Need help with a strange cyclic car vibration

[brainz01]

See my post above on the data from my wife's car for comparison. Seems my app is telling me something directionally correct.

I did measure my car at idle yesterday and picked up a noisy signature from the water pump. It's really bad. Sounds terrible. I cant drive the car for fear of damage. Will retest when the water pump is replaced. I'm betting a bunch of vibrations go away.


But for science, here's the plot I got at idle. Interestingly, the first order vibration of the water pump is 18.5hz, which is 1100rpm. But the car idles at 750, so the pump must be on a smaller pulley than the crank. Check out all those orders! Note that they look pretty low energy at idle - - at least they do in this graph, I've also logged them much higher. I do wonder what they do at 4x the rpm.

[brainz01]

Originally Posted by eastbay

If the 997 suspension parts are the same quality as the 996, I would look at your control arms, ball joints, etc. in the rear hi torque areas. These are garbage components made with nylon on steel balls and bearings with no provision for lubrication. They wear fast and strangley and can cause no end of phantom and fleeting issues.

Yep. More or less the same stuff. I've only got 53k miles on the car, but not unheard of for parts to have failed already. I view the suspension arms as replaceable parts between 60 and 100k miles depending on usage. It's not necessarily that early for me to change them out.

[brainz01]

Also, I looked up the differential ratios. They're not identical front and rear, but close enough to 3.4.

There's no resonance at 3.4x wheel speed on any of my charts, which confirms why my prior test of running the car without the driveshaft resulted in no change in vibration.

Interestingly, all my vibration logs yesterday are lower than previous logs. The key difference versus prior is no right front axle (I ran a stub axle instead). But the seat of the pants vibrations and cyclic beating were unchanged. Weird.

[fanaudical]

Am hoping Sammy chimes in on this.

What I find really interesting about that last snapshot is that the 7or8X spike is still there (limited amplitude) while the car is standing still. That means (at least to me) that there may be a natural resonance of the entire structure at that frequency driven by engine operation.

[brainz01]

Originally Posted by fanaudical

Am hoping Sammy chimes in on this.

What I find really interesting about that last snapshot is that the 7or8X spike is still there (limited amplitude) while the car is standing still. That means (at least to me) that there may be a natural resonance of the entire structure at that frequency driven by engine operation.

I think bad roller bearings can also have this signature. I assume the water pump has a roller bearing, but I suppose it could be plain - - Not sure. New one is in the mail so will know soon. Could also be bearings on other drivebelt pulleys, I'm pretty sure those are sealed roller bearings.

[fanaudical]

The water pump roller bearings change frequency with engine speed.

You've got a spike above 137 Hz with the engine at idle and vehicle stationary; that spike is in the same place (higher amplitude) at higher RPM and ~80MPH. To me, that says something in the body of the car is "rattling" at the same frequency regardless of input/driving frequency.

[brainz01]

Got it. I see your point. Now I'm confused again. Hoping the water pump changes the signature.

[Chocaholic]

At this point...trade it in.




Couldn’t resist. Carry on.

[rusnak]

I wont keep saying it

[DanielDudley]

Yeah. I have been through this wringer with several cars over the years. When it comes to these types of problems, sometimes they get resolved, and sometimes they don't. I had a Subaru with a bad rumble coming from the transmission. It wasn't the transmission, it was a rear wheel bearing. For all I know one of the inner CV mounts is out of whack on your car, and it is going to the wheel, or there is a balance issue on a shaft somewhere out in Alaska that is causing a problem in Florida.

There are different levels of road force balance that an operator can take a wheel to. Some cars need to be balance to 12 pounds or less, and others are happy at 20. Your car is sensitive. If you keep playing with it, and driving the car, you may eventually find it.

If I was shopping for a used car that drove out like yours, I'd assume it was unfixable. I wouldn't stop playing with it though.

[ZOO]

This is a real princess and the pea situation . . .

I feel your pain and frustration.

[sammyg2]

Originally Posted by fanaudical

Am hoping Sammy chimes in on this.

What I find really interesting about that last snapshot is that the 7or8X spike is still there (limited amplitude) while the car is standing still. That means (at least to me) that there may be a natural resonance of the entire structure at that frequency driven by engine operation.

Those smaller synchronous peaks at higher frequencies are most likely harmonics.
Not enough resolution to confirm resonance but it can't be ruled out based on known data.
Raising the idle 200 rpm and re-taking the snapshot would be indicative.
If the higher synchronous peaks are still there, then they are harmonics of engine speed and not natural resonance of structure.

[sammyg2]

Note that Multiple synchronous peaks at 1x, 2x, 3x,4x, etc that are diminishing in amplitude as frequency increases is usually indicative of looseness, which would be normal in sleeve bearing design machines and not necessarily suggest a problem when displayed in acceleration mode (G's) as long as the amplitude is low.

Over the years I developed the habit of starting condition evaluation in velocity (inches per second) with the accompanying waveform, and then switched to either displacement or acceleration depending on what type of machine or what type of characteristic I was evaluating.
That strategy worked fine for me but the downside is that I became dependent on different views and am limited when only seeing spectral acceleration displayed.

Seeing only one view is tough because it doesn't tell the whole picture and can miss higher frequency events.
Example of evaluating rolling element bearings (ball bearings) such as a water pump.
Assuming the noise not from cavitation or the pulley (or belt or internal rub which would both be sub-synchronous), it must be from the bearing.

Rolling element bearing fault frequencies are usually only detectable in velocity when they get to the latter stages of failure, IOW severe. Like looking at the train tracks after the train has already passed.
But by combining velocity with acceleration, and then switching to peak view, a bearing fault can be detected long before it can be heard or felt.
Learned that one the hard way.
When a multi-million dollar machine is down and results in tens of millions in lost production, the bosses don't like hearing oops, I missed.
But the best vibration analysts I have met will tel you that it is not exact, it is interpretive. Best guess.
That's why you can tell the good and experienced ones, they are less likely to know it all and stick their necks out too far.
The newbie will say "oh I know exactly what that is, it's unbalance" or whatever.
The real pro will never be that committed, he'll say things like "the data indicates a high probability that it could be this. Further data and analysis will be required to define".
Doesn't matter how sure he is, because experience has taught him that it can be deceiving and it's easy to be wrong even when you are sure you are right.

I see the future of the vibration analyst as going away soon, to be replaced by computers that do the analysis and spit out a likely cause with probability.
Then it'll be all about blaming the data collection techniques

[sammyg2]

LOL re-reading that post reminded me of when I taught some level one vibration analysis classes a while back.
There is no subject more boring, can you imagine trying to stay awake in a class on that after lunch?
I always kept the air conditioning turned down to 65 or lower to keep people from nodding off.

[john70t]

That phone vibrational record is pretty cool.
(Along w/those g-force meters, lap timers, star identifiers, translators, etc etc all available in one tiny rectangle.)

Just to point out the obvious, but what you brainiacs are doing now could turn into a standard diagnostic app.
It's just a small step to create one tailored to every vehicle via user submission. OBD2 port dongle not included.

[brainz01]

Originally Posted by rusnak

I wont keep saying it

Just to be clear, I believe you want me to double check the suspension tightness, correct? It's in process. Rear bolts and RF are largely done. LF needs a thorough checking, but cursory look at the strut mount looks like it's separating too.

And I need to order the new front strut mounts -- I'm increasingly hopeful this is the issue.

[brainz01]

Originally Posted by sammyg2

Those smaller synchronous peaks at higher frequencies are most likely harmonics.
Not enough resolution to confirm resonance but it can't be ruled out based on known data.
Raising the idle 200 rpm and re-taking the snapshot would be indicative.
If the higher synchronous peaks are still there, then they are harmonics of engine speed and not natural resonance of structure.

I'll do that. But first I've got to replace this water pump.

Not sure this video does it justice, but it sounds terrible:

https://youtu.be/w2kK6qhib5k

[brainz01]

Originally Posted by john70t

That phone vibrational record is pretty cool.
(Along w/those g-force meters, lap timers, star identifiers, translators, etc etc all available in one tiny rectangle.)

Just to point out the obvious, but what you brainiacs are doing now could turn into a standard diagnostic app.
It's just a small step to create one tailored to every vehicle via user submission. OBD2 port dongle not included.

Yeah -- what you can do with a smart phone is pretty incredible these days.

Here's a professional app that helps trouble shoot this vibration stuff for automotive purposes:

https://vibratesoftware.com/

And video of use: https://youtu.be/s1cr3QaOzGY

It's too rich for me at $399, so I was happy to find a free app that logged a 5 second average vibration signature. It's not perfect, but it does seem to provide useful information (for free!)

Unlike the paid app, I have to do all my own math to figure out what speed/hz at which everything is rotating at various speeds. I built a quick spreadsheet for tire size and speed that identified expected 1st order wheel frequency. And using differential ratio, you can scale up for driveshafts.

[sammyg2]

Since it's belt-driven (side load) it'll go downhill fast, good plan to change it now. If you can hear it, then it's way past bad.


On straight coupled pumps, there's a strange phenomenon:

• First stage bearing failure, usually not detectable without high frequency analysis.
• Second stage failure, starts to appear in regular spectral analysis but very low amplitude, likely not detectable by ear without stethoscope (yes I got one, still)
• Third stage failure detectable by ear, temporarily gets quieter with addition of grease, barely detectable by feel. Shows non-synchronous peaks between 4x and 8x coinciding with inner or outer race frequency or fundamental train frequency (cage, ball spin etc). Still low amplitude.
• 4th stage failure easily detected by ear and feel, larger non-synchronous peaks with side-bands and harmonics (haystacks).
• 5th stage may get quiet. May not be detectable under normal diagnostics

Note that is my list and others may see it differently or have different definitions of the stages, there is no O'fficial written in stone version that I am aware of.


On straight (in-line) coupled pumps, assuming the alignment is good, the shaft can be supported by the coupling and magnetic field so when the bearing balls disintegrate into little black pebbles and fall out, they no longer make noise. The shaft is essentially riding on air.
But shut it down and then try to re-start it and all heck will break loose.

Been bit by that one once. Uno.

[brainz01]

Originally Posted by DanielDudley

If I was shopping for a used car that drove out like yours, I'd assume it was unfixable. I wouldn't stop playing with it though.

Yeah, I'll bet most wouldn't notice it unless they really drove the car for a while, or drove the right stretch of [grooved] road. And even if they did feel it, they'd think: wheels are out of balance. Which maybe they are.... but I've got 4 prior balance jobs that would suggest otherwise.

I'm not giving up. But admittedly, it's a lot of work checking all the possibilities -- especially without a lift.