If it helps diagnosing the issue, I'm using Firefox 53 64 bit.
Here are ROI and GM counts since the start of the experiment in a standard image format. There is indeed a rise where you indicate, but there seems to be one earlier too:
MFMP: Automated experiment with Ni-LiAlH
-
-
In the latest plots (Thanks!), the gamma seems to have increased where the pressure was low.
Do you have a second detector that is also showing activity?
-
Yes, there is an SI-8B pancake detector that has a grounded foil shield that is also detecting in addition to the scintillator (ROI shown here is broadband). In the bottom of the scintillator plot (first of the graphs in the images above that can included in his post above, is the plot of the counts for the SI-8B pancake detector.
-
The change in GM counts (in particular) seems interesting. I've made a dedicated graph of its data resampled to 60 second bins, which makes the trend over time a bit more clear:
In addition to the graphs updated hourly on the shared folder on mega.nz (last 14 hours of data plotted), below is attached an additional multipage plot with the latest data since the beginning of the experiment.
-
Ok, thanks for explaining, Bob. Now that I see your setup and Can's graph, it seems to me that (1) your GM detector is going to pick up neither alpha nor beta particles originating from the reactor given the presence of the heat shield and insulating bricks; and (2) Can's (earlier) graph appears to show only noise. (Just to mention two obvious points.) The GM counter appears to be there only to detect gamma photons, then. Have you verified that both the GM counter and the scintillator pick up coincident measurements from a check source that is brought into the area of the reactor and then removed?
How is the 137Cs tracer used in this setup?
-
Can, what does that graph with the trending GM counts look like with the scintillator detector signal alongside it?
-
-
The 137Cs "tracer" is a 1uCi check source whose gamma photons are only seen by the scintillator. Since these photons only go in straight lines, they would have to go through the lead brick to get to the GM detector, and don't really make it through in a sufficient intensity to be detected by the SI-8B.
Alpha would not escape the reactor tube. Only beta of very high energy would escape the reactor tube, but not the heater tubes - only the Bremsstrahlung would escape. The GM detector would detect x- and gamma- photons above about 5keV. I have a 55Fe check source and tested that the SI-8B with the foil shield(s) can detect the 5keV. The scintillator will not detect below about 15keV. I can put the 137Cs source on top of the K-26 bricks and see the count in both - seems like I did that once before.
-
Note that the deep 5s dips in the scintillator spectra are the 5s gaps between starting each new spectrum. During this time the ROI output is 0. The GM detector has no off times.
-
The 5s gaps a lot of noise to the graph. Is there a safe way to filter them out?
ETA: These graphs are probably too in the noise to worry about it. A truly interesting signal will probably appear well above such features.
-
I can replace zeroes with the mean of the series, that would remove some of the noise.
-
There is clearly a correlation between the GM and the scintillator ROI fluctuations. It appears that there is greater variation in the low energy portion of the spectrum based on the relative amounts of variation seen in the GM vs the scintillator ROI.
It is not clear whether the cause is environmental or from the reaction. When we calibrate the gamma spectra, we can look for differences in the spectra as a clue. We can also acquire similar gamma spectra/counts during a calibration run with no fuel. Longer term background acquisition can be planned.
-
Longer term background acquisition can be planned.
I'm going to guess that you'll need to do this at some point, even if you find something that jumps out spectacularly. It would be nice to have it done sooner than later so that people don't get their hopes up prematurely if there's an anomaly.
-
In this last experiment, it is a shame that the temperature didn't stay longer at around 400~500°C. The Ni was loading H. Above 500°C Ni didn't take much anymore.
-
-
Using old forgotten code (thus magic code), I've made this. It could have some additions (e.g. average subtracting) but it's a start. The pdf documents contains other calibrated spectra.
-
I wonder whether the small correlated drift in the counts for the two detectors goes back to a changing of the circulation of air in the room altering the ambient concentration of radon progeny in the immediate neighborhood of the detectors.
-
How old could the code be? You just did this a couple of weeks ago!
Particularly for spectra that may contain information in addition to background, you can't just average all of the spectra. If you are going to use them all, then you need to median combine the calibrated spectra. But this only works if there are just a few instances of information in otherwise null files. You can be pretty sure that below 200°C that all of the files are probably null - unless there were radioactive elements created. I am going to keep running the spectra for a while with a new name - PostExperiment - or something. Tomorrow I may remove the reactor tube and take spectra that don't have even the reactor tube present.
-
Up here in the mountains at 8850' elevation we get enough wind that it is unlikely for radon to accumulate in the lab - I have never seen a signature of radon in all of my spectra (I am not as subject to Fukishima fallout and radon as Alan in Santa Cruze). It is more likely that we would experience cosmic ray effects of some type. We have been under perfectly clear skies for the duration of this experiment. I will run the spectrometer overnight and into tomorrow - we can look for commensurate variations.
-
The data from this experiment is complete and the DAQ is no longer accumulating samples, but the spectrometer is still completing one last spectrum as part of the dataset. After that I will start a new spectrometer sequence (same 5000s live time integration).
