The Great Powertrain Robbery - a classic John Robinson investigation

Published: 14 November 2015

How much power gets lost between the piston and the back wheel? It takes power to make power, and in a series of special tests on a GPX750, Performance Bikes' legendary technical editor tracked down the losses using a special dyno that instead of being driven by an engine, drives the engine itself. 27 years later, it's as relevant as it ever was – and only more mind-boggling the lengths to which John went for the story.

[It's worth remembering at this point that PB – which evolved from the long-running Motorcycle Mechanics in the 80s – was the first magazine to make dyno testing a routine part of bike reviews. With it the mag brought down firms' freedom to bulls*** with impunity, and carved a reputation for the rigour of its technical coverage – none of which would have been possible without great engineer, author and communicator John Robinson, who died in 2001 aged 56.]

or many years now, we have measured bikes’ power outputs. Mostly we use LEDAR’s Heenan & Froude DPX2 and DPX3 dynos, but we occasionally use others, such as the Superflow at TTS, the Schenck run by Mistral Engineering or the Froude eddy-current dyno at Micron. There are slight differences in the figures which these dynos give, but they are within a few percent of one another – and certainly not far enough apart to cause any great arguments.

Where we do see differences is between the measured power – on any of the above dynos – and the power figures quoted by the manufacturers of the bikes. These are bigger; typically, the manufacturer will claim 15 to 20% more than we measure, on occasions it will be as much as 30% more.

Every now and then, we ask them to explain the difference and we usually get sidetracked by one of the following:

-        It’s PS and not bhp

-        It’s SAE (or DIN or whatever) horsepower

-        It’s the difference between gross and net horsepower

-        Oh, that must be the Japanese/Australian/South African/French model . . .

-        They measure it at the crankshaft, not at the back wheel.

And so on.

With the exception of the last item, these are misleading, not to say meaningless, statements which, if you're interested are explained the panel 'Power lines' at the end of this story. The last item on the list had us wondering; given a typical, four-cylinder powerplant, how would you go measuring the power at the crankshaft?

To take a fairly extreme example, would a 130bhp engine lose 30 horsepower in its gearbox? That is just over 22kW, which is the equivalent of 22 single bar electric heaters, all inside one little gearbox. Would it not melt?

There is no real advantage in claiming artificially high power figures. After all, what counts in the end is how the bike performs on the road or at the track, where anyone can see immediately how well it goes. But put yourself in the position of a well-respected engine tuner; he runs a standard machine and sees that it gives, say 100bhp. He tunes it and gets 120bhp, a healthy increase that took a lot of hard work and the only problem is that the factory are already claiming 120bhp for the standard engine. What is Mr Tuner going to tell his customers? If he is honest, the best he can say is that the motor gives 20bhp more than stock. Then a river tuner, who doesn’t have a dyno but who can do simple arithmetic and knows that his machines are just as quick as Mr Tuner’s, decides that his engines must also give 140bhp. But as this is now an approximation and not a measurement, why not call it 145bhp?

The factories are in a similar position. If they know that an engine gives the same power as a rival, they obviously don’t want to quote a lower figure, not matter what their dynos say. They’d rather adopt the same test procedure, and where do the numbers come from?

The Kawasaki clue

Kawasaki Heavy Industries gave us a clue. They publish power curves which are usually the same shape as the ones we measure, but they are shifted a little up the power axis. KHI also publish “running performance curves”, which is the power curve translated into thrust at the rear tyre, in each gear, versus road speed in mph or km/h. Now, if these curves are translated back into power curves again, we get a graph which is almost identical to the one we measured on our dyno. Curious.

One of their engineers gave us an explanation, although in typically Japanese fashion, it raised more questions than it answered. He said that they measure engine output in the same way that we do, by running a chain around the gearbox sprocket and having it drive a sprocket attached to a dyno shaft. These figures are used to produce the running performance curves, so it’s not altogether surprising that they agreed pretty closely with ours.

But what about the claimed horsepower? Easy, he said. They remove the pistons, rods, etc., until only the gear train is left and then they motor it and measure the torque needed to turn it at each speed. These measurements are, in fact, the transmission losses. They are converted into horsepower and are then added on to the appropriate horsepower reading, to give the effective power produced at the crankshaft.

Ingenious. But what does it mean? Well, for a start, it means that Kawasaki can claim 106PS at 10,500rpm for the ZX750F-1 whereas we, and they, only measured 90bhp or 91.3PS at 10,500rpm. And they can do it knowing that it is an honest claim because they have presumably measured losses of 14.7PS at this speed.

Investigations begin

It had progressed from being curious to ingenious. It was now intriguing. How big were the losses? How much is lost in friction? What does it take to drive an alternator? Just how efficient are our engines? Some of us spend a lot of time and money in the search for more power. Wouldn’t it be as easy to try and reduce the losses?

First we had to measure them and try to find the amount of power absorbed by the various parts of the engine and driveline. The Motor Industry Research Association (MIRA) at Nuneaton have an extensive engine testing facility, which includes a Laurence-Scott swinging frame DC dynamometer. In essence, this is like a very large electric motor and it can be used as a motor. Coupled, via a driveshaft, to an engine’s gearbox shaft, electric power applied to the dynamometer can be used to make it drive the engine. It can be motored to a given crankshaft   speed and the restraining torque on the body of the dyno measured in exactly the same way as if the engine were turning the dyno. The same calculations are made to convert torque and speed into horsepower. In this way the power needed to turn the engine at each speed can be found and a graph drawn of power “losses” against crank speed.

Kawasaki Motors UK agreed to help, and supplied us with the GPX750 motor normally used in their service training school. This was a perfect machine – it is a typical, in-line four, with chain-driven cams, four valves per cylinder, shell bearing crank and a roller bearing gearbox which has a pressure oil feed. Its construction allowed us to strip it progressively and still be able to run the motoring test, isolating various groups of components at each step.

This particular engine was more perfect than we knew. It was used at KMUK’s launch and was the first bike on which we’d run our first road test. We’d used it to take national flying quarter-mile and kilometre records, and we had run two separate dyno tests on it, one at LEDAR, the other on the TTS Superflow.

TTS made up an engine stand for it and lent us their drive-shaft adaptor so that it could be coupled to MIRA’s dyno. The technicians at MIRA made up a special sump plug, to carry a thermocouple up into the sump so that we could monitor oil temperature (lubrication and oil drag would obviously play a significant part in these tests, so it was important to keep the same oil temperature and viscosity in each of them).

The motor was ready to go. We ran four tests: 

  1. Complete engine, minus carburettors and exhaust system.
  2. As 1, but with the cylinder head removed, and no coolant in the engine.
  3. As 3, but with the connecting rods and pistons removed (ie gearbox, primary drive, crankshaft, oil pump and alternator in use)
  4. As 3, but with the alternator disconnected.

In each test the Kawasaki was motored at high speed and cooled, if necessary, with a fan until the bulk oil temperature was 85C; all the load measurements were made when the temperature was in the range 80 to 100C. We used an SAE10W-40 synthetic oil to SF classification.

Test 1

This shows exactly what it takes to drive a 750 plus all its running gear. In order to make 90bhp at 10,000 it has to use 34bhp – to overcome friction, to pump the air in and out and to turn its own oil pump, water pump, and alternator (see page 12). It gives an exponential curve: at 400rpm, the losses at 8bhp, at 8000 they are 21bhp and at 11000 they have increased to more than 40bhp – there is obviously a price to pay for high-revving engines. In a real engine the pumping losses (the power needed to draw air through the engine) would be higher, because it would have the added restriction of the carburettors and exhaust system. The next step was to get an idea of what these pumping losses might be.

Test 2

For this test we removed the cylinder head and disconnected the cam chain. Because the head forms part of the cooling system, we also had to drain all the water out. So this test removed the pumping losses, but not in isolation – it also took out the water pump, camshaft friction and the power needed to open the valves.  Cam and valve operation are a part of the pumping losses because without them the air wouldn’t flow in the right direction. The power needed here did not increase so rapidly as the speed went up. At 400rpm it was 4bhp, at 8000rpm it was 7.5bhp, and at 11000rpm it was 10.5bhp. The pumping losses were rising in proportion to engine speed.

Test 3

This step – removing the pistons and rods – was expected to make a big difference. First there is friction between the piston (or rings) and the wall of the bore. Then there is the friction in the four big end journals. On top of that there is crankcase pumping – the pistons displace750cc in the cylinder, and there is the same displacement below the pistons. Finally there is the more subtle loss due to the reciprocating mass. The piston and rod are continually accelerating and decelerating, , requiring a force which has to come from somewhere. 

In running conditions, these losses would be higher, because when the engine is being motored there is no load on top of the pistons and so the side-thrust during the power stroke is less. Also when an engine runs, it accelerates during each power stroke and then decelerates during the following three strokes. Even at “steady” speed, the whole drive train is accelerating and decelerating: this oscillation is worst on a single cylinder and is reduced by having more cylinders and evenly spaced firing intervals. 

The losses due to the pistons formed a large part of the total, increasingly so once the speed went above 7000rpm. At 4000rpm, it was only 1.5bhp, but that accounted for 42% of the “mechanical” losses and 21% of the overall losses. The proportion increased with speed until, at 11000rpm, the pistons were responsible for 17.5bhp, which amounted to 55% of the mechanical losses and 42% of the overall losses.  We couldn’t take the crankshaft out to isolate the transmission losses, because the oil pump is driven from it and the gearbox needs a pressure feed. So at this stage, we were as close as we could get to “power at the crank” – and we were looking at a loss of just under 4bhp at 4000 and a loss of 14.7bhp at 10,500rpm, where the power peaks. Add that on to the measured power and you get 103bhp or 104.4PS at 10,200rpm. Which is very close to KHI’s original claim of 106PS. 

One thing that impressed us during this test was the amount of crankcase pressure build-up, even though there was only the bare shaft turning in the main bearings. We had blanked off the big end journals to reduce the amount of oil being thrown about, and had bolted a strip of Perspex across the top of the block, to prevent oil being thrown out of the engine. This was held by all eight head studs and was sealed to the block with RTV – which we didn’t leave to cure enough. There was sufficient activity in the crankcase to blow oil past this half-formed seal.

Test 4

Without being able to isolate each component, we discovered in broad terms where power is wasted. For anyone looking to save power, there are a couple of lessons to be learnt. The first is that high engine speeds eat up horsepower. The second is that the piston is the largest single source of lost power, even though the power is lost in several different ways. On an engine like this, it would be worth running more tests with extra crankcase breathing capacity, a crankshaft scraper, or even a dry-sumped engine. On engines which have to turn at high speeds, lighter pistons and rods and two-ring pistons might show some worthwhile improvements.We went one stage further and removed the alternator (which had been connected to a battery during the previous tests, to avoid the risk of damage by running it open-circuit). This made a just-perceptible difference – around 0.5bhp at 4000rpm, up to 2 to 5bhp at maximum speed. We’re not sure if the sudden divergence of the graphs about 10,000rpm is due to a spurious result or whether the alternator and its fan suddenly required a lot more power at this speed.

Meanwhile, at least we know how to make our dyno figures agree with the factory claims.

Power lines

Power is measured in several units. British and American engineers traditionally use horsepower (hp) or brake horsepower (bhp, measured on a brake – dynamometer – to distinguish it from indicated horsepower, calculated from an indicator diagram). In Europe they use a similar unit, the Pferdestarke (PS) or the Cheval Vapeur (CV), but now we are all supposed to use the Watt (W) or kilo-watt (kW) which they appear to do in Australia. Everyone else, including the Japanese, uses whatever they feel like.

Of course, the units don’t make any difference to the power which an engine produces; despite what some PR departments would like us to believe. For the record, 1kW is worth 1.34bhp and 1bhp is worth 1.014PS and 1PS is the same as 1CV.

What does make a difference to engine power is the way in which it is tested. Various bodies, like DIN (German industrial standards), SAE (Society of Automotive Engineers), ISO (International Standards Organisation), have set various procedures, so that tests can be repeated and hopefully, give the same results.

When we dyno test machines, we stick pretty closely to ISO4106. This, like other standards, sets out where the power is to be measured (at the output shaft, which is not always the crankshaft), what equipment the engine must have fitted (eg silencers, generators), what temperature it must run at, how long it must be run before taking readings, and so on. You could, of course, choose (or invent) a standard which allowed silencers to be removed (and race cams to be fitted, etc) but if you don’t say what the standard is, the resulting figures don’t mean a fat lot.

In the same way, people sometimes refer to gross or not horsepower, gross meaning before certain losses have been accounted for, net meaning that the losses are included.

What counts, in the end, is the force developed between tyre and road. If the engine delivers P bhp at the back wheel, then the force (in lb) generated is 375P/v, where v is the road speed in mph. (Or, 22.6P/v where P is the power in kW and v is the road speed in km/h. This gives the thrust force in Newtons). For the same amount of power, this will vary according to the gear the bike is in: a low gear will give a low value for v, and thus a high value for the force, compared to the same engine speed and power in a higher gear.

If the power curve is translated into rear wheel thrust versus speed, the result is the running performance curves mentioned in the main story. To put the pumping and mechanical losses into perspective, some of the diagrams show the theoretical power available from the fuel used by the GPX. 

When the power curves were taken, the engine was fuelled through a flow meter, giving the fuel consumption in lb/h at each speed. The calorific content of each pound of fuel is roughly 19,000 BTU (British Thermal Units) and a rate of 1000BTU/h is equivalent to 0.393bhp. So when the GPX was flowing 47.2lb/h at 10,200rpm, the potential heat flow into the engine was 896800BTU/h or 352bhp.

Yet only 125bhp is accounted for – 90bhp is useful power, 35bhp in mechanical and pumping losses. The remaining 227bhp is dissipated in heat losses to the engine metal and coolant, heat and noise in the exhaust, and in imperfect combustion.

The low conversion of chemical energy into useful power explains why engineers are so keen to develop an adiabatic engine (in which there are no heat losses) – as soon as they can produce materials which will withstand the thermal loadings and high temperatures which this involves. 


Power: what they said & what it meant








































THANKS . . . to Alex Dell and Peter Cranstone at Kawasaki Motors UK; John Donald; Martin Jones; Tim Church and Tony Marks at MIRA; Richard Albans and Paul Langley at TTS; Leon Moss at LEDAR.

Words: John Robinson Pics: Phil Masters