz.

The peak around 5400hz may be acceptable considering it's higher frequency and the reduction of the band between 2khz and 5khz, the primary brightness complaint

If this proves to be satisfactory for Jeff, then that part of this work has been successful. But, the story doesn't end here.  My curiosity was put into high gear when Jeff mentioned his willingness to deviate from originality by replacing the ten with a 6" midrange unit. (see figs 6 and 7)

1w1m;  1uf high pass for the tweeter and tweeter polarity reversed

GREY: 10" open back;  Black: 10" chambered

The grey trace is what Jeff could expect with just the 1uf high pass filter and tweeter polarity reversal.

What's interesting is the appearance of the peak at 2200hz.  It may be due to a back reflection as the cylinder is 5" deep.  This corresponds to a wavelength of 2722hz.  Also, the diameter being 10.25" gives a radius of 5" which isn't a good ratio.  It's tantamount to a cubical room.  Also, the magnet assembly is midway between the front and back and also centered in the circular chamber.  This may account for the peak's being some 500hz from the calculated 2722hz.   More wool may reduce or eliminate that.  If so done, the speaker will have to be covered in a mesh bag to keep the wool from pressing against the back of the cone.

The 3dbB to 4dB drop in low frequency below 250hz is attributed to the isolation of the ten from the 12 and also the small volume of the cylinder, just under 400 in^3.

Made from a 2" thin walled mailing tube wrapped with a couple of magazines to fit snugly into a 3" tube.  There are about four dozen such 3" tubes, each 34" long.  The I.D. is 3" and the wall thickness is 3/16".  They were retrieved from Motorola during my service years there in the eighties and were plotter paper rolls.. They're nicely tucked away in the laundry room.