On Wednesday, 7/13/16, 8:42pm, in thread Jed Rothwell on an Unpublished E-Cat Test Report that “Looks Like it Worked”, IH Fanboy wrote:
“To the contrary, cold fusion lost its acclaim when certain hot fusion scientists from MIT altered experimental evidence…” and linked to an Infinite Energy ‘report’ written by Eugene Mallove in 1999 that purportedly describes how MIT scientists ‘fraudulently altered’ their data to be more supportive of the ‘no such thing as cold fusion’ position.
Unfortunately, the analysis by Gene, which is based in large part on comments by Mitchell Swartz, doesn’t hold up to examination. In the Mallove report, he shows data figures showing the excess power curves that MIT used and some preliminary versions of them he obtained as the MIT Science Press Officer. Of particular note are the applied power curves for the H2O and D2O cells. The MIT researchers used two matched calls, one with H2O, one with D2O, for their experiments. The cells were open, so water was lost through release of electrolysis gases to the atmosphere. Since the chemistry is different for the two materials, different results were expected with regards to rate of electrolysis.
The calorimetry was the power compensation type. The cells were held at some slightly elevated temperature with a heater, and any power appearing in the cell from any other source would cause a lowering of the heater power needed to maintain that temperature. So the applied power curves would show a drop if any excess energy was present. But, they also showed a drop due to water electrolysis, so the data MUST be corrected for that before any excess signal can be detected. Also, when the electrolyte is ‘topped off’, the power had to go up to heat up and maintain the now larger volume of water. This would induce baseline shifts in the applied power time plots.
The preliminary data supplied to Mallove showed both the H2O and D2O cell power traces for a timespan indicated as 0 to 80 hours for the H2O cell, and 0 to 120 hours for the D2O cell. The plots looked like continuous, noisy lines. The H2O trace had a baseline shift at about 62 hours, and the D2O had two, one at about 16 hours and one at ~100 hours (with a big noise spike going both positive and negative at ~22 hours). But the plots appear to be basically level, i.e. the dropping from water loss is not seen. Obviously the plots were corrected for this. And further, the correction is obviously not the same for the two cells, since the rate is isotope dependent.
The Y scale on both is -.2W to +.3W. The zero watt line seems to be drawn through the average value of the H2O data prior to the baseline shift. The D2O zero on the other hand seems to me to be drawn a little lower than the average of the values between the two shifts. It seems more to be drawn through the average of the ~16 hr to ~50 hr section. The ~50 hr to ~100 hr data seems to average slightly higher, perhaps at ~0.02W or so.
The final, published data uses unconnected dots instead of continuous lines. The places where the baselines shift are not shown, and the part of the D2O plot prior to the first baseline shift that was centered about ~+0.25 W is also not shown. In other words, the presented data is clipped a little in the time axis.
The Y axes now go from -0.2 to +0.2W. The Y axis zero for the H2O cell is about the same, but the Y zero for the D2O data is now shifted up a little such that it is more in line with the overall average for the whole region. I recall reading somewhere that the dots are local averages computed from the raw data, i.e. they are smoothed a little, but I am not sure that is correct. I can’t see much effect on peak heights due to the smoothing, and normally I would expect a little peak height reduction if smoothing was used.
Instead the Y axis labels may indicate what is going on. The preliminary data used “POWER (WATTS)” and the published used “P-CAL (WATTS)”. This says to me that the heater calibration curve has been used to subtract off the heater input power. In the preliminary data, excess power would be indicated by negative values (since we are looking at the applied power, not the excess), but one could not be sure it was not due to a shift in heater power. That is fixed in the published data; now any rises would indicate ‘true’ excess power. A key point is that there are no positive excursions in the D2O data larger than the ones in the H2O data. Recall that H2O was not supposed to show excess power, This means that no excess power is observed in the D2O cell. This all seems pretty reasonable to me. But apparently the shift from raw applied power to the power difference coupled with the flattening of the curves to take out the mass loss effect sent Mallove over the edge. He resigned his post because the University wouldn’t ‘investigate’ the issue. I don’t see anything to investigate myself.
But Mallove wrote: “The Phase-II Calorimetry curves were later investigated in the outstanding analysis by my cold fusion colleague and fellow MIT graduate Dr. Mitchell R. Swartz. There can be no doubt now that these curves were the end result of a serious lapse in scientific standards in this affair that happened at MIT.”
Mallove also quotes a letter from Dr. Swartz:
“The light water curve was published by the PFC essentially intact after the first baseline shift, whereas the heavy water curve was shifted a second time. The cells were matched,12 and solvent loss would be expected to be greater for H2O.
The Phase-II methodology is flawed because it masks a constant [steady-state] excess heat. Furthermore this paradigm fails to use either the true baseline drift, and may avoid the first 15% of the D2O curve in Types 3, 3B, 4, and 5 curves. What constitutes “data reduction” is sometimes but not always open to scientific debate.
The application of a low pass filter to an electrical signal or the cutting in half of a hologram properly constitute “data reduction,” but the asymmetric shifting of one curve of a paired set is probably not. The removal of the entire steady state signal is also not classical “data reduction.” “
So the first ‘shift’ is the sloping baseline correction, and the second shift is the different zero point in the D2O curves I guess. But both of those are fine. Fleischmann and Pons' calorimetric method also accounts for the baseline shift, albeit in a different manner. And the zero shift would seem to be the result of the heater power subtraction. Note that the subtracted power was probably an ‘ideal’ power, i.e. with no noise. So what’s wrong with that? Not a thing, it’s SOP in most cases.
The ‘masking’ claim could be real, but it requires that the excess power production be constant and unchanging over the span shown, which is unlikely. Most excess power curves show considerable variation. There was one paper by some Italians that showed a baseline shift when the electrolysis power was turned on that was claimed as excess heat, but I publicly criticized that because baseline shifts are notorious and can arise from many sources. In their work, they varied experimental parameters a lot and didn’t get any change in their amount of supposed excess power, another indication the baseline shift arose from a different source.
As well, the Storms data I reanalyzed for my first CF-related publication was his second set of data. He posted a first set in January 2000. I immediately noted a strong negative feedback signal in the excess power curve that tracked the input power and told him so. This was also noted independently by Scott Little and he reported it to Ed as well. To his credit Ed figured out he had some ground loop problems, reconfigured his electronics, and removed about 95+% of the feeedback in the second set (you could still detect a very small signal, which was unimportant).
Furthermore, I haven’t actually read the MIT paper because they only ran for 10 days. This was before anyone used codeposition, and it usually took hundreds of hours of running to form the ‘special active state’ that gives the effect (if they even could get it after that long!). So the expectation, even after the fact now, was that no excess power signal was likely to be observed. However Swartz claims a 62 mW signal. How he gets that I don’t know but I don’t believe it.
Swartz is known for some shaky math. He once got into an argument with Scott Little about data Scott had collected. Swartz claimed he could pick out an 80 nW (as I recall) signal from typical calorimetric data that had a baseline noise of 50-80 mW. When it’s in the noise, it’s in the noise…
The expectation was for zero. The MIT guys did what anyone would do with their data, no faking or falsification, and they found zero, which in the end proved nothing since they hadn’t run long enough. To conclude that the MIT folks removed a constant amount of apparent excess power from the curves, especially just so they could reject cold fusion claims, in unsupported speculation based on the data shown in the Mallove report. If the data covered a longer period and if it showed excess power signals, it would be worth getting the original data and a full description of data workup protocols, but they didn’t do that or get that.
So IHFanboy, the MIT work is sound.
Now the hoopla with press conferences, interviews, press releases, yada, yada, was a zoo and didn’t enhance anyone’s standing…