01-10-2018 12:40 PM
Hello,
currently I am observing a strange behaviour while trying to calculate the vector sum of thre independent channnels. I am using the formula Vectorsum(Channel1:+X;Channel2:+Y;Channel3:+Z) to calculate the vectorsum of three linear processed Autopower spectras. Weirdly if I change the order of X,Y,Z I seem to get different results. Obviously I am doing something wrong or I am overseeing something.
I hope the attached picture provide clarity about my doings. Thank you in advance for any help on this!
Channel1_w:+X is CH13
Channel2_w:-Y is CH21
Channel3_w:+Z is CH16
01-10-2018 01:14 PM
Hi,
I cannot reproduce the issue you're experiencing. Here are my test results and they're matching regardless of what channels order:
Are you processing two different runs/measurements at the same time? Sometime if the channel ID/# are not the same between the two different runs could cause issues. Is it possible to have you send a copy of your processing parameters? You can go to Time data processing workbook:
Click Save Processing Set..., then save the file with unique name and find the file at:
C:\LMS\UserConfiguration\User login\LMS Test.Lab 17\ProcessingSet
Please send a copy of the process setting.
Hong
01-11-2018 05:33 AM - edited 01-11-2018 06:28 AM
Hello,
thank you Hong for your reply. I attached my Processing settings so that you can try those. I am using Test.Lab 16. currently. I tried it again with a new project and sadly I am running into the same problems. I am not processing two different runs/measurements. All data is just from one run. However, I copied the time data of that run from a different project and I am doing a frequency weighting / filtering of the time data before the processing. I now checked the processing without that filtering and then I do get correct results. So I wonder why it is going wrong with the filtering?
Time Data:
Filtering of Time Data using FILTER_ISO5348():
Filter results in Time Data Selection:
Data in Time Data Processing:
Vectorsums in Derived Channels:
Results:
Obviously the result isn't correct because at some point the single channels have higher values than the summed up curves. I've marked two of those points.
Without the frequency weighting I do get perfectly reasonable results:
So, I wonder why this weird behaviour with the frequency weighting is happening and what I can do to solve this? Any help is appreciated.
01-11-2018 11:24 AM - edited 01-11-2018 11:34 AM
Here's my suggestion:
After you applied ISO5349 filter to the data, please go to DerivedFS->Source data as shown in the screen shot above, now Test.Lab rearrange the order of your channels. Personally when I created formula, I do not prefer writing the point ID in. When you write the point ID, you can make mistake. I simpily use the ch#, such as ch2, ch3...etc.
Here's my data in time data selection workbook:
Here's my formula again by using the ch#, can you go to the Source data drop menu and check what ch#s are your newly filtered data? Also, can you create the vectorsum formula using the ch#, writing the poinit ID in some time mistake happen.
Here are my test result:
Everything still the same.
Hong
01-15-2018 09:53 AM - edited 01-15-2018 10:03 AM
Hello,
after having done some extensive testing and having called with Test.Lab support it seems that the problem only occurs after further processing weighted raw time data without saving it inbetween. So in my case I wanted to calculate the vector sum of the Autopower spectra of three longish (>240s) time traces. When I save the data after applying the weighting in "Time Data Selection" and then do the Vectorsum() I get the correct values. But when I don't save them in-between (and the weighted channels remain with yellow background-color) I do get wrong results. After doing an in-between saving in "Time Data Selection" also the weighted time traces have green background-color and I get correct values. With shorter time traces I didn't always observe this behaviour. So I don't know what happens, but I guess after doing this in-between saving it should always be fine.
Here is what I did:
Original time data (3 individual channels that should be summed):
Applying the filter on the time traces and naming the filtered traces with a unique point ID: "...._w" (note the yellow background-color of the weighted channels)
Calculating the Vectorsum in almost every possible way; using the point IDs (Sensor1_w:+X) to create the "SensorSum" channels but also the Channel numbers (CH2) to create the "DerivedChannel" Channels. Appart from this naming difference only the order of the 3 channels is switched in the formulas (X,Y,Z or Y,X,Z or Z,X,Y or...). In the drop-down menu on the right one can see that the used channel numbers coincide with the point IDs. So in fact all those 12 formulas should result in the same Autopower spectrum.
However with my time data this doesn't work:
However, if I do save the data after weighting, all the channels get a green background-color.
The Time Data Processing is just the same:
But now, I do get for all the channels and formulas the same result. After comparing this result to the individual Autopower spectras of the 3 channels it seems it is also the correct sum. So saving before processing seems to be neccessary...
Thank you for all the help!
01-15-2018 10:39 AM - edited 01-15-2018 11:17 AM
Hello,
after some extensive testing and some helpfull contact with the Testlab support I understand that it is recommended to save weighted time data before doing further processings.
My goal was to calculate the vectorsum of 3 autopower spectrums after I applied a filter on those channels. My time traces were kind of longish (>240s) and my attempt resulted in wrong results. After first saving the newly generated time traces and then performing the vectorsum calculation I do get the correct results.
Here is what I did:
My original time data:
The newly generated filtered time traces with unique point-IDs: "..._w" (note the yellow background-color)
After the filtering I try to calculate the vectorsum. In total I am calculating 12 different formulas that should all result in the same Autopower Spectrum. 6 formulas are using the point-IDs (Signal1_w:+X) and 6 are using the Channel Numbers (CH2). In the drop-down menu on the right one can see, that the point-IDs and Channel Numbers used are matching. In the formulas only the order of the directions is changed (X,Y,Z or Y,X,Z or Z,X,Y or ...).
With the result one can see, that sadly not all calculations are the same:
However, when I do save the newly generated, weighted time traces, they get a green background-color:
The vectorsum is just calculated the same as before:
And now all 12 spectra yield the same result:
So it seems that doing an in-between saving after weighting and before further processing is needed to avoid unwanted behaviour. With shorter time traces I sometimes didn't observer this weird phenomenon.