# Driveshaft development

### Help Support HomeBuiltAirplanes.com:

#### Autodidact

##### Well-Known Member
I could not have done any of this without Dan's guidance.
But you did get it done, and I can't say that I don't admire you for that.

Bret

#### wsimpso1

##### Super Moderator
Staff member
Log Member
Help me with my math?
If I happen to have 45 tooth cam drive gear wheel and the engine is spinning at 6000rpm, the tooth signal is 4.5KHz. 10 bits of time resolution should be overkill? That's 4.5Mhz sampling. Ok, that 45 tooth wheel may not be adequate for high harmonics in a stiff system, but it is surely overkill in a soft system? Yes, i could build a system to do that. But there's an easier way:

Wright did not have 10GHz sampling when they found and fixed the 9th harmonic in their Cyclone. They had an analog mechanical system. I believe they had to make it themselves, though.
If I modernised the Wright system, I'd have a tangential accelerometer rigidly mounted to the shaft, along with an FM transmitter and a small battery. An analog radio receiver (cheap constant latency) would pipe that to a PC soundcard. Yes, I'd need at least two systems and a soundcard channel for each.
Recorded waveforms don't need crunching in real time. Modern electronics are wonderful! The 50th harmonic of 6000rpm would generate a 5kHz signal. A cheapo 44kHz sampling soundcard would not be fazed at all by that. A 192kHz commonly available card should do it easily. Some butchery may be needed to extend the low frequency end for startup and shutdown.

What I will say is that if you need to ask how to do measure and fix TV, you probably won't be able to...

As an aside, while my redrive is still in the doodle ideas and see how the numbers look stage, I have just ordered a bunch of data aquisition stuff for measuring IMEP and pressures in an engine intake system you'd likely laugh at.
Cheap USB soundcards. I will convert these to DC input and I have lots of cheap 22kHz analog inputs. I'm a bit dubious about the latency, but that isn't actually a big problem for my upcoming testing as phasing is not critical.
It sounds like you are way above most of us on measurement. After that you will need to do FFT and be able to interpret the results. To also be able to do the design and fab work is a really rare set of skills. Have at it.

Let's remember we are talking about the OP's system. You will need on the order of seven independant channels to run his whole system and be able to catch the interactions. Biggest issue with accelerometers will be noise filtering.

I really like laser tachs. Add zero mass to the wheels as they use printed paper targets, and the laser tach is already low noise. Downside is cost. In the business, we usually just use mag pickups on ferrous splined parts, then get fancy (lasr tach) with non-ferrous parts and really light parts. Rotec and its ilk records the time of passage of each tooth up and down, and so is inhereltly pretty clean on noise. They also have all of the analytical software and lets you analyze any channel either by itself or in relation to any other channel. On a home grown system, well, you had better know how to do all of that and then be a talented vibration system engineer too.

Billski

Last edited:

#### pictsidhe

##### Banned
I think strain gauges is needed. Accelerometers would be overwhelmed by normal vibration.
Another option is strain paint. Paint the crank and run it two hours and disassemble and check for strain.
If any exists, then repeat to find the rpm... or give up
Shaking vibration can be cancelled out with a 2nd accelerometer on the shaft axis. that assumes you have the dynamic range and resolution, anyway. I haven't actually run any numbers on this idea yet, I'm not sure that I'll need it.
Any pickup can have noise. A magnetic sensors output will vary a bit with tooth imperfections and as the distance between sensor and gear change due to case flex. Those optical encoder discs would need extreme accuracy to not supply a lot of jitter, so using their full resolution seems unwise to me. Laser tachs are also going to be prone to vibratory position errors. Some of this stuff can be compensated for, some of it filtered. Expecting and understanding unexpected results is half the battle, though...
On my current favourite system, I could easily measure the angular deflection of my soft link. once I know what parts I will have and what I need to measure, I can pick a suitable measurement system.
For a soft system, a maths analysis and measurement of the lowest orders via cam drive gear and soundcard could be enough. With prop inertia far larger than the engine, a single channel would give meaningful data. While higher orders can play merry hell with new engines, I'm taking an existing engine that works with a variety of small inertias and giving it a softly sprung big inertia headache. So I'm reasoning that I won't need to worry about high orders. This won't stop me finishing my Holzer spreadsheet and seeing what it thinks first, though.
A soft system is most likely to have issues at idle, startup and shutdown. A bit of design effort and measurement of that critical lower end will likely be enough. Ross didn't even measure and has yet to break his redrive. Though I have noticed that his RV hours have been stuck at 422.7 for a suspiciously long time
I am certain that there are numerous people on here smart enough to engineer a redrive with a bit of effort. Design a Wright Cyclone from scratch? That would be less of you...
I ordered Den Hartog a couple of days ago, that should help.
I have other stuff ahead in the queue, so I've just been checking basic feasibility to see if I can get it working on paper at an acceptable weight and complexity.

#### Doggzilla

##### Well-Known Member
HBA Supporter
Using a sensor that operates at only a specific frequency will automatically cancel out other vibrations.

And if the frequency is higher than the latency of the measurement, it will automatically ignore it. Only frequencies lower than the latency can register.

#### rv6ejguy

##### Well-Known Member
A soft system is most likely to have issues at idle, startup and shutdown. A bit of design effort and measurement of that critical lower end will likely be enough. Ross didn't even measure and has yet to break his redrive. Though I have noticed that his RV hours have been stuck at 422.7 for a suspiciously long time
If I operated my engine in the 1050-1200 rpm range for long, things would have broken long ago which is why I did the flywheel mod. This period was really severe with the wingtips and panel, a blur. You moved through it quickly. The rubber drive bushings were pretty rough at the 358 hour mark when examined.

TV was working on other parts when examined...

I don't get to fly much these days, but the figure has incrementally moved up from 358 hours since the big mod session in 2013. I was up flying last weekend in the terrible forest fire smoke that's plagued us here for the last 6 weeks. The flight schools have flown almost zero here for 2 weeks now.

Looking forward to your study. It's a fascinating field.

#### BBerson

##### Light Plane Philosopher
HBA Supporter
The $1500 electronic prop balancer I bought is almost useless on a V-twin. I don't have any interest in the$5000 version.
I will just live with vibration like any Harley owner.

#### pictsidhe

##### Banned
ouch, that's a lot of money to drop on something that doesn't work. Did you try it measuring vibration in the horizontal axis? A 90 degree v-twin is pretty well balanced in that axis, unlike the vertical one.

#### BBerson

##### Light Plane Philosopher
HBA Supporter
I don't know if it makes any difference. Never used the balancer for anything else to see if it's worth owning.
The engine may have been misfiring enough from a carb that was partially clogged by insect daubings. I just haven't got around to trying it again. It became obvious to me the individual power pulses overwhelm even perfect rotational balance. So what's the point?
With no prop the engine runs almost dead smooth unclamped on the tabletop because the torque pulse is low. Put a prop on and it shakes because it's working itself as hard as it can against the prop load. Just like a four cylinder car shudders going up a max load hill at full throttle in high gear. The only real smooth solution is four cylinders or more.

So it will shake some on climb out at full power. Not bad at half power cruise.

#### cheapracer

##### Well-Known Member
Log Member
Try throwing a dart backwards sometime ...

Thank you for explaining why canard pushers can't fly.
That's a terrible analogue for an airplane, and again, if you look at any jet, it has engines anywhere but the nose. GA aircraft with engines in the nose makes sense for air cooled engines using direct drive, but otherwise is merely a convention borne out of historical convenience, and the misconceptions of its contributions should be rooted out, and die horrible deaths.
You said: "I'm not sure where the notion that the weight of an aircraft should be hanging off the nose comes from."

Now you are stipulating engine position. Maybe you should be a little clearer in your writings, not that it changes the fact that regardless of engine position, the weight should be forward of the aerodynamic center.

... just like a dart.

Thank you for explaining why canard pushers can't fly.
They fly with stabilty using the same pyhsics that makes a dart stable.

#### Doggzilla

##### Well-Known Member
HBA Supporter
That's not how balancers work, guys.

They don't stop vibrations, they counter the load being off center and reduce fatigue on the bearings and drive components.

They have absolutely no effect on internal engine vibrations.

#### wsimpso1

##### Super Moderator
Staff member
Log Member
Any pickup can have noise. A magnetic sensors output will vary a bit with tooth imperfections and as the distance between sensor and gear change due to case flex. Those optical encoder discs would need extreme accuracy to not supply a lot of jitter, so using their full resolution seems unwise to me. Laser tachs are also going to be prone to vibratory position errors. Some of this stuff can be compensated for, some of it filtered.

Oh yes, one of the neat features of Rotec and the other commercial products is they have the ability to recognize a skipped or extra tooth and to record a target and get its tooth variation pattern, then correct for all of these errors. Nice to have, and absolutely necessary for a target strip that is glued on a wheel, as there is a discontinuity at the overlap... Without correction for these "features", they may dominate your data and not allow extracting meaningful results...

Billski

#### dsigned

##### Well-Known Member
Again, a dart is a terrible example to illustrate the point you're trying to make, as its trajectory is ballistic. For all intents and purposes, it has no center of lift. Secondly, no one is suggesting having the center of mass behind the center of lift, and having the engine in the nose is clearly not necessary to accomplish that, otherwise the aforementioned canards, and nearly every jet ever produced would be incapable of flight. You're making an inferential leap from "the center of mass must be forward of the center of lift" to "the engine must be in the nose". One of these is true of just about every aircraft I'm aware of, and the other is clearly not.

#### pictsidhe

##### Banned
Oh yes, one of the neat features of Rotec and the other commercial products is they have the ability to recognize a skipped or extra tooth and to record a target and get its tooth variation pattern, then correct for all of these errors. Nice to have, and absolutely necessary for a target strip that is glued on a wheel, as there is a discontinuity at the overlap... Without correction for these "features", they may dominate your data and not allow extracting meaningful results...

Billski
This is why Ford, GM etc will merrily pay big bucks for one of these systems: it doesn't take a week or three to set a system up and start getting hard data.
I'd rather not use skipped tooth. If I did I'd have two sensor physically in pahse, then maybe OR the outputs into my recording thing. That still wouldn't be as nice as a synthesised waverform, though. The correction for off centre targets is going to be pure gold in real life. I may look at the maths that would take. Though I will likely have high quality metal targets to use that won't need correcting.

#### wsimpso1

##### Super Moderator
Staff member
Log Member
This is why Ford, GM etc will merrily pay big bucks for one of these systems: it doesn't take a week or three to set a system up and start getting hard data.
I'd rather not use skipped tooth. If I did I'd have two sensor physically in pahse, then maybe OR the outputs into my recording thing. That still wouldn't be as nice as a synthesised waverform, though. The correction for off centre targets is going to be pure gold in real life. I may look at the maths that would take. Though I will likely have high quality metal targets to use that won't need correcting.
Pictsidhe,

Well, I hope your targets are good... Most of the time, we powertrain geeks are pointing sensors at existing round objects to avoid adding inertia to our measurements that do not exist in the field. At the least added inertia will cause some of the resonant modes to show up at lower rpm than in the real vehicle. They can cause resonance between parts that is not there for real, or worse... That is the problem with adding target wheels - we are changing the inertia of the very inertia we are trying to measure. Another reason we like to avoid adding target wheels if we can make do with what is there is that if there is much vibration, making the target wheel stay put can be a challange. Normally aspirated four cylinder piston engines with a 32 pound torque converter and 5 pound flexplate/ starter ring gear will generate firing pulses of 2500 radian/s/s outside of resonance, turbos and diesels can go much higher. Put a more nominal flywheel on the engine and attempt to isolate the rest of the system from it, and numbers 50-100% higher are possible. Run through a point of resonance during spin up or shut down and it can go quite a bit higher than that, and it is all trying to slip the target wheel on its shaft.

So, the guys who do this a bunch tend to use existing parts with formed clutch housings, rolled splines on shafts, gears that have teeth cut, shaved, and ground with precision but eccentric OD. These all can give one per rev plus gear tooth order plus others that can be strong. The favorite for a clean engine signal is the starrter ring gear, and they are always a little eccentric. Then try putting a radial line target on a flat face perfectly centered. These will all give you a strong one per rev signal. Earing of drawn clutch housings will give two and four per rev signals too. So we love being able to spin the signal wheel at low power and calibrate the computer to correct for it. And a circumferential tape has a discontinuity that you have to correct for...Ugh.

One other problem with putting your own target wheels on something else is that they can allow some other device just below to telegraph, not strongly, but enough to mess up the signal...

So, good luck with your precision target wheels. Maybe they will work.

Billski

#### pictsidhe

##### Banned
I'm hoping to use existing parts, as you suggest, for exactly the reasons you explain... I'll have a steel cam drive gear and likely a steel belt sprocket at the engine side.
Any base error in the pickup won't change with rpm, whereas resonances will, so invariant measurement errors are identifiable and can be considered when looking at FFT plots. Calibrating them out automagically is trickier without your big buck system, though. I probably could do it, but it will be easier to just make manual corrections to the data unless I start doing a lot of engines.

#### Swampyankee

##### Well-Known Member
I'd not be surprised if the helicopter industry has quite a lot of expertise in TV: tail rotors are flying in a highly asymmetric field, and there is a long driveshaft.

One source of vibration I've not seen addressed is airflow through the propeller. Except for the very specific case of perfectly axisymmetric flow, that is that the prop's axis of rotation is parallel to the local airflow and there are no disturbances (for pushers, there are always disturbances), there will be at least one/rev and number of blades/rev vibrations.

#### wsimpso1

##### Super Moderator
Staff member
Log Member
One source of vibration I've not seen addressed is airflow through the propeller. Except for the very specific case of perfectly axisymmetric flow, that is that the prop's axis of rotation is parallel to the local airflow and there are no disturbances (for pushers, there are always disturbances), there will be at least one/rev and number of blades/rev vibrations.
Post 53 of this very thread does mention this topic.

#### wsimpso1

##### Super Moderator
Staff member
Log Member
Post 53 of this very thread does mention this topic.
The sticky that I wrote talked about vibration modes and resonance and forcing functions, but did not talk about what the forcing functions were. Part of the reason for that is simply that there are many forcing function possible.

Four stroke piston engines have big vibe input at firing order. Piston-connecting rod-crank geometry gives next biggest at twice firing order, with reflections at 4x and 8x firing. Then there are differences between cylinders giving you vibe at half the number of cylinders, the number of cylinders, and twice the number of cylinders. And you can have bank to bank differences too. Then there is crank resonance. This frequency is supposed to be above 2x firing at max engine speed, so you avoid the crank going into resonance with firing, but if you have a stiff system, this frequency can be transmitted downstream or a higher frequency down stream forcing function can feed back to the crank...

Go to the other end, and you have prop misalignment to the inflow giving you one per prop rev per prop blade from that, upwash ahead of the wing (in tractor props) and downwash behind the wing and behind the tail (in pushers) gives you one per blade per prop rev per surface's wake encountered. Smart folks designing pushers will not put tail foils at even multiples to each other and will not array them radially to the prop shaft - put some dihedral on them and/or put their mid line above or below the prop centerline. A two or four blade prop in the wake of a vertical and horizontal tail at vertical and horizontal through the prop centerline really gets wanged about as the prop slams through the wakes. These effects are usually small in tractor props, but can be big in pusher props.

Then we have intermediate inputs. Gear mesh frequencies and multiples are out there. If you have Carden joints (Universal Joints) in the driveline, they produce a strong 2 per rev of the shaft.

Then there is impact. If you open and close lash in gears, joints, etc, the impact can excite the all resonance frequencies below and near the strike speed. That is how you experimentally get a read on resonances near a part, instrument and hit it with a hammer.

So there are lots of possible forcing functions that can excite all of those vibe modes. Usually, you worry over engine firing order, twice firing order, misfire orders, prop one per rev per blade in tractor installations, and one per blade per rev per foil wake in pushers, and Carden Joint 2 per shaft rev. If these coincide with vibe modes, you can have resonance.

Have fun guys.

Billski

Last edited:

#### BJC

##### Well-Known Member
HBA Supporter
Wondering if anyone has information on the implementation of the driveshaft systems for the Stemme S10 or the P39 Airacobra?
P-39 drive shaft:

BJC