Hi all.

To begin with, I believe that the real horsepower you have is
what you get to the wheels. Any other number is magical
'patting ourselves on the back' as if we could spend as much
in a year as our salary... That being said, there is some value in
understanding drive train loss, especially for our AWD cars,
when so often for convenience we dyno our cars on a 2WD
dyno, which artificially removes some of the actual loss, and
so gives slightly higher readings than what we actually
can put to the ground..

Drive train loss is a variable function, and is not primarily a
constant. It must necessarily vary with RPM, else there would
be some 'resistence' or force turning the motor the other way
when the motor is off! The resistence is primarily fluid friction,
generating heat in the oil being sheared between moving
parts. There is also resistance caused by air flow in the intake
and outlet tracts, which also varies with speed/volume. Also
the air in the crank case is being forced around as pistons
pump up and down. People 'knife-edge' their cranks to make
them more aerodynamic, and make sure there are big clear
air passages between cylinders.
The resistance due to oil shear grows according to the
relative speed of the two surfaces, and the closer they are
(film thickness). This resistence is superlinear. In view of
most motor's typical ranges of bearing gaps, there is usually
much less resistence from oil at RPM ranges up to 1500RPM,
but then oil resistence climbs quickly.
Transmissions provide different amounts of efficiency for
different gears, based not on the ratio, but on the different
mechanical designs. For instance, one of the top gears is
typically designed to go as directly from input shaft to output
shaft, for mechanical and fuel efficiency, whereas other ratios
may be implemented by the power having to go from input
shaft to a gear on a separate jackshaft, and then to the output
shaft. Ie: You may have 600 real to-the-wheel HP in fourth gear,
but only have 596 in third.
There is an approximate physical measurement that can be
done to see what drivetrain loss there really is. This is simply
done with the same dyno you take your power measurements
with. Once the power has been measured and the engine is
at peak RPM, just take your foot off the gas (and maybe even
shut the ignition off) and have the dyno record the force being
applied by the motor internals to slow the dyno wheels down.
This is just a reverse dyno run. If at the end of a power pull,
you were able to instantly lift the car off the rollers, the rollers
would be able to spin for a long time, as they are designed with
efficient bearings, and minimal wind resistence. With the tires
off the rollers, and no power being supplied by the motor, such
as when you turn the key off, the internal resistence of the motor
and transmission stops the motor very quickly. But with the tires
still on the roller, the spinning enertia of the rollers is resisted
by the motor and transmission friction, as the motor and tranny
is driven by the momentum of the heavy still-spinning rollers.
The difference between how long it takes the rollers to stop,
and how long it would take with no car on them, is the drive
train loss. It will vary across the motor's RPM range, so people
typically use the resistence at the RPM where they saw the most
power to apply to wheel-to-flywheel conversions.
The down-cycle resistence readings won't be perfectly the same
as the real internal friction under power, because of such
differences as the piston-cylinder friction, which will be higher
during the power stroke under combustion, than it would be
during the rundown with no power, just resisting the compression
of air trapped in the combustion chamber... but in total, it's
close enough to know whether drivetrain loss is 30% or 20%
or 16%...
If you really wanted to be technical about it, why stop at
'flywheel horsepower'? Why not go all the way to 'combustion
chamber horsepower'? You could simply measure the fuel
consumption rate at peak power, and measure the air/fuel ratio.
These would enable you to calculate the amount of energy
produced per second from that much fuel, and then you could
calculate the actual power being released. You could then
subtract the power you see at the wheels, and what was left
would be the *total* power loss/inefficiency of the motor. Ie:
*all* the power lost to heat and sound, not turned into wheel
torque. This would be very accurate, but is typically dismal,
showing the total loss from what is chemically available from
the fuel at about 80% *loss*!
So practically, I would go with what a lot of folks are
suggesting, that we stick with to-the-wheel figures. That's really
what we get. Ideally from AWD dynos. We'd all like to see what
different dynos show for the same car, to normalize across dynos.
What you see is what you get. Talking 'flywheel' is like bragging
about how much taxes we paid.