Spectrographic analysis of chassis shimmy
 

I bought this $20 3-axis acceleromter:
http://www.adafruit.com/products/163

and this $29 4-channel USB data acquisition box:
http://www.dataq.com/products/startkit/di145.html

I attached the accelerometer to middle of the top hoop of my rollbar. I started a datalog, then I drove over 6 speed bumps in a parking lot and made a U turn to hit them again. I then stopped, got out, loosened one end of the STB, and repeated the drive.

I then converted the datalog from the left-right axis (the main mode of the miata, torsional vibration), into a WAV file and loaded it into spectrographic analysis software. See below.

https://www.miataturbo.net/attachmen...1&d=1313425425

Left to right is time, (total is about 240 sec), bottom to top is 0 to 30 Hz. The bright columns are each speed bump. The bright red spots are at about 16.2 Hz, which is the chassis torsional resonance. The left half (first 2 groups of columns) is with a tight STB, the right half, is with the STB bolts loosened.

(Due to a quirk of the spectrographic software I had to tell the conversion software that the log is 600 Hz and not 60 Hz sampling, so the frequencies are multiplied by 10 and time is shortened by 10. )

I expected loosening the STB to show a decrease in torsional resonant frequency signifying a decrease in torsional stiffness, but instead what I found was an increase in the strength of the "ringing" (red spots are brighter), and an increase in 'Q' (duration of the ringing), evident as larger islands of red.

The spectrogram software I used is this one:
http://web.archive.org/web/200011180...ams/gram50.zip

It's no longer availalbe on the original website and I found it on the wayback machine (hooray).

what STB are you using in this comparison?

mazda 01+

interesting so it didn't change the stiffness but it changed the damping?

My theory for now is that the engine bay has a resonance that is close to that of the body torsion, when the STB isn't there. Maybe the engine bay vibrates at said resonant frequency and then the body vibrates sympathetically. With the STB, engine bay resonance is moved up, and so it doesn't excite the main torsion resonance.

Mazda wouldn't add a large hunk of metal if it didn't improve NVH or otherwise *do* something.

That the resonant freq isn't changed, but is less prominent, is what my butt says. In contrast, the door bars, and FM butterfly, raised the resonant frequency. But no, I'm too lazy to go try disconnecting the door bars!

BTW I compared the plots also from just driving to and from work with and without the STB connected, with similar results.

BTW played back at 10x speed the chassis resonance sounds like a drum!

Love it!

Go Jason, Go Jason!

I use Spectra for some analysis along with audio design related software. It would be nice to see your data portrayed or displayed in different ways.

Try loading the zipped wav file into it.

This is on the top loop of the roll bar? I'd say they're impressive results for being so far away from the change you're making.

Yes because I figured that would be a good point to measure torsional vibration, and it's easier to tiewrap the accelerometer to it than to the windshield header!

Awesome experiment! Interesting results and analysis too, which I think is spot-on (I'm a mechanical engineer with a *little* bit of vibration experiment experience).

I'd be curious to see if the results are affected with different suspension setups, or even just different damper settings.

The stiffer the dampers' high speed damping, the more the chassis resonances will be excited
And also, the closer you are to the stiff part of your bumpstops (the more you hit them).

take one for the team! :laugh:

How did you determine that the peak around 16Hz is a primary torsion mode? I'm interested in reading any documentation that you have found on the subject. Do you know if anybody has performed any sort of modal testing on a Miata (simulation or real-life impact)?

Assuming the 16Hz peak is a torsional mode, I would expect it to be the torsional mode along the major axis of the frame. If so, the STB shouldn't affect that mode at all since it is tying 2 points together perpendicular to the frame axis which adds zero stiffness and pretty much zero mass.

16 Hz is the strongest vibration frequency of the left-right axis. A long time ago testing with a borrowed accelerometer, this axis had the strongest vibrations, at the top of the rollbar. I suppose if I had a 2nd accelerometer, I could look to see if the left-right movement of the rollbar and the windshield header are in opposite directions.

Next time you are going over a road where the car shimmies, place a hand on the top of the windshield header, or the top of the rollbar. The left-right motion is the most pronounced.

P.S. I just examined the other axes' datalogs and they don't have anywhere near the amplitude of the left-right axis.

And see my theory above about sympathetic vibrations of the chassis, from the engine bay.

16hz, give or take a little is the frequency of wheel/tire rotation at 65 or so mph. I think you just found the 65mph shimmy and conclusively demonstrated that it is related to chassis resonance.

So would adding FM rails and stiffer shocks/springs move the resonance point up or down? Seams like my 65mph shimmy it at about 70-75 now. Keep swearing I have a tire out of balance even tho the tire shop swears I don't. Just curious, with I could get data like that.

Interesting, hmm, 16.2 Hz with 195/50/15 tires is 66 mph.

Note that my car shows a chassis resonant frequency that is higher than even a stock MSM, accdg to my butt-o-meter.

I've shown that the resonance of my chassis is right at 66 mph, but not necessarily that it's the reason for the infamous 65 mph shimmy, although that would make sense.

This is a good idea. The rollbar and the windshield header are stiff and shouldn't have any significant local modes. If you measured 180* out of phase vibration between those 2, I'd agree that a torsional mode is the most probable. But, unless I had tons of experience with the modal dynamics of similar vehicles, I would be hesitant to draw any conclusions on the miata's modal characteristics based on a single accel measurement.

I'd want to see at least 12 accel locations before drawing any conclusions on mode shapes. I'd run the following locations:
Top and bottom shock mount (each corner)
2 points along frame rails (each side)

That should give us enough resolution to figure out the meaningful global modes. You'd need more points if you wanted clarity on any localized modes.

I have the required setup to run a full test like this at work, but there's no way I could swing putting the miata on airbags and running the roving impact test without getting fired.

If the top of the rollbar shows much more high frequency resonant motion left-right, than fore-aft and up-down, what else could it be if not torsional vibration? A convertible would probably be weakest in torsion, and second weakest in bending. Torsion would probably show a much lower resonant frequency, because of the polar moment of the end "masses" (front and rear "boxes") along the long axis. Bending would be weak too but I imagine there's much less inertia exciting a bending resonance, yielding a higher resonant frequency. Also note that most car manufacturers quote torsional rigidity whenever they say "new car is 17% stiffer than outgoing model year". I guess even sedans would have torsion as its main vibration mode.

FM rails definitely will move the resonance up. Shocks and springs won't change that frequency, but can excite it more often and more strongly.

Guys Im no engineer but I dont see how the tire/wheel rotation relates to chassis shimmy in the event of a properly balanced wheel/tire combo. Hertz measures something repeating so when you measure a tire rotating in hertz you are actually measuring a single point on the tire coming around each time . When in theory you actually have a infinte number of point rotating around. When you measure in hertz it seems to me it simplifies a tire into something like a weight on the end of a string being swung around.

See this would work if you have a heavy spot on the wheel/tire combo the it would rotate at x hertz and I could see its affect on the chassis but a balanced tire is always going to be nullified by itself because the force of any point on a balanced tire would be nullified by the point on the exact opposite side of the tire. I know tires/wheels arent balanced at every given point so itd be more of area to area not point to point,but it all averages out, Im guesing.

Ive been in auto repair for along time and experience says that 90% of shimmies that come in at around hwy speeds and sometimes smooth out at a higher speed are a wheel/tire issue. And the other 10% ive seen dont have anything to do with chassis rigidity. i have yet to come across a shimmy I couldnt diagnose and fix.

If you have a 65 mph shimmy I still reccomend checking you balance and wheels tires first.

I have no background in engineer its all in auto and accounting so please let me know if my thinking is wrong.

Agreed, I've never seen a speed related shimmy that wasn't fixed by a proper wheel balance job. I think tho that the reason it's worst at 65 mph is that it excites a chassis resonance of some sort. Or the resonance of the steering mechanism.

agreed

Is the conclusion here that the resonant structure in our cars is the torsion in the connection between the front end and rear end? I.e. the floor pan\frame rails? And the vibration exciting it derives from the wheels rotation?

Since adding mass is not really possible, we are left with either removing the vibration by perfectly balancing the wheels or stiffening the chassis with reinforcing frame rails\frog arms\door bars\roll bar etc.?

Will no amount in change in suspension components like springs or dampening help remove the vibration we feel? I`m thinking along the lines of that the shocks themself in principle act as vibration dampeners (chassi is the platform and one damper in each corner) and could be tuned to remove the resonance or suppress it to an acceptable level.

You stiffen the chassis to improve ride / handling / NVH. This is usually done by using an X-Brace strategy which is most effective torsionally and very seldom done on Miatas as a mod.

lassi, no the vibration exciting is isn't from the wheel's rotation unless you have a wheel balance issue, then it vibrates continuously. This is slightly off topic.

The chassis vibration that bugs me (and others) is the basic lack of rigidity of what you described, the connection between the front and rear via the floor pan / door sills / frame rails. The chassis vibrates in "aftershocks" (undamped) after hitting a one sided bump, and is sometimes excited when hitting a series of bumps "just right". (1) it significantly degrades NVH, and (2) occasionally in a corner or when accelerating I feel the tire skip when some ripples in the road hit the exact resonant frequency. This is loss of grip associated with lack of chassis rigidity.

Dampers (shocks) cannot damp this vibration because it's not relative motion between the wheel and the body.

Adding weight is the wrong thing to do, because it will increase the energy in the vibration which will "raise its Q" - make the vibration last for longer before dying out. It also lowers the resonant frequency, which will make it easier to excite, from hitting bumps.

I have no Idea how to make anything out of a spectrograph. How about some simple PSD plots so you can see the resonance peaks.

Bob

You know those audio / music spectrum analyzers with bars that move up and down in real time? A spectrograph gives you said spectrum analyzer, over a certain amount of time. (in my OP, ~240 seconds)

A spectrum analyzer's X-axis is frequency, and Y-axis is amplitude, and displays frequency content in *one instant* in time. It's a 2D plot. In a spectrograph, Y-axis is frequency, X-axis is time. The color/intensity, is the amplitude. It's a 3D plot.

The reason a spectrograph is better than a simple PSD plot, is that you can see how resonances appear at slightly different frequencies at different times. A PSD plot would simply take the frequency content over the entire 240 sec.

Sorry I don’t work with music I work with automotive NVH and vibration fatigue on a regular basis though. PSD's give me info I can work with and understand the meaning of better.

Bob

Spectrographic results of door opening seam stitch welding:

https://www.miataturbo.net/showthrea...d=1#post761185

bbundy a spectrograph is merely a presentation of hundreds of PSD plots in one plot.