FacebookTwitterDiggStumbleuponGoogle BookmarksRedditTechnoratiLinkedin

Celani Replication

Keep up to date with our replication of experiments based on Celani's cell right here.
Please feel free to add your comments to the blog entries.

LIVE DATA | Experiment Log USA | EuroLatest Video | RSS feed of blog posts | RSS feed of all comments

A Beautiful Beast of an Apparatus (Update #3 - Cool Videos)

on .

Lots to report from the last few days

1- Difference between sensors

We learned from all the additional sensors we installed this week that our temperature control int eh room was very good at controlling the temeprature at the one point of the sensor.  The other temperature sensors shows some variation throughout the day, though.  At least that explains some 12 hour trends we saw during our previous runs.  Going forward we will be more careful about that.  

The graph below illustrates this point.  The cell temps all dropped (no power applied to cell at all), while temperature where the T_ambient sensor is located stays in a tight range.

Here we see the Air temps at the top of the hood run warmer than the cell temps and the T_ambient sensor.

We were suspicious of the new, very inexpensive thermocouples we were using to read the air temps.  This is what Malachi learned about them:


To gauge the accuracy of our new thermocouples we took T_A3 and submerged it into a hot water bath and boiled the water.  This took place around 10:05 AM local time.  The resulting temperature was 100 C (+-0.05 C).

Then, to gauge the precision we took all three ambient thermocouples (T_A1-3) and brought them to the same point. We isolated them from each other in 1 inch by 1/4" diameter individual copper tubes. These three tubes were then put in a larger copper tube with tape on the ends to restrict air flow. Then we waited and the resulting temperatures were within 0.05C of one and other. We can safely say the new thermocouples have an accuracy of +-0.05C and a precision between sensors of +-0.05C.

2 - Installing cells vertically with air flow

Per suggestions from Celani and Ed Storms and a few others on the blog, we are trying some runs with the test cells oriented vertically. The idea is that the air flow outside the cell and the convection cycles inside the cell with make the inside glass and outside glass temps more consistent and usful for indicating actual heat flow. Dr Storms would like to see us using the temperature difference across the glass as an isoperibolic calorimeter. 

Much work was done.  More pictures and video to come below.

3 - Repaired pass through, cracked wire support

Yesterday, as we installed fresh NiChrome wires and Bare Isotan wires into the cells, one of the copper wires glued into the passthroughs broke.  It broke right up close to the pass through so we couldn't solder or attach to it. We decided to attempt a repair.  We heated the passthrough with a propane torch to break down the epoxy.  It worked, but it was a very stinky success and there was black gooey epoxy all over that came off only with acetone and a tiny buffer wheel.  We also cracked a Macor support, but it was only a tiny piece of the corner that came off. We added a dot of epoxy where it meets the flange and called it good.

4 - AFC - Getting ready to test - Insulated stainless cell

Early test are looking promising on the air flow calorimeter.  They are mostly informal, so far.  We have been very busy getting the two cells mounted in the vertical apparatus, but we did manage to put together another core that will serve inside a stainless shell for inside the Air Flow Calorimeter.  The one on the right will be in the calorimeter.  We will insulate the stainless shell of the shell in 2 inch (5cm) of mineral wool insulation so it will require very little energy input to get up to working temperature.

Here are the details of the wire pass throughs.

5 - New wires arriving next week

In the reworking process, we decommissioned the first Celani wire we had used.  We will try to take some pictures of it soon.

Next week we will receive some new Celani wires to put into these fun toys.  More on that when they get here.  We're hoping to have enough calibrations done to be able to pop them in and have some active runs over New Years.

6 - The Monster-Twin-Cell-Vertical-Apparatus and the Calibration Routines

Here is the groovy stuff going on right now.  We just assembled and commissioned this dual cell apparatus.

Cell 1.0 - The Mica/Quartz glass cell with NiChrome on the blue power channel and Isotan wire on the red power channel.

The Macor/Pyrex cell with NiChrome on the blue power channel and Isotan wire on the red power channel.

Isn't that an intimidating mess of sensors?  

The data collectors are set on this new shelf under the hood.  This isolated their heat output from being directly on the cells.

The data is streaming.  We are doing a 7 step up, 7 step down calibration.  Both cells are at 0.5 bar and will do this calibration cycle twice.  Then I will up the pressure to 1 bar and perform that calibration 6 times to get a good idea of the variability and error bars.

Update #1

Happy holidays and happy data watching.

Update #2/3 (now with video)

Just in case the world did end, the EU team wanted to have a bit of fun on their last full day together in Switzerland, so, in a break from the normal schedule, they did a Mizuno type experiment.

On a serious note, we decommissioned the first EU active 360 layer Celani wire from its cell and Mathieu will be taking that back to France to, as planned, see if anything can be seen under the Scanning Electron Microscope and other tests.

Removing the wire

After which, we drank "LA FIN DU MONDE" beer at the university bar - which is French for "The end of the world"

The end of the world

If tomorrow comes - we hope you can join us in our continuing journey.
Happy Christmas Everybody!


Add comment

Security code


0 #56 MaisieJuicy 2018-08-23 08:06
Hi. I see that you don't update your website too often. I know that
writing articles is time consuming and boring.
But did you know that there is a tool that allows you to create new articles using existing content (from article directories or other websites
from your niche)? And it does it very well. The new articles
are unique and pass the copyscape test. Search in google and try:
miftolo's tools
0 #55 73Brittny 2018-07-31 17:24
I can see that your content probably doesn't
have much traffic. Your posts are awesome, you only need more
new readers. I know a method that can cause a viral effect on your site.

Search in google: dracko's tricks make your content
go viral
0 #54 Eric Walker 2013-01-01 06:51
Concerning error, @AlanG, from your graphs I estimate a rough scatter of +/- 3.5C for cell 1.1 (comment #49) and +/- 0.5C for cell 1.0 (comment #33). I believe you did some filtering on the data in order to compensate for the possible mismatch in sampling with respect to the power step? To get a better sense of the actual recorded scatter, it would be nice to see the same graphs without any kind of filtering.

Is there a way to indicate the error bars on the live feed somehow? I'm not sure exactly how this would be done -- perhaps the usual error bars, spaced every one or five minutes?
0 #53 Chuck 2012-12-28 19:13
I'm not a physical chemist, so I don't have any training or knowledge in this aspect, but do we have a way of tracking the hydride content of the wire during operation? Is there a way to do this?
0 #52 David Roberson 2012-12-28 16:24
I need to have a run with the pressure held as constant as possible and at the same level as the calibration to verify these results.

Also, it is important for a calibration run with an inactive wire that is guaranteed to not generate excess power in order to prove that the vertical system follows the same non linear differential equation.

We may be getting the results that we seek, but it is a bit premature to celebrate. It is too bad that we are only seeing 1 watt or so, but the actual level needs to be more accurately determined.

I will be out of the office for the next few days.
0 #51 David Roberson 2012-12-28 16:18
Ok, I took the very latest calibration step test for the cell FC0101 and used that to obtain the latest coefficients. These were then used with the latest full rising waveform step in power and I can say that the results are definitely different than I have seen in the past with earlier time domain runs.

I am unable to make the temperature versus time waveform match the non liinear differential equation solution as before and the difference is not trivial and apparent.

I am concerned about the fact that the pressure seems to be differernt for these two intervals of time so that might be an issue, but if not, there is definite evidence of something unusual with the latest test.

If the Celani wire is loaded, then I suspect that we will be able to verify that extra power is being generated after more careful calibration. I estimate that the excess power is in the vicinity of 1 watt from a quick look at the fit.

Dave Roberson 12/28/2012
0 #50 Ecco 2012-12-28 14:26
I'm noticing that the 1.1 cell is slowly gaining pressure although it's currently turned off. Wouldn't this imply a risk of air contamination when operating the reactor below 1 bar of pressure?
0 #49 AlanG 2012-12-28 04:56
Final chart of cycles 5-8 for cell 1.1. Something changed during the down ramp of the 7th cycle. Looking at the source data I can see that peak T_GlassIn dropped 8 degrees from the previous cycle, but none of the other parameters offer a clue why.

Two adjacent data points are shown for each power transition, to show the range of measurement error with the current setup. The chart source data is from 12/21/2012 21:22
to 12/25/2012 21:32 PST
0 #48 Ryan Hunt 2012-12-28 04:44
The valid calibrations for the vertical cells all started about 21:00 gmt on Dec 21. The first two were at 0.5 bar H2. The next 8 were at 1 bar starting pressure. Yesterday we turned off cell 1.0 and then turned it back on. Today we turned off cell 1.1 and turned it back on. Then, tonight, we started with another 0.5 bar calibration cycle. The details are all in the experiment log.
0 #47 David Roberson 2012-12-28 02:12
@Ryan, I took a quick look at the cell data for cell 102 and it is a bit different than the earlier calibration I obtained from the up/down stepping. The fit was not perfect, but a good start.

Could you list the times during which calibration data is valid for the vertical cell? I am a bit confused about what I have at the present time.

My model is suggesting that the high level drive is at 106.5 watts which I assume is a bit high. The data I obtained did not state the power levels or times of change.

Also, is the Celani wire used during any of this data run?
0 #46 AlanG 2012-12-27 21:45
The two recent charts are extracted from 1 minute data. I then filtered the data by looking for delta power above a threshold and selecting the third sample back from those. I observed many cases where two adjacent samples had significant power changes, and in some cases also significant delta temp change even three minutes back. Some of this is undoubtedly jitter in the thermal environment, but there are enough that I wonder if there isn't something going on with the power measurement as well. Perhaps the problem is in how the raw data is averaged for HUGnet.

Quoting Ryan Hunt:
Thanks for the attention to detail, Alan. Have you looked at the 1 minute data or the 30 second data? The power jumps within one of the 2 second raw "History" reads. I wonder what the same data would look like if you took the previous five 1 minute reads before a power change.
-1 #45 charlie tapp 2012-12-27 18:12
i have a question i realy want this thing to work but even when you are seeing exess power it is very small. could the extra heat be from the wire being wraped around and around turning it into a chock coil? because a choke coil will put off heat. also if i am wrong which i am alot could the wire acting as a choke coil be holding up the results we are looking for?
0 #44 Ryan Hunt 2012-12-27 18:05
Thanks for the attention to detail, Alan. Have you looked at the 1 minute data or the 30 second data? The power jumps within one of the 2 second raw "History" reads. I wonder what the same data would look like if you took the previous five 1 minute reads before a power change.
0 #43 AlanG 2012-12-27 17:54
@Ryan Hunt - I'm convinced that much of the noise in the measurements is due to a timing problem. Many of the power increments appear to span two sample periods (at 1 minute intervals in the graphs). You can see the consequence of this in the discontinuous step in the Down1plot at ~80 watts, for example.

Is there a ramp time parameter in the power control mechanism?
0 #42 Ryan Hunt 2012-12-27 16:14
@ AllanG - Thanks for those nice graphs. We are digging through the data to see what changed in cell 1.1 after the peak of the second cycle. It led us to the resistance data, which did seem to change modes at the peak of the second cycle, but it did it in both cells. Now we are looking into why the first two runs in cell 1.1 started out with such a sag at the beginning as opposed to the sharp clustering at the bottom of the 1.0 cell. One thing is for sure, the Delta T across the glass for cell 1.1 is not adequately consistent at the moment to be a good indication of power out. It is very noisy in these graphs. Even the relatively clean clustering of the lines in cell 1.0 have variation that would be equivalent to many watts.

@ David Roberson - I did a cell power up yesterday on 1.0 and I will do one on cell 1.1 in a little while. That should give you a chance to compare your model in the horizontal to the new vertical orientation on a full big step.
0 #41 Robert Greenyer 2012-12-27 13:28

Thanks for all your analysis people, this really helps us see things.
0 #40 Ecco 2012-12-27 09:28
@Robert Greenyer: sorry for the delayed reply, but this is something I realized only now.

With the 2-layer wire according to the STM experiment, there was about a 20% increase in output power to input at 350 °C (wire temperature or reactor temperature?). If two or them are going to be installed in the MFMP 1.1 reactor, this percentage should increase even if just one of them is going to be heated directly. Perhaps by 10% or so?

A 30% increase with an input power of 50W would imply a DeltaT Out difference of about 20 °C. I'm wondering how much would be needed at the least to rule out any possible imaginable artifact.

This is data from one of the latest calibration cycles for the MFMP cell v1.1 from which I estimated this:


(sorry for the poor graph this time)

EDIT: btw, "Delta T Out" is T_Rise, which is T_GlassOut - T_Ambient.
0 #39 David Roberson 2012-12-27 07:51
My model simulation behaved as expected. The transition between one power level and the next was a perfect fit to the prediction. I did use data from the same run as the input and it appears from the chart below that things might have changed between up and down runs.

If this is true, then the difference must be rooted out if the results are to be reliable.
+1 #38 AlanG 2012-12-27 05:34
Delta T vs Power In for Cell 1.1

It's not as consistent as 1.0, and it looks like something changed at the start of the third down ramp. A closer look at the data there might show what happened.

Spreadsheet is at
+1 #37 David Roberson 2012-12-26 23:54
I want to congratulate you gentlemen on doing a remarkably consistent job of measuring the input power versus outside glass temperature.

I get a R^2 value of .999957 for a curve fit to a quadratic equation. From experience I can expect my model to yield near perfect results in the time domain response.


Here is your generous contributions so far towards our $500,000 target, thanks everyone! : $45,020   Please Donate
See the current state of our booked costs here