THE STAR SAPPHIRE ENGINE

An article by John L. Lumley
Willis H. Carrier Professor of Engineering, Cornell University
We are grateful to Professor Lumley for his permission to publish his excellent article.

THE STAR SAPPHIRE ENGINE by John L. Lumley 29th June 2001 The Star engine, of course, evolved from the 346 engine. In the late '40s, W. O. Bentley was hired as a consultant on the design of this engine. Bill Smith has verified this for a talk he gave to the W. O. Bentley Society. Bentley was at a loose end, since Lagonda had gone bankrupt (again) in 1947, and been bought by David Brown (along with Aston Martin). Bentley had been Technical Director at Lagonda since 1936, when the receivers had brought him in after the first bankruptcy (he had been at a loose end since 1932, when his own company had gone bankrupt.

The Star Sapphire Engine

The thirties were not a good time for the automobile business. In any event, it is said that at Armstrong Siddeley he was pushing for double overhead cams, which he had used for the Lagonda 2.6 L engine. The word on the street is, that a prototype engine was made in this form. So far as I am aware, no trace of this engine has been found, not drawings or castings or scrap. However, the story goes on, that the engine with DOHC was determined to be too expensive, and the chains driving the cams too noisy, for production. Mr. Bentley is supposed to have been politely told that his services would no longer be required, and the engine was designed by Fred Allard and his team at Armstrong Siddeley. Fred Allard was responsible for the design of the Siddeley Special engine. As we know, Fred used a high-mounted camshaft and short, angled push-rods as an economical way to get a nice hemispherical chamber and angled valves without the complexities and noise of a chain-driven DOHC layout.

The Star engine was designed by A. Rice, Chief Designer of Armstrong Siddeley Motor Car Division, Bristol Siddeley Engines Ltd. Using the 346 engine as a starting point, he had two major objectives: (1) the Sales department had to have a genuine 100 mph capability for marketing reasons, and (2) he wanted high mid-range torque to give good overtaking capabilities. Reliability and longevity would not have been stated as design objectives, but would have been regarded as implicit in the design of any Armstrong Siddeley engine.

The version of the Star engine that was finally manufactured was a relatively minor modification of the Sapphire 346 engine, using as many of the same parts as possible. The crankshaft, cylinder head casting (with minor modifications) and valve gear are common; the valves are larger in diameter, the displacement is greater, because the bore is larger, and so the block was new, as well as the pistons, with slightly heavier connecting rods. The cam has a lower lift. To accommodate the larger diameter bores, while using the same valve gear, the cylinder centerlines were displaced 0.1 inch from the crankshaft centerline to provide clearance for the camshaft. The larger bores have no water passage between adjacent pairs (1&2, 3&4, 5&6, called Siamesing, like Siamese twins joined at the hip), but the walls of the block were pushed out a bit, to provide larger water passages on the other sides.

We should keep a few numbers in mind, which roughly characterize the Star, the 346 and the Jaguar XK engines (which we will use for comparison). Note that the differences may appear to be small, but they are significant:

ENGINE	BORE	STROKE	DISP.	COMP. RATIO	P.POWER	@ RPM	P. TORQUE @ RPM
346	90mm	90mm	3.4L	7.0	150hp	5000	194ft lb @ 2000
STAR	97mm	90mm	3.99L	7.5	165hp	4250	230 ft lb @ 2000
XK	87mm	106mm	3.78L	8.0	220hp	5500	? ?

In order to understand the Star engine, so that we can compare it with other contemporary engines, and more recent ones, we will have to talk a little about engine design in general. Much of this material can be found in more complete form in Engines: An Introduction (J. L. Lumley, Cambridge University Press, 1999).

First, let's talk about friction. Roughly half of engine friction is due to the rubbing of the piston and rings against the cylinder walls. This scales with the average speed of the piston in the cylinder; that is, if we double the piston speed, it doubles the friction from the rubbing of rings and piston against the cylinder walls. The work required to pump the gases into and out of the cylinder is also included in the frictional losses, by convention. This is very small at wide open throttle, but is large at idle, when the throttle is nearly closed. This also scales with the average piston speed. The other parts of the frictional losses (the valve train, about 1/4 of the total, bearing friction, auxiliaries, etc) scale more with rotational speed, but can be satisfactorily scaled with piston speed (which is proportional to rotational speed in a given engine, but the proportionality depends on engine geometry). Because all pistons, rings and cylinder walls are made of the same stuff, regardless of the size of the engine, all engines operate in approximately the same range of average piston speed, between zero and 4000 feet per minute, or 20 meters per second. This includes tiny single cylinder model airplane engines (that sound like an angry hornet) and enormous marine diesels with pistons 3 feet in diameter.

Although 20 m/s is not a magic figure, hardly any engine exceeds it, and then only by a little. This is because the mechanical efficiency (the ratio of power available after subtracting all the frictional losses, to that available without the losses) which is in the neighborhood of 0.85 at low piston speeds, has fallen to roughly 0.6 by 20 m/s, and is falling rapidly at that point. Note that we are talking here of an engine with wide open throttle, whose speed is being adjusted by load - imagine going up a steep hill with the pedal on the floor, in too high a gear, for example, to achieve the low piston speed. Idling, with no load and the throttle nearly closed, is a different story - it corresponds to a mechanical efficiency of zero; the power output is small, because the air pumped in is small due to the closed throttle, and the amount of fuel correspondingly small, and all the power produced is expended on friction and pumping losses - no useful power is produced, only enough to keep the engine turning over. Idling in a passenger car engine usually corresponds to a piston speed of roughly 1.5 m/s.

The output of an engine is often given in terms of horsepower per liter. 50 HP/L used to be regarded as a good figure, although some performance engines now produce in the neighborhood of 100 HP/L. The Star engine produces about 49 HP/L. This way of comparing engines arose because of the formula limitations in racing, which specified an upper bound to engine displacement. There was thus a push to get the greatest power per unit displacement. However, power/displacement is not a very useful way to compare engines. You will have to take my word for it (I cannot convince you without boring you with equations), but I can increase the power/displacement of an engine by shortening the stroke and increasing the bore, leaving everything else (the displacement, in particular) unchanged. This is why racing engines have changed from a stroke/bore ratio of nearly 2 in 1912 to our current value of approximately 1. A value of one is referred to as "square," and less than one as "over square."

A much more sensible measure of an engine's performance is the power/piston area. I will justify this in a minute, after I have introduced another quantity.

The brake mean effective pressure (bmep) is a measure of engine performance used by designers. The bmep is the steady pressure which, applied to the piston crown throughout the power stroke, would produce the observed power. The bmep is determined by the product of all the efficiencies (volumetric, mechanical, combustion, thermodynamic) and the inlet density, the heating value of the fuel and the fuel/air ratio. Hence, it is independent of the
displacement of the engine, the number of cylinders, the engine speed, and so forth. This is a figure used by engine designers to compare quite different engines. Current values of the bmep are of the order of 140-160 psi (0.7-1.0 MPa) for most normally aspirated engines (i.e.- no turbocharger) at wide open throttle - it falls in proportion to throttle opening.

The various efficiencies deserve a mention. We have already mentioned the mechanical efficiency. The thermodynamic efficiency is the ratio of the work actually produced (in the cylinder) to the energy available in the fuel. A lot of this energy leaves in the exhaust gases, and in the cooling water. The volumetric efficiency is the ratio of the mass of gas that actually enters the cylinder to the mass it could contain if filled at the density in the inlet manifold. The first is less than the second mostly because of the pressure drop across the valve aperture. The combustion efficiency is the ratio of the fuel that is actually burned in the cylinder, to the fuel present in the cylinder - some of the fuel present escapes burning for various reasons, and goes out with the exhaust gases.

Now, why power/piston area? Again, no equations, but bmep, piston speed and power/piston area are not independent - power/piston area is proportional to bmep times piston speed (with a numerical factor). Hence, any two out of the three are sufficient to specify engine performance.

I am including here a figure showing a large number of engines from aircraft, railcars, and automobiles. The original of this figure is from C. F. Taylor's wonderful book The Internal
Combustion Engine in Theory and Practice (The MIT Press, 1977, 1985). I have brought the axes up to date by including piston speed in meters per second, and bmep in mega Pascals, and I have added the various points indicated by numbers from 1-13, which are the US performance engines for 1997. In addition, the small 0 corresponds to the Jaguar XK engine in 3.8 L form, and the larger star and circle belong to the Armstrong Siddeley Star and 346 respectively. The version without the Star and 346 appeared in Engines: An Introduction.

Note that values of constant piston speed run vertically, and bmep horizontally. Values of constant power/piston area run diagonally. Any two are sufficient to locate an engine. For example, the numeral 3, which is a 1997 Ford Mustang, is found at a piston speed of about 13 m/s and a bmep of roughly 130 psi, corresponding to perhaps 2.7 HP/in2.

The numerals 8 and 11 are turbocharged; all the other 1997 engines are normally aspirated. Numerals 3 and 9 have two valves per cylinder, while all the others have 4. The engines labeled "passenger car" are the US 1954 domestic fleet.

We can see that the Star performs a little better than the 346. Both of them perform about as well (by these criteria) as the best of the 1954 US family cars, and about as well as the 1997 Mustang. To produce performance significantly outside the dashed line designating "passenger car," you need a high compression ratio and manifold tuning, which most of the 1997 performance cars have. The Armstrong Siddeley engines have neither.

A word about manifold tuning: as the valves open and close, the gases in the inlet and exhaust manifolds start and stop, and compression and expansion waves bounce back and forth. By proper design of the manifolds, you can arrange to have the gases bounce at just the right time to cause high pressure at the inlet valve just as it is closing, pushing into the cylinder perhaps 10% more gas than would fit at the average manifold density.

So, is that the end of the story? The Armstrong Siddeley engines are equivalent to a good 1954 US family car engine? No, there is a lot more to it than that - this is just the beginning.

Note the Jaguar XK engine, which is over at the right edge of the figure, at a piston speed of about 19 m/s. The bmep and power/piston area may be a little exaggerated by the factory figures, but it is clear that there is some basic difference between the A/S engines and the Jaguar engine. In fact, the difference is that the Jaguar engine has enormous valves that are opened nearly as wide as possible, and the Armstrong Siddeley engines do not. To understand why this matters, and why it was done, we have to talk about the gas flow through the intake valve.

The performance of any engine is limited when the speed of the gas flow through the intake valve aperture reaches the speed of sound at some point during the intake stroke. The ratio of the gas speed to the speed of sound is called the Mach number, and this condition at the intake valve opening corresponds to a Mach number of 1. When the Mach number in the valve opening reaches 1, we say the flow is choked. This is important because information travels in a gas as a pressure disturbance, which travels at the speed of sound. When the gas is flowing through the valve opening at the speed of sound, so that the flow is choked, information cannot travel upstream against the flow. For example, the flow in the intake manifold will not know if the pressure in the cylinder has dropped, and that more gas could be accommodated there. When the flow in the valve aperture is choked, the flow in the manifold will not change no matter what happens in the cylinder.

The engine designer arranges the size and lift of the valves so that choked flow does not occur at any point in the intake stroke until the engine speed at which the designer wants peak horsepower to occur. When choked flow begins to happen at some point in the intake stroke, as engine speed increases it will happen for a larger and larger part of the intake stroke, the volumetric efficiency will fall rapidly, and the power produced will fall. This choked flow is the usual reason for the occurrence of peak horsepower where it occurs (although various other effects are designed to happen at about the same time).

Now, the open area of a poppet valve increases as the valve lift increases, up to a point. For small lifts, the valve opening is the tightest constriction in the intake flow. At a lift equal to about 0.28 of the valve diameter, however, the open area stops increasing, because it is then equal to the area of the intake valve port. Now the port diameter is the limiting area, and there is no point in lifting the valve more, since it will not increase the available flow area.

Recall the XK engine. The intake valve diameter, as a fraction of the bore, is 0.536, while the Star has a value of 0.471, and the 346 a value of 0.480. The valve lift in the XK, as a fraction of the valve diameter, is 0.214, while in the Star it is 0.161 and in the 346, 0.188. It is clear that the XK has much bigger valves, that open farther. However, none of these is the truly significant value. What we want is the valve open area divided by the bore area (The Mach number through the valve opening is directly proportional to this). For the XK engine it is 0.246, while for the Star and the 346 respectively it is 0.143 and 0.173. The ratio of the XK to the Star is 1.72, and to the 346 1.42. That is, relative to its bore area, the XK has 72% more valve opening area than the Star does. That makes an enormous difference, and explains why the XK engine reaches its peak horsepower at 19 m/s instead of 13 m/s. Other things being equal, an increase in piston speed for peak power corresponds to a proportional increase in peak power. This is not quite true, because the mechanical efficiency is dropping, so that the peak power rises a little less than proportionally.

If the Star camshaft were modified to provide a lift of 0.504" instead of the stock 0.29," which would be the maximum 0.28 of the valve diameter, the Star would have had the same valve open area per unit bore as the XK, and the engine would have produced nearly the same horsepower as the XK engine at a similar piston speed. It does make one wonder why someone has not ground a high-lift cam for the Star engine. The wide-open valves would probably have missed the piston crowns in normal operation, but would have required a cutout in the crown to be immune to timing gear failure. What a racing engine this would have made at the time (it would be regarded as a little heavy by present standards)!

Why did the Star engine designer deliberately keep the valve lift small , limiting the engine's maximum performance? The same question can also be asked about the 346, which is similarly restrained, although not as much as the Star. After all, in other respects the engine is designed for high speeds. The 346 is square, and the Star is over-square, to keep the piston speeds low.

Most of the information about flow through valves was developed during and just after the second world war from about 1942 to 1947 by the US NACA (National Advisory Committee for Aeronautics), in connection with aircraft engine development. It would presumably have been made available to the designers at Armstrong Siddeley as part of Lend-Lease. We have to assume that the choice of valve lift is deliberate.

There are two basic reasons to keep the valve lift small. The first is to limit stress. With engine maximum performance limited by the valve lift, the maximum piston speed and bearing loads are kept relatively low, and the useful life of the engine is greatly extended. Although piston rings operate in hydrodynamic lubrication through most of the stroke, and hence never touch the bore, at the top of the cylinder at the beginning of the power stroke, the rings are operating in boundary lubrication, so there is metal-to-metal contact, and wear takes place. Keeping the piston speeds low keeps the connecting rod and piston, which is proportional to the square of the engine speed. Limiting the engine speed limits the bearing loads. However, the connecting rod big end bearing area is a about 6% larger on the Star than on the XK, so that the Star could have stood up to the higher engine speed of the XK. We have to conclude that the limitation of performance by a small valve lift results in very conservative loading of the Star engine, leading to very long life. This is why the engines last 250,000 miles without complaint.

In passing, we note that the main bearing area of the XK engine is nearly double that of the Star. Inertial loading on main bearings in a six-cylinder engine is relatively small, since the loading reverses every sixty degrees of crank rotation, and the crank is heavy. Hence, the primary main bearing loading is due to the combustion, and will be approximately the same for the XK and for the Star; it will scale with the bmep. It will also not change much with engine speed, since the bmep does not change much with engine speed. We have to conclude that the XK main bearing area is excessive and unnecessary, and that the Star main bearings would probably hold up well if a high lift cam were used. The A/S designers were aircraft engine designers, after all - they knew what they were doing.

We have examined one reason for small valve lift - conservative loading. The other reason for small valve lift concerns turbulence in the cylinder just before ignition, which has a profound influence on flame propagation, resistance to knock (pinging in the US, pinking in the UK) and reliable ignition. To understand this, we have to talk about air flow in the cylinder during the compression stroke.

It was known from the first decade of the century that a hemispherical combustion chamber with overhead valves with a large included angle, like the 346 and Star, produced better performance, but no one had much idea why. This form of combustion chamber allowed larger valves, and we have seen that that is a good idea. However, the improved performance went beyond that - such engines had slightly higher thermodynamic efficiency, were resistant to knock when run on fuels of low octane rating, and almost never misfired, even on very lean mixtures.

It was not until 1939, on the eve of the second World War, when the US NACA was gearing up for aircraft engine production, that the mechanism was understood. The flow through the intake valve (for the usual orientation of the intake manifold) comes in across the cylinder and at a downward angle, so that it travels down the opposite wall, across the piston crown, and up the adjacent wall. This is known as tumble. This tumbling motion, formed during the intake stroke, continues as the piston rises on the compression stroke. In fact, it gets stronger, drawing energy from the compression of the gases, in a mechanism related to the spin-up of a figure skater when she pulls in her arms. As the piston nears the top of the stroke, just before ignition, the shape of the volume in the cylinder is getting very flattened, and finally the tumbling motion known as turbulence. This motion increases average flame speed, reducing the time for combustion (which increases the thermodynamic efficiency and hence fuel mileage), reducing the time that the end gases (the gases not yet ignited) are sitting waiting for the flame front to arrive, and hence reducing the time for the autoignition reaction to take place (reducing knocking), and greatly improving reliability of combustion in lean mixtures (when the flame speed is otherwise quite low).

What does all this have to do with low valve lift?, I hear you say. Well, the amount of tumble (and hence, the intensity of turbulence) depends on the velocity of the gas entering the cylinder, and that depends on the lift of the valve. If the valve is open wide, the gas velocity will be low, while if the valve is not open very much, the gas velocity will be high, the tumble velocity will be high and the turbulence level will be high. All these gas velocities are measured relative to the average piston speed. An average hemi head has turbulent gas speeds about equal to the mean piston speed, its small valve lift, might have turbulent gas speeds 72% larger(based on the area ratio), with corresponding improvements in all the effects. These things were probably not understood in detail in 1956.

The effect of turbulence on flame speed and its related effects are primarily important at relatively low engine speeds and partial throttle. At wide open throttle, near peak power, fuel consumption, knock and misfiring are not usually a problem. Hence, in addition to a conservatively loaded engine, A/S designed an engine that is particularly reliable and well-behaved at lower speeds and partial throttle.

A/S were concerned about the behavior of the Star engine at low speeds in another respect, also. Because the Star was to have a Borg Warner automatic gearbox, it was essential that the engine have good torque at relatively low speed. The gearbox has only three forward speeds (although it does have a hydraulic torque converter). If the low speed torque is not high enough, the gearbox will have to shift often. The Star engine has an only slightly increased peak bmep of 142 psi at 2000 rpm, compared with the 346 peak bmep of 140 psi at 2000 rpm (with twin carburettors). We will return to this small difference in a moment. With the larger displacement, however, the peak torque (proportional to bmep times displacement) rises to 230 ft lbs at 2000 rpm for the Star, against 194 ft lbs for the 346 (with two carbs) - almost a 20% increase. As we have seen the peak power is not up much (165 against 150 with twin carbs) because the Star is choked at high speeds, due to the small valve lift.

To explain the slight increase in bmep, we should talk about valve timing. With the small valve lift in the Star, lower than the 346, if the valve timing had not been changed from the 346, the peak bmep would probably have been lower than the 346, due to the pressure drop across the valve aperture, leading to a reduced volumetric efficiency. To understand this, we have to discuss the effect of valve timing. Let's look first at the 346.

Here is the standard 346 valve timing: the inlet opens (IO) 8 CAD (crank angle degrees) BTC (before top center), and closes (IC) 62 CAD ABC (after bottom center). The exhaust opens (EO) 46 CAD BBC and closes (EC) 18 CAD ATC. There are two things to notice about this: first, the inlet and the exhaust are both open at the same time around TC (called overlap), in fact for a total of 26 CAD. Second, the inlet doesn't close until the piston is well on its way up the cylinder on the compression stroke.

The overlap around TC is designed to make use of the outrushing exhaust gases, which can drag fresh charge in through the intake valve. The late closing of the intake valve makes use of the inrushing fresh charge, which does not want to stop - leaving the valve open longer lets the gases continue to rush in until the rise in pressure stops them, ghetting the most charge possible into the cylinder.

Both these effects clearly work best when the engine speed is higher, and the gases are moving faster. At low speeds, the slow-moving gases do not have much inertia, and the effects are small. Worse, at partial throttle, the pressure in the inlet manifold is low, and when there is overlap, the low pressure drags exhaust gases back into the cylinder, leading to poor idling. The amount of overlap, and the lateness of closing, is usually adjusted according to the speed at which the engine is expected to operate; a racing engine willtypically have very large overlap and very late closing, and will not idle at all. Modern engines, with variable valve timing, or cam phasing, can change the amount of overlap and lateness of intakeclosing according to engine speed.

The 346 has moderate overlap and moderately late closing, consistent with reasonably good (though not perfect) idle. This is effective probably in the range above 4000 rpm. In the Star, the low lift with this valve timing probably resulted in somewhat low bmep at 2000 rpm. The Star valve timing was changed to IO 6.5 CAD BTC, IC 48 CAD ABC, EO 48 CAD BBC, EC 6.5 CAD ATC. This way, the overlap is reduced to 13 CAD, less than half the 346, and the late closing of the intake valve is reduced to 2/3 of the 346 value. Since the designer is much more concerned with low rpm performance, this does just what he wants - the bmep at 2000 rpm comes up just above the 346 value; the bmep drops quite a bit between 4000 rpm and 5000 rpm, but the engine is not intended to run there. And the engine idles much better.

That is essentially all that can be said about the design of the Star engine from the point of view of power production. However, the cooling systems of the Star (and the 346) are notable. In most automobile engines, the cooling water arriving from the radiator enters the block and circulates around the cylinders, flowing up into the head through holes in the head gasket, and out to the radiator again from the front of the head. This is not the most effective sequence. The hottest part, in greatest need of cooling, is the head, the cylinder walls being relatively cool. Look at a motorcycle engine - the size of the cooling fins indicates the relative need for cooling of the various parts. The fins are biggest around the exhaust valve seat, somewhat smaller elsewhere on the head, and taper down the cylinder barrel. Bear in mind that in a water-cooled engine the cold water returning from the radiator gets warmer as it circulates and absorbs heat from the engine. It should go first, when it is coldest, to the exhaust valve seats, then to the rest of the head, then to the cylinder walls, and thence back to the radiator. Most automobile engines are connected backwards. The manufacturer's position is, that this is cheap, and works well enough. What the hell. Armstrong Siddeley, bless them, send the cold water first through a distribution pipe at the top of the block, which squirts cold water on the exhaust valve seats through holes in the head gasket; the water then circulates through the head, down into the block and then back to the radiator, exactly as it should. Such a cooling system improves knock resistance, because the head is cooler (and the reaction in the end gases proceeds more slowly), and reduces friction because the cylinder walls are warmer. A cooling system like this is almost unique. The recently (1992) completely redesigned 5.7L small block V8 intended by Chevrolet for the Corvette has a much-touted "reverse flow cooling system," which does what the Star cooling system does. The designer goes to great lengths explaining the advantages of such a cooling system, but gives no indication that he is aware that Armstrong Siddeley preceeded him by some 42 years.

In summary: The Star is a conservatively designed engine intended for long service, highly resistant to knocking and misfiring, relatively efficient for its day, optimized to produce high torque at low speed rather than high power at high speed. A high lift cam would probably turn it into a monster.