Having a pair of rollers pressing against the road wheels really does introduce some unrealistic losses into the picture; in that you have an unnatural degree of compression of the tyre wall brought about by a relatively small diameter roller. What's more you have two contact patches giving double the loss. If the operator is really keen and puts a strap on the back of the car to pull it down onto the rollers it gets really interesting! If a car needs that level of grip it generally has a very powerful engine which in turn means big, (wide) tyres. The bigger/wider the
tyre then generally the more energy it takes to compress the side-wall. All basic stuff.
You don't have to be a genius to work out that the faster you go, the more effort it takes to keep deforming the side-wall (same deflection but in an increasingly smaller time-frame) hence "total rolling losses" increase with road speed.
Where it gets interesting is in the way you set the rolling road up. On my Sun RAM12 it has the facility for holding a given road speed - regardless of engine applied load. Straight forward, you dial in the road speed and when the car gets to, for example, 60mph, it won't go any faster. You can increase throttle until the tyres spin and smoke, but the brake holds the rollers at 60mph - regardless of what gear you are in. This is very useful for mapping - where you need to hold a steady load and speed to play with the fuel and ignition settings. On my rollers I have a digital read-out which goes to 0.1 of a bhp. When you tweak the timing on the lap-top you can see small changes which, while not deadly accurate, do give you an indication of what's going
Taking a power curve is a different story. Steady state running is not the way to do it on a rolling road. For an acceleration run you need a computer with software that captures the torque at a given sample rate - usually based on roller (mph) speed. You drive against the roller in one gear, from low revs to maximum, and the computer records the load and speed - obviously calculating the bhp from these numbers. This is not as simple as it sounds.
On the front of my machine I have three dials. These allow me to adjust the rollers for simulated load. I can dial in the weight of the vehicle, the power needed to overcome wind resistance (at 55mph) and the rolling resistance (again at 55mph). This allows you to accurately simulate road conditions for fuel consumption tests. If you think I go to all that trouble for a single power run you are sadly mistaken!
There are a lot of reasons why a rolling road is not as accurate as an engine dyno. One of them is that the power of the engine is constantly changing. As combustion chambers heat up, power fades. As oil and water temperature changes, the power coming off the flywheel also changes. If you are involved in accurate research you have to carefully monitor all these parameters. I'm not. I don't.
What I have done is adjust the rolling road acceleration rate to give me a very close approximation of known cars. Example: I had a BMW showing 177bhp on the Magnex (of exhaust pipe fame) inertia roller. This machine is pretty accurate since it's just a lump of mass and a computer calculates the work done in accelerating, and decelerating, it. The computer doesn't know where the power is coming from, it's just work done - or undone. On my rollers I got exactly the same figure from that same car, in both cases corrected for standard temperature and air pressure.
For cars making between 130 and 200 bhp it seems to give realistic numbers and it's repeatable to within a brake or two, given the limitations of the system and the changes in actual power between runs. For smaller engines, like a full race 2CV, I have to increase the acceleration rate of the rollers, otherwise it could take all day to get to maximum rpm! On big 300 bhp V8 engines they woosh up so fast you can't be sure it isn't wheel-spin.
For these I increase the roller load to slow everything down. I have to admit there's an element of "feel" to it. We've done over 5000 power runs now and I think I can tell when it's all over too soon - or not.
What you get from all this is a power figure at the wheels across the rev range - totally meaningless, and here's why: The rolling resistance of the tyres depends on road speed. If I run the car up in third gear the road speed reaches maybe 90mph. If I run it in fourth gear the rollers might make 120 mph. In third I have less power loss through the wheels than I have in fourth gear, which means that in third (at for example, 6000rpm) I have 80bhp at the wheels. At 6000rpm in fourth I have only 70bhp at the wheels. That missing 10 bhp is being lost in the road wheels rotating at 120 mph instead of 90 mph.
The clever bit is measuring the rolling losses on the run down and adding these to the "at-the-wheels" figure. I can run in any gear and get the same power curve - but only after the rolling losses have been measured and added to the wheel figure. If I look at the road wheel power in different gears at the same rpm I get a totally different reading.
To prove the point we ran a Golf GTi at 3000rpm and recorded the power at the wheels. I then put another 10psi into the tyres. This higher pressure reduces the amount of distortion of the side-wall and hence reduces the rolling losses. We "gained" nearly 5 bhp at the wheels. Moral: if you want more power at the wheels, or improved fuel consumption, keep your tyres up to pressure. From this you can see that the idea of calculating a transmission loss common to all cars is patent nonsense - you need to know the road speed, the tyre size and its rolling resistance. If you get big variations at the wheels from week-on-week tests then you may well have been very careless with your tyre pressures.
From my limited experience the major loss is the tyres. If I dip the clutch and keep the car in gear while I measure coast down losses I get some added drag - possibly from the gear oil in the box but more likely due to the clutch dragging plus the load on the release bearing. To state that you get a given percentage (I think it was something like 15% or 17%) only shows a total ignorance of how the whole thing works. No shame in that. I used to happily measure power at the wheels and quote it in features - because I was ignorant of exactly what I was measuring. Ignorance isn't stupidity - it's just a case of not knowing what you are talking about.
Where ignorance becomes dangerous is when you set yourself up as a "professional" and present your ignorance to the world as proven fact. I do not claim that my rolling road gives deadly accurate flywheel figures, but it does give a repeatable number that bears some relationship to the real world.
In a contest to get the most power at the wheels I would run my car in first gear with 60 psi in skinny tyres fitted to bicycle rims. Does that make the engine more powerful?