I've attached an exception report and spectral profile data of the same point. One showing 3.nnnn g's and the other 16.nn g's??? Go three places and get three different readings! Are others experiecing this? CSI 2120 w/RBMware 4.9. Or, does anyone have an explanation or can stear me in the right direction.

Spectral_Analysis_Scan_Report_2226.doc (33 Kb, 43 downloads) OA from 2 sources

"For overall mode in analog, the frequency range for data acquisition is 2 Hz to 20 Khz. For overall mode in digital, the frequency range is 2 x delta f to the F max specified in the analysis parameter set."

This is from the CSI paper on Digital and overall.

I don't know if this is an answer to your question, but seems like I read somewhere that overall value stored in analog is higher than when taken in digital. The value on your report may be for the value within Fmax of the spectrum and the A-TP overall value on the spectrum may be from 2 to 20K Hz.

Can't find where I read this exactly.

Maybe someone else knows the exact wording for this or it might be an error in the software. I store my overall in digital so I am not totally familiar with the analog overall.

The 16 g's is identified as true peak. You can see it on the time waveform.

The 3g's is some kind of overall. As was mentioned, there are two kinds of overalls:

"analog overall" - includes the effects of all frequencies. You can in theory calculate from TWF as sqrt(sum(x^2)/N) where N is number of sample points and x is value of each sample.

"digital overall" - includes only frequency content up to Fmax. You can in theory calculate it from the spectrum something like sqrt(2*sum(x^2)/N) where N is number of bins and x are the value at each bin. There may be some window factor required as well.

It occurs to me you probably are already familiar with these terms and may just be asking how to get them to show up the way you want in the report in a specific place. That I don't know. Sorry if I told you what you already know.

Hello Sam.

The analysis parameter listing at the top of the attachment called "Peak Total" is a calculation of the peaks found in the SPECTRUM above a threshold level that is controlled in the spectrum analysis scanning report configuration.

The waveform has impactive energy and that is driving up the noise floor of the spectrum. The floor noise is probably not included in the "Peak Total" calculation and that's why the peak total (3.364) doesn't match the spectrum peak (6.83). Check to see what the "Threshold Value is set at -- it's under the "spectral options" tab.

The overall = 16.32 is an analog overall calculated from the waveform from 1 Hz. to 80,000 Hertz. It will be slightly different than the peak (+/-) value displayed to the right of the waveform. I don't remember what the "TP" stands for.

Thanks Sam.

TP = True Peak = 16g's as can be seen on TWF. Right?

This message has been edited. Last edited by: electricpete,

quote:

TP = True Peak = 16g's as can be seen on TWF. Right?

Pete,

Where is this reading "16g's" on the waveform?
I see on the right where it list the + and - peaks to be from +14.63 to -14.28. Does TP mean "true peak" in CSI terms?

I don't use CSI. Everything I posted is based on my guess from looking at what you posted.

I think TP = true peak here. No other parameter will be that high here.

I don't know the reason for slight variation in true peak identified next to spectrum and true peak on TWF. I suspect there may have been some variation between measurements which caused variation true peak 16.32 at the time spectrum was collected and 14.6 at the time TWF was collected (apparently two different sequential sets of data).

Looking at the time waveform knowing true peak (14.63), crest factor (3.04) and analog overall expressed on peak basis (6.8), we can see that these three measurements agree:
crest factor = true peak / rms
crest factor = true peak / [peak/sqrt(2)]
6.80 = 14.63/[3.04/SQRT(2)]

This confirms the meaning of 6.8 as analogue overall on the TWF expressed on a peak basis.

There is no way the analog overall on a peak basis could be anywhere close to the true peak on an impacting waveform such as this. So no way 16g's can represent an analogue overall.

The Analog Overall has been mentioned a couple times. An Analog Overall cannot be taken with Sam's setup. His waveform is in Acceleration. An Analog Overall cannot be taken within an Acceleration waveform using CSI software. Please refer to Dr.Know technote 97-00791.

David Eason

Thanks gentlemen; the TP is True Peak and over 40 kHz I think in the RBMware 4.9 some versions was 20 kHz and 80 kHz.

And the Hoyle way is the square root of the sum of the squares in the given baseband frequency as Pete mentioned. The A-TP = acel true peak or V for IPS and there's a DG for digital and digital goes to 80 kHz I believe in the CSI way of doing things. But depending on where you look at data; it changes and changes with where and how your setup is. That's why I only use one setup so all will be somewhat consistent.

The noise floor is the noise floor and above that should be 'real' data. Hopefully, the noise floor remains the same and based on electronics and specifications of instrumentation/tranducer/etc...

But, I'll accept the square root of the sum of the squares for the Fmax for now. I'd go ahead and calculate it by hand if the software would list the peaks or could get it via cursor I imagine.

Anyway, I appreciate the feedback and thank you for your time.

quote:

The Analog Overall has been mentioned a couple times. An Analog Overall cannot be taken with Sam's setup. His waveform is in Acceleration. An Analog Overall cannot be taken within an Acceleration waveform using CSI software.

David,

Is it possible that Sam converted his waveform from velocity to acceleration after it was collected and would this convert the overall value also? Or did he acquire a special waveform using acceleration? Will any of these or some other override allow what is showing?

Thinking some more, the formula for digital overall that I gave above is wrong. The digital overall should simply be square root of sum of squares ie sqrt(sum(x^2)) with a possible window factor.

I tried out a digital overall calculation myself. In Emonitor you can export the data to excel by right-clicking.

Example calc using sqrt(sum(x^2)) for 800 line spectrum is attached (graphic spectrum included as well). The results of the calculation was digital overall=sqrt(sum(x^2))=0.142. But Emonitor tells me it's 0.115.

I was thinking the window factor if it is required would be greater than 1. So, I'm not sure why my answer came out higher than E-monitor's. Any suggestions?

digital_overall.xls (110 Kb, 9 downloads)

Digital Overall is calculated as:

OA = [sqrt(sum N^2)] / sqrt (1.5)

1.5 is noise Bandwidth factor for hanning window.

So everithing falls into place
0.142 / 1.22 = 0.115

Thanks David, that explains it.

I must have been mistaken about the factor being greater than 1.