Wayne Cox (w...@erinet.com)
from "Motorcyclist" magazine and posted on DIY-EFI (copyright - what's that?)
Motorcyclist June 1996: Cylinder Head Tech
At first, the task of clearing and recharging the cylinders in a high-speed,
four-stroke engine seems impossible. Such processes need time, and it's hard to believe
there's enough available for this one, which faces many impediments and is crowded into
the merest fragment of a clock's tick.
The intake stroke lasts for 180 degrees of crank rotation, which is only three-thousandths
of a second at 10,000 rpm. Camera shutter openings are as brief., but light has no mass
and moves at 950 million feet per second. Air's mass makes it lag, and it hits a sonic
wall about 1100 feet/second, with localized shock waves further blocking the intake ports
at much lower air speeds.
Yet cylinders get filled-with efficiencies sometimes exceeding 100 percent-without
mechanical supercharging. This is possible because the intake process actually begins in
the preceding exhaust stroke and extends far into the following compression stroke. We've
methodically learned to make the pesky effects of inertia work for us; and minimized the
bad effects of problems that cannot yet entirely be solved.
On a cylinder head's intake side you have only atmospheric pressure, 14.7 pounds per
square inch at sea level, working to stuff air into the cylinder. No matter how hard the
descending piston tries it can't pull air in behind it. It can only create a space for
atmospheric pressure to fill.
It's a different story over on the outlet side, where a pressure close to six atmospheres
exists when the exhaust valve opens to begin the event called "blow down".
Further, after blow-down, pistons mechanically force exhaust products from the cylinders,
and do so against the resistance of undersized valves, badly designed headers or steel
The more important exhaust event is the high-velocity shove the rising piston gives
exhaust gases during the exhaust stroke. The shove peaks at maximum piston speed (in most
engines occurring a little less than 80 degrees of crank rotation before the piston
reaches top dead center), where it suddenly gets yanked to a stop. But the momentum of the
gases in the exhaust pipe continues, leaving behind a partial vacuum. This starts the
air/fuel mix above the part-open intake valve moving into the cylinder before the piston
begins it's intake stroke.
Engines benefit from exhaust-augmented intake flow in two ways; an obvious advantage is
that it gives the too-brief intake period an early start. The second effect, less obvious
but also important, is that combustion chamber cross-flow during valve opening overlap
(the period during which both intake and exhaust valves are open) clears residual exhaust
gases, which slow combustion, depress power by displacing part of the fresh charge, and
can require some weird kinks in the ignition advance curve.
Exhaust systems primarily aid intake flow by their manipulation of the combustion
"sound wave". A sound wave creates a disturbance ahead of it and leaves one
behind; such "positive" waves bursting from the exhaust port are followed by
negative pressures. When the strongly-positive exhaust wave emerges from the end of a
pipe, it leaves behind a negative-pressure tail, which then reflects back toward the port.
If the length of the pipe is right, the negative wave will arrive back at the exhaust
valve as the piston reaches TDC, thus further assisting in clearing the combustion
Sound waves are reflected by any cross-section change in the duct in which they are
traveling. The sawed-off end of a pipe is one such change; the closed end of a pipe is
another. The difference is that increases in section invert the wave while reflecting it,
changing positive waves to negative and vice-versa; section reductions reflect the wave
with the same sign.
While speaking of sonic waves, I should caution you about confusing their behavior with
that of the media in which they travel. Like all sound-conducting media, air has mass and
the other properties of matter. Sonic waves are by contrast, purely energy and thus follow
an entirely different set of rules. Such waves make zero-radius 180 degree turns and
reversals without delay or loss of strength.
Plain pipe ends do a poor job of returning the energy of an emerging sound wave, which is
why horns have flared open end-to get better energy recovery and thus amplitude.
Megaphones, the exhaust pipe horns known in engineering as diffusers, are vastly more
efficient in this regard. Racing two-stroke engines expansion chamber exhaust systems have
elaborate blow-down diffusers, because of their heavy reliance on this vacuum-cleaner
effect to pull air through the transfer ports.
Four-stroke engines seem perfectly happy running with plain parallel-wall pipes, though
engines developed for megaphones have to be reworked to function well without them.
Harley-Davidson's famous racing chief, Dick O'Brien, never was totally convinced that the
megaphones used on the "low Boy" KR's did anything but make noise. At the time I
was sure he was missing something, but now I believe his reservations were valid.
Oddly, the 45-degree cut-off at the end of KR straight pipes did coax a tad more power out
of H-D's cranky old side-valve engine; O'Brien was at a loss to explain this oddity. I
tried a 90 degree cutoff once, and found the KR didn't like it. No coherent theory I've
heard or conceived explains why that should have been so.
It now appears exhaust pipe diameter, meaning gas velocity in the exhaust system, is more
important than sonic wave activity. Actual gas velocities vary in ways tough to grasp and
impossible to calculate, but the nominal speed is easy to figure and provides a useful
rule-of-thumb: simply multiply piston speed by the ratio of cylinder bore and pipe areas.
Nominal gas speed were well below 200 feet/second in most vintage bikes, but in the AJS 7R
of the 50's it was up to 220 feet/second. By 1972 the small diameter pipes on H-D's XR750
raised that engine's exhaust velocity to just above 300 feet/sec. The Triumph 650 TT
Special I used to set a Bonneville record (and acquire an abiding dislike of Wendover,
Utah) years ago also had small pies and 300-plus exhaust gas speeds. It had 1 3/8-inch
pipes, which almost everyone thought too small. My slide rule said they were the right
size, and the larger-diameter pipes we tried slowed the bike.
Gas velocity is even more important over the engines intake side, where it packs air into
the cylinder between the intake stroke's ending and intake valve closing. This is crucial,
since with high-speed engines there is a significant lag between the piston beginning the
intake stroke and the flow of air into the cylinder. Outflow in the exhaust can pull air
across from the intake to give the intake process a head start, but cylinder pressure
still precipitously falls through the first half of the intake stroke. Air simply can't
keep up with the piston, which at 9000 rpm in the XR750 goes from it's stop at TDC to 80
miles per hour in 1.5 inches, reaching that speed in 0.0014 seconds.
Fortunately, the air inertia that delays air/fuel inflow causes it to crowd in at the end
of the intake stroke, and beyond. The XR750's intake ports are small enough to raise the
nominal gas speed to 370 feet/second, which gives it plenty of momentum. This is why
intake valve closing is delayed for many degrees after the piston has finished it's intake
stroke and begun compression. Closing the intake valve while air is still flowing into the
cylinder, or closing it after flow reverses, gives less the best power. You have to close
the intake valve(s) just as the inflow slows to a stop, thus trapping the greatest weight
of air/fuel mixture in the cylinder.
Serious tuners need some means of shifting cam timing ( in increments no coarser than 1.5
degrees) to let them experiment their way to the optimum intake closing. This is usually
done with multiple oversize bolt holes in the driven cam sprockets and offset bushings,
although my old Aermacchi required woodruff keys with a sideways-jog at the shaft and
timing gear join to shift camshaft phasing.
High-performance engines' intake valves close typically 60 to 80 degrees after the intake
stroke ends and the compression stroke begins, so you know gas inertia is playing a major
role in cylinder filling; if it didn't there'd be no need to delay intake closing, and no
sensitivity to the timing of that event. None of the other valve actions-exhaust opening
or closing, or intake opening-are nearly as important.
Flow benches can be used to blow a lot of smoke up your shop coat when you're looking for
horsepower. You can always make air flow numbers rise by increasing valve head diameter,
or by enlarging the passages leading from the atmosphere. But higher air flow numbers do
not necessarily translate into more power, as many in the engine development field
(including yours truly) have discovered.
Mercedes-Benz made the big-port mistake with the design of its awesomely complex
eight-cylinder M196 GP car, which had desmo valve actuation and intake ports the size of
drains. They found themselves being out-horsepowered by the British Vanwall, with an
engine that was virtually four Norton 30M Manx Cylinders and heads bolted to an aluminum
Rolls Royce armored car crankcase.
Ford's 1960's four-cam V-8 also had huge intake ports, and while it turned more revs than
the Offy four-banger engines then dominant at Indianapolis, it was no better than a match
for them. When given an early peek at the Indy Ford's cylinder-head castings, I expressed
the thought that its ports might be too big. Ford's engineers were too polite to tell me
how absurd they considered my remark to be, but their expressions made it plain. I was too
polite to send them an "I told you so" note after Dan Gurney sent one of the
engines to Weslake Engineering in England, where it's intake ports were made smaller and
its output got bigger.
Ford's engineers were then vastly ignorant of the world beyond Michigan's borders. They
had no idea Harry Weslake and Wally Hassan (who created the very successful
Coventry-Climax racing engines) had learned years before not to take too literally what
the flow bench said. They were narrowing intake ports to provide nominal gas speeds in the
range of 350 to 400 feet-second, making good use of the fact that kinetic energy packing
air into the cylinders increases with the square of it's velocity.
Harley-Davidson's experience with the highly successful XR750 should have kept it from
making the big-port error in the CR1000. Yet, that's exactly what it did: the VR's intake
ports were made so big, nominal intake velocity was down at 200 feet/second, which may
explain why it's proved sadly inferior to engines that do not test nearly as impressively
on the flow bench.
Grand prix car engines represent the pinnacle of four-stroke development. Formula One's
designers are spinning 3.0 liter V-10 engines up to 15,000 rpm's and getting close to 800
horsepower. Ford's GP Zetec V-8 is doing the same with 375cc cylinders, which implies that
it's possible to build a 750cc V-twin that will make nearly 200 horsepower.
Cosworth Engineering's Keith Duckworth was the creator of the modern high-output
four-stroke. Casting aside tradition, Duckworth combined large-bore short-stroke cylinders
with narrow-angle valves and a compact combustion chamber. He didn't originate the use of
high-intake port velocities to ram-charge cylinders, but he and those he's influenced now
design for nominal intake speeds approaching 450 feet/second.
Of course, there's a lot more to cylinder gas exchange than port velocity. But unless
you've spent eons dragging air through ports, manifolds, etc.,, at a flow bench, you
probably have no real understanding of what aids flow and what slows it. If there is any
rule for the inexperienced to keep in mind it is that everything a reasonable intelligent
person should intuitively believe to be right will probably be totally wrong.
Take valve shape for example, these days typically an unstreamlined disc on the end of a
stick. Your eye will tell you the shape is horrible, an example of how we've fallen into
decadence since the days of those British power plants with beautiful, deeply tuliped
intake valve. Then you hit the flow bench and find that the one with all the loveliness of
an overgrown nail is better at all lifts. And then you repeat the experiment with another
port and find it responds better to a tuliped valve. Some ports are like that, by virtue
of slightly different interior contours or different valve angles.
Or you can try valve seating surfaces-maybe someday you can tell me why sharp edges are
better here than rounded ones. The worst valve I ever tested was one I made the mistaken
belief my eye could judge how air would behave between the valve and seat. I ground a
valve head with a radius instead of a flat where it seated, along with a similar-shaped
grinding stone for the seat. Testing this idea required tons of work, yet my streamlined
valve and seat combination was worse at all lifts than the typical series of abrupt,
You'd think that getting the valve completely out of the way while flow-testing ports
would let the air really whistle on through. But peak flow almost always occurs with the
valve in place, at a lift equal to about 30 percent of valve diameter. And this is with a
manifold and carburetor in place, and a cylinder between head and flow bench receiver (
the cylinder's adjacent walls can significantly influence flow around intake valve heads).
Multiple valves (more than two per cylinder) actually offer little or no real valve-area
advantage. You can prove this to yourself by drawing circles representing valves inside a
larger circle signifying the cylinder bore. Unless you fudge the whole thing with
unrealistic provisions for valve seats, clearance around the valves, etc., the total for
valve head areas is about the same for two, three or even five valve layouts. The benefit
lies in the fact that total head area counts only at or near full lift: at lesser lifts,
flow is largely limited by the valve seat ring area, really more a function of the total
of valve circumferences than area. Viewed this way, multiple valve layouts are better,
though only Yamaha has found any gain with more than four valves.
Air flow in ports takes paths totally unlike those you would normally envision, unless you
happen to have an abundant knowledge of compressible fluid dynamics. In your imagination,
air may move in orderly lines of travel, with particles marching along the roof of the
port staying high, those on the floor staying low, and all traveling in neat, linear
streams. The reality is a very different matter.
When flow in a duct (an intake port, for example) arrives at a bend, it loses any
semblance of orderly behavior. Particles on the inside of the bend travel the shortest
distance (offering the least resistance to flow), so they tend to maintain speed in the
downward turn to the valve seat. But flow in the top of the port slows relative to the
floor, creating a large velocity gradient. Pressure in a moving fluid varies inversely
with it's speed, so the velocity gradient creates a lower pressure at the port floor than
at it's roof. This differential causes air at the sides to move upward and the midstream
air to move down, with the resulting flow stream made to divide into to contrarotating
vortices where the port bends. Add to this the invisible "smoke ring" vortex
forming beneath the opening intake valve and you have enough disorder to confound even the
best of minds (or computers).
Port and valve configuration (both shapes and angles) can profoundly influence combustion
efficiency as well. Jack Williams AJS 7R made it's best power with an intake port shape
that compromised flow in favor of creating more combustion chamber swirl and redirecting
incoming fuel droplets away from the cylinder walls. I am reliably informed that Keith
Duckworth has settled on the intake valves leaned 15 degrees from the cylinder axis, and
ports at 30 degrees from the valves in a similar trade-off between flow and combustion.
Intake flow influences combustion because both carburetors, and fuel-injection nozzles
deliver fuel in liquid form. The best you can hope for is a fog of droplets small enough
to stay suspended in the air while evaporating; big drops are centrifuged out of the air
stream, splatting against the intake port and cylinder walls, which is bad for power, fuel
efficiency and emissions. Fuel can't burn until it evaporates; if you have raw fuel still
trying to burn when the exhaust valve opens, it goes out the pipe, wasting your money and
polluting the air.
My experience (not the final word on anything even for me) is that the biggest improvement
in flow from a change in port shape- with the least port enlargement and resulting
velocity loss- is obtained by widening the port floor upstream from the valve seat. Air
likes to take the most direct route, and the more you ease that route the better flow
becomes. Shaving metal out of the lower sides of the ports bend (making a D-shaped
cross-section, with the port floor on the flat side has in my tests shown big flow
improvements in sharply bent ports.
Smoothing intake flow (thereby minimizing the turbulence of the main flow stream) is best
accomplished by making sure the port's section area decreases all the way from the carb
inlet to the bend above the valve seat. The small diameter, high-velocity section of the
port needs only a slight convergence of 1.5 degrees included angle, which doesn't sound
like much. But a 12 inch section of aluminum pipe taper-bored for a 1.5 inch inlet and a
1.498 inch outlet flows better than a parallel-wall pipe, and a lot better than air going
from the cones' small end to it's beg end. Sound waves love a divergent duct, air flow
I'm not convinced that polishing a port's interior surfaces to a mirror finish does
anything but look good. The problem here is that while we know there's a degree of
roughness beyond which flow suffers, we can't agree on the limit to which polishing helps.
One those rare occasions when I do porting myself, I settle for a smooth but not polished
finish. If I were in the head porting business like my long-tie friend Jerry Branch, I'd
put a spit shine inside the ports and combustion chamber, just as he does. The way Jerry
does it, his customers never have to wonder if the ports are smooth enough.
Jerry has discovered that some ports flow better if he cuts tiny slots across the floor of
the bend upstream from the valve. The slots apparently act as turbulence generators that
energize the air and make it stick to the port floor, following the bend more closely.
That's the theory anyway, though like so much we believe about port air flow, it's
arguable because air hides is secrets behind a cloak if invisibility.
In time, we will know a lot more about the details of flow in and out of cylinder heads.
For decades, researchers have used smoke, pinwheels, dye droplets, etc. in their attempts
to see what air is doing. The water-anaolgy method, where water substitutes for air and
flow is made visible with fine bubbles or aluminum particles, is still used in many labs.
But the growth of mystery-dispelling technologies has recently brought doppler-laser
metering and computer imaging to the field. Maybe one day soon we'll learn why the things
a century of experience has taught us actually do work, and why others do not.