Helium Test Results

Are the Celani wires operating comparably in Helium to the bare Isotan control wires we used to calibrate?

- No - Cell 1.0 was 10 degrees cooler at 40 W on all 3 sensors than it was during the previous helium calibrations.
We checked the instrumentation thoroughly, checking the raw thermocouple voltage and tracing it through the A/D converter, and through the math  and found that everything is working correctly.  The thermocouples are, indeed, reporting that the cell is cooler. 

On the second test run we did last night, the measurements lined up with the first one from yesterday pretty closely.  That would suggest that something is significantly different between this wire and the last one.  It is worth noting that the mica was off-gassing during the high temperatures in the helium calibrations and black coatings were growing on the wire during many of those.  We chose to keep the data from the two runs immediately after installing fresh wire on the belief that the contamination did not affect the wire till higher temperatures and on subsequent runs.  

It has been suggested by a Robert Hunt in the comments that there might be more resistance at the terminals near the ends of the cell, which would cause the heat to be generated near the cooler flanges.  We did test the wire as we installed it and it appeared then and still appears to be in the same total operating range.  It is possible that the resistance varies over the length of the wire and changes and makes hot spots that are farther away from the sensors.  These could be kinks in the wire, more heavily oxidized parts of the wire, or something near one of the terminals at the ends.    The total overall resistance is still in the range we expect, though, so any high resistance spots are on the order of 1 or 2 ohms at most.

We may have to heat the cell up and look at it with the IR camera and see if a thermal hot spot is evident on the glass.  At these temperatures, I think it would be highly unlikely that we would have thermal wavelengths short enough to penetrate the quartz glass and make this big a difference.  The difference between cell 1.0 with Quartz glass and cell 1.1 with borosilicate glass isn't even that big.

-No - Cell 1.1 showed that the T_glass_out was roughly 10 degrees cooler, but the T_mica and T_well were spot on.  We figured out on that the glass assembly bolted back on 120 degrees rotated from it's previous position.  That put the sensor on the windward side of the glass instead of the leeward side.  We loosened the glass and put it back to the orientation it had been in.  On this cell we had 5 pretty consistent Helium runs, so we were able to calculate 95% confidence limits.  As of this morning, it appears that moving it back, again, put it back within the confidence interval, but still at titch low.  I am surprised that there is that much variation from one side to the other.

The lower black line is the run yesterday when the T_Glass Out sensor was facing the wrong way.  The line is from the second run (last night) after the sensor was turned.  I think it is close enough.  Being a little low better than being a little high because it is more conservative for this experiment.

It is interesting to note that Cell 1.0 is still running warmer than cell 1.1 at the same powers.  That lead us to compare the hydrogen calibration data of cell 1.0 and 1.1 to each other in hydrogen.  In the graph below, we have the hydrogen calibration data with confidence limits for Cell 1.1.  Then we applied the average calibration data for cell 1.0 in black lines.  We can see that cell 1.0 usually runs about 6 to 11 degrees hotter than Cell 1.1 and the T_mica and T_Well are farther apart.  Then we added the average of the two Helium runs with active wire for cell 1.0 in orange.  The internal lines generally agree with the control curves in Hydrogen for that cell, but the T_Glassout sensor was much cooler.  Any insights from the crowd?

Resistance Behavior in Celani Wires

For this Helium run with live Celani wires, we ran the first four steps of the calibration script we have been using for all the previous calibrations.  This brought the current up to a presumably safe maximum of 1.5 amps in each wire.  The two wires behaved significantly different.

Here are the power steps taken:

--Cell 1.0--
The 14 layer wire in cell 1.0 was most interesting.  There was one known instrumentation artifact at 11:08 in the graph below.  After that the current rose on the next step and the current dropped off substantially, followed by some strange noisy oscillations.  The following step was also interesting. 
The overall trend was decreasing as the temperature rose, just as the bare isotan had done in that temperature range, but this wire also showed an overall drop in resistance from the start to the end of the cycle from 18 to 16.5 ohms.

On the second test run in Helium, resistance was less noisy and demonstrated a relatively clean decrease in resistance with increasing temperatures.  There were still a few interesting but smaller shapes on the plot, though.


--Cell 1.1--
The 2 layer wire in cell 1.1 also showed a dramatic drop when the current increased from roughly .45 to .9.  There was much less noisy changes, but there are smoother changes and offsets.  The overall trend was much more downward.  The large rise at the end was when the voltage went to a clean zero and the values for current and voltage were too small to get as accurate a reading for resistance.  The whole range of this wire was much tighter, though, ranging from 15.85 to 15.65.

This all leaves a lot of questions.  Was this wire already hydrogen loaded somewhat?  Was there some source of hydrogen in the cell causing some loading?  Is the resistance drop more to do with oxides being reduced by current mobilizing oxygen in the wire?  Is doing this comparison step and passing current through the wire before it is loaded with hydrogen altering the performance of the wire? 

On the second run, this wire demonstrated an increasing resistance with temperature - opposite of the wire in cell 1.0.

Are we ready to move forward and begin a loading?

Not quite.  I want to get a better grasp on cell 1.0 first.  Once I have an explanation I have more confidence in I will move on to loading and running the wires.

What is the calibration we are using going forward for the loading and live run with the Celani Wires?

Based on the calibration data, we have a great deal of confidence in using the T_Mica sensor to calculate power out.  It has shown to be very consistent at constant pressure.

When we took the 99% confidence limits in temperature and turned them around into watts, this is the results for each power level of the calibration runs.  The resulting precision of the calibration maxes out at under 1.5 watts out of 100 watts input.  At the 48 watt starting point for the live run, we'll be at a confidence interval under half a watt.  Not bad for this apparatus.  We achieved that by doing 27 transits of these steps.  And those were split between before and after the rebuild, even.

While we chose T_mica for the automatic calculation, we will be linking the data directly into a spreadsheet here and calculating the P_out based on all 3 sensors simultaneously.  Once we get that smooth, maybe we can make it work on the outside data server and share it with anyone interested.

Calibration curve for the Thermal Shunt Calorimeter

Someone was asking for the calibration data for the calorimeter.  Here is the calibration data we have collected so far on the Thermal Shunt Calorimeter.  As a heat source, we put in electric band heaters. We did two step up and back down cycles.  The first cycle was 2 hours between steps.  The second cycle was 1 hour between steps.  

The output of the calorimeter is a voltage that is relatable to the heat flowing through the fins.   When we averaged the ups and downs, we ended up with this calibration curve.  The one hour and two hour curves fell almost right on top of each other.   Then, for each cycle we subtracted the downs from the ups to see how settled the points were.  Surprisingly, there wasn't much difference between them.

 Like I mentioned, before, though, the device is susceptible to thermal noise like small air currents.  Below is a graph of P_xs calculated from the calibration curve above over a couple days of constant power input at 29.41 W.  We see quiet periods at night and very noisy periods during the day. This gives us a resolution of +/- 0.5 watts even while it sits on a table in the middle of a busy shop.  Not bad.

This is not a definitive calibration study, but instead, a first glimpse at it.  We intend to build a constant temperature chamber to remove as much thermal noise as possible and then run the calibration a few times with heat sources in different locations to see how it responds. 

This should be enough to stimulate a lot of thoughts.  Keep the suggestions coming.