Like several others I know, by the 1970s I was collecting tech papers on two-stroke design published by learned academies. Most of them are unsubtle advertising for consulting services, but one that stood out was by Alfred Jante and dealt with small-scale methods for evaluating scavenge flow in two-stroke engines. Years later, Bud Aksland, who over a long career in racing had prepared and dyno-tested untold numbers of two-stroke cylinders and exhaust pipes, would find himself using a semi-automated Jante flow-mapping machine. It had been built in Eastern Europe and promoted in the West by a mysterious-looking gentleman wearing a dashing eye patch.

Jante had found a fairly straightforward way to develop good scavenging by mapping the upflow at the level of the cylinder-head gasket. To make a Jante map, install an upper crankcase half on the flow bench. Put a cylinder in place, having a piston located in it at BDC, and no cylinder head. With the flow bench blowing on “exhaust,” air will enter the transfer ducts in the upper crankcase and emerge through the transfer ports into the cylinder. The flows will in some manner join together as they hit the non-exhaust cylinder wall, then turn upward.

To measure the upflow velocity, an impact probe, whose opening was at head-gasket level and pointed straight down, would be moved about on a repeatable X and Y pattern. With an attached sensitive pressure gauge, the ram pressure of the flow could be measured at all points and then converted into a velocity. With this data in hand, a map could be drawn of the upflow velocity at any desired scale and contours of constant velocity penciled in.

Jante had found by correlative testing that good scavenging produced a map that was half upflow. This in itself was intuitively satisfying, for we like to imagine a D-shaped “piston” of scavenge flow, rising up the non-exhaust half of the cylinder, while in the other half, the outrush of exhaust forms a roughly D-shaped downflow. Naturally, this lovely picture is disturbed by the inevitable mixing of scavenge air and exhaust, and by the certain fact that some flow emerging from the transfer ports—that closest to the exhaust port, for example—turns from the “true path” and rushes out the exhaust port.

We know this must be true because, while a four-stroke needs roughly half a pound of fuel per horsepower per hour, even a highly developed two-stroke racing engine needs 0.58 to 0.65 pound/hp/hour, or 15 to 30 percent more. As mixture emerges into the cylinder from the transfer ports, parts of the flow feel the inviting low pressure of the exhaust port and turn toward it to leave the cylinder with the exhaust flow (this is why, in the late two-stroke period, Yamaha began to move the “A” transfers away from the exhaust port, leaving as much as 10mm of unused “real estate” on either side).

Possibly worst of all was unsymmetrical flows, such as one wall tongue wrapping nearly to the exhaust port or a center tongue that points off to one side. Lack of symmetry was clearly bad because such flows could easily short-circuit fresh charge right to the exhaust.

In Jante’s ideal D-section upflow, the highest velocity would be right on the non-exhaust cylinder wall, with layers of progressively lower velocities filling the space between it and the diameter dividing the upflow half of the cylinder from the downflow (exhaust) half, where the velocity would become zero. We can therefore think of the flow in the cylinder (with cylinder head in place) as a loop, emerging from the transfers, then flowing up the non-exhaust cylinder wall, across the underside of the head, and down to the exhaust port.

Jante also described various kinds of defective flows. One was “center tongue” flow, in which the map made at head-gasket level has a projecting tongue, pointing at the exhaust. Another was “wall tongue” flow, in which the upflow splashes onto the cylinder walls, wrapping around toward the exhaust. Possibly worst of all was unsymmetrical flows, such as one wall tongue wrapping nearly to the exhaust port or a center tongue that points off to one side. Lack of symmetry was clearly bad because such flows could easily short-circuit fresh charge right to the exhaust.

Eagerly I constructed the basic arrangements I needed to begin mapping flows in race-engine cylinders. Since there was a J-head Bridgeport milling machine in the shop, I decided to use its X and Y feeds to move the test cylinder while the impact probe was fixed to the stationary head of the machine.

Imagine my delight when, upon mapping a Kawasaki H2R B-type cylinder, I saw that the velocity contours closely resembled Jante's ideal: a D-shaped upflow occupying the non-exhaust half of the cylinder, with zero upflow on the dividing diameter and progressively faster layers of velocity all the way to the non-exhaust cylinder wall. Those guys at Kawasaki were okay!

Kawasaki had, I knew, used its own quite different development method. The scavenge test setup used an upper crankcase half and a cylinder and piston but also had a Plexiglas cylinder head. Instead of blowing air up through the crankcase half, they connected the suction side of the blower to the exhaust port. With the blower at speed and a bright light shining down through the plexi cylinder head, a technician would toss a spoonful of fine light powder up under the transfer openings and simultaneously start a stopwatch. The watch was stopped as soon as the last particle had left the cylinder. The shorter the time, the better the scavenge process.

Then I asked myself: "Okay, this cylinder produces an ideal Jante scavenge pattern. But there has to be more to it than a nice pattern because the Suzuki triple makes at least as much power but at 1,500 fewer rpm. That means the Suzuki must be filling its cylinders about 20 percent better." Having put many hours into flow mapping, I didn't want to abandon it, but I couldn't make that into a paying job; I had my hands full as it was. So I set Jante mapping aside.

Now I got a look at a cylinder from Suzuki's RG500 production racer. It stunned me with a quite different aiming of its five transfer ports. The "A" pair were conventional, being aimed back at about 35 degrees, with quite flat roofs. But then the "B" pair were aimed up at 45 degrees, and the "finger" or rear port was aimed up at 60 degrees. I'd seen this in Paul Schweitzer's 1949 book, Scavenging of Two-Stroke Cycle Diesel Engines—it was referenced from an article by Charles Curtis in a 1933 copy of Diesel Power magazine—but there was something else interesting. People who'd ridden the RG500 racebikes talked about their quite-sudden narrow power to the point of making the bike significantly harder to ride. So hard, in fact, that the engineers had deliberately tried to soften its "hit" by closing its rotary intake valves quite early at 55 degrees ATDC (Aprilias closed its intake discs as late as 84 degrees ATDC). To put this in perspective, that is the same closing timing Kawasaki was giving its runabout low-performance singles. Hmm, more stuff to think about.

I had made it a habit, upon being given access to the season’s new racebikes, to remove the top end, take it home, and walk up and down, drinking coffee and staring at all the parts—cylinder, piston, head—in turn.

Initially, Suzuki’s RG had overcome Yamaha’s early reed-valve 500cc inline-fours on sheer horsepower, Barry Sheene becoming 500cc world champion in 1976 and ’77. Yamaha fought back, not with more power but with more drivability as a result of developing its “power valve” system of variable exhaust-port height. This helped Kenny Roberts on the 0W48 Yamaha to throttle up sooner and with a more controllable power delivery than was possible for Sheene on the Suzuki. Roberts was 500cc world champion 1978, ’79, and ’80. That made it seem there was something about Curtis scavenging that gave sudden torque, but the jury was out.

Then something else interesting happened. I had made it a habit, upon being given access to the season’s new racebikes, to remove the top end, take it home, and walk up and down, drinking coffee and staring at all the parts—cylinder, piston, head—in turn. I seem to be slow to notice things, and so need this extra “staring time.” And I like coffee. In 1978, Yamaha had done something very different in aiming the new TZ250’s “B” transfers. Engineers had begun to swing the rear vertical walls of those transfers more toward the exhaust port. They started with a small change and then increased it a year later. I decided to see if this would work on the TZ750, whose transfer design had for years changed not at all from that of the 1969 TR2 350cc twin.

The amount by which I could turn those “B” transfers depended on how much wall thickness there was. Would I quickly cut right through into the rear cylinder stud tunnel, or up into the water jacket? Only one way to find out. Not only did I want to turn the “B” pair as much as possible toward the exhaust side, I also wanted to flatten their upward-aimed roofs. It worked. Riders said the engine was more rideable and made more torque (Who, me, own a dyno? I wasn’t Bill Gates). In the 1981 Daytona 200, Rich Schlachter had the lead and was pulling away when his chain began to skip on the rear sprocket. He managed to salvage third place. The cylinders on his Bob MacLean-sponsored TZ750 had the re-aimed “B” transfer ports.

Honda came up with the same idea in 1984 on its NSR500, gaining 4 hp per cylinder. Did it do so with the aid of a giant electronic brain running KIVA code? No, Honda did it the old-fashioned way, without engineers. Two experienced engine technicians—one of whom was called “Lobsterman” for reasons unknown—had the duty of actually finding the horsepower that the engineers hoped they had designed into the new parts. They did it on a single-cylinder dyno test engine (just as Kel Carruthers had done on Don Vesco’s dyno in 1971), by preparing endless cylinders, pipes, and heads, and by testing everything imaginable to come up with a best combination. Just as Aksland did years later for Roberts’ Grand Prix team.

In 1997, as I walked through the paddock at the German GP (made possible by a pass slipped to me the night before), I was asked if I’d like to see a cylinder from Suzuki’s new factory 250. It’s good to have friends. My eyes went straight to the “B” transfers and there it was: Their vertical back walls had been turned significantly toward the exhaust port. Clearly there was something interesting going on: three manufacturers, all aiming their “A” and “B” transfers at a common point. Could it be that colliding the transfer streams in this way caused them to form a somewhat slower, fatter rising column of mixture that would take longer to reach the exhaust port?

It’s okay to suddenly sit up in bed at 4 a.m. if you have something interesting to think about.