The desmodromic idea—using a positive drive to close as well as open valves—is nearly as old as the internal combustion engine itself and has taken many forms. The Mercedes-Benz W196 Grand Prix car won Formula 1 championships in 1954 and '55 with desmo valve drive. In 1957, Norton applied for a patent on a four-cam desmo system, later built and evaluated on its Manx 500 race engine. Both Cosworth and Ferrari are said to have prototyped desmo cylinder heads for possible F1 application. BMW, too, gave the idea a try in its 1994 liquid-cooled boxer R1 prototype, which was never produced.
So far, Ducati is the only maker to have stayed the course with desmo. In a 1943 thesis, its future designer, Fabio Taglioni, proposed such a valve control system for his engineering thesis of 1943 but was unable to complete his education until after the war. After spending 1950 to '54 at Mondial, Italy's foremost builder of high-revving GP singles, Taglioni went to Ducati, where, at last, he was able to realize his system in Ducati's three-cam 125.
In desmo’s modern form, a conventional opening cam pushes the valve open through a pivoted finger follower, with the cam lobe acting almost directly above the valve stem. A second and complementary cam lobe, looking much like a half moon, then pulls the valve closed by acting on an L-shaped closing lever. One end of this lever carries a pad against which the closing cam acts. The other end is forked to bear against the underside of a fitting attached to the valve stem.
To maintain valve clearance at all points, the two cams must be made with considerable accuracy. Clearance is necessary in valve systems because, otherwise, the heat expansion of the valves would hold them slightly open.
All my adult life, I have read that, because the desmo system eliminates springs, it can slam the valves open and closed when engine airflow demands it, avoiding compromise imposed by valve inertia or spring dynamics. Turns out, this is false.
Numbers reveal that such compromise was, indeed, considerable in the case of the classic Manx Norton race engine, among many others. To control the motions of its two large valves with metal springs, those engines required extended valve timings of up to 350 degrees. When combined with a tuned exhaust system, the very large valve overlap that resulted prevented Manxes from accepting full throttle below 5500 rpm, leaving a usable rpm band only 1000 to 1300 rpm wide. At the time, this was termed “megaphonitis,” but the real cause was the need for extra time in which springs could overcome the inertia of such large heavy valves. This imposed large overlap let exhaust pipe waves stuff the cylinder with exhaust gas and provided long intake closing delays (after bottom center) that allowed low-speed back-pumping of fresh charge, weakening midrange torque.
When John Surtees went to MV Agusta in 1956, he discovered a quite different characteristic. Valves in the MV's small 125cc cylinders were much more easily controlled by springs, so they had been given much shorter timings, resulting in a strong, wide powerband. That wide power allowed Surtees to steer the MV with the throttle in long, fast bends, which was impossible on the peaky Manx.
Mercedes was motivated to develop its “hard valve drive” by valve float encountered when it ran a number of 310cc single-cylinder test engines with metal valve springs. Manfred Lorscheidt began work on the desmo concept in 1952.
I finally sought out the actual valve timings and lifts used on various Ducati desmo racing models. What I found was the very antithesis of slamming the valves open and closed when engine airflow demanded it. Instead, I found that Ducati was quite conservative for a long time, giving the system plenty of time to accelerate and decelerate the valves. Its four-valve Superbikes began life with ultra-long, Manx-like valve timings and quite moderate valve lifts.
Thanks to the site www.bikeboy.org/duccamspec.html, anyone can look at the timings and lifts. They show that the "Ottovalvole" 851 race engine started out with a whopping 346 degrees of intake duration back in 1987, which is practically indistinguishable from Manx Norton two-valve timings of 30 years before. Ten years later, Ducati had moderated this to 296 degrees, and eight years after that to something positively ZX-10R-like 254 degrees (for comparison's sake, intake duration on Mercedes' desmo engine was always 256 degrees).
Why the change? Very long valve timings weaken midrange, and as Ducati bent every effort to winning World Superbike races, engineers gradually shortened valve timing to deliver the acceleration that Carl Fogarty (among others) demanded. Opening the valves faster generated larger forces, so there were rocker failures in 1994 and cam belts were replaced more frequently. By the late ’90s, the race department accepted that there was little point in making the valve train last longer than the crankcases.
In 2003, when I asked current Ducati CEO Claudio Domenicali how much valve acceleration the closing levers were capable of transmitting, he answered obliquely, “That depends on how long you want the levers to last.” Because the L-shaped closer levers are loaded in bending during valve closure, they are in a real sense “valve springs,” with all that implies: They have a natural frequency, they bend and deflect, and the valve may not be where it should be at every instant.
The above are not criticisms of the desmo concept. Desmo, like any other means of controlling valves, requires continuous development to keep up with rising engine rpm and other changes. Mercedes, after finding its "Zed-drive" desmo system reliable, boosted lift by 26 percent to increase power. Ducati's early competition in World Superbike were Japanese fours with chain-driven dohc. As those fours were revved ever higher, they needed a more stable cam drive, so gear-drive kits appeared (Honda's RC30 and RC45 were designed with gear-driven cams). Steve Scheibe showed me cam-chain guides from Harley-Davidson's VR1000 Superbike development—thick steel, twisted into amazing shapes by cam-drive instability.
In Ducati's most recent eight-valve iteration, the Panigale 1199, the previous fiber-reinforced rubber cam belts are out and steel chains are in.
The Internet has more to tell us. I had heard from Eraldo Ferracci that, in the late ’90s, Ducati race-engine horsepower had stalled. Friction rose steeply above 12,000 rpm, and motored friction tests performed with partial engine builds had shown the friction was in the heads—in the valve train.
An Italian university site provided more. It turned out that the cams and valve levers were bending, cam ball bearings were providing inadequate rigidity, and parts were not rotating and reciprocating smoothly, but were instead chattering and vibrating, turning power into heat.
As a result of extensive dynamic FEA study, camshafts were made larger and hollow. Plain bearings, with their high radial stiffness and inherent damping, took the place of ball bearings. Valve operating levers were reshaped and lightened. The result was the “Testastretta” head. Engine friction returned to normal.
This need for continuous engineering is not unique to desmo. Have a look in detail at Honda's progress through the RC30 and RC45 V-4s, both of which are steel-spring engines. All we can see is progressive changes, but engineers don't change things because they enjoy it; they do it to defuse a crisis, to solve specific problems. Why did Honda put flywheels on RC30 cams at certain times? Why were gear drives changed as they were? Don't you wish you had the language and budget to seek out those retired test supervisors in Japan and hear about the flying parts that buried themselves in dyno-cell walls and ceilings? Everyone in racing has to do continuous engineering. Good design is just a starting point, not a foreordained answer to every problem. Coping with what reality throws in its way is the job of development.
It is very interesting to compare the appearance of desmo closing levers from one of Ducati's air-cooled street engines with those of their D16 racer replica. Big difference: The D16 levers have wide webs and are heavily braced (and D16's cams, like those of Ducati's MotoGP engines, are driven by gears).
Ten years ago, Domenicali asserted that Ducati’s desmo system is not just an historical curiosity, arising from the long-ago unreliability of springs and continued as a quirky mechanical tradition. Ducati’s continuing development has made it the equal of any system today.
Part 1: Metal Valve Springs
Part 3: Pneumatic Valve Operation