Harry, I use a blindfold when the data is being optimized. :-) The LMS routine takes the raw data and makes my simulated curve match it. I do not have any idea what the result will be and it could be either positive or negative. An earlier calibration sets the rules that the data is compared against.
Most of the time I download the live data from the MFMP site and evaluate one of the power steps. If there is a problem with the collection of the data, then that shows up in such a way as to generate a visual flag which I can review to determine why it is behaving in a strange manner. That is a rare occurrence. How would you handle the evaluation in the absence of a calorimeter? Time domain transient analysis is the best that I can do under the current restrictions. I am open to suggestions provided they are possible to achieve, but do keep in mind that I can only request special tests by the MFMP group and I have no control over their decisions. I believe that it is preferable to do something instead of wait for someone else to spoon feed me. I chose to post the results of my program runs to ensure that the vortex group is aware of any progress. If you are serious about blind analysis being useful and not kidding then I will answer. Of course it is important and is essentially conducted every time I run a set of data through my program. Initially, I was expecting to see positive results, but that is not what the program produced. Any new data that I download might demonstrate either positive or negative excess power since I do not have a clue about what will be found. I must admit that after so many runs with no excess power being determined, I am becoming biased toward that expectation, but I do not modify the way the program operates to achieve that result. Dave -----Original Message----- From: Harry Veeder <[email protected]> To: vortex-l <[email protected]> Sent: Thu, Feb 7, 2013 3:52 pm Subject: Re: [Vo]: MFMP Null Result On Thu, Feb 7, 2013 at 1:07 PM, David Roberson <[email protected]> wrote: > The questions that are being asked are important and the MFMP guys are > working very hard to answer them. A number of additional measures have been > taken at various times to root out unusual behavior and to improve the > accuracy of the results. Everyone realizes how important this is to get > right. > > One test that they ran last month was per a request I made that is quite > similar to what is suggested by Jack. First the cell was stabilized with > all of the power being applied to the test Celani wire. At a specific point > in time, the power was quickly shifted to the heating wire which is NiCr. > The input powers were matched to a close degree. I noted that the apparent > excess power changed by about .4 watts if I recall. That actual value for > this discussion is not important, but if you need I can look it up. The > source of the difference was not determined at that time, but both wires > were exposed to helium instead of hydrogen for that test. A vacuum and > other attempts had been recently performed to remove any LENR activity that > might be normally there. The details are written in a log on their site. > > This lack of power output correlation concerned me then and still does. > There are numerous variables to contend with and it is apparent that control > of the accuracy is not trivial. Everyone is getting a good education as to > how difficult these tests are to confirm. > > Lately, I have been worried that the excess power being shown on their web > site(~5 watts) with the current technique that they have been using to > calculate it is far too large to be real. I do not want to see too many > folks let down by reality when the calorimeter does is miracles. Another > guy, Ascoli, used a technique to adjust their results that compensates for > the density changes of the hydrogen. The final curve he determined matches > my steady state program output closely. I use the outside glass temperature > minus the ambient to calculate the instantaneous power which is more immune > to changes within the cell such as gas density. Of course my program takes > into account the delay associated with heating of the glass and monitor. > > The amount of direct hot wire generated IR that escapes through the glass > envelop is a potential contributor to inaccuracy. If this drifts, then the > power captured and monitored on the outer glass test point will vary. There > has been evidence of this effect in the past when goop collected upon the > test wires leading to changes in emissivity. That is the current theory I > apply to calibration drift. Amazingly, the recent calibration factors > appear to be holding well after many days of burn. > > This is a learning experience for all of us. Experimental science is a form > of bondage! Does it ever get better? > > Dave Doesn't S&M include blindfolds? ;-) Early you also said you believe in "letting the data speak for itself." In that case, you should also be blind as to whether the data set contains an expected positive or negative signal. In other words you should be analysing data sets without knowing what exactly is being tested in each set. Do you think in principle a blind analysis can be informative even without calibration data? One could choose any data set as their baseline and see how the data sets *compare*. Harry
