"Spectrum Peak Hold" 1

[TylerJO]

Hello all,

I'm working with some accelerometer data and I'm wondering if anyone can give me a brief walkthrough of what's happening when I am looking at a "Spectrum Peak Hold"? I've tried googling the phrase and I get nothing useful. When I apply the SPH what comes out is just a frequency bandwidth with spikes at certain frequencies...I'm just not 100% sure what the amplitudes of these spikes actually represent. I know it's not an RMS value, and I assume it's, as the name suggests, it's just the "peak" acceleration at that particular frequency, but I can't find any information to back up this assumption. Any help would be great. Thanks.

-Tyler

 

[Strong]

Spectrum Peak Hold for most analyzers is the spectrum of the highest amplitude in each bin/line/filter for however long the "averaging" takes, or number of averages. The amplitude depends on the selected scale, and it could be RMS or Peak or Peak-to-Peak or dB; same as would be useed to scale the normal averaged spectrum.

Walt

 

[TylerJO]

Ok, definitely helped. Could you explain "averaging" though? My data comes from one test so I don't think there should be anything to average...unless I'm misunderstanding what you mean. Sorry if this seems like a basic question, I'm just trying to be as thorough as possible.

Alittle more info about the process I'm using too, if it helps: I'm using an "Autopowers Linear" function, a hanning window, no weighting, and "peak" formatting...I tried switching it to "RMS" formatting and the numbers were all reduced by 1/sqrt(2) so I'm a little more confident that my original numbers are in fact "peak" accelerations.

Thanks again

-Tyler

 

[Strong]

Typically each block (say 1024 samples) of time domain (waveform) is converted to FFT (spectrum), so there would be an FFT spectrum for each time data block. For vibration data one might take 4 to 32 blocks (FFT) and average them into one averaged spectrum. The normal average process is ensemble averaging where amplitudes are summed for each FFT and divided by the number of averages. The Peak hold spectrum is simply holding the highest amplitude value that comes along at each frequency for the number of FFT spectra being "averaged". The "Peak Value Spectrum" would be in Peak or RMS amplitude units, depending on your selection. There is nothing like learnig by doing it!!

Walt

 

[GregLocock]

So in your case where you have one frame of data in your ensemble, peak hold is identical to the spectrum for that frame of data.

Peak hold in general is the sort of thing that real engineers find useful sometimes while the theorists tear their hair out.

For instance, if you put an accelerometer on the hub of a wheel and drive down the road for a long time on different roads and different speeds, eventually you will get a nice peak at the wheelhop frequency.

Cheers

Greg Locock


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

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

 

[GregLocock]

Oh another way of looking at it is that if you were to overplot each individual spectrum in your ensemble, the peak hold would form the skyline of the spectra.

Cheers

Greg Locock


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

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

 

[Strong]

The "Skyline Spectrum" indeed!!

Walt