From Scientific American, December 1964, pages 80-88
Scanned, then OCR'ed with Tesseract, and images edited with Gimp on Puppy Linux.
Miriam, 2009-05-15.

The Cover
The photograph on the cover shows a fluid switching device called a "nor-or gate," which is used as a logic element in an all-fluid digital computer (see article "Fluid Control Devices"). The device consists of a solid block of glass in which shallow channels have been chemically etched to allow the passage of a fluid. (Ordinarily the working fluid is air, but in this case colored water has been used for the sake of clarity.) The channels were etched by a special optical fabrication technique, developed by the Corning Glass Company. The empty channels are wet and appear gray. The blue water at top is the main power stream. The narrower channels at left are the control jets: either one "or" the other, on being activated, can switch the power stream from the left to the right output channel at bottom. If neither one jet "nor" the other is activated, the asymmetrical construction of the interaction region, in addition to the atmospheric pressure exerted through the open channel to the right of the power-stream nozzle, will switch the power stream to the left output channel. In the photograph the control jet at top left, which contains yellow water, is on, causing the power stream to turn green and to be deflected into the right output channel. Circular holes are air vents.


The Illustrations
Cover photograph: William Vandivert
Photo of divide-by-10 circuit: William Vandivert
Diagrams: James Egleson
Schlieren photograph: Harry Diamond Laboratories, U.S. Army



FLUID CONTROL DEVICES

They perform the operations of amplification and switching mechanically rather than electrically. For some purposes they are more reliable than their electronic counterparts

by Stanley W. Angrist


FLUID CIRCUIT performs the operation of dividing by 10 in an all-fluid digital computer: for every 10 input pulses circuit delivers one output pulse. Input pulses enter from above the plane of circuit through circular, bumplike hole attached to straight channel running from top to bottom just to right of center. The 10 identical logic elements, or modules, are arranged in a series of five pairs, three at left and two at right. Each pair contains two steady input streams (small sausage shapes), two outputs (small circular shapes attached to short straight channels), eight control jets (small teardrop shapes) and eight open vents (large circular shapes). Shallow channels were chemically etched in Dycril plastic by a special process, called Optiform, developed by Bowles Engineering Corporation. For this photograph the channels were filled with white paint to increase the contrast. Normally the working fluid is air.

In modern usage the word "amplification" ordinarily suggests "electronics." It is on electronic devices that we mainly rely for all sorts of amplification jobs, from the boosting of radio signals to the control of the most massive machinery. The classic example of the principle employed is the simple three-element electron tube, or triode. Any weak, fluctuating voltage applied to the grid of the tube acts on a stream of electrons flowing through the tube, and the resulting modulation of the stream serves to amplify the signal. The tube can do more than merely amplify: a voltage on the grid can also shut off the electron stream or turn it on. On this simple principle an entire technology has been built, including the electronic computer.

Suppose that instead of a stream of electrons one used a stream of liquid or gas and tried to perform these operations mechanically rather than electrically. In place of the grid, precisely controlled jets of fluid directed at the main stream from the sides would modulate or deflect the stream. Thus one might achieve amplification or switch the stream "on" or "off." The idea occurred to engineers many years ago, and several patents on such devices were issued in the 1930's and 1940's. No one showed much interest in exploiting the method, however, because it is so much slower than electronic amplification that its possible uses seemed very limited.

Four years ago the concept was revived in a serious way. B. M. Horton, R. E. Bowles and R. W. Warren of the U.S. Army's Harry Diamond Laboratories and J. R. Greenwood at the Massachusetts Institute of Technology independently made separate studies of techniques of fluid amplification, and they found them promising for certain important applications, including computation and operations in logic. Fluid amplifiers offer some distinct advantages over electronic devices. They are simple in construction, have virtually no moving parts, can be powered by almost any source of energy to pump the fluid, are immune to damage by heat or ionizing radiation (which electronic equipment is not) and would be comparatively inexpensive if mass-produced.

The reports by Horton, Bowles, Warren and Greenwood set off an almost explosive surge of new research on fluid amplification and control. More than a score of large companies and research laboratories are now exploring its development and uses. Some believe that the fluid amplifier will be a far-reaching innovation in technology. It is hardly likely to match the spectacular rise of the transistor, which within 15 years has grown from a laboratory curiosity to a multimillion-dollar industry, but in the long run it may prove to be just as important.


FLUID AMPLIFIER consists of a solid block of material in which shallow channels have been cut to allow the passage of a fluid. If the high-energy power stream is not disturbed and hits the splitter head on, the stream will be divided in two, half of the fluid passing into one outlet and half into the other. If either of the two control jets is turned on, however (left and right), it will deflect the power stream in the direction of the opposite outlet channel and all of the stream will pass out through that outlet. The heart-shaped chamber in front of the power-stream nozzle is designed to prevent the stream from clinging to one of the channel walls.


TWO-STAGE FLUID AMPLIFIER can be constructed simply by using the output stream of one stage as the control stream for the next. By making each successive power injection larger than the one before, one can build up almost any degree of amplification.


COANDA EFFECT, named after its discoverer, Henri Coanda, is the basis for a wide range of fluid switching devices. At left a high-speed stream of air is injected into a wide container of air. As the stream moves along it entrains air from both sides and becomes broader, carrying more air toward the open end of the container. This process causes the pressure to drop in the zones between the stream and the walls, creating an unstable situation. Any disturbance or change in the shape of the container (right) will cause the equalizing return flow of air to push the stream toward one wall, where it will remain locked as long as the stream keeps flowing.


Two different techniques of fluid control are being investigated -- one quite simple, the other more complex and more versatile. Let us begin with the elementary system.

Consider a solid block (of steel, plastic or any other suitable material) in which shallow channels have been cut for passage of the fluid [see illustration, FLUID AMPLIFIER, above]. A high-energy stream of fluid, called the power stream, is pumped in through the inlet at one end. The stream passes across a widened chamber (designed, for reasons we shall consider later, to prevent the stream from clinging to one of the channel walls) and arrives at a fork consisting of two outgoing channels separated by a pointed structure called the splitter. If the power stream has not been disturbed and hits the splitter head on, the stream will be divided in two, half of the fluid passing into one outlet channel and half into the other.

As the power stream enters the system, however, it runs the gamut of two control jets, one on each side. Let us say that one of the jets is turned on and this control stream hits the main stream with a certain momentum. It will deflect the power stream by a certain amount, and the stream will then have a net momentum that is the sum of the original momenta of the power stream and the control stream, taking direction into account, as the geometric diagram below shows. Given just the right amount of nudge by the control stream, the entire main stream will be bent in the direction of one of the outlet channels and will pass out through that output.

Obviously the output represents an amplification of the energy applied by the control jet, and the gain is equal to the ratio between the power stream's momentum and the momentum of the control jet. The degree of the deflection of the power jet is proportional to the momentum of the control jet. The system is therefore called "proportional control." The control jet's "signal" can be measured in terms of either momentum or pressure, depending on the way the apparatus is set up; if the control nozzle is placed close to the power stream, the main factor deflecting the stream will be pressure rather than momentum.

The elementary setup I have described is, of course, only a single-stage affair. It is easy to see that it can be expanded into a multistage system simply by using the output stream of one stage as the control stream for the next [see illustration, TWO-STAGE FLUID AMPLIFIER, above]. By making each successive power injection larger than the one before, one can build up almost any degree of gain, or amplification, one wishes. That is to say, the small nudge, or signal, applied by a control jet in the first stage can be amplified by successive steps to control a very large power stream indeed in the final stage.


This, then, is the basic outline of the first of the two fluid-control techniques under study. As an instrument of proportional control it performs amplification in the usual sense of the word, and it promises a number of interesting applications, including its use as an element in an analogue computer. The other technique of fluid control does not produce proportional amplification; rather, it corresponds to a triode that turns a flow of electrons on or off. It is essentially like an element in a digital computer.

In this system the fluid power stream, left to itself, locks onto one wall of the channel in which it is flowing, and as a result the stream exits through the outlet on that side. An injection of fluid from the control jet on the same side will cause the stream to swing over to the other side and lock onto the wall there, so that it then flows out of the other outlet. In either case the stream maintains a stable position, flowing only to a particular outlet unless it is switched by a control jet. The reasons for this are inherent in the mechanics of the situation.

Consider a high-speed stream of air injected through a nozzle into a wide container of air. (I call the fluid air to make the example specific; the results to be described would apply to any gas or liquid.) As the stream moves along it will entrain air from both sides and become broader, carrying more air toward the open end of the container [see bottom illustration, COANDA EFFECT, above]. The stream's pickup of air along its sides, however, causes the pressure to drop in the zones between the stream and the walls of the container. The resulting pressure difference creates an unstable situation: the higher pressure of the surroundings ("ambient pressure") will push air back into the low-pressure zones on both sides of the stream to equalize the pressure.

Suppose now there is some disturbance or asymmetry (say in the shape of the container) that causes the equalizing return flow to push the stream toward one wall. As the zone between the stream and that wall narrows, there is less room for the admittance of counterflow to replace the air being entrained by the stream on that side and therefore the comparative pressure in the zone drops further. Very quickly the stream moves over against the wall. It stays locked onto that wall as long as the stream keeps flowing, because on the wall side a region of low pressure persists near the nozzle, whereas on the opposite side of the stream the ambient pressure pushes the stream toward that region. The phenomenon is known as the Coanda effect -- so designated because it was first studied by a Romanian engineer named Henri Coanda in the 1930's.


PROPORTIONAL CONTROL, the principle on which the fluid amplifiers at the top and middle of the opposite page are based, is illustrated geometrically. The vector representing the momentum of the control jet (light colored arrows) is added to the vector of the power stream's momentum (solid colored arrows) to obtain the momentum of the output stream (black arrows). Thus the gain, or amplification, produced by the fluid system is equal to the ratio between the power stream's momentum and the momentum of the control jet. The degree of deflection of the power jet is proportional to the momentum of the control jet.


SCHLIEREN PHOTOGRAPH, made by an optical technique that detects density gradients in a gas flow, demonstrates Coanda effect. A high-energy stream of air is locked onto the chamber wall at right after being subjected to a short blast of air from control jet at left.

Obviously the stream can be moved away from the wall in either of two ways: by increasing the pressure on the stream from the wall side or by reducing the pressure on the opposite side. This can easily be done by means of control jets on the sides. If the stream is locked onto the right wall, injection of air from the right jet or removal of air through the left jet will cause the stream to swing away from the right wall as soon as the pressure on the right side becomes higher than that on the left. Once the stream has crossed the center line of the channel it will go on to attach itself to the left wall for the same reasons that it originally locked onto the right one. The control jet can then be shut off, the stream will stay locked onto the left side without help.

In the two-outlet type of device the locking of the stream onto the right wall means, of course, that the entire stream exits through the right outlet, and when it is switched to the left wall, it goes to the left outlet. (The Coanda effect explains, by the way, why the proportional-contro1 amplifier must have the channel widened into a heart shape near the power-stream nozzle; this configuration prevents the stream from locking onto a wall.) .It has turned out that the behavior of the stream depends a great deal on the distance of the splitter from the power-stream nozzle. Warren of the Harry Diamond Laboratories has investigated this matter in detail [see illustration, DISTANCE OF SPLITTER, below].


DISTANCE OF SPLITTER from power-stream nozzle in the standard, two-outlet type of fluid switching device has an important effect on the behavior of the stream. When the splitter is very close to the nozzle (top row), the stream does not lock spontaneously onto either wall. The stream can be directed into one outlet or tl1e other by a pulse injection from a control jet, but as soon as the control injection stops, the stream resumes its normal behavior of dividing at the splitter and flowing into both outlets. lf the splitter is located at a distance amounting to three to five nozzle widths from the nozzle (second row from top), most of the stream will begin to flow spontaneously to one outlet. When the splitter is six nozzle widths or more away from the nozzle (bottom two rows), the stream will lock onto one wall and flow only to the outlet it has chosen spontaneously or the one to which it is switched. By blocking one of the outlets (second column from right) one can divert the stream from the outlet it naturally prefers to the opposite one, but when the block is removed (column at extreme right), stream will return spontaneously to preferred outlet.

When the splitter is very close to the nozzle (within a distance not more than about twice the nozzle's width), the stream does not lock onto either wall, because the distance is not sufficient for a significant pressure difference to develop. The stream can be directed into one outlet or the other by a pulse injection from a control jet, but as soon as the control injection stops, the stream resumes its normal behavior of dividing at the splitter and flowing to both outlets.

If the splitter is located at a distance amounting to three to five nozzle widths from the nozzle, the stream begins to be influenced by the Coanda effect: in the presence of an asymmetry most of it will flow spontaneously to one outlet. When the splitter is six nozzle widths or more away from the nozzle, the Coanda effect takes full charge. The stream will lock onto one wall and flow only to the outlet it has chosen spontaneously or the one to which it is switched.

Some very interesting effects can be produced by blocking one of the outlets, as the illustration on page 85 shows. In one arrangement such a block will divert the stream from the outlet it naturally prefers to the opposite one, but when the block is removed the stream will return spontaneously to the preferred outlet! Thus the device exhibits a property that amounts to memory, and it can be put to use for that purpose. Of the various possible uses of fluid control devices the most intriguing is their application as the basis of a digital computer. The type of device in which the stream locks onto one wall or the other is precisely suited for functioning as an all-around element for such a computer. It gives a binary digital response: the stream can be directed to either of two exit ports, one representing 0, the other representing 1. It can be shifted back and forth from one port to another like a flip-flop switch. Indeed, it can even serve as an oscillator. This can be accomplished by connecting the control jets on the opposite sides of the main stream by a tube that will transmit sound. VV hen the main stream flops from one side to the other, it generates sound waves -- a compression wave on one side of the stream and a rarefaction wave on the other. If the sonic path through the tube connecting the two sides has a length that is about half the wavelength of the full sound wave, the compression and rarefaction waves, traveling through the tube in opposite directions, will cross to the opposite sides in the same time. This change in pressure at the respective sides (from compression to rarefaction on one side and from rarefaction to compression on the other) is sufficient to switch the main stream from one wall to the other. The stream will oscillate back and forth as the sound waves travel back and forth.

The fluid device can also be designed to act as a logical gate expressing the concept "and" or "or." For the "and" gate the jets are so arranged that jet A alone will go to one output, jet B alone will go to a different output and the two jets combined will go to a third output that signifies A "and" B [see top illustration, LOGICAL "AND" GATE, below]. For the "or" gate there are two jets, A and B, on one side of the main stream; either jet A or jet B, on being activated, will cause the main stream to go to a certain exit port [see middle illustration, LOGICAL "OR" GATE, below].

The fluid-control concept lends itself to a great variety of arrangements. The device can be designed, for example,. to reset itself, with the main stream always returning to a given exit port in the absence of a deflecting pulse; this can be achieved simply by placing the main-stream nozzle closer to the right wall, say, than to the left wall. By building a network of channels of varying cross section and with varying conntrols it is possible to create a system that will carry out a series of different operations. And the characteristics of the system can be varied by using various fluids. In most of the computational elements constructed so far the working fluid is air, because it is so readily available and easy to handle.

Let us look now at a couple of specific applications of fluid control on which considerable research has been done. One is the steering of rockets in flight. This is managed at present by the use of devices such as jet vanes and swiveled nozzles through which pulses of exhaust are discharged to apply small thrusts correcting the trajectory of the rocket. These devices are far from ideal, because it takes a great deal of force to operate them, their response to commands is fairly slow andf they involve moving parts that are subject to being put out of action by the hot exhaust streams. A fluid-control system now under study would avoid those difficulties. The control would be applied to the main exhaust stream itself, that is, the thrust that drives the rocket. A small power stream would be bled from the main exhaust, and this small stream would be used to deflect the main stream when necessary to change the direction of the rocket [see bottom illustration, ROCKET CAN BE STEERED, below]. The small stream would be controlled by a pair of jets; when the right jet was turned on, it would turn the stream so that it struck the main exhaust from the side, thereby changing the direction of the driving thrust slightly with a resulting change in the trajectory of the rocket's flight.

The idea has been tested in a small land vehicle powered by a turbine engine. By means of a five-stage proportional amplifier weighing less than 10 pounds, an extremely small fluid signal amounting to a flow of less than five thousandths of a pound per minute controlled a jet of 33 pounds per minute that was used to deflect the power stream driving the vehicle.


LOGICAL "AND" GATE can be built using fluid-control components. Jets are so arranged that jet A alone will go to one output (left), jet B will go to a different output (middle) and two jets combined will go to a third output (right) that signifies A "and" B.


LOGICAL "OR" GATE contains two control jets, A and B, on one side of the main power stream. With neither jet on, the stream will go to one output (left). Either jet A or jet B, on being activated, will cause the stream to go to other output (middle and right).


ROCKET CAN BE STEERED by the fluid control device depicted here. A small power stream is bled from the main rocket exhaust and used to deflect the main stream when necessary to change the direction of the rocket. The power stream is controlled by a pair of jets. When one jet is on (left), the stream simply adds to the main driving thrust of the rocket. When the other jet is on (right), it switches the stream so that it strikes the main exhaust from the side, thereby changing the direction of the driving thrust and hence the trajectory of the rocket's flight. Fluid system appears to be more efficient and reliable than other systems currently in use.


The other application I want to describe is an artificial heart pump developed jointly by workers at the Harry Diamond Laboratories and the Walter Reed Army Institute of Research. This device uses a fluid amplifier of the wall-locking type with air as the working fluid. The stream of high-velocity air emerging from the power nozzle locks onto one wall and then flows on to a chamber containing an artificial ventricle that is made of a flexible plastic and is filled with blood [see illustration, ARTIFICIAL HEART PUMP, below]. The rising air pressure in the chamber squeezes the ventricle and thus forces blood through a valve into the circulatory system of the patient. In effect it acts like the systolic contraction of the heart.

After the ventricle has contracted to a certain volume, the air pressure in the chamber opens a port through which the air then escapes and returns to the place where the power stream is flowing toward the splitter. Hitting the power stream from the side, the returning air causes the stream to switch from the original channel to the opposite channel. This results in a pressure drop in the chamber containing the ventricle; the ventricle therefore begins to expand under the weight of blood that is now able to flow in by gravity through an inlet valve at the top. Meanwhile a control valve on the other side of the main stream in the amplifier is feeding in a certain amount of airflow to prevent the pressure from falling too low. By the time the ventricle has expanded to its full volume this valve admits a flow of air sufficient to switch the main stream back to the original channel, and so the cycle begins again.

The valve admitting air from the side is, of course, the device that controls the duration of the heart pulse. Another valve in the channel on the same side of the splitter controls the rate of the pulse, and the valve that admits the main stream controls the amplitude of the systolic beat. All these controls are adjusted beforehand so that the pump supplies blood at the rate and in the amount the patient requires. The models built so far are capable of producing blood pressures of up to about 500 millimeters of mercury and blood flows of up to 10 liters per minute. They can maintain pulse rates within a wide range, from as low as 30 to as high as 180 counts per minute. They have proved to be able to adjust their output automatically to changes in the patientís blood-vessel resistance to blood flow. Apparently they cause less destruction of red blood cells than some of the commercial artificial heart pumps now in general use. The chief advantages of the fluid-control pump are its simple, rugged construction, its compactness (it is a small box only a few inches long and weighing only 10 pounds) and its low cost.


ARTIFICIAL HEART PUMP uses a fluid amplifier of the wall-locking type with air as the working fluid. The power stream of high-velocity air locks onto one wall and then flows on to a chamber containing an artificial ventricle that is filled with blood (top). The rising air pressure in the chamber squeezes the ventricle and thus forces blood through a valve into the circulatory system of the patient. After the ventricle has contracted to a certain volume (bottom), the air pressure in the chamber opens a port through which the air then escapes and returns to the place where the power stream is flowing toward the splitter. Hitting the power stream from the side, the returning air causes the stream to switch to the opposite channel. This results in a pressure drop in the chamber containing the ventricle; the ventricle therefore begins to expand under the weight of blood that is now able to flow in by gravity through an inlet valve at top. Meanwhile a control valve on the other side of the main stream is feeding in a certain amount of airflow to prevent the pressure from falling too low. By the time the ventricle has expanded to its full volume this valve admits enough air to switch main stream back to original channel, and so the cycle begins again.


The fluid-control concept is only in its infancy -- still in the research stage with little practical use made of it so far. One of the problems that must be solved is that of miniaturization of the fluid element so that it will compare in compactness with an electronic or a mechanical component. The most promising technique for doing this seems to be optical fabrication. This entails photographing the design (in this case the pattern of channels, outlets and so on), reducing the negative, projecting the pattern onto a sheet of material that is sensitive to light and then etching away the exposed portions with chemicals. By timing the duration of the treatment with the solvent the channels can be etched to precisely the desired depth. The material may be a plate of sensitized glass; after the photographic exposure, treatment of the glass with heat will transform the exposed parts of the glass into a crystalline phase that is about 20 times more soluble in dilute hydrofluoric acid than the unaltered glass. There are other materials besides glass that lend themselves to similar optical sensitization and chemical machining.

How reliable are fluid amplifiers compared with their electronic counterparts? A number of comparative tests have been made under various conditions. In the range of what may be called ordinary temperatures -- between 150 degrees and minus 50 degrees centigrade -- electronic amplifiers give a more reliable performance. But at higher or lower temperatures the fluid device is more reliable; for example, at 250 degrees C. or minus 150 degrees C. Its reliability is five times greater than that of an electronic device. When they are compared in various environments of ionizing radiation, the story is much the same: the electronic systems are more reliable at ordinary radiation levels but the fluid systems surpass them under exposure to high radiation levels, such as in or near a nuclear reactor. The fluid systems also stand up much better under vibration -- an important factor in a rocket, where the guidance system is subjected to the severe vibration accompanying the rocket's takeoff and acceleration. Strong vibration tends to weaken the solder joints and other connections' of an electrical system; it has little or no effect on a fluid ampliHer. One test model has come through 3,500 million cycles of high-speed vibration at 260 cycles per second without any detectable impairment of its performance.

Research is going forward toward three main goals: (1) understanding the basic processes that underlie the operation of fluid amplifiers, (2) developing methods of producing them in compact size and in quantity at low cost and (3) finding the applications for which they will be most suitable.



BIBLIOGRAPHY

FLUID AMPLIFIER HANDBOOK.
Diamond Ordnance Fuze Laboratories, Ordnance Corps, Department of the Army, 1962-1963.

FLUID JET CONTROL DEVICES.
Edited by Forbes T. Brown. The American Society of Mechanical Engineers, 1962.

Proceedings of the FLUID AMPLIFICATION SYMPOSIUM.
Diamond Ordnance Fuze Laboratories, Ordnance Corps, Department of the Army, 1962.