Ok, now I understand. The drift in the energy scale is exactly why I included the tracer sources. Even within the 5000s integrations, the energy scale will drift some and cause a variation in the bandwidth of the lines. I have tried to correlate this drift to temperature measured near the center of the NaI crystal on the outside (it is in the dataset), but I have not found a correlation. It probably is due to variation in the temperature of the internal dynodes of the photomultiplier tube, which I cannot measure. Perhaps this variation wouldn't exist when using a temperature controlled solid state photomultiplier instead of an electron tube.
MFMP: Automated experiment with Ni-LiAlH
-
-
I have started a background capture with both the GM and ROI counts and the gamma spectra. I will take 24 hours of files with the reactor cold. It is a useful thing to do while I am working to prepare the next experiment. I will create a new folder for this background data in the Google drive, but I am not going to bother putting the data in the folder as they come in - I will just dump them all in tomorrow.
-
Great idea, Bob. Since it's probably inconvenient to tie up those detectors for a full month, for example, perhaps you could accomplish the equivalent by keeping them on and capturing the data during intervals between experiments? My layman's guess is that a month worth of background data capture should put any anomaly that stands out above that background on a strong footing, especially if it's repeatable or recurring.
-
That should also eventually work as a background to use for later operations/calculations. If the spectrometer has a channel drift problem it's better to have a background spectrum composed of many different files instead of a single long one, so that they could be individually calibrated and then averaged together.
-
(And as I think Bob has mentioned, the variance in the background is important to analyze and take into account as well, in addition to the mean and median.)
-
I chose 5000s as a compromise integration. It is short enough that there will not be big energy shifts in any single integration and long enough to acquire the tracer sources with sufficient counts to determine their energy locations. It is not so long as to contaminate a single integration that might have an outburst with long term noise. Taken together with the time domain ROI, these should pinpoint the nature of an outburst in time and energy. I am going to collect as many days of this data as possible between experiments. There is nothing that should require the data to be taken continuously unless it is a lunar cycle that is having an effect. Daily should capture any diurnal variations.
Also, in the last pre-calibration, no gamma data was taken (they were just OFF). In the next pre-calibration, I will leave on the detectors. I am still optimizing the neutron detector threshold level, but that is coming along nicely. It may be possible to integrate that count in the next experiment and in radiation meteorology that follows.
-
I added a folder called "GammaMeteorology_20170507" to the experiment folder, "Experiment_HClEtch_20170503". It contains 24 hours of cold data that includes GM and gamma scintillator ROI counts. The corresponding gamma spectra files are also included.
-
The signal seems to have overall changed less than last time, but I'm not sure if to any significant manner.
EDIT: GM and ROI do still appear to be correlated to each other. Quick graph below.
-
Very nice analysis and graphs can . I would expect the GM to have less S/N than the ROI because these random sources have something like Poisson statistic noise. The S/N goes about as the square root of the counts. With fewer counts in 60s, you would expect the GM to have more noise.
It is interesting to speculate what causes the variation. Could it be the moon's location blocking out particular cosmic sources? I will check the sun and moon's phases for this dataset. The skies have been clear. We don't get aurorae at this latitude.
-
It could be a coincidence, but I wonder if it has to do with solar wind density.
http://www.swpc.noaa.gov/products/ace-real-time-solar-wind
-
-
The data is here:
ftp://ftp.swpc.noaa.gov/pub/lists/ace/
The files of interest are the ones ending with "_ace_swepam_1m.txt", column "proton density".
EDIT: from a quick check I think the correlation will not be perfect/simple because the data comes from a satellite in orbit at the L1 Lagrangian point, or to put it simply always facing the sun, while your location is subject to night-day cycles.
-
So, maybe the correlation would have to be a function of the solar flux density times maybe the magnitude of the cosine of 1/2 of the local solar angle where the local solar angle is zero when the sun transits.
-
I think it could also be a function of the status of the Earth's geomagnetic field; perhaps there are better indicators for the actual signal reaching the Earth than what I might have found.
EDIT: for example, there's a USGS magnetometer in Colorado which measures the magnetic field of the Earth at that location and the data seems to somewhat correlate with the GM counts from your latest run, but it's not perfect; the signal is likely a combination of many things.
https://geomag.usgs.gov/plots/
-
Probably the most valuable thing is not to find the perfect correlation (which would be interesting in and of itself), but rather get an idea of the magnitude of natural variations so we have a reasonable metric to say when a measured signal during the experiment has exceeded natural variations and by how much.
-
The field coil is done. This will be used for electromagnetic excitation of the reactor by having this suspended around the insulated reactor tube k-26 bricks. The coil is 4 layers of 15 gauge heavily enameled copper wire with a total of 130 turns epoxied together while on the form. The length is 2.25" (57mm) and the inside square dimension is 3.5" (89mm) with rounded corners. The inductance is expected to be about 1.2mH and should produce a magnetic field of about 28 gauss on the axis per ampere of current. I plan to add a series 1uF capacitor and drive it at resonance with a high power audio power amplifier. I should be able to get at least 5A peaks for 140 gauss peaks at 4600 Hz. Here is a picture:
The script driven Labview control program is being modified to accept a new script command, "W", which will play a specified .wav file from the computer to drive the audio PA and the coil. Initially this .wav file will only be a 5-10 second pulse of a 4600 Hz sine wave, but there is nothing keeping me from having the .wav file be a chirp, or any other waveform. The coil will be series resonated for sine drive at resonance to provide a greater current and magnetic field. If I drive the coil directly, the reactance of the coil will limit the current and broadband magnetic field magnitude, but then almost any waveform can be used.
-
How do you get resonance Bob? By changing the capacitance in some way - an L-C circuit?
-
From what I remember reading, rumors are that there is a specific load-matched frequency, possibly between 500 and 2500 Hz, that somehow is able to induce excess heat in a suitably prepared cell. Reportedly, every fuel load has a different optimal frequency and apparently a square wave works best. So, instead of a fixed signal I would go if possible for a long, slow frequency chirp and observe any effect on the reported temperatures.
It's not mentioned from these tips when exactly this signal should be applied. -
and drive it at resonance with a high power audio power amplifier
This morning I received an update from me356.
He tells the research goes forward and that he will try to find time to answer some questions soon. Here are some pictures.
Interesting that you are using a audio amp for the power source. When the photos of me356's controller was published, I commented that the large cable connector was a "Speak On" audio cable connector used in newer higher powered PA systems. I use them with my 3000watt RMS amp. (I play acoustic music so, the 3000W is a little over kill! )
These connectors are rated at up to 40 amp. I wonder if me356 has designed a similar system? Although since he is connecting the Speak On connector to his controller, he would have to have some integrated amp, which seems unlikely. Interesting none the less!
As always, thanks for sharing your adventure with us!
-
I will put a 1uF non-polarized capacitor in series with the coil. Once it is in place (not that the reactor will make much difference), I will put a 100 ohm resistor in series with the C-L and sweep it in frequency with a series resistor of about 100 ohms so that it can be driven by my audio generator. Here is a simulation of what I am describing:
In this simulation L1 is the big square coil. R1 is the DC resistance of the coil. C1 is the non-polar capacitor that I will add in series. R2 is only put there for measurement. Measure the voltage amplitude from the top of C1 to ground while you sweep the frequency of the audio signal generator. What you will see is a big dip in voltage as C1-L1 go into series resonance. At series resonance the reactance of C1 cancels that of L1 and you are only left with resistance, R1 at that resonant frequency. It becomes an ~100:1 resistive divider at the resonant frequency. Once you have found the resonant frequency, just create a .wav file with that frequency as a tone.
The C1-L1 WILL resonate at some frequency, approximately 1/(2pi sqrt(L1 C1)). The trouble is that you only know L1 and C1 approximately and need to find the exact resonant frequency to use in making the driving .wav file. Making the .wav file with a tone at the measured resonant frequency is trivial with an audio editor.
