The Whizzer

aka: A Walsh driver variation



A friend made mention of a loudspeaker with a whizzer instead of a typical cone.  A whizzer is that little cone mounted in the center of a conventional loudspeaker.  It claimed function is to disperse the higher frequencies.  However, since this whizzer is mounted to the main cone at its apex, it also oscillates with the main cone.  It therefore acts as a Walsh type cone.  Being mounted at the apex of a larger cone and being lightweight, it has little to no effect on the alignment of the loudspeaker's voice coil since the larger cone is supported at its outer circumference by the annulus.  Removing the larger cone also removes this outer support, making the whizzer best operated in a vertical direction with the magnet on top.  With the magnet at the bottom, the moving part of the system becomes mechanically unstable.  This can be stabilized with two spiders spaced about a half inch apart but this adds an additional load upon the cone.  Dual six spoked spiders would reduce that effect substantially but this would require a longer voice coil bobbin to accommodate the two spiders.  Dual spiders were used on Cetec-Gauss woofers.

The idea behind the Walsh driver was to obtain a cylindrical radiation pattern.  This was considered decades ago but no one has yet been able to make a reliable pulsating cylinder.  I suppose a system like a flat diaphragm with the voice coil affixed to the inside in a zig-zag manner like the Magnepan with then diaphragm curved into a cylinder but this poses a a few problems.   First, the diaphragm would have to be flexible enough to be corrugated in some manner so as to expand and contract.  Second, the low frequency limit would have to be limited rather high, like at least 400hz to avoid the effect of the chamber inside the cylinder.

The closest to a pulsating cylinder seems to be the Radialstrahler MBL 101 X-treme




The first two studies on the cone.  The black is the first in which the speed of sound through the paper wasn't considered.  This was an oversight.  After measuring the speed of sound in the treated paper used for the cone, corrections had to be made. (red)

The height of the cone was arbitrary and worked OK as the system does well to about 1khz.  Taking it lower may be attempted later as 400hz would cover most of the midrange.  A 1 inch diameter voice coil will suffice but an underhung coil would may be preferred as this would keep all the wire in the magnetic field and improve efficiency.  To maintain close to the same number of turns or more, a thinner wire will be used. This coil is wound with AWG 36; the next will be wound with AWG 40



Construction and Assembly


he cone has already been made.  The measurements for the supporting ring and base are on the drawing board or, in this case, the dining room table.

The aluminum disc is the bottom of a pie plate.  The idea is to make a cone with it.




The voice coil is one of two left that were hand wound years ago to repair Wharfedale Super 8 units.

The spider is one of several ordered from China along with several cones ranging from 12", 10" and 8"




The level was used to determine how true the cone and coil assembly was to the magnet gap.  It was referenced to the level of the table. It was again checked after the magnet was placed into the recess of the board, the purpose of which is to prevent the magnet from sliding.



PHOTOS 4 and 5

The foam annulus isn't glued to the wood ring. A close look will show an inner layer of veneer to hold the foam by friction only.  This was done for two reasons.

One, it's more than adequate for this purpose, to keep the cone and coil parallel to the magnet gap and two, it simplifies assembly and disassembly.

The extra cone was made later as was the addition of the barrier strip.  Initially, this base was used in the assembly seen in photo 8 but that proved awkward.






It may look crude but it doesn't buzz at any frequency above 800hz when driven with about 8 watts, after which a buzz was detected and stopped by lightly touching the wires close to the cone apex.  Further operation was ceased below 800hz to avoid breaking a wire.  Subsequent response tests were done with a 1.2khz second order high pass filter, seen in photo 11










The two mics in the same vertical plane parallel to the vertical plane in front of the loudspeaker and of equal distance of 100mm, about 4 inches.  A difference of 1mm will present an error of 0.7% in arrival times.

The speaker was pulsed with a half cycle at 500hz.  Both mics received the signal at the same time and at the same amplitude of 40mV, peak.  See FIG 2, next.

This was to ensure that the mics had equal sensitivity and output.




The black and red traces are those of the two mics.  The initial peak is the half cycle.  the subsequent peaks are cause by oscillation of the cone after the initial half cycle ends.  There may be some ambient room noise also, mostly generated by computer cooling fans.





The two mics and the paper strip.

The magnet was borrowed from a Wharfedale Super 8 FS/AL which has a phenolic spider, making it very easy to disassemble and reassemble

The voice coil is one of my hand wound units for the Super 8



Two photos showing the proximity of the mics to the paper strip.  Here is where a timing error can occur.  the distance of the mics to the strip is critical.  Measuring the gaps was a royal pain in the butt (rpitb) not only due to the flexibility of the strip but also due to the crude clamp adjustment.

The shim used was a strip of cardboard of thickness 0.052".  A light was used behind the mics while the shim was inserted.  Any gap between the shim and mics or paper was easily detected. Making the adjustment was essentially guess work and had to be repeated more than a dozen times. (rpitb)







distance from strip to mic about 0.050" (1.27mm)  Thickness of shim used 0.052" (1.32mm)

It was tried with the mics farther away but to get a decent signal, the gain had to be increased to the point where a buzz was heard.  This may have been due to a weak glue joint or possibly the voice coil touching the pole piece, or both. A longer distance between the mics and paper would diminish the effect of unequal distances of the mics to the strip.  That was considered but there would have been the possibility of errors caused by room reflections. The results below seem reasonable especially when compared to a similar test being done on the cone itself.

RE: PicoScope file 20221108-0001 17cm 3khz half cycle

distance between mic centers  17cm (0.17m)

BLK=bottom mic; red=upper mic

red is 147uS (0.000147S) behind black

speed through strip  0.17/0.000147=1156m/S

The wavy pattern is probably due to the strip not being taut like a piano string.  The zero point is at the center; the yellow square on the black trace at the zero points of the X axis and Y ordinate.  the red trace crosses the X axis just under 150uS to the left of the 0,0 intersection





FIGURE 4  Theory of Operation

The angle at A is 72o and the angle at B is 108o

This represents the cone.  When a signal is applied to the voice coil at B, two things happen.  The pressure wave moves outward towards C at 343m/sec.  It also travels along the cone at 1156m/sec. As it does, it also produces a pressure wave radiating outward in the direction of C.  By the time the wave travels through the cone and reaches A, the pressure wave at C has already traveled 23mm.  The radiated pressure wave at A will be in phase with that at C.

Since this pressure wave is radiating in a circular manner, we have a cylindrical wave front.  This is the principle behind the venerable Walsh loudspeaker.  However, compared to that, this is simple.  The cone of the Walsh loudspeaker is made in three sections of three different materials.  This complicated things as the speed of sound through each of these three media will be different.

It's unknown if the cone of the Walsh loudspeaker is a true cone with a constant angle.  To get a perfect cylindrical wave front owing to the different speeds of sound through each section would require different cone angles for each section.  However, the differences may be negligible.





The Testing


The photo on the right shows the 1.2khz second order filter.  It was needed as the CLIO doesn't allow adjusting the sweep bandwidth and this device isn't capable of handling low frequencies and wasn't designed to.  A more robust design is being considered.

Figs 5 and 6 show the response at 1w1m. It needs more work as a response within 10dB above 1.2khz is unacceptable.  THD isn't all that bad at just under 2% in that passband.

Fig 6 is a pink noise response. 








Details, Baby. Details


Figs 7 through 17 are spectrum analyses at 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 khz, resp. taken at 1w1m

These don't look as good as most tweeters, cone or dome but considering the crudeness of this design, they aren't that bad.

Another model is under consideration   photo 12 below


Comparison among a few other speakers can be seen at the following links.

Horn Loading the Faital Pro 3FE22-16   figs 8 through 19

Wharfedale Super3-Focal T90 Comparison    fig 3 at the bottom of the page

Unfortunately, I don't have any data on two ribbon tweeters here - YET.



















A very small blip, the second harmonic of 9khz can be seen at 18khz, just left of the 20khz marker



The second harmonic of thos 10khz tone is obscured by the 20khz marker but careful inspection will reveal its presence



There is no second harmonic shown here for this 11khz tone because CLIO doesn't measure harmonics beyond 20khz




The mic 1/2 meter from the open end, the front.



RED is response from the side; GRN is response from the front.

Both traces were run at 1/4 watt, 1/2 meter which is equivalent to 1w1m re SPL.  There can be and usually is a slight difference in the shape of the curves taken at different distances due to radiation pattern from the loudspeaker as well as room reflections in a non gated response.

Since measuring from the top would have required placing the speaker closer to the floor thus increasing room reflections from paraphernalia around the room on the floor as well as this response not being gated, it was decided to make the measurement at 1/2 meter with the speaker midway between floor and ceiling.  With the mic at 1/2 meter from the source, reflections are minimized due to their greater distance from the mic.

Anyway, there isn't much difference between the curves but the red one taken from the side is a little smoother.  How much of an aural difference this would make may be open to debate.  An A-B test could be performed if the speaker were mounted on something that would allow the speaker to be rotated 90 degrees to point at a listener or the ceiling during an audition.  However, the 6dB peak between 1khz and 2khz should be noticeable.

The response is anything but smooth with a large dip in the middle or two large peaks at the ends, depending on one's point of view.  While that isn't something into which will be delved with this unit, it will be given consideration with the next unit, which will be quite larger.  Photo 15




Another subtlety recklessly forgotten was the effect of a football sized ball of wool on top of the large end of the cone.  The idea was to reduce or eliminate the ceiling reflection caused by energy from the inside of the cone.  A measurement was made from the side of the cone with and without the wool.  No appreciable difference was observed.  Figs 19 & 20

Of course, such a ball of wool in front of any conventional loudspeaker will have a substantial effect on its output.  This cone with its acute angle won't act as a piston.  It relies on longitudinal waves propagating along the cone. This will cause pressure variations in the air and those will move at 343m/sec.







Two not gated responses

light green- no wool   dark green-wool



Two gated responses

black- no wool   red-wool







The Next Unit


This is the magnet from a 12" Wharfedale Super 12 CS/AL that was shattered during shipment.  The shipper fully compensated the seller and the seller fully refunded me.  I still have the aluminum voice coil and phenolic spider which will be used for this project.

The magnet assembly measures 4.75" (121mm) in diameter and is 3.25" (83mm) in height.  It weighs 18 pounds (8kg) and certainly isn't something ya wanna drop on yer toes.

The nails are filed to fit the gap.  They were used to measure the inside and outside diameters of the gap.

The voice coil is 1.755" inside diameter with a coil height of 7/16" (11mm) as well as a top plate thickness of same.  Considering the small diaphragm excursions above 400hz, this ought to behave like an underhung voice coil.  The field strength is 17000 gauss (1.7T)

A six inch long cone is being considered.  It will have a 4" diameter large end and a 1.76" diameter at the voice coil end.  This cone, actually a frustrum, will have a surface area of 97  A 12" unit has an effective (projected) piston area of about 76 The actual surface area of the cone is 86 (not including the area of the dust cap)





Back to the loudspeaker main page

Table of Contents