Low-throttle driveline oscillations

[pmbrunelle]

Sometimes, when driving slowly with near-zero torque (torque converter locked up), there can be an unpleasant bucking sensation in the car as the engine inertia oscillates with the driveline lash and compliance.

I do not think that consumers would accept this behaviour anymore.

How are these oscillations normally dealt with? Suppose a new-ish vehicle with a DBW engine, and a CVT.

 

[GregLocock]

I've forgotten the term for this, basically the referred inertia of the driveline is bouncing on the engine mounts and the spin compliance of the tires (etc), torsionally. It can get very complicated. Obviously you need to eliminate backlash where possible, and consider the tuning of the engine mounts. It may be that someone has a posh model of the driveline that can investigate this, typically in the bad old days I'd have headed down to the track with a box of bits. If this is something that has developed over time I suggest your vehicle may be being used by someone overly fond of jack rabbit starts, or journalists, replacing the engine mounts might fix it.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[pmbrunelle]

I think this is called "shuffle".

The nearly brand-new vehicle did not oscillate with its factory transmission.

When we fitted a prototype transmission (also a CVT) to it, the oscillations appeared. Nothing else changed; same engine/diff mounts, same tires, same half-shafts, etc. I have full access and control of the software algorithm which controls the prototype transmission.

I have read papers online that describe an anti-shuffle controller contained within the software of an engine controller, which modulates the torque output of the engine, probably via the fast path (ignition timing). The anti-shuffle controller is perhaps tuned to damp a specific oscillation frequency, or may be designed with a specific driveline model in mind.

The engine controller on this vehicle is a black-box to me, so I don't know if it implements an anti-shuffle controller. If there is an anti-shuffle controller, maybe we changed the driveline too much for it to work as designed?

If a different engine mount would need to be installed in order to use this transmission, I don't think that would be a showstopper.

 

[GregLocock]

I'm guessing 3 Hz or so? Your CVT is a direct torque coupling between the engine and the driveline, rather like a manual trans (as opposed to a proper fluid flywheel which can absorb various indignities). I like the idea of suppressing this by comparing say ABS tonewheel and crankshaft tonewheel. However back then we had no such luxuries.


Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[pmbrunelle]

Good guess! It's happening at about 2.5 Hz.

I don't have direct access to the tonewheel sensors, only through their respective modules and the CAN bus, but since the oscillation is a fairly slow phenomenon, the speed information available through the CAN bus might be fast enough.

Suppression by the engine controller sounds like a good idea, but that's out of scope for this project.

It might be possible to suppress this by varying the CVT ratio, but I'm not sure. The CVT is next to the engine; the elastic and backlash-producing elements are downstream of this.

 

[GregLocock]

I don't think the CVT ratio will help much - but it is a quick and dirty experiment. What is the configuration of the powertrain - FWD, AWD, transverse engine, etc?

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[pmbrunelle]

What are you getting at with the powertrain configuration?

The number of gear reductions, shafts, joints, and opportunities for backlash?

Whether the torque that acts on the engine mounts is equal to engine torque, or if it is also multiplied by the transmission and/or final drive?

I don't want to identify the vehicle on a public forum, so I don't feel at ease to state the powertrain configuration.

 

[Tmoose]

My 2013 Mazda CX-5 FWD with manual transmission loves to buck at small throttle openings and low rpm in first gear.
It also spins its tires easily at pretty modest 1st gear starting acceleration.
That has been a little surprising with the front wheel biased weight distribution.
I'm attributing both in part to such a direct non-compliant link between engine and tire

My Manual transmission skills undoubtedly are not what they were 30 years ago.
(The older I get, the faster I was)
Throw in some foot neuropathy to deaden clutch and throttle feel and I wonder how good I'll get in the next couple months.

I'm fairly convinced that the drive by wire throttle is biased with quicker opening to make the car feel peppier in normal driving.

I would not put it past one of them computers making decisions about how quick to let revs drop during shifting.

 

[GregLocock]

"
I'm fairly convinced that the drive by wire throttle is biased with quicker opening to make the car feel peppier in normal driving."

Yep that's called performance feel and replaces the old snail cam on a carbie

"I would not put it past one of them computers making decisions about how quick to let revs drop during shifting."

Again yes, this time for emissions reasons and smooth gear changes.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[pmbrunelle]

In a semi-related topic, an overly sensitive throttle and strong engine can cause driver-induced oscillations, where acceleration causes the driver's foot to lift off the pedal, then the wheel torque is reduced, the driver's foot falls backs on the pedal, and so on. This can be worse on bumpy roads.

However, the phenomenon I was talking about in this thread happens with fixed accelerator pedal position.

 

[gruntguru]

Is the powertrain configuration really so unique that someone here might guess what make/model we are talking about?

Shuffle huh? Intuitively there must be an excitation - something that adds energy to the system like the tiny impulses delivered from a clock spring to the escapement. If it was all inertia vs compliance it would surely damp away to nothing. Something to do with the slope of the engine torque curve at the given throttle opening? Perhaps that torque characteristic has a slight hysteresis?

Not suggesting a solution here - just curious to learn about the phenomenon.

Perhaps add an engine "steady" in the form of a damper rather than a spring? Link



je suis charlie

 

[gruntguru]

After further thought:
[ul]An engine damper is probably not a fix - probably only small displacements there.
Excitation is most likely hysteresis in engine torque. MAP (and therefore torque) lags rpm changes due to plenum volume. Throttles at the port would probably not experience the problem.
There may be an additional excitation due to the lambda feedback loop - it also has a resonance around 2 - 3 Hz. Might be worth opening the loop to check for any effect.[/ul]

You haven't mentioned whether this vehicle has DBW throttle - that would surely be a factor, with the factory controller possibly changing throttle opening with rpm changes - even with constant pedal position.

je suis charlie

 

[3DDave]

OP said "DBW engine" which is the same thing as DBW throttle.

I would not expect to figure this out without instrumenting the crank position, the ignition at each cylinder, the fuel delivery pulse at each cylinder, the CVT position, and CVT command, and one or more accelerometers. I'd start with a 1kHz sample rate, but those who do engine design likely have a better suggestion for time base. Probably need a full chassis dynamometer.

Something is either exciting the 2.5Hz natural frequency or is driving it in spite of it not being a natural frequency.

 

[pmbrunelle]

I don't think the CVT ratio will help much - but it is a quick and dirty experiment.

After a few weeks on another project, I was able to return to this topic!

Varying the CVT ratio didn't have much effect on the shuffle.

Something to do with the slope of the engine torque curve at the given throttle opening? Perhaps that torque characteristic has a slight hysteresis?

Yes, it seems to be related to this!

The engine calibration has an Eco/Sport mode selection. Changing between those two calibrations (presumably with some effect on the engine torque curve slope) affects the tendency to shuffle.

Also, with my transmission calibration, I was targeting lower cruise RPMs (for passenger noise/comfort reasons) than what the vehicle originally had from the factory. I think that the delay/hysteresis phenomena of the engine are more apparent with lower RPMs.

From the factory, I suspect that the engine was never calibrated/expected to run properly in the low-RPM regime as I tried to do. So, if I really want to cruise at lower engine RPMs, then maybe the engine calibration would need to be adjusted, if there is not a mechanical limitation (camshaft timing?) that cannot be worked around.

 

[BrianPetersen]

You haven't told us what engine it is, in case someone here knows something about it.

Other drivetrain layouts can do something like this, too. We called it "snatching", and my carbureted gen 1 Honda Civic was terrible for doing that when puttering around in first gear (manual) at parking-lot speeds.

What's the frequency of the engine rocking in the engine mounts? I'm thinking it's going to be in the same vicinity, thus prone to being excited by something/anything else.

It takes at least one engine revolution for any movement of the throttle to show up as a change in a firing impulse, and that's on top of the delays associated with the volume of the intake plenum filling and emptying and the delay associated with the mechanical inertia of the engine. Control loops with a response delay in them are hard to tune. At 1200 rpm that's 0.05 seconds just for the one-revolution delay between throttle opening (leading to an increase/decrease in charge during an intake stroke) and the increase/decrease in crank speed associated with the stronger/weaker power stroke. And the intake charge mass is extremely sensitive to throttle angle change at low engine revs and light load.

The frequency of not enough - just right - too much (command closing) - just right - not enough (command opening) would suggest around 4 four-stroke cycles per oscillation cycle, 8 crank revolutions per oscillation ... plausible for this to be in the 2 - 3 Hz range.

It's highly likely that the OEM calibration took this into account and set minimums on the engine output torque and transmission input shaft speed (combination of road speed and CVT ratio), below which they unlock the torque converter rather than trying to fight with the calibration.

 

[pmbrunelle]

You haven't told us what engine it is, in case someone here knows something about it.

Well now it's two engines (on two different platforms), but the engines are similar. They are 1L, 2 cylinders, 4-stroke, idle around 1500 RPM, rev limit around 8000.

With the OEM calibration, cruise was in the 4000+ RPM range, but some users could perceive this as buzzy/unpleasant. I am trying to operate down around 2500 RPM, but apparently this is challenging.

What's the frequency of the engine rocking in the engine mounts?

That's an interesting point... I guess I could estimate the stiffness fairly easily by applying a torque, and measuring deflection with a dial indicator or something.
Mass moment of inertia is probably going to be tricky... I guess I would have to pull the powertrain and hang it from a rope to do a torsional pendulum experiment (right now there is not much tolerance for vehicle downtime).

It will be simpler for me to modify the engine mounting to change the frequency to... something else.

 

[BrianPetersen]

Oooo-kay, that changes things a little. This sounds like a snowmobile or ATV application. I'm rather familiar with big-ish-twin motorcycle applications - but only in hire bikes. I don't own one.

All of the ones that I have rented, mostly of the BMW F800 or F900 series, have been quite unhappy about being under any appreciable load down low in the revs. You start feeling every cylinder firing, and it starts feeling coarse below 4000 rpm or so, and I don't like it. I recall the F800 wanting to be in third gear at around-town speeds, and forget about sixth unless you were on the motorway. My own four-cylinder bike is fine at 2500 rpm in city traffic.

Is this CVT a snowmobile type with centrifugal clutch (or a centrifugal belt-clamper) and all mechanical controls, or is it an automotive type electronically controlled with fluid torque converter and electronically controlled lock-up clutch? I want to be clear about terminology, because the snowmobile community tends to call the belt-and-cone-pulleys mechanism the "torque converter" and the centrifugal mechanism that clamps the input cone-pulley together the "clutch".

 

[BrianPetersen]

Also, for now, don't get too scientific about the engine-mount frequency. Grab the (not running) engine by hand and give it a good shove against its mounts, and watch what happens. If they're soft enough to matter, you can see the motion and get a visual estimate of the frequency. If you can barely get it to move at all (solid mounted, or nearly so) take a video of the engine while driving (camera solidly mounted to nearby chassis works) and see if it's doing anything.

 

[pmbrunelle]

There may be an additional excitation due to the lambda feedback loop - it also has a resonance around 2 - 3 Hz. Might be worth opening the loop to check for any effect.

I unplugged the oxygen sensors, and then a bunch of oscillations smoothed out; the driving became better. It looked like this was a big factor.

Then, I reconnected the oxygen sensors, power-cycled the ECU, and the system remained non-oscillatory

Intermittent borderline problems...

Is this CVT a snowmobile type with centrifugal clutch (or a centrifugal belt-clamper) and all mechanical controls, or is it an automotive type electronically controlled with fluid torque converter and electronically controlled lock-up clutch? I want to be clear about terminology, because the snowmobile community tends to call the belt-and-cone-pulleys mechanism the "torque converter" and the centrifugal mechanism that clamps the input cone-pulley together the "clutch".

The CVT ratio varies with an electric actuator (computer-controlled). I can't say much more

Also, for now, don't get too scientific about the engine-mount frequency. Grab the (not running) engine by hand and give it a good shove against its mounts, and watch what happens. If they're soft enough to matter, you can see the motion and get a visual estimate of the frequency. If you can barely get it to move at all (solid mounted, or nearly so) take a video of the engine while driving (camera solidly mounted to nearby chassis works) and see if it's doing anything.

I didn't remember to give the engine a tug yesterday, but by visual inspection, the mounts are rather far apart, so they're probably stiff enough in rotation.

When my colleagues were operating the vehicle, and I was running alongside looking at the engine, the engine didn't appear to rock. A chassis-mounted camera would have been better, of course.

********************************************************************************

I have more or less come to the conclusion that I'll only be able to really fix the shuffle if I can modify the engine's calibration to suit the new low-RPM running conditions.

 

[BrianPetersen]

Re the O2 sensor unplug and reconnect ... When you unplugged them, it would have gone into an open-loop mode. If the engine controls have any similarity to OBDII, it probably has a fault code stored, and remembers that it's unhappy through a power-down, and it may take a few drive cycles to forget its unhappiness, even after you've reconnected the sensors.

Or, maybe, it contains adaptation "self-learning" logic, whose learned settings (or lack thereof) were unsuitable for what you've done, and by doing what you've done, you made it forget and re-learn.

The automotive ECUs contain short-term and long-term correction factors in maps in volatile memory which they update based on what they see from the O2 sensor when the engine runs at each speed and load setting in the map, so that the next time it sees that speed and load setting, it can jump right to the correction factors it previously saw. Probably if you got the engine to run at lower revs but higher load than the stock calibration allowed, you may have been "off the map" (of what it previously saw). Unplugging the O2 sensor may have had some effect on this, and possibly made it reset the short-term and long-term correction factors.

Anyhow, if it's stopped acting up, the easiest thing to do for the moment is to just keep driving it for a while, and get a few cold-start and warm-up cycles on the books, to see if the problem comes back.