Time Dependent Intensity (TDI) Corrections

[John Donovan]

Hi Andrew,
It's because the program is recording the elapsed *real time*, not the specified/calculated time.

Due to instrument/software latencies and other delays, we have to use the actual elapsed time to get an accurate TDI slope.
john

[AndrewLocock]

Ok, thanks!

[John Donovan]

Dave Adams (USGS Denver) recently sent me an MDB file (probe run file) that he was having a problem with and in sorting that out we noticed (because his run had over 500 unknown samples in it!), that loading the Standard Assignments dialog was taking longer than one would like.

The problem turned out to be due to the fact that the app scans all unknown samples in the current run when loading the Standard Assignments dialog, in order to list all unknown samples that could be utilized in the quantitative blank correction for the selected element (matching element, x-ray, spectrometer and Bragg crystal):

http://smf.probesoftware.com/index.php?topic=454.msg6694#msg6694

So we optimized the code and now that dialog loads about 5 to 8 times faster, which is nice when one has a very large run like Dave had.  Update Probe for EPMA from the Help menu and all will be good.

[Ben Buse]

Hi,

Is it possible to export the errors for each TDI point - I'm using Output - Save TDI

Thanks

Ben

[John Donovan]

Hi,

Is it possible to export the errors for each TDI point - I'm using Output - Save TDI

Thanks

Ben

Hi Ben,
When you say "errors for each TDI point", do you mean the % TDI correction and the % TDI variance for each data point in a sample?
john

[Ben Buse]

Hi John,

Sorry I should have made myself clear, I mean the error bars as shown on the figure below



Thanks

Ben

[John Donovan]

Oh right.  Doing that will disturb the format for those already depending on the existing output, so I can do that but I hope no one minds...
john

[John Donovan]

Hi John,

Sorry I should have made myself clear, I mean the error bars as shown on the figure below



Thanks

Ben

Hi Ben,
I was able to add output of the TDI intensity one sigma error values from the Output menu as seen here:



Let me know if this works for you. 
john

[Ben Buse]

Thanks John, this works well

Ben

[Ben Buse]

Hi John,

How do you calculate the error (% rel) for analyses using TDI, do you simply calculate the counting statistics error, or do you calculate the standard error for the intercept for a linear regression

Thanks

Ben

[Probeman]

Hi John,

How do you calculate the error (% rel) for analyses using TDI, do you simply calculate the counting statistics error, or do you calculate the standard error for the intercept for a linear regression

Thanks

Ben

In the TDI output to file?  It's counting statistics calculated the same way for the error bars in the TDI plots, but always 1 sigma.

[Ben Buse]

Sorry for the slow reply. No I meant in the output of results - (Save analysis output, or save user specified format output)

Thanks

Ben

[John Donovan]

Yes, they are all 1 sigma counting statistic errors.

[BenH]

Hi John.
I'm not sure you are calculating the uncertainty for data collected with the TDI correction switched on.  Are you taking in to account the uncertainty of each point on the TDI curve and propagating uncertainty through the extrapolation to time=0 seconds?  We ran 30 points on a beam sensitive glass and compared the standard deviation to the uncertainty reported by the software.  With TDI off the standard deviations compare very well to the reported "% ERR".  The absolute uncertainties for these analyses with TDI switched off are shown below (they agree quite nicely):

TDI off                   ZnO      Al2O3       SiO2      
Standard deviation   0.09       0.07       0.16           
Reported "% ERR"   0.09       0.05       0.18           

With the TDI switched on the reported standard deviation values are quite a bit larger than the %ERR as expected if the uncertainty on each TDI point isn't propagated through the extrapolation to time zero:

TDI on                   ZnO      Al2O3       SiO2
Standard deviation   0.46       0.25       0.62
Reported "% ERR"   0.09       0.05       0.18

In this data set we used the log-quadratic fit to the data.  The uncertainty calculated by the software might be misleading if I have this right.

[John Donovan]

Hi Ben,
Which TDI output are we discussing here?  The Analyze! window/log window output, the Output | Save Time Dependent Intensities menu, or the Output | User Specified output menu, or some other place?
john