Sine sweep vs Random

[izax1]

If I do a sine sweep from, say from 20-500 Hz, and measure or analyze acce at one point, will I see the same frequncy peaks with a PSD test measuring or analyzing at the same point?

Just to clearify a discussion point with collegaues.

Thanks.

 

[IRstuff]

It's unlikely, since the instantaneously energy available for exciting a resonance is statistically low. That's precisely why a sine sweep test exists in the first place.

In other words, a 2-g rms random has its energy statistically distributed across the entire spectrum, so the amount energy available for exciting any mode is statistically low. Contrast with even a 0.1-g sine sweep where the entire 0.1-g acceleration is applied directly into a mode.

TTFN

FAQ731-376
Chinese prisoner wins Nobel Peace Prize

 

[GregLocock]

I think he's asking - will the frequency of the peaks in the spectrum be the same in the two tests? Yes, if your system is linear.



Cheers

Greg Locock


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

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

 

[FeX32]

Yes, the spectral energy will be highest at the resonant frequencies (same peaks as the sin sweep)


Fe

 

[izax1]

Yes Greg, that is what I am asking. Not the Acce level or PSD level but the frequencies at the peaks.

Thanks

 

[SomptingGuy]

If you focus on a peak, you may find the amplitude goes a bit AWOL, possibly breaking some parts of you kit. I drove a vehicle BIW at its first mode once. My ears still hurt.

- Steve

 

[IRstuff]

Again, random vibe results in random excitation, so output correlation to what you get from sine sweep may be quite low. You may see "peaks" that are off from the actual simply because the exact frequencies never got excited, only the off-peak ones.



TTFN

FAQ731-376
Chinese prisoner wins Nobel Peace Prize

 

[MikeyP]

Hmmm. I think the dynamic range of today's data acquisition systems would be enough to pick up all the peaks due to a random excitaiton.

"Random" implies measuring autospectrum and hence averaging, so given enough time, the random excited response should look like the sine excited response in terms of where the peaks are and their relative heights.

M

--
Dr Michael F Platten

 

[IRstuff]

The standard resonance search in MIL-STD-167 requires 15 seconds of dwell to detect a resonance. I'm not at all sure that a random vibe ever generates sufficient dwell time near a resonance to make it stand out above the background random vibe.

TTFN

FAQ731-376
Chinese prisoner wins Nobel Peace Prize

 

[MikeyP]

I suppose that might be the case if there is sufficient background noise, but there is no mention of background noise from the OP. There are plenty of us who use random exitation for modal analysis for example - it is much quicker and is useful for papering over friction-type non-linearities and getting a "linearised" response.

However, your comment about dwell times does make me think of a reason why you might get a difference between the position of the apparent peaks in a swept sine and the peaks in a random test. If the sweep rate is too fast and the damping very light, then you could get a situation where the resonance doesn't have enough time to build up to it's maximum response. In that case, the peaks in the swept sine test could appear in slightly the wrong frequency. The peaks in the random test will always be in the right place. Of course all this assumes linearity as others have already said.

M

--
Dr Michael F Platten

 

[izax1]

Well, since this is a purely theoretical question, we assume that the dwell time for the sine sweep is long enough to excite the resonances.

I guess we'll just have to do a test next time we're at the shaker table and see.

Thanks all for your input.

 

[GregLocock]

On a practical note, when testing cars, I always used to use band limited gaussian white noise for experimental modal analysis.

However, this can give very poor coherence at the peaks because there are so many non linearities (both deliberate and accidental) in a whole vehicle.

So on problematic systems I would use sine sweeps, sometimes with amplitude feedback, and sometimes not.

This has the advantage of giving great coherence, but can disguise problems with test setup. It also only tests each non linearity at one amplitude for each frequency, which is not necessarily representative.

Well that hasn't made life any simpler has it?



Cheers

Greg Locock


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

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

 

[IRstuff]

And, don't try to do sine sweeps on top-heavy devices on a slip table. Goes with Greg's comments about test setup problems...

TTFN

FAQ731-376
Chinese prisoner wins Nobel Peace Prize