Thooper Dooper Whizzer

AKA Walsh type driver

 

This project was instigated by a previous project, THIS ONE

It is followed by another whizzer project using a stiffer paper, 65# cardstock.

A link to that page is at the bottom of this one.

In fact, this idea came to mind before the previous one was complete.  Photo 1 here is the same as photo 15 in the previous project.  If this crazy idea works well, I'm in trouble because I have no spare magnet.  There are two available but they're part of two working loudspeakers.  They're not hard to remove, just 4 bolts but .......

Those two 12 inchers can be seen hiding in the background of photo 11 in the above given link

There is another option.  Field coils.  This will involve having two pole pieces machined, a cylindrical or U shaped case and two plates and the purchase of enough enameled copper wire.  There is a perfectly good all original pair of Jensen F12Ns and all it takes to remove the field coils is 4 bolts and the unsoldering of two wires.  The bolts aren't the easiest to access with the cone in the basket but they are accessible.  However, if the ratchet slips, there goes the cone.

 

 

 

PHOTO 1

The 18 pound magnet assembly.

It measures 4.75" (121mm) in diameter and is 3.25" (83mm) in height.

 

PHOTO 2

This is the plate onto which will be installed the spider. The shims are to raise the coil to keep it vertically centered with the top plate.  These shims were specifically made to achieve that specification.

 

PHOTO 3

This is the coil and spider salvaged from that damaged Super 12.  The coil wire is aluminum.

 

PHOTO 4

The red shim is 0.007" and it's a tight fit. Removing it will be awkward to put it mildly as I'll have to reach into a 6 inch opening on an 8 inch cone.

 

PHOTO 5

A preliminary assembly.  The support beams are not glued or screwed - yet.  The octagonal ring on top is part of the support system as such a long and top heavy cone will be unstable.

The inside of the ring is lined with a strip of 1/8th foam.  The cone will not be glued to it.  The cone compresses the foam by about 1/3rd its thickness.

 

 

PHOTO 6

The 18 pound magnet assembly is nested in a hole in the thinner upper piece of plywood to avoid shifting.

The cone isn't glued to the coil assembly yet and there's good reason for that; the acquisition of plastic model  cement, aka model airplane glue.. Titebond will do the job but it doesn't break down readily even with acetone should disassembly be required.

 

 

 

 

PHOTO 7

The previous upper stabilizer didn't work too well centering the cone.  This temporary method works well.  the hard rubber tabs were gently placed on thwe cone and beam after glue was applied.  The voice coil is shimmed.  When the shims were removed, a 10hz tone was applied to the coil and driven to 2mm in each direction.  No rubbing.

After correcting this cone (see fig. below), a center beam fixed to the polepiece will be used with a 4 inch diameter six spoked spider on top.

 

PHOTO 8

A closer look at the rubber stabilizers

 

PHOTO 9

Nail polish was used to hold the cone to the coil. This will allow very easy disassembly. Since the system is second order high pass filtered at 500hz, this held very well.

 

PHOTO 10

I friend was curious about how much acoustical energy was radiated from the open end.  This was measured at 1/4w, 1/2m to avoid having to place the system on the floor.  1/4w 1/2m is the same as 1w1m with exception of possible differences in radiation pattern.  Comparison tests made over the past few years showed this to be of little to no concern.

The result of this is shown in fig 2, further down.

 

PHOTO 11

Side position measurement.  One thing nice about this system is that there is no high frequency beaming.  One is always on axis, horizontally.

 

PHOTO 12

Making sure the mics are in the same vertical plane.

 

PHOTO 13

Eliminating rotational skew not that this would have made much difference.

 

PHOTO 14

The two mics in position to compare the arrival of the upper and lower signals.  This was confirmation that the cone angle was in error.

 

PHOTO 15

Despite the above mentioned errors, it sounded better than expected.  There are some peaks in the response that were easily suppresses with the digital parametric eq.  The cone angle being erroneous may also contribute to response irregularities. The signal from the bottom will be 0.0498mS (49.8uS) behind the upper signal. This is 4uS lower than that which was measured with the two mics in photo 14 above.  This is pretty damn good considering the calibre of the equipment used.

Despite the 50uS differential, the human ear can't detect that.  The best it can do is detect a delay of about half a millisecond, 0.5mS, 500uS.  Check this link.  https://www.aalto.fi/en/news/the-human-ear-detects-half-a-millisecond-delay-in-sound

(See figs 3, 5 and 6 further down)

 

 

 

FIGURE 1

The dimensions here are in meters.  The reasoning is that the speed of sound in air is given as 343m/sec at 20o C.,  Most rooms are closer to 75o F, which is 24.4o C.  The speed of sound at this temperature is 345m/sec or 13,583 ips.  The 345 is easier to handle during calculations that the 13583.  Also, when making measurements, it's a lot easier to interpolate a third of a millimeter than a third of a sixteenth of an inch.  Anyway, to convert the metrics to inches, simply multiply by 39.37.

OK.  The red figures in this drawing refer to corrections made after making the cone.  During testing after a listening test, it was noticed that the cone angle was in error.  The sound from the top radiated into the room before the sound radiated from the bottom.  This was due to an incorrect assumption.  This cone is made with construction paper which is thicker than 20 pound bond.  The construction paper was used to accommodate the larger cone and that it resembled the typical paper used for midrange units. I did have intentions of making the speed measurement but somehow I was distracted.

The larger cone in the drawing is the original, the smaller is the corrected model.  The speed of sound in that paper was assumed to be 1141m/sec but was later found to be 1806m/sec.  This allowed the longitudinal wave from 1 to arrive at 3 before the airborne wave reached P6.  It was clear that the distance between 1 and P6 had to be shortened.

The time between 1 and 3 has to be equal to the time from 1 to P6 in order for this device to radiate in a cylindrical manner. Let this time be t.  Let the distance 1,3 be d1 and the distance 1,P6 be d2. The speed of sound in air is 345m/sec and in the paper it's 1806m/sec.  Since time, t, is distance divided by rate, we have d1/345 = d2/1806.(eq.1)  This moved P1 to P5.  By subtracting this P1P5 distance from the upper radius, we get P4.  This is the new radius of the top of the cone, the diameter of which is now 0.104m. Now, the device will radiate closer to a pulsating cylinder.

There is some error here.  First, the determination of the speed of sound through the paper and the angle change at P3.  The above relationship in eq1 will lengthen the height of the cone because the length of the side (P3P4 will have to remain the same as P3P2.  Maintaining the same cone height will shorten the side by 0.006m, from 0.178m to 0.172m.  The difference in time that it would take a wave through the paper to traverse 0.006m is 3.33uS.  Our ears cannot detect that difference.  In fact, the ear cannot detect a sound arrival time of less that half a millisecond, 0.0005 sec. or 500uS and that's with percussive sounds..  This was measured at the following link.  Aalto University

The article is short and well worth reading.

A larger drawing is available by clicking on this one

Oh, one note worthy of mention.  If the above confuses you, don't feel bad as it has the same effect on me and I wrote the damn thing.

 

 

 

 

FIGURE 2

RE: photo 10.  Response at 1/4w 1/2m.  RED-open end   BLACK-side.

The overall SPL is much the same despite the different curve shapes.  This may be due mostly to the different mic position.  A better result may have been obtained if the system were rotated 90 degrees but it assembly prohibited that.

 

 

 

FIGURE 3

RE: photo 15.  These are the two signals picked up by the two mics.  The blue is from the lower mic  The peaks are used to get the difference.  The red peak is at 93uS and the blue one at 147uS.  The signal from the lower mic is 54uS behind the upper signal. The calculation from photo 15 gives 50uS.  Most of that error is due to finding EXACTLY the high point of the waveform.  Unfortunately, this PicoScope doesn't have the feature of moving a cursor on the curve to find the high point.  Of several attempts with enlarging and narrowing the trace using Pico's software, 4uS appeared most.

The corrected cone should, ideally have these two traces superimposed. An error of less than 10uS will be acceptable.

Usually, the points at which the traces cross a horizontal axis is used to measure time but the slopes of some traces differ at different frequencies, especially if they are far apart in frequency, like 500hz and 5000hz.  It is uncertain as to which is the more accurate but I look for consistency.  For example, the down slopes of these two marked traces are different.  Similar differences have been seen in the initial rising slope.  This  may be due to frequency transmission through the cone material.  This phenomenon has been noticed when the half wavelength of the frequency is greater than the distance between the mics.  Half wavelength was used because the pulse tone is a half wavelength, no negative half.  The negative cycles here are due to the cone's continuing to oscillate after the signal stops.  Such behaviour is measured by impulse tests.  Ideally, the cone would stop when the signal stops. If this were the case, a guitar string would stop when it reaches its rest point after being plucked.  In short, the guitar wouldn't work very well, the string is infinitely damped.

Since it works, Mother Nature is a damn good engineer.

 

 

 

FIGURE 4

Impedance curve of the cone speaker, nominal at 14W  

 

 

 

Living Room Responses

FIGURE 5

Curiosity won here despite that any difference was considered to be moot.  Black is normal polarity and red is reversed polarity.

These responses were measured at 1 watt from a distance of 8 feet and a height of 64 inches, ear level when standing.

Usually there's a substantial peak(s) and or dip(s) around the crossover point, the lack thereof being attributed to the upper register above 500hz radiating in a cylindrical manner.

It should be noted that in a non symmetrical room, which is usually the case, the combined response of two speakers each with a flat response will be anything but flat. 

 

 

 

FIGURE 6

The red curve is without equalization; green is with eq; grey is THD.  In both cases, the L-pad was full clockwise (FCW)  normal polarity.

THD between 500hz and 6khz varies between 1% and 3%

The response between 300 and 12khz is within 7dB.  It could have been made much smoother but that seemed pointless.

 

 

 

PHOTO 16 & 17

The EQ screen with description.  Note that the bands are numbered from right to left.

This is because I started at the high end of the audio spectrum where most of the equalization was desperately needed.

The other three lower EQ bands were added just to flatten the response; they didn't have as much effect on the overall sound.

 

 

 

The second thooper dooper whizzer

Back to the loudspeaker main page

Table of Contents