hi,all.I am learning a course about csi2130.
and it has below selection for averaging:
1:normal average
2eak hold
3:synchronous time
4rder tracking
5:negative average.
6:exponential average.

why is order tracking in averaging set?
what does negative average mean?
what does synchrounous time means?

can someone give an advise?

1 normal average
2 peak hold
3 synchronous time
4 order tracking
5 negative average.
6 exponential average.

Negative averaging is having four consultants in the same room using four different data acquisition systems - or I could be wrong.

When one measures distance, it is impossible to have negative distance! Negative is a created number from a reference point.

Don´t be so negative Sam, I have once had the same opinion as a fellow consultant in the same room and the other guy was also a professor at a university and one worked for the supplier and one for the buyer so we blamed a never again found designer of mounting.
In my humble description:

1. Add data as you measure them, divide by the number of meas. Normal average to suppress random noise.
2. Hold the max value at each collected data, used for knocktest.
3. Sync. average, trigger data collection from tacho to filter out all vibrations that does not sync with the RPM of the Tacho, used to find the vibrating roll in a press nip.
4. Order tracking is that an average? I see it as collecting to be able to present a run up/down waterfall eg. speed and vibration is collected as the machine RPM differ and you plot vibration against speed in RPM or order/(time) by stacking the FFT´s in a 3d fashion.
5. Negative, you take a reference spectrum and start collecting data and each new data is sort of subtracted from the reference so if you start doing a knocktest and is lucky you may extract the peak from the knocktest and you can see how close it is to the machine vibration freq´s even during operation.
6. I guess it adds the data not linear but in a exponential fashion, never used it IRL.

Pls. clarify and add what I missed out in this simplified trial to explain.
Olov

If I understand CSI equipment correctly, negative averaging is where you do a bump test on running equipment. It first collects data while you whack it. Next, it collects data without you whacking it. It now subtracts out everything that is alike and you end up with just the whacking, which is a lot like when I was a kid. Many times I ended up with the whacking.

David Eason

Sorry, you're right Oli - having a bad hair day.
But only one hair.

You have a reference spectrum and subtract it from the whacko data. I really don't see negative averaging though. You may want to Peak Hold avg on both for obvious reasons.

Word of caution - a good place to get in trouble and may violate safety ordaniances in a given plant. Will whacking a machine shut it down or worse; cause damage?

Oh, subtracting must be negative compared to adding ? I remember the good old days when you could haul up a diesel compressor using the oh crane and put it between 2x 800MW steam turbine generator and shaking the whole with the largest pneumatic shaker you could find while in full operation on full load and collecting data from inside the endwinding with full voltage on. (In a nuke plant) I would have needed some negative average than but only had positive so it went well. It was fun fun fun until the diesel fuel was finished and some exhaust in the room. It was a resonance naturally. People are so fuzzy these days and they do stupid things anyway. We did stupid things w/o so much fuzz. Like cooling the brushes in the same generator with the fire hose, but you heard that one already maybe and it´s not Friday until tomorrow. It´s end of the month and all work and things are reported, delivered and invoiced so it feels like Friday almost. Olov

As explained, negative averaging is most often used when doing impact testing to determine resonances while a machine is in operation. If a resonance is "near" a running speed (or harmonic), this can work. If the resonance is too close to the running speed, then the resonant frequency will most likely be subtracted during the negative averaging, depending on the resolution of the spectrum.

Here's how it works on the 2130. You select "negative averaging" and then set the number of "averages" you want. Say you set this to 6, then when you push the Enter key, the meter takes 6 averages, just as it normally would. Then it stops, displaying the average spectrum. You have two choices at this point. You can press Negative Averaging (F3) or Store Data (F9). If you press F3 the meter resumes collecting data, but instead of 'adding' the new collections to the existing average spectrum, the new data is 'subtracted'. It will keep collecting (and subtracting) data until you press the Stop key.

I usually press the F3 key to Store Data on the first pass, to store the 'initial' or 'reference' spectrum. Then I run the collection again and use the Negative Average function. When doing a "running impact" test, you must also set up a trigger.

There is a "Bump Test Equipment Running" Expert for this, if yours has the Expert features. It automates all this.

hi, all, thank you for your explaination.
May I comprehended it like this:
it use to test the impact of a running machine,
and fisrt we whalmed the machine, and catch the data block 1, then we let the machine free run, and get the data block2;
finally, we get data block3 which is (block1-block2). then we do fft to get FFT_result_mag.

Am i right?

or we do it like:
FFT_result_mag=fft_mag(block1)-fft_mag(block2);
but I am not sure about do substraction in freq domain, it may get the minus of magnitude.

I would do it in freq. domain as suggested and let all negative results be replaced with 0. But that´s me and it´s Friday. Olov

So you take two linear summation averages, subtract them and the results become negative averaging ------- in a word; impossible!

It is just subtracting one spectrum in one data block from another. RTA's did this in the 1980's from cascades or stored spectra and would add, subtract or, or Block 20 from block nn or. I never heard it called negative averaging. Coming up with peculiar nomenclature confuses the issue in my opinion.

Sam,

One spectrum is taken while impacting the structure and then another is taken without impacting. The operating psectrum is averaged out and what is left is the result of the impacting.

Just because you don't understand it doesn't make it impossible.

With the machine off, a trigger mode is set. You do a regular bump test.
With the machine running, the trigger mode is off and more averages are used (more than 8 is recommended). The reason for the extra averages is that you only want to get 1 impact per average. If you miss one, that's ok. If you get 2 or more, that's bad.
Art Crawford's opinion is that Negative Linear Averaging is the best thing since sliced bread!

But I do understand it. Been using it back in the '80's. Averaged out or subtracted out? Does your impact hammer have an accelerometer attached and are you getting two read-outs?

You've been using it since the 80's but it is impossible?

No accel in the hammer, it doesn't even need to be a hammer.

Art Crawford (you know who that is, don't you Sam) was one of the original instigators of "negative averaging", at least with CSI. They've had it in their equipment for at least the last 20 years, so it's not new at all, and not new nomenclature.

It is quite possible and works very well. I use it all the time. I don't understand all the math, but people like Ken Piety certainly do. If it's good enough for Ken & Art, then it's certainly good enough for me.

I do not know Art Crawford, but I know Ken.

Ono Sokki gave such capabilities as did B&K and HP I believe plus others. I would set the analyzer (Ono Sokki) and it will capture automatically upon a preset trigger at a level and put into memory block 1 and bump 2 into blk 2 and so or and/or capture averaged data into memory blocks and one can then add, subtract or perform whatever mathematical function one wants.

Whacking a running machine involves many cautions even apart from instrumentation. Liability insurance paid up?

Ken and I entered dicussion in 1990 about buying my company and software; so I may know who he is. In my own software you could get print-outs automatically of frequency and magnitude in Hz, g's, IPS & Mils with digitized data for labeled analyses or dump the whole of all data to a printer. And with cepstrum and balancing.

I'm using an inheirted CSI 2120 presently (but only shortly) w/v. 4.91 as I recall and put it at 1985 technology or behind. So, things are difficult with it and very busy.

Taking an averaged spectrum and to be fully consistant, you need ~32 averages to be stored into block nn? Then triggering? Clock or key phasor, or internal? Now a spectrum collected from a bump test that you intend to store into block n2 and substract block nn? Or, are you averaging the two blocks together?

From this you are going to determine? So, Danny and Rusty - educate me. Please expound on this so all will know exactly how this functions works internally, mathematically and how it is used in analyses. I'm sure all would like to hear and all would like to know.

This message has been edited. Last edited by: Sam Pickens,

Sam,

I have explained it a simply as I can. Any further understanding for you will require an effort on your part. There is a good article fully explaining negative linear averaging on CSI's website and I suggest you download it.

I have heard your complaint about the 2120 and RBMWare being 1985 technology so many times that I frequently recite it as an example of technological ignorance. If you really understood the technology, you would not make such claims but for some reason you can't admit your ignorance. You are certainly entitled to your opinions, but in this area, they are apparently not shared by anyone else.

The reason I'm saying 1985 or older is due to the fact that others had more advanced features in 1980 that few if any have 'caught-up to' today. B&K's RTA, HP's, Ono Sokki's to name some had far more capabilities in 1985 than CSI does now to the best of my knowledge. Would that not place it at 1985 or before. Maybe it is your lack of knowledge of the other systems that are out there. There are more; I just didn't name them.

Thanks for your explaination. But does not your explanation stating how the negative average is accomplished come out as subtracting linear summation averages? Averaging an orange and an apple doesn't make or does make ______________?

Regardless, I will have a look at the link you provided and glean what I can.

Sorry I put you in the same tub as the ilearn guy. I know you are in the real world and do and perform and have a track record. We may not always agree. I have always had respect for your opinion and read your comments. I don't think you have ever heard me say or write anything poor about you either personally or professionally.

You can mail me at PdMSamPickens@gmail.com or picksam@juno.com.

Well I still use my old 2120 vintage 2000 for balancing since I got so used to the interface it´s hard to swap and I have 2100 as a "spare". I can only read the ever reoccurring discussions for every new PC software version release, it echoes in my memory from the approx 15 years I spent supporting that software and answering the same things all the time with ever new fixes of the same old problems, seems they never got fixed? For me the lack of window function support and the trouble to fix it for every new windows generation is setting the time stamp on the vintage around what Sam states at least for the PC software. Have you seen any new innovations since the turn of the century? Any new or old toys implemented, Cepstrum anyone? I have not used a 2130 that much but I would not be surprised if it´s still an Z80 churning along inside it as the other´s had, it would also validate Sam´s estimate. Correct me if I am wrong. Olov