Maybe measuring the heat across the glass isn't a good idea
We did get a chance, today, to look at the possibility of correlating the T_GlassIn-T_GlassOut vs Power In. The results are not encouraging.
The data shows a very large spread between calibration runs, as well as interesting knees and bends in individual lines. Some of those would be interesting to dig into at some time, but for the moment, the other two calibration curves are much better choices.
Comments
However you do the experiment. the problems are the same. The current error spreads are due to diagnostic errors - or an as yet un-quantified variable affecting the results. These always initially have to be ironed out.
Quantifying energy transfer to a water bath is no more simple than the current setup.
For example if the spread is much worse for Glassin then you know that the problem is not with ambient room temperature variation.
This helps in determining the error - what parameter has changed consistently in one direction for these curves?
You don't need to boil water to prove excess energy.
Ambient room temperature might be one of the problems - but it can not be the only one as 8 Celsius room temp spread would be easy to spot - and it does not show up as a random error in each curve.
There must be some variable that accounts for the spread of the curves. Perhaps time of day (or ambient room temp) ??
By the way, the work you guys are doing, this whole project: It's truly wonderful, totally amazing and very inspiring. There are a lot of us rooting for your success!
A random error would affect the smoothness of the curves as well.
A systematic error affecting both temperature readings would tend to be cancelled out subtracting one from the other.
It is very odd that subtracting such thermally close coupled points does not reduce the error spread.
And it seems fishy to me that each run is so smooth (low standard deviation from the fitted curve) but such a big spread run to run.
This must give some information as to where the error may lay.
My opinion is that since we already know that the gain is potentially around 50%-75% more than the input power (a relevant and scientifically macroscopic achievement), then focusing too much on the absolute precision at this stage might be counterproducti ve to the replication progress. The focus should be on triggering the reaction.
But I can understand why the MFMP team might want to focus on accuracy first, and then attempting to retrieve a large signal from the system. Having to backpedal on apparently successful results wouldn't be very fun.
The results suggest that you may prefer to use a heat flow transducer rather than spot thermocouples across the glass.
It might also stabilize the reading if you could temperature-con trol the environment your device is operating in. What I think you are seeing are variations in the temperature of the outside due to ambient temperature changes and perhaps air currents. The idea is good but I think without temperature control of the external surface, there will be too many variations.
@Rats
Even though it's simpler, I don't think a water bath would be as flexible and efficient as an external water jacket and flow calorimetry. I agree that if the effect is very strong, for example a 50% increase in heat flow in the active wire compared to the control, then almost any calibration method will work including any of those already used.
Water boils at 100 degrees.
I am sure there must be a simpler approach to testing for excess energy. From memory the COP of the Celani cell is at least 1.5 (as high as 2.8). That is pretty substantial and you don't need micro-measureme nts to determine this. So why not simply insulate the cell as best as possible and dunk it into a bucket of water. The temperature rise should be enough to determine the amount of excess energy produced. Am I missing something here?
RSS feed for comments to this post