amdx said:
http://arxiv.org/abs/1305.3913
Anyone care to read and review the paper.
I'll play!
Before I started, I went and read the Wikipedia entry on the device
under test, but I didn't read any of the other criticism or alternate
explanations for what it may be doing.
The numbers are page numbers in the PDF. I have also given the section
titles from the PDF.
---
"Introduction"
3: I guess the researchers had to infer how many resistors there were
and how they were arranged because they were not allowed to see the
device disassembled, either before or after the test. Seems a
little strange. I would think the inventor would at least be willing
to show people the outer housing and resistors, without the "secret
sauce" installed.
"Part 1"
"Device and experimental set up"
3: The internal construction of the device is described, but then it
says they couldn't weigh parts of it because the device under test
was already running when the test started. So how do they know that
the description is accurate?
3: Why is the AC input waveform a secret? Does plain old sinusoidal AC
(at whatever frequency) not work? How about DC? DC would be a lot
easier to measure accurately and cheaply.
4: The seal of the end caps is asserted to be hermetic, without any
further proof. Couldn't you take a similar cylinder, hammer an end
cap into it, and work on it with pumps and pressure gauges to test
that assertion? Hydrogens are pretty small, so they will leak
through tiny places. If the hydrogen is essential to the process,
leaky end caps would make the device work worse. On the other hand,
oxygen, nitrogen, or other stuff leaking *in* might make it work
better.
4: I'm sure the 1200 C black paint has a product name, part number, data
sheet, etc. Where is it?
4: In so many words, "We weighed one that was just like the one that was
already running". How do you know? Why not 1) look at the shell of
one, weigh it, measure it, x-ray it, whatever you want to do; 2) make
a unique mark on it somehow; 3) let the inventor put the secret sauce
in it; 4) observe that the mark is still there; 5) do whatever
additional measurements you want; 6) THEN turn it on? In super
perfect world you'd get to watch the inventor putting the secret
sauce in it, but that seems unlikely to happen here.
4: The whole idea of working out the heat produced via an IR camera
seems goofy to me. I know you can use IR cameras for this purpose,
but as far as I know they tend to get used for relative measurements
(this IC/circuit breaker/whatever is warmer than the one next to it),
or for things that don't have built-in temperature sensors, like
maybe electric motors or engine exhaust pipes. In the latter case,
you usually only care about the instantaneous temperature, not the
heat produced. Also, you don't usually care about a 1 C or a 5 C
difference; the minimum step you care about is 10 C or more. The
application here seems to require much greater precision than that.
In high school physics we did calorimetry with thermometers and
polystyrene cups. There was a picture in our textbook of an
industrial-grade version with a vacuum for insulation and all
that. Why couldn't you stick this device in a calorimeter like
that?
The device is going to have some amount of thermal mass. Presumably
the input power gets diddled at 50 Hz or 100 Hz or so, but the camera
is only updating at 1 Hz. I'd think you'd want a little better
update rate on the camera.
5: They decided that they only cared about the apparent power. I think
another way to put this is that they assumed the power factor was
1.0. Why not at least look at the active and reactive power just to
make sure that's a good assumption?
Can the meter they used deal with the presumably goofy waveform
they used? The meter specs say it can autorange from 45 to
65 Hz. If the waveform is goofy enough, it will have components
outside that range. The meter specs say it can measure harmonics,
but is it looking at *everything*, or just at 60, 120, 180, 240,
etc?
Why not hook up a couple of different meters and see if they agree?
Given the levels of power involved, the voltage inputs to the two
meters shouldn't load anything down excessively.
Filming the meter and a wristwatch is kind of a cute idea (I've
proposed it myself for other applications) but that meter can also
log data to its internal memory... so why not use the internal
logging?
Again, it sounds like they used 1 Hz update on the power meter,
which seems like it might not be enough.
6: Why not include the results of the radioactive monitoring in this
report?
"Data analysis"
6: They assumed conduction was negligible. It would have been really
complicated to stick a couple of thermistors or thermocouples on
the steel framework periodically to see how hot it was getting, I
guess.
"Calculating the power emitted by radiation"
7: Why couldn't they measure the emissivity? This seems like something
you could easily do if you had the outer shell, devoid of the secret
sauce.
7-8: They only had one IR camera looking at the bottom of the device.
Why not have another one or two looking at the sides? I understand
that if you had one on top, convection would tend to heat up the
camera itself, but I would think you could tell the camera to factor
that in when computing a temperature.
"Calculating power emitted by convection"
11-12: Well... OK. I guess you can do it this way, but I think I've
read that in the real world, this depends a lot on the fluid,
currents in the fluid, the exact shape of the devices involved,
and so on.
Again, wouldn't it be simpler to stick it in a calorimeter, and
know that you've captured *all* the heat the thing is putting out?
"Performance calcuation"
13: Is there a rationale for the 10% error number?
"Ragone chart"
14, Fig. 9: Why not get a chart that has both conventional and nuclear
sources on it, so the results for the device under test can be
plotted with some context?
"Part 2"
"Device and experimental set-up"
15: This time they at least give a brand name for the paint. Still
no part number or spec sheet. They also admit that somebody can't
paint evenly. Still no emissivity measurement.
15: Magic power supply again. I wonder what the output waveform looks
like?
16: Using two IR cameras now, which is maybe somewhat of an
improvement.
16: Assertion that the video recording is non-falsifiable. Got a
chain of custody for that recording? Did you buy the camera at
random from a shop?
17: Trying to measure the emissivity. Again, why couldn't you do this
in a more accurate way by testing the outer casing without the
secret sauce in it?
18: Again, why not include the full report on possible radioactivity?
"Analysis of data obtained with the "dummy""
18: They did try their power meter on the input line to the resistors.
No mention of what the waveform looked like, or what the meter
thought the distortion was, or anything like that. If the control
box really was dissipating 100 W, it must be a fairly beefy thing.
"Analysis of data obtained with the device"
20: The 35%/65% on-off time is kind of interesting, but over what
time period?
20: Any rationale for using 5 divisions instead of the 10, 20, or 40
used in the previous test?
22: They seem to be using the 2% "error" number for radiated energy
as the error number for the convective energy also. I am not sure
this is justified.
"Ragone Chart"
22: Why not just do 37.58 kWh / 116 h = 0.324 kW for the average
consumption, and not have to refer to the (estimated? calculated?)
35%/65% duty cycle?
22: Is there a rationale for the 10% error number?
23, Fig. 15: Why not get a chart that has both conventional and nuclear
sources on it, so the results for the device under test can be
plotted with some context?
"Remarks on the test"
25, Plot 3: Would you care to label your X axis? It is presumably
seconds, but this is not stated. For extra credit, tell Excel to
scale it so that the major divisions are 60 seconds or 100 seconds
or something reasonable.
26: Trying to prove the heat output is not due to a resistor alone.
How about a resistor inside the same steel can but with no secret
sauce?
27, Plot 7 and Plot 8: Again, label X axis please. Scale X axis to
even units. On the other hand, this does seem to show that the
on time is roughly 2.5 minutes (assuming the X axis is seconds),
which answers a question I asked above.
27: The argument seems to be "the emitted power keeps going up after
the resistors are switched off, so something else must be
happening". What if it's just the heat from the resistors finally
working its way through to the outside of the cylinder?
---
Before anybody asks, I don't work for an oil company, mining company,
or anybody in the power generation and transmission industries (coal,
nuclear, hydro, wind, geothermal, whatever). The investments I have in
any of these companies, if any, would be through a 401(k) plan, nothing
direct. I also don't work for or have any investments in the
manufacturer of the device under test.
Matt Roberds
