Zen and the
Art of Gear Drive Design :
Part 2.
This is
a continuation of last issues’ discussion of gear reduction design choices
and the difference between my drive and the Ross Aero drive.
We’ll
start with the choice of the coupling system between engine and gear drive
input shaft. The Ross drive
uses what is widely referred to as a clutch damper plate. It uses
pre-loaded springs to absorb shock loads between the engine and output
spline. The actual part
used is a marine engine damper but it works the same as an automotive
one.
Damper used in the Ross
re-drive. The springs are
pre-loaded so that they are not active during normal engine
operation.
The term
‘damper’ is not really accurate here. Springs do not actually absorb or
dissipate energy, they just store it. This would be fine if clutch dampers actually worked the
way most people assume. It is
widely believed that these parts work by compressing the springs during
the engine’s torque peak and then expand when the torque declines after
the power stroke. This would
presumably level the torque output and provide a smooth power delivery to
the gear drive. There is not
room here for a thorough engineering analysis, but a look at the basic
design of the damper is enough to reveal that this view is erroneous.
In order for the spring damper to
work in the manner described, the spring rate and travel would have to be
carefully selected to work with the other components. The weight of the crankshaft,
flywheel and propeller would drastically affect the design. The engine used would also be a
critical factor. A six
cylinder engine would require a much different “damper” than a 4
cylinder. After all this
careful design, you would be very disappointed to find that it works only
at one particular rpm and power setting. Clearly, this is not an acceptable
design for an automotive application where the engine speed varies by a
factor of 10 and power by an even larger amount. Aircraft applications require a
smaller range of operating conditions but still much too wide for this
approach to be workable.
At this
point it should be obvious that our original assumption about the intended
purpose of the damper is wrong.
The engineers in Detroit are not stupid and they don’t put in
relatively expensive parts for no reason, so why do manual transmission
equipped cars have a damper?
The short answer is to absorb unexpected torque overloads. This happens only on rare occasions like when someone
gets overly aggressive with the throttle and suddenly releases the
clutch. The springs store the
energy of the shock load and release it in a more controlled fashion in
order to avoid breaking drive-train parts.
I’ve
heard some airplane builders with a bit of engineering background comment
that something is wrong with the design of some redrives because the
torque required to cause any compression of the clutch damper is far more than the engine is capable of
putting out. Their
observation is correct, but
their conclusion that a lower spring rate is needed is all wrong, due to
the factors already mentioned.
So the next question is, why put a clutch damper in an aircraft
gear reduction drive?
You may have noticed that there is no clutch pedal in an airplane,
so we need to look a bit deeper.
The
redrive designers I’ve talked to were all relatively knowledgeable people,
but most of them still
thought the damper’s function was to absorb engine torque peaks. It did not seem to me that they
had thought this through very thoroughly, because they had no answer when
confronted with the evidence about the miss-match between engine torque
and damper rating. I suspect
the real reason they used it was because it was a readily available part
that could be used to couple the engine to a splined shaft. Still, there is a pretty good
record of success while using the clutch damper in redrives. For what its worth, here is my
explanation of how and why it works.
As I
have explained in the past, it is not torsional vibration per se that
causes props, crankshafts and other driveline components to break. It is torsional resonance
which can cause the problem.
The resonant frequency
of a driveline is determined by the spring rate and mass of the entire
system, along with any play such as spline or gear lash. It may seem odd to think of
something like a crankshaft as a spring, but it actually is. Keep in mind that the spring rate
of the damper is NOT a factor under normal circumstances because
the springs are pre-loaded and the input and output plates are against
their stops. It acts like a
solid disk until the damper is loaded enough to further compress the
springs.
At some rpm, engine torsional
vibration (or some harmonic of it) is going to coincide with the system’s
resonant frequency. If the
amplitude of the vibration is higher than the losses in the system, the system starts to “ring” at
higher and higher amplitude until something breaks. It is during this condition that
the damper has a chance to
intervene and stop this destructive sequence. When the amplitude of the ringing
reaches the point where the springs start to compress, conditions change
radically. When the damper
plates are no longer against the stops it no longer looks like a solid
disk. The springs are now
part of the driveline and affect the system resonant frequency. The additional spring rate lowers
the resonant frequency and halts the build-up of vibration.
This
method of resonance control prevents catastrophic damage to the
drive-train but if the system were to be operated at this point for any
significant period of time, another factor would rear it’s ugly head to
ruin our day. The clutch
damper was not designed to operate continuously with its springs being
compressed and released.
There is considerable sliding contact between metal parts when the
springs are active, which would quickly wear them out. It is imperative that resonance
not occur anywhere in the normal operating range of the engine if the
spring damper is to have a reasonable life.
I lack
the expensive equipment needed to accurately measure the resonant point of
the drive-train but through careful inspection and
audible clues I have concluded that the resonant frequency of a typical
13B installation using the Ross drive occurs at an engine speed of around
1300 – 1500 RPM. At this speed, the signs of resonance can be heard as a
rattling noise as the gears and input shaft are slammed back and forth.
(Note - resonance at higher engine speeds would not necessarily be
audible.) To avoid resonance at these low speeds, it is a simple matter of
idling the engine above this point.
The other conditions that cause resonance to occur are rough
running during engine start and any condition causing the engine to run on
just one rotor. Running on
one rotor causes an alarming amount of shaking and vibration and once you
have experienced this first hand, it will eliminate any fantasy you may
have of using “one rotor operation” as a backup mode.
Inspection of the Ross drive damper
shows that it very rarely compresses the springs. The tell-tale signs of sliding
metal to metal contact can be seen but are very minor. I accepted this as proof (after
856 hours of flight testing) that resonance is not occurring at normal
operating speeds.
The
BMW Seduction
OK, we
have now pretty well analyzed the operation of the clutch damper and
concluded that it is a workable solution, so why go looking for a
different way? Part of the
answer has to do with my own evolving tastes and preferences when
flying. When I first
“tasted flight” almost 30 years ago, it was in a Benson Gyrocopter. That instant when the wheels
first lifted off the ground
is still clear in my mind and forever hooked me on flying. Just the mere fact that I was
controlling a machine capable of traveling through the air made me feel
like every wish I’d ever made had been granted. I wanted nothing more. At least I wanted nothing more
until six hours later when
the engine seized a piston.
Thus started a never ending quest for more reliable engines and
other refinements to my flying machines.
Cars
have never been a high priority for
me, but it was a short drive in one that got me to thinking about
human perception of “quality”.
At the time, my personal car was an aging Ford Escort that cost me
$600.00. Its paint was faded
and the hood was severely wrinkled where I had inadvertently felled a tree
on it while clearing the land for my house at Shady Bend. In spite of its 130,000+ miles, it
was reliable as an anvil and I could not understand why anyone would pay
50 times more for something that did exactly the same job. The new cars that I sometimes
rented on business trips only reinforced my opinion.
A friend
and I were working on a project one day when I needed some supplies from
the local hardware store. My
car was blocked in the driveway by his and rather than move it he threw me
the keys and said to take his.
I’d never driven a BMW before but I was immediately impressed with
the “feel” of it. I use the
word feel because I didn’t know what else to call it. It wasn’t particularly fast or
quiet or superior to my Escort in any definable way (well OK, the paint
was nicer), but there was no denying that the driving impressions were
worlds apart. BMW was
probably counting on that feeling to motivate drivers to go out and buy
one for themselves, but I just wanted to know how they did it. I don’t know everything they
did to create the “BMW driving experience”, but one important factor was
the use of a “gebo”. This is
a rubber isolator mounted between the transmission and drive shaft. Its sole function is to modify the
noise and vibration signature of the drive-train and even though the
change is fairly subtle, it has a major effect on the human perception of
smoothness and quality of the car.
Lately,
the carmakers in Detroit have gotten the message as well. When the decision is made to
design a new engine or transmission, the most important factor has nothing
to do with horsepower, torque, reliability or any of the other factors
that we gearheads care most about.
The first topics to be discussed at a Detroit engineering meeting
are the NVH (Noise, Vibration
& Harshness) factors.
Power to weight ratio and reliability are still at the top of my
list but I do have a better appreciation of these other factors after my
short drive in the Beemer.
So, there you have the
long winded version of why I chose to use an elastomeric (fancy word for
rubber) damper in my redrive.
To borrow an old Mazda advertising slogan, “It just feels
right”.
This
damper has worked well, but I did have to make one change from the
original design. The rubber
isolators used in the first production unit were cut from shock absorber
bushings. This was acceptable, since I assumed the drive would be built in
relatively small numbers. The
number of drives ordered has far exceeded my expectations so the isolators
were among innumerable details which had to be re-thought in order to make
the drive easier to produce.
I was forced to have a custom molded rubber part produced in order
to eliminate the labor intensive job of accurately cutting and trimming
the donuts from shock bushings.
RD-1 Damper assembly installed on
flywheel. The capturing
hardware from the top-right damper has been removed to show the rubber
isolator.
Input Shaft Thrust
Bearing
The decision to
incorporate a thrust bearing was already discussed in the last issue and
was totally vindicated by the results of my engine teardown. I have done two internal
inspections during the 175 hours of flight testing on the new drive and
everything, including this thrust bearing, looked perfect.
One thing still
bothered me however. The
bearing I used was of the same type and similar in size to the thrust
bearing in the engine (which wore out while using the Ross drive). Since the maximum thrust rating on
the bearing was well above what it sees in this application, I took a
closer look at what was going on.
The problem turned out to be RPM, not the thrust. The rating of this bearing (2900
RPM) was only half of what it was being asked to endure. This is not a problem in the
automotive application, where the bearing has no thrust load at all except
when the clutch is disengaged.
Using the bearing life formulas from SKF Bearing’s engineering
guide, it looked like the life expectancy would only be about 300 hours
under the conditions expected in our application. Since I had already bought parts
for the redrives, I was reluctant to change my design. I reasoned (rationalized ?) that
changing the bearing is a relatively easy job requiring only about 2
hours. (Changing the bearing
in the engine is a different story).
After waking up several times in the middle of the night agonizing
over this (Note from Laura: I can vouch for that! It is not always easy living with
an engineer..), I finally got up, turned on the computer, and designed
in a different bearing. I had
to eat the cost of the unused bearings and my production schedule was
wrecked again, but I (we) slept soundly afterwards.
This is the original
thrust bearing used in the first production RD-1 re-drive. It is similar in size and type
to the one in the Mazda 13B engine.
The surface velocity is significantly different at the inside and
outside ends of the needles. This causes a sliding contact and limits the
RPM at which this bearing type can be used.
This
ball bearing replaced the needle bearing in the RD-1. It does not have the RPM
limitation of the needle bearing but it’s greater thickness required a
number of design changes. Life expectancy is over 4000
hours.
Engine Adapter
The adapter between engine and the
gear case is traditionally a cast bell housing and this is what is used on
the Ross drive. As in most
cases, the design of any
given part is affected by all the other parts. I had decided early on that I
wanted more distance between the bearings of the prop shaft in order to
reduce the stress on them and to hold prop shaft wobble to an absolute
minimum. This required a
longer gear housing than on the Ross. If the overall length of the drive
was to remain the same, this meant that the adapter had to be
shorter.
Another decision was to use
the automatic transmission flywheel instead of the custom aluminum part
used on the Ross. It was an
attractive choice because it’s available, the price is reasonable and some
builders already have them.
(The RD-1 price is lower if you supply your own.) Many builders
were concerned about this choice because of the many reported cracking
failures on Blanton’s Ford V6 re-drive, which also used an automatic
transmission flex plate. On
the Blanton drive the flex plate is used only for the starter. In the car, it is bolted to the
torque converter which damps any ringing so cracking is not a problem. On the RD-1, the flex plate is
bolted to the damper which will also eliminate any ringing and subsequent
cracking (another example of resonance induced failure). In any case, the flex plate is
larger in diameter than the manual transmission flywheel (and 24 pounds
lighter) which made the
design of a bell housing adapter more difficult. The section thickness near
the mounting holes came out a little too thin for a sand casting. Mazda uses a die casting for their
bell housing which gives higher strength and better tolerances but is far
too expensive for low volume production.
In addition, the
shallow depth of the bell housing required for the RD-1 meant that it
would resemble a large frying pan and have a flat face rather than the
conical form of most bell housings.
To achieve sufficient strength with a sand casting, the flat face
would have to be either very thick or have a lot of reinforcing ribs. Either way it would add excessive weight.
Since the design
called for a short, flat adapter, the obvious choice at this point was to
make it from structural plate.
This material is significantly stronger than cast aluminum and
there is no question about possible casting flaws. Half inch 6061 T651 was
chosen for the plate.
Spacers made from ľ” plate (same alloy) are used to mount the plate
at the correct distance from the engine.
Torsional rigidity is
provided by two wide spacers which capture two bolts each. These spacers also have
counter-bores that mate with locating dowels on the engine to precisely
align the adapter. Two additional spacers capture single
bolts.
The resulting structure does
not totally enclose the flywheel and damper assembly. Some objected to this arrangement
because they felt that loose parts could be launched out of the adapter at
very high speed and do considerable damage. My position is that if there are
loose parts in the adapter, the damage has already been done. Having the ability to visually
inspect and re-torque the fasteners in this area is the best way to insure
that parts don’t get loose to begin with. During my testing it was also very
nice to be able to inspect the rubber damper parts without removing the
drive from the engine. I now
think of the lack of total enclosure as an advantage rather than a
draw-back. If the idea of an
exposed flywheel is unacceptable to you, it would be easy to fabricate
sheet metal panels to close the openings between spacers. There is plenty of material in the
spacers to allow drilling and taping screw holes to attach the
panels.
RD-1
adapter plate & spacers.
The two longer spacers provide plenty of torsional rigidity. The large hole on the right is for
mounting the starter. (A starter for 1986 – 1988 RX7equipped with manual
transmission is required).
Gear Housing
Adapter plate, Main Gear &
planet carrier housing for the RD-1.
The threaded holes in the gear housing are for oil return
lines.
The
decision to use a casting for the main gear housing rather than machining
it from solid billet was based on weight and man-hours of machining. This housing is secured to the
adapter plate by 12 long bolts which go all the way through the outer wall
of the big end. I didn’t feel
comfortable with a relatively thin flange to do the job. This probably cost a couple of
extra pounds in weight but it is still lighter than a solid billet housing
would have been.
Design
Compromises
These
two articles were intended give you an overview of the major design
considerations but there are innumerable design details left out due to
space considerations.
I also
want to make it clear that this was not a “cost is no object” design. The best design in the world is
worthless if you can’t afford it.
The goal for the RD-1 was always to design the most reliable drive
possible for what I perceived as the price range that most builders using
the 13B engine were willing to spend. Of course it also had to meet one
other important criteria. It
had to be something I was willing to fly behind while carrying my own wife
and kids. I know this
does not sound specific enough for many builders. Some have called and asked if the
drive would meet the requirements of Part 23 of FAA regulations regarding
certification of aircraft. They emphasized that this was important to them
if they were to be comfortable
flying their own families.
I must make it clear that I have not studied these requirements and
have no intention of certifying this or any other product that I
make. The only guarantee I
offer is the one printed on our sales brochure - “We Fly What we Build.”
Operational
Limitations
The results of this effort was an
unqualified success. The
parts are beautiful and even the best machinist cannot achieve the
accuracy and consistency that computer controlled machine tools can
deliver. I must admit that I
was very nervous about committing thousands of dollars to parts that would
be useless scrap if my drawings were wrong. I just got finished inspecting the
parts and still haven’t gotten over how incredible it seems to send off a
bunch of data and get back parts that fit. My sincerest thanks to Gaylen
Lerohl for his invaluable help with this.
PREVIEW
of the RD-1(A)
This is the new propeller shaft for the RD-1(A). The shaft
diameter is 45mm as opposed to the 35mm shaft on the original design. It
is a single piece rather than the separate hub and shaft used
previously. Material is 4340 which is thru heat treated. It is
machined from a forging which is stronger than if machined from solid
billet due to the better grain flow in the forging. It is an
expensive part but it gives the capability of using metal props and
competition level aerobatics. The shaft is hollow to save weight and
will make a constant speed version possible in the future. All
drives ordered after 4-1-00 will use this shaft. This was the reason
for the price increase on the RD-1.
Update 8-4-99
The prototype RD-1(A) was installed on our RV-4 test mule (the
RVotter) in early July, 1999 and flown to Oshkosh and back (about
2200 miles round trip). The drive performed as expected and
production has begun.
For more details on the RD-1(A) and other redrive development news,
see Zen and the Art of Gear Drive Design part III.