Belted Air Power Chevy V6 ...Indirect Update

[rv6ejguy]

Dan makes important points.

In flight testing with the proper equipment is really needed to explore the TV question as Hartzell/ Lycoming does with various combinations. Even these direct drive engines often have avoid rpm ranges at certain throttle settings. The engineers have determined that these are relatively benign in most certified installations and just recommend extended operation in these ranges be avoided.

Rotax 912 engines have bad TV in the lower ranges below 1200rpm depending on prop MIs. You must not run these engines here for any length of time or the reduction gears will be chewed up in short order. Eggenfellner's latest setup apparently has an avoid range below 1200rpm as well.

What would be very helpful is to have an actual plot like Dan's above for all engine and prop combinations that we fly to see where the bad ranges are, either to avoid them or correct them through design changes in the engine/ drive combo. This is unlikely to happen any time soon for most drives due to the sheer amount of time needed to collect data for multiple propellers and engines. It could never be afforded by companies like RWS or Marcotte which cater to several different engines fitted with dozens of different props.

If Eggenfellner only supplied one prop type with all H6 engines and Gen 3 drives, the task is much smaller and probably well worthwhile as the entire package stays the same. One instrumented flight test would reveal the entire TV spectrum.

In the meantime, we all fly along blissfully unaware that a serious failure due to TV could be around the corner at any time.

[DanH]

<<This is unlikely to happen any time soon for most drives due to the sheer amount of time needed to collect data for multiple propellers and engines.>>

Perhaps not quite as true as common knowledge would have you think. Consider the simple two-element model, two inertias conected by a stiffness (think two disks connected by a shaft). If the two inertias are equal they will oscillate in equal opposition. We're a long way from that; propeller inertia is much higher than crank/flywheel inertia. The vibratory motion can be thought of as an oscillation of the small inertia while the large one remains more or less stationary. Put another way, the large inertia is the anchor against which the small inertia oscillates. It makes little difference if the prop inertia is 5x, 7.5x, or 10x the small inertia. An anchor is an anchor.

Here's the frequency equation for a two element model. Plug in a few variations for J1 and see how little it changes F. Use 1, 1.5 and 2 slug-ft^2 for J1, 0.2 for J2. Use 10,000 ft-lbs/radian for K, a realistic shaft number.

F= 9.55 {square root of [ K (J1 + J2 / J1 J2)]}

F is in cycles per minute, so divide by 60 for hertz. Hertz to RPM for a 4-stroke (tells you critical RPM) is [(hertz x 120)/# of cyls].

Aw heck, I'll just tell you. The answers are roughly 39, 38, and 37 hertz. Assume a 4-cylinder 4-stroke. Switch props for one with twice as much inertia and you move the F1 critical to a point in the RPM range a whopping 60 RPM lower. You figure amplitude is gonna change very much?

I just hosted my whole EAA chapter at my shop and I've had two beers, so do check my math <g>

[rv6ejguy]

Always enjoy your posts Dan, makes me think. I buy the math and the basic premise.

I have heard some people who have changed prop types from wood to metal or composite and either reduced or increased perceived vibration considerably. I have seen this also recently on a 912 Rotax with three different props on my test stand with all props balanced reasonably well. Do you have some other angle on this observation?

Why is Hartzell doing so much testing on relatively similar props on relatively similar engines if there is theoretically so little difference? A quick calc shows a pretty big difference in MIs for a 60lb metal prop and a 12 lb wooden one as related to say a 25lb. flywheel turning at twice prop speed.

I'm still planning to try measurements with an RF chip and accelerometers- have to get the whole thing done first.

[DanH]

Ross,
<<changed prop types from wood to metal or composite and either reduced or increased perceived vibration considerably.>>

I believe that. The previous example was a simple two-element model, fine for illustrating the point that you have reasonable latitude in prop substitution as compared to a baseline propeller. For example, if a vendor recorded acceptable shaft telemetry numbers with a one-piece wood prop, he could safely assume that any other wood prop of similar construction that would actually fly the airplane will be about the same. That cuts the volume of "required testing" considerably and builders would have safe leeway in prop selection.

BTW, I have some live telemetry plots comparing back-to-back runs of two wood props, one a classic eliptical blade form in hard maple and one a tapered "toothpick" in mahogany. Ain't much difference.

You will start seeing more difference as you swap for propellers of different construction. The key here is construction; the difference in construction is not just a matter of material or inertia. Our real systems have many inertias connected by many stiffnesses, not all of which are shafts. One of them is blade root bending, which is assigned a shaft stiffness equivelent for analysis (same for belts, BTW). Go back and run the two-element equation again, substituting 5000 ft-lbs/radian for 10,000, the same half-as-much or twice-as-much as we used for inertia. You'll see that the frequency changes more than it does with a half-as-much inertia change. Stiffness is a bigger deal.

Let's compare a one piece wood prop to something like a GSC ground adjustable wood prop that happens to have the same inertia. The GSC has small round blade roots to allow pitch changes in the hub clamping, while the one-piece prop has fat, thick roots. Which will bend easier?

Now combine a change of stiffness and a change of inertia. A heavier prop will drive frequency down, but if it has a neatly compensating increase in root stiffness you may not see a change. A heavier prop with really stiff roots might push frequency up, perhaps the substitution of a carbon Warp Drive w/ the aluminum hub for that wood GSC. Substituting a one-piece metal prop with thick roots (a really, really stiff root) might push it up even more...or down. Depends on the balance of stiffness and inertia.

Pushing frequency up leads to the first factor in vibration perception, as well as an important point about actual vibratory amplitude. Remember what I said in the previous post about amplitude being roughly proportional to manifold pressure? The above calculations told us the system's natural frequency. Resonance happens when that natural frequency is matched by an exciting frequency. The primary exciting frequency is gas pressure oscillation, ie, firing events. You've seen lots of torque curves, so you know the force powering the exciting frequency is rising with RPM, sometimes sharply, in the lower end of the engine operating range. Moving the resonance point up the range as little as 200 RPM can increase the resonant amplitude enough to be noticable with weak human perception.

Weak human perception plays out in lower RPM resonance too. We perceive lower frequencies far better than higher frequencies. Might still be the same measured vibratory torque amplitude, but a lower RPM shake is more noticable.

Last, consider the natural frequency of airframe components. If, for example, the natural frequency of the horizontal stabilizer happens to be the same as the resonant vibration shaking the engine, your perception will be that the vibration is much worse. Heck, the whole airplane is shaking like a wet dog!

Human perception is a terrible measure for this work.

<<Why is Hartzell doing so much testing on relatively similar props on relatively similar engines if there is theoretically so little difference?>>

Well, don't confuse propeller life with drive life. Aluminum, unlike steel for example, has no "knee" in it's S-N (load vs cycles) curve. In the words of my engineering mentor, aluminum is frozen mush. If you stay well below the knee with a steel component, theory says it's fatigue life is more or less infinite. Aluminum has no knee; with enough cycles all aluminum components break, as the curve eventually reaches zero. In simple, Hartzell wants to make sure that all points on the blade have loads very low on the S-N curve so the thing will stand a lot of cycles before it turns to mush and ruins your day.

Like that horizontal stabilizer, a prop blade has a natural frequency or frequencies. An aluminum propeller's natural frequency is somewhat higher than the system's F1 frequency we've so far discussed, and by hand methods is a bit difficult to calculate. The primary reason is centrifugal force, which stiffens the blade with increased RPM. The second reason is that blades don't have uniform sections; it is difficult to accurately calculate bending. The higher exciting frequencies that might make it resonate stem from events other than just firing events; note that the conventional aircraft engines known to be easy on props have pendulum absorbers tuned to 5th, 6th, etc orders. I'm sure the propeller wizards have good computer tools for calculation these days, but they still find it sensible to pull live telemetry. A failed drive is bad enough; you find yourself flying a glider. Propeller failures can result in the loss of the whole engine, and the occupants wind up dead.

I've not studied propeller vibration very much, but I think bolting an aluminum propeller to an auto conversion with unknown torsional amplitudes is akin to playing Russian roulette. You put two bullets in the cylinder if you choose a used propeller; some of it's available S-N life has already been used up.

Did you ever buy that copy of DenHartog I recommended? I think it has a good section on propeller vibration.

<<I'm still planning to try measurements with an RF chip and accelerometers- have to get the whole thing done first.>>

Still think you would do better if you went straight to shaft telemetry, but so many guys have talked about using accelerometers that I'd like to see somebody actually try it. You're smart enough to do it. The problem will be sorting through all the other shakes to find the ones that stem from torsional events. You'll need some software.

[DanH]

Ross,
A follow up on your question.

It has been a while since I was into this stuff, so I dug around on my hard drive for an old tool, a simple DOS program for Holzer calculation written by Tom Irvine. Has no inputs for engine output so it won't tell you vibratory amplitude, but it will generate pretty accurate numbers for the natural frequency of as many elements as you wish. They can be used to go further, but I have no plans for that. What I wanted to confirm was the statement about changes in prop inertia making very little difference, but this time with a more realistic 4 element model and some real data from the old Suzuki drive.

All torsional systems have a natural frequency for every connecting stiffness, or put another way, inertias minus one. Thus a four element model generates 3. Generally they increase as you move further along the string away from the propeller. We already looked at F1 with the blunt two element model. The specific issue I wished to review was what happens to the higher frequencies with shifts in prop inertia and blade root stiffness. Could they be the source of your observations?

Answer is "no". Plugging in real inertia and stiffness data from the old bad Suzuki drive generated an F1 of 46 hertz, or a resonant RPM of 1840 with a 4-stroke I-3, exactly as you saw in the graph posted previously (which was generated with far more elaborate software). The F2 is 190 hertz or 7600 RPM, which is why you don't see it as another peak on that graph. F3 would be of concern only if we reached 16,680 RPM <g>

Ok, so triple the prop inertia and cut root stiffness in half. The answer is an F1 of 38 hertz and an F2 of 156 hertz. They would be resonant at 1520 and 6240 RPM with the Suzuki I3.

So, only the F1's are of great interest in this case. The others are out of the operating range, which is highly desirable. An additional 320 RPM (1840 vs 1520) would probably increase resonant vibratory shaft torque because, as previously stated, engine oscillating torque output (the forcing vibration) is higher. "How much" would need amplitude calculations or telemetry, but more is safe bet. Even so, it is entirely possible that an observer would claim the lower frequency combination to be worse, as it may well be reaching down into the range of a lot of airframe natural frequencies. Consider: a steel tube fuselage has a torsional stiffness and mass at each end. What do you have?

Anyway, an anchor is still an anchor. Changing propellers makes a difference but it ain't really a big deal in the strict context of shaft stress in the PSRU. Vibration of the prop blades themselves is another matter.

[rv6ejguy]

Without proper measurement equipment at this time, I can only make inferior human based observations. If your example is typical with F1 happening at relatively low rpms and low frequencies, this is what I typically see in my Sube installation and the Rotax on the test stand and apparently on the new Egg H6 packages. Perhaps coincidently, most of these have serious TV below 1200 rpm.

What we don't know in the Sube packages is where F2. F3, F4 etc. are as they are at higher frequencies that we can perceive without instrumentation. The Rotax has millions of hours of flight time so either they have done testing or there is no problem at typical operational rpms. My sources at the Rotax overhaul facility sees plenty of messed up drives and in almost every case, this has been attributed to excessive low rpm idling, contrary to Rotax directives. Engines which are idled over 1400 seem to have few gearbox problems.

I have very noticeable periods at 600-950 and 1100 to 1600 rpm so idle is set to 1000 and taxi above 1600. With the geared prop, this works out fine. In flight stuff is a total unknown at this point.

I'd be interested to hear from any others flying auto conversions with redrives and any noticeable rpm zones.

[DanH]

Ross,
<<Without proper measurement equipment at this time, I can only make inferior human based observations.>>

Of course.

<< If your example is typical with F1 happening at relatively low rpms and low frequencies,..>>

Very likely. Prop size, shaft size, and engine size tend to be relative, yes?

<<What we don't know in the Sube packages is where F2. F3, F4 etc. are..>>

Calculating all the frequencies isn't hard, IF you can take the time to gather accurate stiffness and inertia values.

<< have very noticeable periods at 600-950 and 1100 to 1600 rpm so idle is set to 1000 and taxi above 1600. >>

You mentioned that lower range issue in a previous thread and I didn't think much about at the time. Sorry. Ain't no mystery.

We spend most of our worry on natural frequencies that intersect with the firing frequency. The variation in crankshaft angular velocity due to firing events is by far the most powerful exciting frequency, but it is not the only one. Next on the list is the recip frequency, a variation in crank speed due to decelerating and accelerating pistons. Intersecting F1 with this exciting frequency will give you a resonant RPM just like intersecting the firing frequency, but it usually doesn't generate nearly as much amplitude.

With your 4-cyl Sube, firing events are a 2nd order frequency, meaning they happen twice every crank revolution. Recip is a 4th order frequency. I'll assume a 48 hertz F1 based on your "between 1100 and 1600" observation; that will be close.

The easy way to visualize all this is to plot a Campbell diagram. I've attached one below.


Shot at 2007-07-15

[DanH]

The thread started with reference to a BAP drive, so let's take a quick look at a Campbell diagram for the 6-cylinder.

A 6-cylinder has exciting frequency major orders of 3, 6, 9, and 12. Only 3 and 6 are of interest here. I'll assume an F1 of 48 hertz like the previous example.

As you can see, the system should be resonant at 480 and 960 RPM. You may be tempted to say "Eureka! That's why the BAP drive works!", since the 6th order intersection is pretty much below the operating range and the big, bad 3rd order is merely on the edge of it. There are a few details to keep in mind.

(1) Resonant ranges typically extend across a range from 0.8 to 1.2 times the critical speed, more or less. The peak isn't always in the exact middle.

(2) I've assumed a 48 hertz F1. We don't know the exact F1 of the real system. A little lower would be better, a little higher makes things worse.

(3) You would still have a starting transient, and an operator who sets idle speed too low would still have problems.

As Ross asked, is there anybody here you can report actual operating experience for this system with an eye toward the resonant range? The physics don't lie; it is there. Question is, where?

BTW, you might wonder why I posted this diagram when it seems in favor of the BAP drive. If you have that idea, you may also have a misconception about my goals. I'm not here to knock Jess's work. I'm here to understand, and perhaps spread a little understanding in the process.


Shot at 2007-07-16

[rv6ejguy]

Good point, I think we almost assume something bad may be happening in the flight ranges since we don't know without instrumentation. In many cases there may be nothing to worry about but it is not knowing that is scary.

I see that vendors may be reluctant to admit that they have not tested this aspect of their packages but seems to me it would be a selling point and costs would not be too bad and results reasonably applicable to the typical narrow range of generally composite prop types recommended. I was always under the impression that the prop would make a larger difference in range.

With 75 of Jess' packages flying and what must be thousands of hours on them, it would be interesting to get feedback on reliability to date.

[DanH]

Ross,
Again back to a previous conversation;

You now understand why I had reservations about switching to harder urethane bushings in your drive. A lower torsional stiffness is necessary if you wish to reduce shaft load by eliminating the need to pass through a resonant peak every time you come up off idle. That may not be possible with urethane bushings, but they are a miserable choice of soft element anyway.

Lowering F1 would probably lower F2 also. The goal is to drop F1 as far as reasonable without allowing F2 down into the top of the operating range. I don't know where it lies right now.

An accurate model might let you pass on telemetry. You gather stiffness and inertia data, then compute frequencies. If you're sure F1 is below idle and F2 is above redline, you might reasonably decide there is nothing critical to measure. If there are no critical intersections, even amplitude calculations are moot.