Santa gave us a mountain of data for Christmas. We have had a lot of data to crunch through in the last few days (2 cells, 10 runs, up and down steps, and extra sensors) To help get our hands around it, I had to invite Wes, an intern we had last summer, to come back for a bit.
Our goal with this calibration experiment was to help us better understand the repeatability and predictability of this apparatus as we try out the vertical orientation.
Here are some key questions we asked of this data set:
1 - What measurements in the cell correlate cleanest to the input power and how much uncertainty is there with that measurement?
We explored 4 possible indicators to correlate with Power In:
- (T_Mica -T_Ambient)
Visually, we can see that some measurements were less consistent than others. These nice graphs by Alan G, from the comments of hte last bog article illustrate that nicely. Cell 1.0 clustered nicely with data points falling close together.
Cell 1.1 had much more variation. If we were running an experiment with a live wire based on this data, we would not know for sure if we were making 20+ watts.
To get a grasp on which measurement was cleanest, we processed the data down according to the following diagram
From the confidence intervals calculated in temperature, we multiplied that by the slope of the temp vs power line at that point to come up with an uncertainty in watts. These are the confidence intervals we attained for each measurement:
And Cell 1.1
We put the T_GlassIn-T_GlassOut data on separate charts. For Cell 1.0, they clustered well, as you could see from the tightness of the curves on the graph from Alan G way up above.
For Cell 1.1 it was very different, though. It was an order of magnitude higher. In the comments, Alan G pointed out that this may be because the T_GlassIn sensor may have changed it's thermal contact to the glass.
So, we settled on ignoring the T_GlassIn - T_GlassOut, but paying attention to the other three simultaneously. This is what we will use for our calibration curves. The lines are on this along with the confidence limits. If we get points beyond those confidence limits, we can be 99% certain that we are seeing excess energy.
We will settle one of these lines to use to calculate a representative P_xs calculation, but we will be checking all 3 data points against these baselines and uncertainties.
2 - How did the 0.5 bar runs compare to the 1.0 bar runs?
- For the most part, it was close enough that the 0.5 bar runs fell within the confidence intervals in the graphs above. T_mica, T_well are very close, T_glass out was at upper confidence limit - Both runs at .5 Bar were higher than the 1.0 bar runs. So, we did a third run done yesterday to see if we will need to adjust our T_Glass_out baseline, but that made it fit in even better. All these 0.5 bar lines are included in the graphs up above.
3 - How well settled were the temperatures after the 45 minutes?
The reason for doing the step ups and the step downs is to be able to measure the difference between them and see if the settling was adequate this is what we found by taking the difference for each cycle and then graphing the average and then each point as well, to give us a good idea of the range of variation. There is a real positive bias which means that there are long time constants involved.
Separately, when we looked the thermocouples on the flanges of cell 1.1, it looked like it had a good 60 minute settling time, at least. Doing the ups and downs and averaging them works pretty well to take care of this and any other long time constants in the system.
4 - How stable was the environment around the cell?
The ambient was relatively tight (27.6 +/-0.4C)
The other air sensors mounted higher in the air stream in the vented hood showed more variation - about 1.5C.
Both displayed very similar behavior. I thought we had the new cell pretty tight. Perhaps it is in the plumbing. I guess we will have to sniff for leaks and see if we can fix it.
Here are some questions we didn’t ask at first, but the data voluntarily revealed:
Interesting Wire Resistance Behavior
The behavior of the bare Isotan-44 wire changed from the first cycle to the second cycle. The Isotan wire started out actually decreasing resistance as the temperature rose. On the first run it decreased all the way to the max temperature of the test.
On the second cycle, though, the resistance dropped till the temperature hit about 190C on T_mica, or so. Then the resistance started rising as the temperature increased up to roughly 270C.
All subsequent runs also had this dual behavior of the wire.
This says to me that the metal of the wire changes behavior as the hydrogen is loaded. The behavior of the untreated wire may be able to inform us about the behavior of the treated wire. Obviously, the basic curve is one of reducing resistance, presumably as the wire absorbs hydrogen.
Note: The resistance data showed a large rise when the power went to zero. That is because we were at such a low level of voltage and current with that power supply when it is set to zero, that we are at the limits of precision for the measurements.
My goal is to decide very soon whether we are satisfied with these points as an adequate baseline or if we need to do more testing. IF we think we're satisfied we will install new Celani wires in each cell and load over the next few days with the goal of triggering on Dec 31. One major question is whether we can get the Celani wire installed in a similar enough manner to the isotan wire to avoid having to test it in helium. I would like nothing more than to have an excuse to drink twice as much champagne to ring in the new year.