Application of fan units to a hovercraft
The lift system
Designing a miniature air-cushioned vehicle
The side pods and the bow
Development of the lift system
Re-thinking the cushion
Postscript- Programming the transmitter
I started thinking about this project in 2003 and it was eventually built in 2009. This may seem like a long time but I started with a rather mixed bag of conflicting information. Let me list what I knew.
The first thing was that I had never seen a model hovercraft that worked although I had seen a few pictures in the model press. The impression I had was that all model hovercraft use skirts and they use airscrews to power the lift and the drive. The airscrews seemed to be fitted with shrouds that probably doubled as guards although on swamp buggies the airscrew is in a mesh and no shroud is fitted. I am highly suspicious of shrouded airscrews because the blades are too far apart to interact and, once the shroud is fitted, the blades operate with a pressure rise between front and back and the blades in the centre of the airscrew cannot generate a pressure rise. The flow in the centre reverses and at least one and possibly three “toroidal” eddies form and the efficiency of the airscrew drops alarmingly and the input power required increases in consequence. The airscrew becomes very noisy and is clearly in trouble.
In the past I have had a few encounters with fans and the use of fans and I had formed the view that fans were not well understood. I did not know whether this is true but I did know that it is quite easy to buy a full-sized fan unit that is simply a fan/motor combination in a short section of pipe work that is complete with flanges ready to fit into some system of pipe work. The blades are likely to be as-cast, the motor is likely to be mounted on radial spokes of circular cross-section, and there will be no blades other than those of the fan. The fan will be incorporated in some pipe system designed without any reference to the requirements for the fan to work efficiently. Generally fans are not taken seriously and, if they are, all the effort is directed towards tidying up the fan itself and not to the pipe-work upstream and downstream of the fan.
I know that when an aircraft gas turbine engine is tested on the ground it is fitted with a large bell-mouthed intake to serve air to the engine as it might be fed with air when it is in flight. I also know that there would be a large forward-acting force on this bell-mouth because I used to operate a wind tunnel with such a bell mouth and, left unlocked, it moved. I know that a bell-mouth is essential if a ducted fan is to work efficiently yet I have never seen one on a model.
I had looked inside model aircraft propelled by ducted fans and the ones that I saw made no attempt to serve the fan with air properly and the exit system seemed to be fitted for convenience rather than the essential engineering/scientific requirements of the fan. It occurred to me that hovercraft modellers do not use fans because it is unlikely that they will have the necessary science to be successful.
This was a great deal to think about. The trouble was that, whilst I knew the “theory” of fans, I also knew that any device that sets out to create a flow against a pressure gradient is much more difficult to design than say a turbine that extracts energy from a flow that goes with the pressure gradient. I came to the design with the intention of making fans to my own design. I could see that this would not be easy as it involved the design and manufacture of sets of twisted, aerofoil-shaped, blades in wood and fixing them to hubs. It seemed like a big investment of time that might all come to nothing and I had no baseline figures of airspeeds or fan speeds to help with the design. The design process was not straightforward and eventually I came to the conclusion that I would have to make something just to get started and get some figures that might help me with my design. In the event I made a lift system that worked so well that I had to continue to create a complete hovercraft. It is shown at rest in figure 1 and in motion in figure 1a. It is easy to see how much it lifted out of the water.
I kept a log of the design study but it is not suitable for publication here because it was coloured by my preconceptions and not at all systematic. I have decided to write up the design in what should have been the sequence because this at least does not create red herrings that have to be superseded.
When I started I did not actually want to own an air-cushioned vehicle as a recreational boat because I thought that once I had found out what can be done with it, it would not be much fun to operate. In the event, there still seems to be a long way to go in learning about this vehicle and its behaviour. It has a further attraction for an old man in that it is light and easy to handle.
What I really wanted was a project that might make a change from a solid year of writing my textbook. I fancied trying to come to understand this fan business once and for all and it might find its way into the textbook. I decided that I wanted to use the available science to design and build a system to lift a miniature craft and to design and build thrusters to drive it. The returns came when the vehicle first lifted and then when the craft first moved on its cushion of air driven by the thrusters. The fact that a craft weighing 7½ pounds will run at its maximum speed for 15 minutes on 3,300 mAh of charge says that there is not much wrong with this design.
I had other ongoing projects that I wanted to pursue and I reduced the investment of time by buying fans that I thought might work. They were cheap but proved to be satisfactory.
An ordinary boat can be driven by an underwater propeller and it is well known that the bow wave builds up with increasing speed and becomes so large relative to the boat that the only way to go faster is to have bows that will lift the boat up and on to the surface and plane. There is a definite transition phase between the boat running as a displacement hull and lifting up on to the plane. Considerable power is needed to make the transition but, once it has been made, the power requirement falls but is still quite large.
There is an argument for lifting the boat physically by creating a volume of air under a suitably shaped hull so that the hull is lifted clear of the water and then driving it in some way, for example by water screws or airscrews. One can see how this seemingly simple and different idea might attract the attention of those who like inventing things but it seems to me that there should be some criteria for success. It all becomes rather a waste of time if the power required to drive the air cushion vehicle is greater that that to drive a similar displacement hull at the same speed. It is also desirable that the air-cushion vehicle should have acceptable sea-keeping qualities that are at least as good as the displacement hull. I think that this latter criterion might be very difficult to achieve so I set my sights on designing an air-cushioned vehicle that would run on low power over smooth water in light winds. If it turned out that it behaved acceptably on a surface with wind and wind-generated ripples that would be a bonus.
It is obvious from the outset that lifting a hull out of the water on a cushion of air must involve a leakage of air because the water cannot be one part of an airtight seal. Given that open water is seldom free from surface motion the leakage must be significant. So the cushion must be fed with a steady, and possibly large, flow of air at some pressure in excess of atmospheric pressure by a device that is light and efficient. The only suitable device that we have is a fan mounted in a duct. (I discount the use of airscrews because by the time the modifications needed to let the airscrew produce a pressure have been made it is a fan in a duct.)
This means that we need an understanding of how fans work. I do not think that it is easy to explain so let me develop an explanation.
Figure 2 shows a miniature fan of 64 mm overall diameter. I used this fan in my thrusters. I think that whoever designed it did not make too bad a job of it. The hub is of substantial diameter compared with the diameter across the blades. (It is slightly greater than the commonly used 40%.) The hub is of an acceptable shape and the profile is not critical. I am not impressed with the blades. They all appear to be of the same shape even though they are very thin and are just flat and sharp at leading and trailing edges. It is hard to see why the blades are not thicker and have a camber. It has six blades. They look as though they were designed by someone who did not fully understand these fans.
As purchased, this fan is fitted to an electric motor and runs in the cylindrical shroud shown in figure 3. The mount for the motor is on three blades that are “straight” radially but the strip of masking tape on one of the blades shows that they are curved from leading edge to trailing edge. These essential components are typical of fans of all sizes. But the fan is not complete because it needs an intake and an extension of the duct downstream of the fan. These two components have shapes that depend on the application of the fan. A proper explanation of this statement must follow an explanation of how the core of the fan actually works.
So I need to set out how fans work. Figure 5 shows the core element of a fan unit. It matches figures 3 and 4. The fan fits snugly inside its cylindrical duct and is driven by an electric motor. As they rotate the blades induce a flow through the cylindrical duct. What we would like is for there to be an intake duct that is designed to act with the hub to make the flow of air just in front of the blades be along the axis of the fan and at the same velocity at every radius. I have drawn such an intake in figure 6. I have also drawn flow lines and surfaces of uniform pressure. These flow lines start a long way in front of the intake where the air moves very slowly and only as the intake is approached does the speed increase significantly. As the speed increases the pressure must fall and, in the bell mouth, the pressure falls quite rapidly to its lowest value just before the blades of the fan. This means that the bell mouth is subject to atmospheric pressure on the outside and to pressures below atmospheric on the inside. It will have a net force on it acting forwards. I have shown, in blue, the cross-sections of surfaces of uniform pressure in order to complete the flow net.
It is prudent to think about the flow into this fan through the bell-mouth. Throughout, the flow is converging and convergent flow is known to be stable so we can reasonably expect this flow pattern to occur. What is not so certain is whether the flow will rotate. There is every reason why it should especially with a rotating hub but it is unlikely that the speed of rotation will be high enough to seriously impair the performance of the fan. Large fans often have fixed “hubs” supported by untwisted, streamlined, radial vanes to prevent this rotation.
In figure 6 I have shown an intake that is shaped like a 90° arc rotated to give a surface of revolution that fits on the duct. This shape is satisfactory for stationary fans but, if the fan moved forwards at a speed that is a significant fraction of the airspeed produced by the fan, the bell mouth would have to be changed to something like the shape shown in figure 7 to give smooth flow into the bell mouth at the inlet rim.
So now we have blades rotating and air flowing at a steady and uniform velocity towards the blades. At any radius the tangential speed of the blade is where is the angular velocity in radians per second. (Note where is the speed of rotation in rpm.)
At any radius there is a velocity of the air over the blade relative to the blade and I can draw velocity diagrams. The small fan shown in figures 2, 3 and 4 has an outer diameter of 2.5² and a hub that is about 1² in diameter. Its maximum speed of rotation is 11,000 rpm. So the linear speed at the root of the blades is 48 feet/second, at the tip is 120 feet/second and the speed at the mean diameter is 84 feet/second. (120 ft/sec = 82 mph.)
At this speed the air enters the blades at about 66 feet /second.
This permits the drawing of velocity triangles for the root, the tip, and the mean diameter.
In figure 8 I have drawn the velocity diagrams for the flow into the fan. The diagrams are for the root, for the tip and for the mid radius.
Combining the air speed with the blade speeds gives the speeds and angles of the air relative to the fan blade. These are shown as the red arrows.
I have already said that the blades will have an aerofoil shape and, if they are to function as aerofoils they must operate at an angle of attack. I do not know the precise value of this angle but 8° would not be abnormal. If I redraw the three velocity triangles on a common vector representing the air speed I can represent the aerofoil sections of the blade on the three vectors representing the velocity of the air relative to the blade.
I have done this in figure 9. I have shown an under-cambered aerofoil section set at 8° to each red vector. I have also show, in dotted, the chord line for the mid radius aerofoil.
Note that the blade must twist from root to tip and if the blade were to extend from the centre to some very large radius the angle of the relative velocities would change from 90° to very nearly 0° to the plane of rotation. Here we have a root set in a large hub and a tip radius that is only 2.5 times that of the hub.
Now we must decide what these blades do to the air and, whilst we know the flow pattern over one aerofoil in an extensive stream of air, here we are dealing with six aerofoils acting together in a common and contained flow and they interact. I measured the chord of my small fan (2.5² diameter) at the mid point and as I know the diameter I can draw aerofoil sections that are at the correct pitching to be the same as this fan.
In my view there is much to be gained by drawing, by eye, the flow pattern over these blades. Of course in drawing the pattern the normal rules must be obeyed and it is in obeying the rules that the pattern emerges. The outcome is in figure 10. The blades are at the correct angle for the blades to be moving vertically downwards in my diagram but, of course, the diagram is for the air relative to the blades so mentally we must hold the blades still.
I have supposed that the flow line that will come to rest on the nose of the aerofoil (the stagnation line) curves upwards just as it does for isolated aerofoils and it is shown in red. I have also supposed that the flow line from the trailing edge curves away as I have shown again in red. I have inserted a set of flow lines that are not totally consistent but will not be too far away from reality. We can draw some conclusions from them.
The first thing to notice is that the blades interact even though, initially, they seemed to be a long way apart. This is the flow pattern for the mid radius and the blades would be significantly closer at the hub and much farther apart at the tip. The designer of the fan has chosen to make the blades taper towards the tip and so reduce the interaction between the blades. There is always a compromise to be made between the number of blades that tend to obstruct the airflow at the root and the interaction between the blades at the tips.
Flow diagrams can be interpreted if it is recognised that the spacing of adjacent lines decreases as velocity rises and pressure falls. Where lines diverge velocity falls and pressure rises.
Clearly the pressure on the concave face is high and over the convex face is low to produce a net force on the blades. This will be a lift force on the blades as in normal aerodynamics. The fan must be driven to sustain this force.
A very important further observation is that any two adjacent red flow lines are closer together as they approach that they are as they depart. This means that there is a net rise in pressure as the air flows through the blades.
Now we might look at the overall effect of the rotating blades. In this case they divert the flow by and angle of about 13°. This seems to me to be realistic but I think that it should be interpreted as a typical angle and not an accurate one. This information makes it possible for us to draw a velocity diagram at exit.
I have extended figure 9 to give figure 11. The extension is the addition of the velocity triangles at exit. In order to construct the triangles I have used the fact that the axial component of the velocity of the air will not change during its passage through the fan. So the axial component of both the velocity of the air relative to the blade at exit and of the absolute velocity (velocity relative to the duct.) of the air at exit will be the same as the velocity at inlet. In each diagram the change is that the velocity of the air relative to the blade is diverted. I have let all three be diverted by 13°.
We see that, in each case, the velocity of the air relative to the blade is smaller at exit than inlet but the absolute velocities of the air have increased significantly.
The fan has done two things to the air. It has increased its pressure and it has given it a significant kinetic energy. Unfortunately part of that kinetic energy is of rotation and this cannot be avoided. However some of the kinetic energy of rotation might be recoverable as a rise in pressure.
I have now set out the role of the ducted fan unit comprising the cylindrical duct, the bell-mouthed intake and the fan. It can now be employed in some engineering application. What we must not do is allow this pressure and the kinetic energy to be frittered away in bad design.
Application of fan units to a hovercraft
I have said that my goal was to make a craft that would move on a cushion of air and be driven by ducted-fan thrusters. This means that I have to adapt fans to produce a pressure rise in a flow of air that need not be at high speed for the lift and to adapt the fan unit to produce the maximum thrust for the thrusters. The two adaptations are not the same.
The lift system
We have to decide how a lift system actually works. My miniature hovercraft is built on a piece of plywood that is 32² long and 15 wide. The craft weighs about 7 pounds and the weight per square inch of underwater area is 0.0146 lb. This is equivalent to 2.1 pounds per square foot. This in turn is equal to 0.4 inches of water gauge. (Difference in level of a water manometer.)
So, if this craft is to be lifted, air at a pressure of at least 0.4 water gauge must be held in place under the craft. One can imagine systems to contain this air but, necessarily, they must all include some sort of seal between the water surface and the “hull”. Such a seal will leak and so air must be supplied steadily to make up the leakage. The flow of air that must be supplied will depend on the way in which the seal is designed. It seems to me that there is no reason for the air that leaks out to move at high speed, what is needed is a steady flow at a pressure greater than 0.4 inches of water gauge to meet the pressure drop where the air actually leaks and that leakage may fluctuate in position and magnitude with movement of the free surface of the water over which the craft moves.
All in all the design goal must be to maximise the pressure produced under the vehicle for a given input of power to the ducted fan and then vary the input power to match the lift to the conditions under which the lift system is operating.
So how can we produce the maximum pressure?
The first thing to deal with is the motor. We cannot just leave it as a cylindrical can. It must have a streamlined fairing. A plastic fairing was supplied for the fan used in my lift system but, for the thrust fans, I made a wooden fairing as shown in figure 12. It is just typical of the required shape. Motor cooling was by inducing a flow of air through the motor.
The next thing to do is to try to exchange the unwanted kinetic energy of rotation for a rise in pressure. This does not contravene any aspect of science but the only mechanism at our disposal is a set of curved blades that are fitted immediately downstream of the fan whatever its application. Given that the rotation does not correspond to any known vortex the design of such blades is not really possible but an attempt must be made to use curved blades. In both sizes of fan that I used there are only three blades and that is almost certainly too few. They are curved but not twisted and they are really the supports for the electric motor. Some further attempt must be made to augment these blades. For my lift fan I fitted cruciform blades around the motor and, for the thrusters I fitted three extensions to the original blades. They are in figure 12 and are attached to the motor cover.
It is hard to know how successful these measures have been but nothing must be left entirely to chance. In both cases I end up with in-line flow ready for the next stage.
The speed of the air in the lift fan will be of the order of 40 mph and this was measured using a pitôt tube and a water-filled manometer. The difference in level is about 0.8 inches of water. This is kinetic energy that is available to be recovered as a rise in pressure. There is only one way to do this and that is to use a diffuser. A diffuser is just a divergent duct in which the velocity of the flowing air can be reduced in a controlled manner and, in doing so, recover kinetic energy as pressure energy. We need to know more about the performance of diffusers.
The function of a divergence when used as a diffuser is to recover kinetic energy at the inlet at the small end as a rise in pressure energy at exit.
Figure 13 is a diagram of a diffuser. The kinetic energy at section 1 is and at 2 is . It follows that the maximum possible recovery of kinetic energy in flowing from 1 to 2 is It will be less than this for a variety of reasons. However it is clear that the shape of the circular diffuser depends on the two diameters and the angle of the divergence.
Others have experimented to find out how this angle affects the performance of the diffuser. The outcome was presented as a graph of the loss versus angle but it is of more interest to us to see a graph of the recovery versus angle.
It is figure 14. It is important. It tells us that, in a round diffuser, the greatest recovery is at and that the maximum is well defined. It shows that the recovery at 20° is only 20% so, if we want to recover kinetic energy as pressure, we have a small range of to work in.
It is unlikely that the divergence used for a hovercraft will be circular. It is much more likely to be rectangular. The friction losses in a duct of any shape are dependent on the ratio of the area of cross-section to the perimeter. So we must expect that a diffuser of rectangular section will behave in a different way to one of circular section. The graph for a rectangular diffuser of 4 by 1 in cross-section has been plotted against a notional angle calculated by finding the equivalent circles and deriving the angle. It takes its maximum at about 7°. The graph suggests that a recovery up to 80% is possible at something like 8°.
It would obviously pay to attempt to get this notional angle when designing but other factors, such as the practical layout of the craft, may require the angle to be greater or smaller.
A further important message from this graph is that, if the angle is greater than, say 35°, the recovery goes negative and, instead of the pressure rising in the diffuser, it falls. This accounts for the abysmal performance of lift systems where no attempt is made to control the flow and what little rise in pressure is produced in the fan is largely lost in the subsequent disorderly flow.
Clearly I require a diagram of a fan unit fitted with a diffuser. I have drawn it as diagram 15 where the diffuser is a conical divergence. The basic fan unit is shown in black and the additions in red. I have drawn a suitably shaped cover for the electric motor and this is shown in black because it should be fitted to every fan unit.
The additions that make this fan unit suitable for powering the lift system are the extensions to the flow straighteners that form part of the fan unit and a divergence that, in this case, is a simple cone.
It is now possible to plot the way in which the axial velocity and the pressure of the air change as the air flows through the whole assembly.
I start with the velocity graph. Here I have plotted axial velocity because when the air comes off the blades of the fan its axial velocity is unchanged from that at inlet but it has a significant angular velocity that cannot be shown on this graph. We have seen in figure 6 that the flow converges as it flows first through the bell mouthed intake and then through the space between the duct and the hub. It follows that, between A and B, the velocity of the air will increase in some way such as that I have shown in the graph of axial velocity versus length. At A the air has a low velocity. Between B and C the air is flowing through the fan where there is no change in axial velocity but there is a rise in pressure and the creation of angular velocity. This angular velocity has kinetic energy that may be recoverable as a rise in pressure during the flow through the flow straighteners and their extensions. There is no change in axial velocity between B and D. Now we have wholly axial kinetic energy that can partially be recovered as a rise in pressure if the velocity of the air is reduced carefully in the new divergent cone. This takes place in the divergent cone that has an exit area that is twice the inlet area and the velocity will fall to one half of the inlet velocity. This is shown in D to E on the velocity graph.
It is the pressure graph that is interesting. Broadly the pressure in the duct falls between inlet to the bell-mouth and entry to the fan where the pressure rises. Between the fan and the exit the pressure goes on rising to give a useful pressure above atmospheric pressure at exit. My lift fan gives about 1.1 inches of water gauge. The sequence is as follows. Between A and B of the pressure graph the pressure falls below atmospheric pressure in the way that I have shown. In the fan, that is, between B and C, the pressure rises above atmospheric and the air acquires angular kinetic energy. Some of this angular kinetic energy can be recovered in the flow straighteners between C and D. The process is likely to be quite inefficient for several reasons so I have shown only a small recovery. Then, in the divergence between D and E, the pressure rises by as much as 0.3² in my miniature lift system to give about 1.1² at exit at a modest fan speed of about 7,000 rpm.
I need to justify the choice of a 2 to 1 increase in area in the divergence. The kinetic energy of a moving air is proportional to the square of its velocity. If the velocity of air is the kinetic energy is proportional to the area of a square of side c. I have drawn such a square in figure 16. If the velocity of the air is reduced to its kinetic energy is represented in figure 16 as the small square to the same scale
Clearly a reduction in velocity of one half reduces the kinetic energy to one quarter. This means that, if the divergence has an area ration of 2 to 1 the kinetic energy will be reduced by ¾ and, if the divergence has an angle of 11° the recovery of kinetic energy as a rise in pressure can be as high as 80% according to the graph in figure 14. If one thinks of a larger area ratio it soon becomes obvious that we are in a region of diminishing returns. An increase in the area ratio to 3 to 1 would only give the possibility of recovering 895 of the kinetic energy when compared with 75% for the 2 to 1 ratio.
I have drawn figure 15 with an area ration of 2 to 1 with an angle of 11° and the cone is quite long. In the cone the velocity will fall and the pressure rise and we might expect to recover about 60% of the inlet kinetic energy. We need the air to enter the cushion at a significant speed so that it can meet the air leakage so any further increase in the area at exit gives little extra recovery of kinetic energy, an unwanted increase in length and possibly too low a velocity at exit.
The thrusters are there to drive the craft once it has been lifted out of the water by its lift system. We need our fans now to produce an axial force. The adaptations that were fitted to the fan used in the lift system are no longer appropriate because the thrusters take in air at atmospheric pressure and deliver it at exit at the same pressure. It is common to calculate the force by calculating the momentum per second of the air leaving the thruster and equating it to the force. This is correct but not much use when one is trying to design a thruster. We need to know where the force is actually acting on the thruster.
There will be a forwards force on the bell-mouthed intake. This can be seen from the fact that the pressure over the whole of the inside of the intake is below atmospheric pressure and all of the outside is at atmospheric pressure. There is no way that a forwards or backwards force can be exerted on the parallel part of the duct. The major force acting forwards on the fan unit is that on the fan caused by the increase in pressure of the air as it passes through the fan.
We must consider the possible ways of increasing this pressure difference across the fan. It is clear that the pressure at inlet to the fan is fixed for a given intake and fan speed. If we want to increase the pressure across the fan we must look to the exit. We know that the air leaving the fan has gained kinetic energy of rotation and we can possibly recover some of that using flow straightening blades. In the process a forwards force will be generated but there will also be a small increase in pressure as we have seen in the lift fan.
Now if we fitted a convergent cone to the exit it would increase the velocity of the air and produce a pressure drop to exit. As the exit pressure must be atmospheric pressure the consequence will be an increase in pressure at exit from the fan and an increase in the pressure rise over the fan and therefore an increase in the forwards force. The fan must impart more energy to the air to create this desirable effect. In passing we might note that there is a net force acting backwards on the surfaces of the cone but the pressure difference between inlet and outlet produces a greater force acting forwards.
It is appropriate for me to draw a diagram of the fan unit when incorporated into a thruster. I have done so in figure 17 where the additions to the basic fan unit are shown in red. The major difference between this and the fan adapted for lifting is that the tail cone is convergent instead of divergent. I have included extensions to the flow straightening blades as before
Again I have drawn the graphs of axial velocity and of pressure above and below atmospheric pressure along the fan and tail cone.
Between A and D the axial velocity varies in this combination of fan and tail cone just as it did for the lift fan but between D and E the velocity now rises.
The big difference between the pressure graphs for the two combinations of fan and tail cone is that where the pressure at exit from the divergence is above atmospheric pressure, for the convergent cone it is at atmospheric pressure. The graph of gauge pressure along the fan and tail cone for the thruster is very different. Here the pressure varies in the intake as it did for the lift fan but in the fan the pressure rises to above atmospheric pressure between B and C, rises by a small amount in the flow straighteners from C to D and then falls to atmospheric pressure in the convergent cone.
We must think very carefully about the design of this divergent cone. It is not straightforward.
A fan is a device that relies on the blades behaving as aerofoils for its efficient operation. When I say that the blades must behave like aerofoils I mean that the flow over them must be smooth and “attached” to the blade. (It would pay a reader to look at my article on “Making sense of aerofoils” on this web site.) In some ways it is easier to explain the meaning of “attached flow” by explaining “detached flow”. If the flow of air cannot follow the shape of the blade it will break away from the blade at the leading edge and follow a path that is not determined by the shape of the blade. It becomes detached and when the blade is the wing of an aeroplane it is said to stall. In an aeroplane the process of stalling normally starts at the root or the tip and progresses along the wing. The same thing happens on the blades of a fan
In the thruster we have a fan and we know that we can increase the rise in pressure and hence the force exerted on the fan by fitting a convergent tail cone. What we cannot do is make the air converge so much that the blades of the fan become overloaded and cause the flow to become detached. It is possible to hear this process of becoming detached. My thrusters were designed to work on 8.4 volts without detachment. They work properly almost to the maximum speed and run very quietly in no aural difficulty but, just before the maximum speed is reached, the noise of the fan changes and it starts to screech as detachment starts at root or tip, usually the tip. I cannot run the fan fast enough to detach the flow completely. I got my convergence right.
So how can we design a divergence? We can start by looking at the behaviour of air flowing through a convergence.
Figure 18 represents a convergence in a pipe. At entry to the convergence the diameter is and the velocity of the air is . At exit from the convergence the diameter is and the velocity is . This change in velocity causes a pressure change from to .
We can use the Bernoulli equation to write :- from which where is the density of air.
Now we can also say t hat or where and are the areas of cross section at inlet and outlet. As the area is proportional to this leads to :-
and then to :- .
This is a simple expression but we must decide how to structure it for design. I have been expressing pressure in inches of water gauge and I will continue to do so. Then, if the velocity is expressed in feet/second, the expression above can be written :-
where the pressure drop is in inches of water gauge and the velocity in feet/second.
This will have to be explored using a computing package but it can only be explored if we use practical figures. The most important is to have an idea of velocity of the air entering the convergence. For the thrusters that I used, the speed of the air must be less that 60 mph or 88 feet/second and so it would make sense to look first at speeds of say 20, 30, 40 and 80 feet/second. If is fixed graphs of the pressure rise in the convergence in inches of water gauge can be plotted against exit diameter for the four air speeds.
This graph has no value in design unless we have an idea of the air speed at exit from the fan unit. I measured this with a simple pitôt tube and manometer and it was about 40 mph or about 60 feet/second.
This means that it is the green trace that is relevant. Then one must make an estimate of the increase in pressure in the convergence that the fan can sustain. I thought that it could not exceed 0.5² and opted to make the exit 2.22² in diameter with the option of trimming the length of the convergence and so increasing the exit diameter if it proved to be too small.
The determination of the dimensions of the exit diameter of the convergent cone for a thruster is not to be guessed. If it is too small the fan will run wholly or partially stalled.
Designing a miniature air-cushioned vehicle
It must not be supposed that this design just proceeded remorselessly from inception to fruition by logical sequence. I had no knowledge of other miniature hovercraft nor for that matter of the workings of full-sized machines. I knew that there was something difficult about the hovercraft and that it had had a short life in service. I recall a friend saying to me, when I showed him my partly-built lift system, that I should throw the electric fan away and get an i.c. engine.
In my introduction I have said that I do not regard the hovercraft that I have seen as viable engineering devices but experience tells me that that does not mean that the principle cannot be realised in hardware if some science and engineering know-how can be brought to bear. I have had earlier experience with the use of fans by those who do not practice much in the way of science. I had a professionally-built, wind tunnel for student use in my care. It was rated to give 25 metres per second in the working section. I thought that the arrangement of the fan and its duct work could do with some improvement and when the lab technicians has reconstructed this part of the tunnel the speed in the working section was 36 metres per second. I now realise that had I exerted myself a bit more that might have been 40 metres per second or even higher. As it was this 50% increase in speed could have been achieved by the maker for little extra cost. That tunnel also showed me that there was a very real forwards force on the shaped intake to the tunnel. Subsequently I designed a new tunnel from scratch and it worked very satisfactorily although, in retrospect, I did not pay sufficient attention to the exit from the fan as can be seen in figure 20 where I stopped designing when the air left the fan. I knew that there was more to understand about fans.
When I started this design for a hovercraft I had very little to go on, no useful figures of airspeeds or flows nor of the power required to drive small fans. I thought that there was no alternative to just using what knowledge I have to construct a lift system and see how well it performed. But even so, just to make a lift system required a considerable investment of time and money and I felt that it should have the potential to become working air-cushioned vehicle. This required a decision about what the vehicle should do.
I knew that there is no chance that I might find a large, smooth, level, hard surface that I could use at any time to test a miniature hovercraft so the only alternative was to use a pond or a lake. Then there was the problem of wind and wind-driven ripples to consider and I decided to set my sights low and aim to use my craft on water on wind-less days. This gave me a clear goal but I did not envisage a high-speed craft just one that would go faster than I can walk and fast enough to prove the concept.
I looked at the configuration of the SRN 1 and saw no future in a craft that had no axis that could be aligned with the direction of travel. I considered the use of a skirt but rejected that because I did not fancy air bubbling out everywhere. This left some design where the air flowed from front to back under a flat board that could be the bottom of the hull. This leads to the provision of side plates to prevent leakage of air outwards from the sides of the board. I am told that this is called a sidewall hovercraft.
I thought that I would use this concept, make a lift system to test it and, if it worked, go on to convert it into a hovercraft. I thought that the control of the air as it flowed under the flat-bottomed hull could not be predicted in advance and that the stabilising of the hull when “floating” on its air cushion could be left till last. However I did think that it was essential to prevent the air from moving from side to side.
Figure 21a is my first diagram of the lift system and, had I anticipated this article, it would have been much more tidy but the principle of the lift system that I actually made is in this sketch.
The intention was that air would enter the vertical-axis fan through a bell-mouth, turn through 90°, and diverge as it flows to the front in a suitably shaped duct before turning through 180° to go under the hull and then from front to back in channels formed by fitting plates or strakes to the flat bottom. I foresaw the need for a dam at the back to lift the hull but did not contemplate the possibility of making the dam adjustable remotely.
I thought that the reason that most hovercraft are not very effective was because the air was just left to find its own way about and I was not going to make that mistake. So I had in mind dividing the flow in the divergent duct into six separate rectangular sectioned tubes running side by side as in figure 22.
This is as far as I went in the design before starting to build. I had this to say about it after it had been constructed and seen to work.
“I am sitting here as pleased as punch. I tried my lift system and it WORKED!!!”
Figure 23 shows what it looked like. It was just a trial horse made mainly from materials that I had left over from aero-modelling.
It worked much better than I had expected. Let me explain.
I pondered this whole problem back in 2004 and it seemed to me that what is needed is a useful rise in pressure to lift the craft and a decent flow to ensure the continued lift. I realised that this was all about having a well-designed fan and a well-designed divergence to recover the kinetic energy and to feed the cushion. The outline sketch was realised as a practical device.
I started with a fan unit that cost £20 and included a motor. It is a GWS GW/EDF-150. It had no intake but the blades seem to be reasonably well designed with a sensibly sized hub and a fairing over the motor to streamline it. The motor was mounted on three blades that were also designed as flow straighteners. It will run on voltages up to 12 volts. It seemed to me to do quite well on 6 volts but whilst it seemed to produce considerably more thrust on 12 volts the sound was awful. Something is amiss at this voltage. I now think that the flow was so bad that the blades were ripping into the air in all directions.
I was certain that the fan needed an intake duct and I made one. It was too big for me to turn on my lathe so I constructed it from ply and balsa. It is shown here partly constructed as figures 23a and 23b. The completed intake fits snugly on the fan unit.
I have learnt over the years that vertical intakes are easily disturbed by cross flow and seemed to me that it would make far more sense to let the fan point forwards rather than be mounted vertically. The forward force on the intake would help the drive and there would be a small ram effect from the forward motion. It would involve another 180° bend but I did not see that as a great building problem nor was it something to cause a serious additional resistance to flow.
However the first requirement was a transition from round to square and that transition could be the start of the long divergence. I made the transition from balsa and ply to have a notional angle of 11°. In order to do this I simply chose a length for the transition and, as the area of the duct at the exit from the fan is settled by the design of the fan it is simple to find the area that the transition on the supposition that the transition is conical. Taking the square root of this area gives the length of the side at exit from the divergent transition. The transition went from 3² diameter to 3 175² square in 4² and this gives a notional angle of divergence of 11.8° when the area of the hub is taken into account and an increase in area of 1.7 to 1. So, in the transition, the speed of the air drops by about 40% and this will reduce the friction loss in the rest of the duct-work.
I had to have some figure for the overall increase in area in mind and I started with an area at the fan of 6 square inches and ended with an area of 15 square inches at exit. This could give a maximum recovery of kinetic energy of 65%. I thought that this was a useful figure for a modest increase in area.
Of course I would only get some useful fraction of this if the divergence was long and the inside surface smooth and free from steps and lumps and bumps. I thought that this divergence could not just be an unobstructed duct especially as I had to make two 180° bends. I used guides and splitters to control the flow.
Figure 25 shows the fan unit with its intake and the divergence fitted and the 180° return bend that followed it.
Note the way that the bend is divided into four and the splitters extend into the transition and round the fairing over the motor to act as flow straighteners.
I had made it all in balsa and ply.
Then I needed a hull, if that is the word, and to mount the bend in it in preparation for building the divergence into it. Ply comes in sheets 48² by12² and I wanted the width to be about 15². So I cut the ply obliquely and scarfed it back together to have a width of 15² and then I could cut a length of 32² from it. Two vertical sides of 1.25² and simple rectangular formers at back and front gave a basic tray on which to build. I fitted the inside surface of the front return bend that I intended to make the full 15² wide and have a depth of 1² so that the exit area was 15 square inches.
Once the hull was made, the aft 180° bend had to be located and set in the hull. Figure 26 shows where I put it. I have been an aero-modeller and keeping track of the centre of gravity is second nature and I was aiming for a centre of gravity about midway back from air inlet to transom. The divergence was going to be quite heavy because it would have 5 sheets of 0.8 mm ply in it and there would be ply forming both the outside of the front return bend and the fairing that would go over it. I fitted a ballast box in the front fairing but it was not needed. Figure 26 shows the basic hull and the rear 180° bend and the divergence under construction.
Figure 27 shows the front of the hull at the same stage of construction as in figure 26. The inside of the front return bend is visible. The half cylinder that forms this bend was made by laminating three layers of 1/32² balsa sheet round a broom handle and fitting semicircular formers inside it.
Figure 28 shows the bottom on the divergence in place and the formers for the front return bend in position ready for the sides and the internal splitters.
These splitters for the long divergence are shown in figure 29. They were awkward to make. Note the cut-outs in the outer pairs to avoid obstructing the transition from the bend to the long divergence. They are shown in figure 30. It was all coated with sanding sealer, smoothed, and varnished because it will get wet.
Fitting the top plate was a nightmare because of the need to work with the finished surface on the inside but it eventually went on.
It ended up like figures 32 and 33.
Somewhere along the way I had fitted the strakes to the underside of the hull. They are shown in figure 34. This was the stage at which I tested it in my fishpond. It proved to be quite capable of lifting itself and eventually it lifted 11 pounds on 6 volts.
When I tried this I simply put it on the pond so that it could not escape and connected up to a 6 volt battery with no speed control. Immediately the craft lifted until air could escape very vigorously from the sides and the back. At the same time it set off forwards looking very much as though it might go quite quickly. I tried pushing down on the fan unit and a big force was needed to push it down at all. In fact a force of 5 pounds would only sink it by about ¼ ². The flow of air and water from the back was quite impressive.
I had a lift system that could lift but not a working hovercraft. I had to think where we go next.
On 17th September 2009 I recorded the following. “The euphoria has died down and now I have to think more seriously. It was clear that the lift system was quite capable of lifting quite a heavy weight and, on 6 volts, there was air bubbling out of the sides and the whole thing bobbed about as air formed bubbles under the hull. It was not at all what I thought it should be and I thought that I was in for a long haul getting the cushion to work.”
It is generally accepted that the lift system needs all the power that it can get and I had rather taken this on board. I decided to use a speed controller to see what would happen at lower speeds. I turned on the fan and watched the craft bobbing about and then reduced the fan speed. To my surprise, at quite a low fan speed, the craft set off forwards quite smoothly with no bubbling, just leaving a trail of popply water. The lift system works at probably less than ¼ of the maximum speed on 12 volts. I suspect that it could work on 4 volts.
This lift system is very effective. It can easily lift 9 pounds and I have since measured the power consumption and it is less than 30 watts and probably as low as 24 watts.”
I had clearly built a lift system and it was capable of propelling itself so the obvious thing to do was to see how it behaved on open water. This involved just letting it run freely with speed control but no directional control. I was agreeably surprised at the speed that this lift system could make.
The photographs in figures 35 a, b, c and d show the lift system on the move. It produced a bow wave and a system of waves generated by the return bend at the bow. I found these interesting in the context of article on “The Bow Wave” on this website. The photo 35a is, to me, quite remarkable and only possible because there was virtually no wind. We can see two seemingly separate wakes but in fact they are both generated by the bow. We can see in 35c and 35d a much reduced version of the bow wave produced by a swim headed bow in my article on bow waves on my website. I reproduce it here as figure 36. The lifting of the water ahead of the swim headed bow is directly similar what happens at the cylindrical surface that is the bow of this hovercraft. As much water goes down at the bow as goes up as shown for the swim headed bow but, for the hovercraft, there is no solid bottom in contact with the water and, instead of lines of alternate high and low pressure we get little waves that come out into the open at the stern.
The lift system is not lifting the craft high enough and the dam at the stern will have to be extended so that the craft lifts its bow clear of the surface. Figure 37 shows an enlargement of the lift system under way. All the air from the lift fan is flowing out of the gap between the dam and the surface and the dam will run at a height that suits the rate of flow from the fan. I thought that the dam could wait until I had some thrusters to drive the boat.
Note that the craft had acquired a deck. This was necessary because the bubbling that occurred during testing splashed water into the open hull and into the radio bay.
The side pods and the bow
Whilst testing the lift system it became evident that it was short of buoyancy. This was hardly surprising as the hull is only 1.25² deep. It needed side pods.
The photographs of the lift system running made it evident that the semi-cylindrical outer cover for the front return bend would not be a suitable shape for the craft when it has to cope with a ruffled surface to the water. Some sort of fairing over the return bend was needed. I fitted these to give the hull its final appearance as in figures 38a, b and c.
I intended to steer the craft by fitting two thrusters as far apart as possible across the hull and changing the speed of the fans to give an asymmetrical thrust to steer. This can be done by mixing on the right hand stick of the transmitter so that up-down gives speed control of both thrusters simultaneously and reversing and left-right speeds up the left and slows the right to give a left turn and vice versa.
For this system to be effective the thrusters need to be set wide apart. Figure 39 shows where they ended up. The thrusters were mounted on lengths of 1/2² by 5/16² beech bearer wood that were glued inside the sides of the hull. The fairings for the beech bearers were fitted before mounting the thrusters.
Having bought a GWS fan for the lift system it was an obvious next step to look for thrust fans in the GWS range. I knew the sort of thrust needed to drive yacht hulls and this is in the range up to say 16 ounces. So, when I found a fan that, at 8.4 volts produced 5 ounces and I thought that I might be able to increase this by perhaps 2 ounces, I opted to use two GW/ED64-150 fans. The only worrying point was that the manufacturers are not very confident that their fans will run on 8.4 volts without “fatal damage” if no heat sink is fitted. In the event they appear to run satisfactorily in my set-up with ventilation.
I have described here in great detail what must be done to get the best out of ducted fans used as thrusters. Now I had to do it. The decision to mount the thrusters on beech bearers meant that the unit as purchased needed a ring round it to take the bearers and to hold the intake and the tail cone. This I made from laminations of 3 mm ply by boring the ring and shaping the outer profile on a milling machine. I did not photograph a ring but one is shown in figure 40a with the intake glued to it and the frame of the unit inserted. Figure 40b shows the ring with the whole fan unit in place.
I rather like turning wood on a metal turning lathe and it was possible to make the intakes and the tail cones in wood using my lathe and making the support ring on my lathe and milling machine.
Figures 41a and b show a completed intake mounted on the ring.
The motor fairing is just a turning job but fitting the extensions by gluing plates to the fairing was not easy but manageable.
This left the tail cone and I have already said that I opted to make the exit diameter 2.22². This gave a long tapered bore to make with the need to make it accurately. It needed care and I bored it in lots or small steps and sanded the inside smooth. The result is shown in figure 12.
The thrusters were mounted and with the radio gear and the speed controllers fitted it was ready to test.
Development of the lift system
This took place in two stages. The first was a trial of the craft that led to modifications that, in their turn led to the need to think the whole design of the cushion again.
I took the vehicle to the lake and balanced it fore and aft and tried it. Figure 42 shows that it lifted and moved reasonably quickly but it was a displacement boat. The bow wave is clearly visible and the bow has not been lifted above the surface. It was clear that the dam at the stern that I had cut back by 1/2² would have to be restored to its full depth.
The steering worked but needed adjustment. I have to say that, in practice, the method of steering by changing the speed of the thrusters asymmetrically reduced to steering by slowing one motor or reversing it because the most likely running mode is for full speed. The result was a turn that was associated with a drop in boat speed although as I get used to steering this way the speed drop is not so great.
I reinstated the dam and tried the craft in the fish-pond. Air leaked out from under the boat everywhere but the bow lifted clear of the surface. My design had lifted everything to the surface and there were no seals anywhere It was clear that the underside would have to be modified. The craft would need deeper side-walls and deeper strakes.
Figure 43 shows the changes that were made. They amounted to sticking balsa strips to the existing strakes. The new arrangement can be compared with the original arrangement in figure 34. It can be seen that if the craft is lifted just enough for the outlet edge of the return bend to be at the surface the air will be trapped in the channels. I had found that, if the lift fan runs too fast, air escapes at the front and the side-walls are deeper at the front and taper towards the back.
On the strength of the performance with the high dam I settled to run at 8.4 volts and bought two NiMH batteries of 3,300 mAhr capacity.
It was a matter for another trip to the lake when on a windless day. The conditions were very gloomy.
The boat ran straight away with no ballast and lifted clear of the surface. There was no wind as is evident in figure 44 where is can be seen that the boat is not dipping its bow into the water, not making a bow wave and running quite cleanly.
Figure 45 shows the stern.
With the new set up the boat ran at full speed for 16 minutes on one battery. I regard this as satisfactory for most purposes but I thought that it should be possible to make other refinements.
The attitude of the craft varied with the speed of the lift fan and the speed of the thrusters. This may be inevitable because there is no conventional buoyancy to offset the moment of the three fans in trying to bury the head. It led me to think me that there would be something to be gained from having a moveable dam to alter the attitude of the hull although I did not see a mechanism for this and did not stop to analyse it. I did think that it might even be possible to mix the thrust stick with the dam to make the trim alter automatically. I thought that I might then be able to find the best combination of fan speed and dam height.
The moveable dam is shown in figure 46. It is just a ply plate cut to fit between the strakes and hinged with servo mounting grommets and screws. The operating mechanism is shown in figure 47 with its servo buried in the transom.
I thought that ideally this new dam should be tested in windless conditions. In the event I had one opportunity to test it in a wind high enough for half the wind-powered turbines about two miles away to be stopped. On the lake there was some shelter from the wind from houses and trees but it was not ideal for testing model hovercraft. Nevertheless I learnt a great deal.
The fitting of the moveable dam involved the addition of about 2.5 ounces at the stern. I fitted ballast to counter this but it was not necessary and had to be removed. Then the moveable dam turned out to be a major improvement and it was quickly evident that there is a best height for it.
The craft coped with the conditions very well. I had expected to hear the effects of the wind across the intakes but the craft runs so quietly that the noise of the fans gets lost in the general wind noise. The craft was designed for use on windless days and so this visit gave an opportunity to see how the lift system would cope with a swirling wind. I think that it did very well. The lake has vertical walls and this gives up/down popple and not waves. It seems to me that this should affect the seal of the cushion but it seemed not to matter although, when turning, the craft tilted into the turn and the “bow” started to dip into the water. This is shown in figure 48. I will think about this and try to find a cure for it. It may be tied in with another problem. The boat leaks air under the side-walls near the aft end as is evident in figure 49. This throws the whole design of the underwater parts of the craft into the melting pot. I need to stop and think about the side-walls, the strakes and the bow all over again.
Re-thinking the cushion
In order to start I really need to specify the conditions that have to be met if possible. The requirement that is central to the whole concept of the hovercraft is that the craft, when up on its cushion of air, should have the minimum contact with the water. For this design this means that, when the craft is moving on its cushion, the several strakes and side-walls should dip into the water by the least amount consistent with having an air seal. The finish of these wetted parts should also be good.
So we have to consider the way in which the air is trapped under the hull. First the bow must be clear of the surface so that it produces no bow wave yet, at the same time, there must be no leakage. The sideways leakage should be prevented by the side-walls. Clearly there must be leakage at the stern, indeed, all of the air should leave from under the hull through the gap between the moveable dam, the surface of the water and the side walls.
The first thing to consider is the bow. Figure 50 is of the craft before the moveable dam was fitted and it shows that it is possible to have the craft up on its cushion with no leakage and no evidence for contact between the bow and the water. The only surface wave to the side of the craft is caused by friction on the side-wall. It is hard to know how the seal at the bow is achieved but I have some pointers. As I have said, when I fitted the moveable dam I re-balanced the craft using lead in the forward ballast compartment. When I tried to sail it the head was obviously in the water and producing a bow wave so I removed the added ballast. Immediately the head lifted and the craft ran properly. So I had added about two and a half ounces at the transom and this had not upset the trim. I suspect that the false bow is resting very lightly on the surface and making a large area of contact as indicated in figure 54 and I will add to this argument later in this text.
Figure 51 shows the craft with its moveable dam on a disturbed surface and with leakage under the left hand side of the bow. In close-up it is evident that the air leaks forward carrying water with it and the water splits into four streams and surface tension pulls those streams into the shape like the web of a web-foot. It seems that cushions do not work as well on uneven surfaces as they do on smooth surfaces but that is not surprising. What is surprising is that the craft runs so well.
Now I need to look at the lift system more carefully. Figure 52 is drawn to scale and it shows the essential features of the lift system. Air enters from the fan through the divergent duct and passes round the return bend to emerge as a stream of air under the bottom of the hull. The flow is split into six flows in the divergence and the return bend and these six streams are kept separate by the five strakes under the hull that are of uniform depth from end to end. The side-walls are deeper than the strakes to give a sideways seal and I have made them taper in depth from front to back. There is a moveable dam at the stern.
I need to re-assess this design in the light of the most recent testing. Figure 49 shows a definite leakage under the side-walls towards the stern and this suggests that the tapering of the side walls was mistaken. I did it because I had noted that, when the craft is stationary with the lift system running, the main route for leakage is where the air first emerges under the hull. The taper appeared to give enough depth to seal off this leak and reduced the wetted surface until I fitted the moveable dam.
Now I think that I need to revise the shape of the side-walls. Look at figure 53. We can see that the hovercraft is leaving a hollow wake in the free surface and that the wake quickly fills as the surface rises. I have used figure 53 to help me to draw figure 54 where I have drawn a cross section of the water and the hull at the stern. In both views I have shown the level of the lake in blue and the likely shape of the actual free surface. It is clear that the hovercraft displaces water downwards just like an ordinary displacement boat. This is inevitable because Archimedes principle still applies and the craft must displace its own weight of water. However, in the case of the hovercraft, the shape is not determined by the shape of the hull. It is determined by a complex interaction between the inertia of the water, the way that that the pressure of the air changes as it flows under the hull and the consequent change in the depression of the free surface. The change in the depression alters the area through which the air flows and this slows the air and increases its pressure. The steady running condition is when the average depth of the depression is about 0.4². I have attempted to draw a profile for the water surface under the hull and it is shown in red and the maximum depression must be towards the stern and that maximum must be about 0.9². The side-to-side shape of the displaced water is like the red profile in the lower cross section and the air flows out under the dam and above the depressed surface.
We can now see that the addition of the dam is a major change. When I tried it in the fishpond I thought that the lift system needed even less power to operate it than it did with the fixed dam and that the operation of the dam appeared to alter the trim. The most noticeable change was that the craft was anxious to move forwards with only the lift system switched on. I am not sure why this happens but the change is that the simple fixed dam has been replaced by an angled plate. It is tempting to suggest that the fixed dam was offering an excessive resistance to motion but there is another point to be noted. The presence of the extended dam creates a convergent nozzle that exhausts at atmospheric pressure and this means that the inevitable pressure drop in the nozzle makes the pressure before the nozzle greater that it would be without it. Indeed the greatest pressure of the air flowing in the lift system occurs at the start of the dam. This means that the pressure under the whole hull increases and this accounts for both the improved lift and for the leakage shown in figure 49. In fact I did not think the lift system through to its end when it discharges at the stern. (This is a repeat of the same error that I made with the wind tunnel in figure 20 where I just let the air leave the fan unit.) The facility to adjust the dam is now seen to be important.
All this leads to the need to deepen the side-walls at the rear ends.
I can now return to the fore and aft balance of the craft. I said that the craft now runs without forward ballast and that the bow appears to rest on the surface to give a seal. The moveable dam, when set in the best position, alters the shape of the depression of the water under the hull and moves the centre of lift aft. As it does so the hull tends to topple forwards to give a seal at the bow. All three fans produce forward-acting forces that also tend to make the bow rest on the surface so somewhere there is a best running condition that might not be that same for all speeds.
There is more work to be done to find out how to get the best out of the craft on both smooth surfaces and on surfaces disturbed by the wind. I am not in a position to do this work so my design must stop here.
So what have I got? I have a model air-cushioned vehicle of modest weight of 7½ lbs that is about 36² long and runs at a very acceptable speed on about 110 watts of power for about 15 minutes. It steers well, runs acceptably in a stiff wind that has been made fluky by trees and buildings, and is altogether more interesting than a conventional displacement hull.
I have said that I cannot undertake the fine-tuning of the craft as it stands but further development is possible. I think that the next big step is to increase the thrust of the two thrusters just to find out how the lift system behaves at higher speeds. I would be looking in the first instance to construct new fans to go into the existing ducts to increase the size of the blades especially at the tips in order to increase their areas and hence the force on the blades. This may have a knock on effect of needing new motors of the same physical size. As a second line of development larger fans are an option and I suppose that a doubling of the thrust may be practical.
I simply have no idea how well other model hovercraft perform because I have never seen one that actually worked. I have seen silent video footage but saw nothing that I fancied. I simply started from scratch to see where a comprehensive knowledge of the relevant physics and the necessary skill to realise a design in hardware would take me. It has been a rewarding exercise and it will be useful in writing my textbook. It may be that others might want to use this design as a guide to designing a craft of their own. In my view the three bell-mouthed intake ducts are essential. The divergences in the lift system cannot be shortened to give a greater angle of divergence. The inside surfaces of the lift system must be smooth and joints must not create steps or ledges to disturb the flow. The splitters are essential. The thrusters perform efficiently because the convergent exit duct has the correct exit area and taper. It cannot be guessed at or be the subject of trial.
The square-sectioned bend on which the lift fan is mounted is hideous and I did consider giving it a fairing. There was no simple way to do this with the electrics in a box behind it and, in truth, at the speed of this craft, it would just be cosmetic. If I were to build another model I would re-think that bend and the back of the bell-mouthed intake to the lift fan to fair over the construction.
Ivor Bittle December 2009
Postscript -Programming the transmitter
As a result of fitting the moveable dam I found that the Spektrum transmitter has no proportional channel that can be used to control the dam. I was looking for a rotary or linear potentiometer like those on my ancient JR MC20 Tx or my old 40 MHz Futaba. This forced a change to the JR. It also forced me to look again at the mixing arrangements for the thrusters so that they could go ahead together and also be used for steering by slowing one thruster. It seemed to me that I should explain this system.
The system requires four channels and four free mixes. For those who put speed control and steering on the right hand stick, two of the channels will be aileron and elevator. The other two channels are really dummy channels and can be any two spare channels after two channels have been allocated to the lift fan and the moveable dam. I used the rudder channel and the gear channel.
The two thrusters will be plugged into the rudder and gear sockets of the receiver and the aileron and elevator sockets will not be used. We must now arrange for the rudder channel and the gear channel to respond only to commands from the aileron and elevator channels through the four mixes. You must either use the MIX ONLY command on both channels or set both the travels to zero using the ATV command.
Now we can feed signals to the rudder and gear channels from the aileron and elevator sticks. We need four mixes. They are :-
MIX 1 Elevator ® rudder with mix value 100%
MIX 2 Elevator ® gear with mix value 100%
This effectively makes the elevator control both the rudder and gear channels. If you set up the receiver and connect two servos to it in the rudder and elevator sockets you can check that these mixes are working properly. Moving the elevator stick should make both servos travel between their normal limits in the same direction. If the receiver were to be connected to the thrusters the speed controllers could be set up to give from full ahead to full astern for the full travel of the elevator stick.
Now we have to feed in signals from the aileron stick to reduce the speed of the right thruster to turn right and the left thruster to turn left. This means that we have to feed in a negative signal on one side only to each of the rudder and gear channels. This requires the other two mixes :-
MIX 3 Aileron ® rudder with mix value –100% on one side only, the other side being zero.
MIX 4 Aileron ® rudder with mix value –100% on one side only, the other side being zero.
This can be checked using the servos and, if the elevator stick is pushed right forward, operating the aileron stick to, say, the right should bring the appropriate servo back to centre and not affect the servo if the stick is moved to the left. The second servo should do the same but be a mirror image of the first. It may be necessary to alter the side of one of the mixes.
This is fairly straightforward so far and would be the same for all computer transmitters with four free mixers. But now we have to connect the thrusters through their individual speed controllers and speed controllers vary in their arrangements for setting up.
However the speed controllers operate, put the elevator stick in mid position and set the both thrusters to zero revs. Now move the stick to fully up and set both speed controllers to maximum revs forwards. Finally move the stick fully back and set the controllers to give maximum revs astern.
The set up is then finished. You might have to swap the thrusters over in the rudder and gear sockets.
My thrusters are virtually useless in reverse, they just make a hideous noise, but at least I have a puny force available should I make some silly mistake.