1983

I found this described in a book, but I had to do “reverse engineering” on a real one (take it apart, trace the circuit and measure the components) to satisfy myself. The heart of it is a triac, which is two silicon controlled rectifiers (SCR’s) in parallel but with polarities opposite, in a single chip (Fig. 1a and 1b). An SCR is triggered into its ON (anode to cathode conducting) state when a small positive voltage is applied to the gate. Once ON, the gate has no further control; the SCR will continue ON until the anode-cathode current is interrupted or reduced nearly to zero. In the dimmer a triac is used instead of an SCR, because current has to be controlled in both directions. To reduce the average current, the triac is triggered to the conducting state late in each half cycle of the 60-Hz ac. The lagging gate voltage is obtained from a resistance-capacitance phase shifter, R and C₂, as shown in Fig. 1c. After a given SCR of the triac is triggered, late, it conducts only until the line voltage next crosses zero, so current flows through the lamp for only part of that half cycle. The phase lag of the gate can be varied (by the resistance) from zero to nearly a full half cycle, so the light can be dimmed almost to zero. The other components, L and C_1; serve to prevent electrical “hash” from going back over the power line. The triac goes into conduction suddenly, so L and C₁ are needed to “round off” the sharp current changes. The phase shifter in the specimen I took apart had R = 0 − 500 000 Ω, and C₂ = 0.05 μF.

The triac dimmer is remarkable. My specimen is the smallest size, yet it is rated at 600 W, and it dissipates only one watt per ampere. Contrast this with dimming a light by connecting a resistance in series. There you would be running an electric heater as well as a light! Or think of a variable transformer such as a variac. For 600 W, that would weigh 10 lb or more and cost a small fortune. Triac controls are used for fans, drills, etc., but one should not use a light dimmer for those purposes. Such loads are inductive, and they require a somewhat different control circuit.

Interesting questions arise. Does the electric meter charge honestly for a dimmed light, when the current is far different from a sine wave? What if all lights and other loads in a city were reduced by triacs? The load on the generating station would be absent for the first part of every half cycle – which surely would make a problem.

As a kid you probably sailed flat stones and tin can lids. They didn’t stay horizontal; they slowly turned toward, or into, the vertical plane and fell to earth more or less edgewise. The frisbee stays horizontal, and that is an important reason it is such an ingenious invention. Let’s see where the difference lies. Because the tin can lid (or other disk) is spinning, we suspect right away that the turning is gyroscopic precession. But for that to happen there must be a torque. So we look at the two forces acting on the disk, those of gravity and of the air, to see if together they give a torque. Imagine we are riding with the disk and can see the air flowing by, above it and below it (Fig. 1a). The disk is shown inclined just a little to its direction of motion, so it will have lift. It will have the character of a flat “wing,” which is well understood. Most of the deflection of the air occurs ahead of the center, so the resulting upward force acts on a line that passes ahead of the center of the disk (and center of gravity) as indicated by the arrow. That force and the gravity force together give a torque around an axis in the disk that runs in and out of the paper, as we look at the figure. That is all that is needed to make the flying disk act like a precessing gyroscope. The relationship among the axes of spin, y, the applied torque, z and the precession, x, true for any gyroscope, are shown in Fig. 1b. As related to our flying disk, x is also the approximate direction of flight. So as seen by the thrower the disk turns slowly (Fig. 1c), counterclockwise if the spin is in the sense shown.

Consider now the frisbee. The one thing certain is that torque is absent, for there is no precession. The net force of the air must act on a line very nearly through the center of gravity. And that condition evidently holds over a large range of velocity, for even as the frisbee slows almost to a stop, precession does not seem to appear. That has been accomplished through the special contour of the frisbee. It is an “airfoil,” and as such is in the realm of empirical design, best done in a wind tunnel. With only a little imagination, the air flow around the frisbee and the net forces can be sketched, as in Fig. 1d.

Besides staying flat, the frisbee sails remarkably far, and that suggests that the airfoil shape gives it a very favorable lift-to-drag ratio (a term that is self-explanatory, the drag being the backward force that slows the frisbee down). We asked an aeronautical engineer about the foregoing. He answered only by telling how to make some tests! We should find the “glide angle.” Sail a frisbee off the top of a building at such a downward angle that the velocity stays constant. Then the path is a straight line and there is no acceleration, so the force analysis is simple. Doing the comparison test with a flat disk might not be so easy, particularly because of precession. But the questions might interest students on a fine summer day, since they will be throwing frisbees anyway. We would like to hear of any results.

Related devices come to mind. The “clay pidgeons” used in trap shooting are shaped much like the frisbee. They come out of the mechanical launcher spinning, convex side up, and fly straight and far. Is there more than a chance relation? Consider the discus, surely designed without much help from physicists. Because of the large mass and angular momentum of the flying discus the gyro effects and the lift and drag due to the air are not expected to have much effect on the trajectory. Nevertheless, no effect is too small to win or lose in the competition. If allowed, could our knowledge about frisbees lead to greater discus records? Another example is the boomerang. There precession plays an important (and somewhat complicated) role in making the object return to the thrower. The physics and aerodynamics of the boomerang are now quite well understood.¹  That they can even be thrown successfully by persons other than tribesmen is suggested by the fact that they are a commercial item in this country.²  The effect of the frisbee on concepts of UFO’s is left to the reader!

Physicists often while away the time of an elevator trip speculating as to what is behind the little square that lights up when it is touched. But the trip is over before the answer is found. The part that is touched is solid: it is not a mechanical push-button. Does it depend on conduction through the finger? Or is it capacitive coupling? It works through a leather glove, but not a mitten – unless you have come in out of the rain. Only one guess seems safe: the color indicates a neon discharge.

Having heard that many years ago Peter Franken¹ had taken the direct approach and disassembled a live touch panel, I tried the same. But everything I did called the elevator – sometimes with people. So I quit. Rescue came when Dave Shalda, the University’s expert on elevators, supplied a circuit diagram and a spare “glow tube” with which to experiment.

The glow tube proved to be just a modified version of a neon lamp. The characteristics of it that are essential for touch operation are present also in the garden variety neon lamp, and that makes it easy to study and demonstrate touch operation with materials readily at hand. I connected up the circuit of Fig. 1a, using a variable dc power supply. The lamp was an NE45, Fig. 1b. The resistor shown in the diagram is the one that is inside the metal base of the lamp. (Most neon lamps with screw or bayonet bases have built-in resistors.) As the voltage of the power supply was turned up, the discharge did not “strike” until about 80 V was reached. Once started, it did not go out until the voltage was turned back to about 60. Thus in the interval between 60 and 80 V, the discharge is capable of continuing if started, but it will not start by itself, and that is the first requirement for making a touch switch possible.

The second, and only other, requirement is that there be a simple way to trigger the discharge, when the voltage is set within the interval 60 to 80 V. This happened very reliably with the NE45 when the glass bulb was touched, the area of greatest sensitivity being that nearest the anode. (The anode is the piece that stays dark; the cathode the one that becomes sheathed in the orange glow.) I tried a few other things. With ac instead of dc voltage the discharge could be triggered, but could be kept going only by holding the finger on the glass. That’s because it goes out and has to be restarted every time the voltage passes through zero, 120 times a second. The familiar “wheat seed” neon lamp, Fig. 1c, could be triggered, but not as sensitively as the NE45. An external resistance of 20 000 Ω was used. Caution: The above makes a good demonstration, but fingers will wander! Make sure the high-voltage metal parts and wires are taped.

Moving from here to controlling an elevator is engineering. The current is routed through the coil of a relay, which starts the chain of events that calls or stops the elevator. Switches activated by the elevator interrupt the circuit to extinguish the discharge when the elevator arrives. The special design of the gas tube increases the sensitivity, the light output, and the interval between the spontaneous striking and extinguishing voltages. The specimen at hand is sketched in Fig. 1d. The anode, l, is a wire that extends to within a millimeter or two of the inner surface of the glass. It is shielded, except at the tip, by a glass tube, m. The cathode, n, over which the orange glow resides, has a large area, annular in shape. On the outside of the glass there is a spot of a transparent conducting film, o. (There is a third electrode, a grid, consisting of a ring of wire just above the cathode, not used in systems having only one elevator, and not shown here.) In operation the supply voltage is 135, which is midway between the spontaneous striking and extinguishing voltages. The external resistance is such that the current is about 30 mA; enough to give a bright glow.

Of course the elevator rider does not touch the gas tube itself. The inner square, p, in Fig. 1e is of conducting material, and is connected to the conducting patch on the outside of the gas tube by a coiled wire. The surrounding frame, q, is translucent plastic, through which the orange light comes.

Unanswered questions remain. How does touching the outside disturb the electric field inside enough to start the discharge? Is it the capacitance-to-ground of the body? The static electricity we always carry around? Or the 60-Hz ac we pick up by being antennas? If an elevator is handy, experiment. Touch the panel with a fine wire that is held in the hand, or grounded, or connected through a small capacitor – or a resistor. Stick thin plastic on the touch area. But don’t leave it there!

Non-mechanical phonograph or video disk pickup. No record wear.

Cordless telephone extensions.

For most of us the mention of liquid crystals calls to mind the display on a watch. But that association is recent. Liquid crystals have fascinated experimenters for nearly a century. There are several classes of them, each class having its distinct kind of ordered structure, and each showing a variety of types of behavior.^1-4  One of the classes of liquid crystal, in a special kind of display “sandwich” is commonly used in watches and calculators, and may soon become familiar in automobile instrument panels. We have a limited objective here: to describe the workings of the sandwich that is used in the above applications. Most of what will follow was communicated by Dr. George W. Smith of the General Motors Research Laboratories, where, as one might guess, the practical interest in liquid crystal displays (LCD’s) points toward the instrumentation of cars. The Corvette has already been so converted.

In simple terms, the LCD works on light-polarization effects, at the heart of which is the remarkable ability of a liquid crystal to rotate the plane of polarization of light on command by a small electric voltage. A layer of the liquid, sandwiched between pieces of the familiar Polaroid film as polarizer and analyzer, forms a controllable light valve, to stop or pass light, and therefore to make a given area appear light or dark.

We follow Dr. Smith’s sketch of the sandwich, Fig. 1a. The liquid crystal is confined between glass plates G. Next are Polaroid films P and A, in crossed orientation. Finally, a reflecting surface R. Transparent, electrically conducting films, E are applied on the glass plates. Postponing for the moment the matter of forming numbers or letters, how does this configuration turn the light off and on? In the liquid crystal the long thin molecules prefer to align themselves more or less parallel to one another. At the boundary, which is the coated glass plate, they align themselves with any “grain” that exists in the surface. That grain is provided simply by rubbing the surface, before the sandwich is assembled. The rubbing makes micro-grooves, and they are sufficient to cause the molecules to align parallel to them at the boundary. The directions of rubbing of the two facing surfaces are at right angles. To conform to the rub-directions at the two boundaries, and also be more or less parallel to one another throughout, the direction of the molecules executes a gradual 90° twist, in the distance between the plates (Fig. 1a). Now comes the even more surprising fact: the plane of polarization of the light follows the twist, turning 90° in the short distance of a hundredth of a millimeter, which is typical of the separation of the glass plates. The light therefore can traverse the crossed Polaroid films and emerge (L). The area looks light.

The molecules of the liquid crystal are active not only optically, but electrically. The application of an electric field by a small voltage between the plates orients them parallel to the field, perpendicular to the plates (Fig. 1b). The twist is gone, so the light is stopped at the second Polaroid film. The area looks dark. The field strength required is considerable: while only a single-cell battery is used (in a watch) the small separation of the plates results in a field of over 1000 V per cm.

All that remains is to see how the numbers and letters are made. The transparent conducting film on one plate is etched to form 7-segment characters, with a connection out from each segment to the solid-state chip (Fig. 1c). The opposing plate has only a rectangle, which covers the segments, but not the connections.

There are variations on the configuration described. If the Polaroid films have parallel, not crossed, pass directions, the characters will be light on dark background. Crossed Polaroids give dark characters on light background. The reflector can have a color filter on it. For viewing in darkness as well as in daylight, the rear coating can be made half-reflecting, with a light source (white or colored) located behind it. When applied to a driving or flying machine, the LCD will be backed up by a microprocessor chip, so that calculated information, like estimated time of arrival, as well as primary data like engine temperature, will be displayed. A given LCD can give various messages, in automatic rotation or at the call of the operator; the limit will lie only in the variety of sensors that can be planted in the machine to feed data in. (The Corvette panel reads out 14 items, in English or metric units, in 9 display areas, in numbers and graphs, in color.) Who will need the car radio any more for entertainment?

In conducting a column of this sort, the temptation is to report on the amazing gadgets invented in the last decade or so, mostly based on integrated circuit chips. For a change of pace I would like to discourse on a clever physics device that comes from millions of years ago – that which the housefly relies upon to maintain equilibrium in flight. A larger scale example of it came to the attention of M. L. Foucault in the mid-1800s, when he was making parts for a clock that was to be used for guiding a telescope. He was working on a thin steel rod, held in the chuck of a lathe. He noticed that if the rod was “twanged,” set into vibration in a plane, and the chuck was then rotated, the plane of vibration did not follow the rotation of the chuck, but remained fixed in the laboratory coordinates. That set him to thinking along lines that led him to the invention of the Foucault pendulum. (He says so in a footnote to his first article on the pendulum.¹ ) From there he followed the same line of thinking and invented the gyroscope. But back to the housefly. The fly uses, for continuously sensing changes in his (excuse me for omitting his/her) orientation in space, a version of the twanged rod.

Flies belong to the order Diptera, the term signifying that they have only one pair of wings. In place of the hind pair of wings that other flying insects have, they have a pair of halteres – stiff stalks, loaded at their ends by knobs, as shown in Fig. 1a. The stalks are hinged at their bases, and during flight they oscillate in arcs of about 50°, at a frequency of several hundred hertz. The hinge is such that the motion is constrained to a plane that is fixed with respect to the fly. (The plane of motion, the arc, and other features are similar enough to those of the wings to suggest strongly that the halteres evolved from wings.) When the fly changes orientation during flight, the halteres, like Foucault’s twanged rod, would “like” to maintain their planes of vibration in the laboratory frame. But because of the constraint of the hinge, the planes must change with the fly, and that results in torques at the hinges. The torques are sensed by nerves, and those signals are used by the fly to execute maneuvers, or to correct orientation if he is trying to fly straight. How do we know all of this?

As early as the beginning 1700s experiments were reported in which it was found that if the halteres were cut off, the fly became disoriented in flight. In the same era it was found that if, on a fly without halteres, a thread was attached to the back end of the abdomen, like a kite tail, the fly could fly straight – interesting, but explaining little about the halteres! An understanding of the physics of the problem did not come until well into the present century – more than 200 years later. (It seems unlikely that Foucault knew of the problem, for if it had come to his attention after his observations on the twanged rod in about 1850, he probably would have seen the solution.) In the present century many hypotheses were put forward as to the action of the halteres, leading nowhere, and not until about 1938 was the dynamic (called by workers in the field “gyroscopic”) action hit upon. In 1948 a monumental piece of work was published by J. W. S. Pringle,²  which seems to have answered most of the questions.

I will recount a few of the highlights of what Pringle found, but reading the original (37 pp) article is recommended. A fly was fastened to a platform that could be rotated, and electrodes were placed in the various nerves at the base of the halteres. While the fly’s wings and halteres were buzzing and the platform was rotating, the torques on the halteres caused certain nerves to “fire” during each stroke so that the signal was a series of pulses at the vibration frequency (or twice that) rather than an average of the torque. The most effective signals were produced by yaw (rotation about the fly’s vertical axis). That finding was supported by a rapid series of flash photos of a fly whose halteres had been removed. The disorientation was seen to be mainly in yaw. I have attempted to sketch the result, from the figure in Pringle’s paper (Fig. 1b). Pringle says that (as of 1948) he could not find a mathematical analysis of the dynamics (he calls it gyroscopic effect) of an oscillating rod (and presumably not of the twanged rod either), so he presents the analysis in his paper. That, taking into account the known planes of vibration of the halteres, shows that yaw is sensed more strongly than either of the other rotations, pitch and roll. Pringle proposes a mechanical model made of a rod, hinge, springs, etc. One of the features in it (which he found in the fly) is that the hinge is elastic, so the motion is at the resonant frequency with a moderately high Q, kept going by a muscle pulse at one (not both) end of each stroke. Pringle remarks that, with one or two dubious exceptions, there are no other examples in the animal kingdom in which the gyroscopic effect is used.

The experimenter may feel the urge to try something. Flies are still expendable.

David G. Stork of Swarthmore College writes to me “as one bird to another” to explain a gyroscopic effect in frisbee sailing that I did not mention in my recent column,¹  and which in fact I had not seen. It is the maneuver of skipping. As he describes it, if the frisbee is thrown at just the right glancing angle toward a smooth horizontal surface (such as blacktop), it will touch briefly, skip up, and continue in flight. He says it is quite easy to skip a frisbee in that way – the trick lies in insuring that the proper edge of the frisbee touches the blacktop, and that is where a knowledge of gyroscopic motion comes in.

We recall that in the earlier column we identified the three principal axes as the line of flight, the axis of spin, and the transverse axis which is horizontal and perpendicular to the line of flight. At that time we were concerned with precession about the line of flight, as caused by a torque about the transverse axis. But for the skip, we need precession about the transverse axis, so the frisbee’s leading edge will tilt up and cause it to “take off” and fly again. That requires a torque about the line of flight, and in the correct sense, so the nose will go up, not down. David Stork tells how to accomplish that. If you are right handed and throw in the standard way, the spin will be clockwise (viewed from above). You must throw it a little tilted, so that the left side (as you view it) will make contact with the blacktop. Then the clockwise torque around the direction of flight, from the force of the blacktop will give precession that will tilt the leading edge up. Stork shows this with a diagram of angular momentum vectors, which you also can do easily. Next time I see an expert sailing a frisbee, I’ll ask for a demonstration.

This started when Albert Bartlett of the University of Colorado, and former president of our AAPT, agreed to look into the “bulb savers” (devices for extending the life by lowering the filament temperature) that are now widely advertised and available in hardware stores. He enlisted the help of Gary Geissinger of the Aerospace Systems Division of the Ball Corporation, who made measurements on one kind of saver. That got me started measuring, on a saver of a different kind.

Why the interest in savers anyway? Simply that there are many applications in which greater light-bulb life, at the sacrifice of some efficiency in converting electric power into light, is a good trade-off. Reasons might be high labor cost or nuisance of replacement (ceiling lights in an auditorium), or the involvement of personal or public safety (warning or exit lights). Closer to home, I would gladly waste a little power in the tail light of my car if it would never have to be replaced.

In ordinary use the main cause of demise of a bulb is evaporation of the filament, and that is a very steep function of the temperature. Bulbs manufactured to run a little cooler than normal are available (e.g., from Sears); typically about a 50% increase in life is claimed. The saver devices on the market go for much greater – in one case spectacular – extensions of life.

What we will need to know about tungsten filaments is readily found in handbooks, conveniently given in terms of voltage above or below normal, rather than temperature. The solid parts of curves A and B in Fig. 1 are reproduced from one such source.¹  Curve C, the life factor, is plotted from an empirical formula from the same handbook. For voltages near normal, the rule of thumb is that each 5% change doubles or halves the life.² 

The saver examined by Bartlett and Geissinger was a wafer to be placed in the socket under the bulb base. It turned out to contain a thermistor, introduced in series in the circuit. The reason for using a thermistor instead of a resistor was puzzling at first, but the cleverness of it later appeared. The problem is to lower the voltage by about the same number of volts for a high or a low wattage bulb. A thermistor can approximate that. It is a semiconductor,³  whose resistance goes down as its temperature rises. It heats internally when it carries a current, as anything having resistance does. Higher current (say when a higher wattage bulb is used) causes the equilibrium temperature to be higher, the resistance to be lower, and therefore (ideally) the voltage drop to be constant. In a test with bulbs of several wattages the voltage on the bulb was lowered by the amounts shown in Table I, column A. (It must be noted that the equilibrium temperature of the thermistor will vary with the type of socket, the ventilation, etc., so the data are only approximate.) For comparison, column B shows results that would be found by replacing the thermistor by a resistor. Clearly one resistor would not serve for all the bulbs, while the thermistor does so fairly well. The results for the thermistor indicate less life extension than the factor 4 claimed by the seller, but within the ball park.

I was able to examine three brands of quite a different kind of saver, a kind that employs a simple diode – a half-wave rectifier – to reduce the power delivered to the filament and therefore to reduce its temperature. Two were external wafers; in the third case the diode was concealed inside the bulb base. We expect the effect to be the following: For any resistive load the power delivered is the time average of V²/R. Half-wave therefore will heat the filament the same as would full-wave ac if the nominal (or peak) voltage were reduced by the factor $1/2$ or 0.71. Curves A and B from the handbook do not go down that far, but they were easily extended. A light bulb, a variable voltage transformer, a photographic light meter and a meter stick did the trick. The reading on the light meter was made always the same, and the inverse square law was used to find the light output relative to that at normal voltage, which was taken as 120 V. The small circles and the dashed lines are the results.⁴ 

All the diode-type savers came out essentially the same. (The base of the third specimen was removed, to test it with and without the diode.) The light output was about 25%, as compared to operation of the same bulb without a saver, in good agreement with the curve at its 71% voltage point. The power consumed was measured to be 60%,⁵  giving a relative efficiency of 40%. The claim of a factor of 100 in bulb life is not extravagant; it is consistent with curve C.

I was curious as to whether a half-wave load would confuse the watt-hour meter. To test that I turned off all other loads in the house, and ran two 250-W bulbs, first with the diodes in the same direction and second, with one reversed. The total load was the same; but in the first case half wave, and in the second case full wave. To my surprise, the meter charged the same to within about 2%, which was my accuracy. Smart meter!

The thermistor lowers the temperature enough to extend the life by a factor of 2 to 4, without much reduction in light output. That might be a good trade-off, for certain applications. There may be a dividend from the fact that the temperature comes up to equilibrium slowly (over a few seconds), which reduces expansion stresses and hot spots. For ordinary lighting there is a question as to whether the thermistor saves money. Besides lowering the efficiency, the wafer costs as much as about five 100-W replacement bulbs. The diode saver is a clear case of overkill. To obtain the same light as from a 100-W bulb in normal operation, a 400-W bulb would have to be substituted, which would consume 240 W.⁵  Thus to avoid having to replace a 100-W bulb at the end of 1000 hours, one would run up a bill in wasted power sufficient to buy a basket full of replacement bulbs. On the other hand, if you are using four times as much light as you need, you can install diode wafers and save both power and bulbs. Or you could just buy smaller bulbs!