Mizuno Airflow Calorimetry

  • NOISE NOISE NOISE

    The noise THHNew propose in his vagueness is indeed noise.

    Electrical noise?? Audible noise?

    Unless defined and accounted for

    noise is just typical THHnew noise and blather


    RB: this is a very strange approach - that if you cannot exactly understand what causes noise, then it does not exist?


    I don't want to seem difficult about this - the noise here is quite interesting and perhaps someone (maybe ascoli, who is good at that sort of thing) can think of a source that fits it well. Probably the best preparatory work would be to look at approx 500s segments of time in this waveform and work out the noise by regression fitting a line to (only) the segment data and then taking mean square error as a noise amplitude measurement. In addition one could look at the noise statistics of how it deviates from the fitted (linear) curve. Note that the gradient discontinuity when the power is switched off cannot be fitted by a linear curve so should be an end-point, not in the middle of one of these segments. Having this noise amplitude measure throughout the curve we could look at how the amplitude varies with other parameters. I don't guarantee this would lead to any gems of extra info, but you never know.


    A job for RB?


  • Noise is a disturbance signal which overlaps a basic signal, reducing its intelligibility.


    Noise is exactly what you are doing in this thread trying to interfere and reduce the intelligibility of what is emerging from a in-depth survey of the information available on Mizuno's tests.


    Of course, your noise is easily identifiable and can/should be ignored by those sincerely interested in the subject. But it is also a precious indication of the speciousness of the reasons which are brought in favor of the Mizuno's results. Thanks for your contribution in revealing one of the peculiar ways-of-doing of the field.

  • About masses of reactors


    Cruciform 2017

    "

    The reactor is made of SUS 316. Its volume is 2740 cm3 and its weight is 20.3 kg. "



    yes - I know that. It is just that the 2019 paper, as I quoted before, says something different.


    1.1. Old method
    A reactor with a cruciform shape was first used in this project (Fig. 1). It weighs 50 kg. Later, 20-kg versions of this
    reactor were used, as well as cylindrical reactors. All have palladium rods in the center. The rods are 250 mm long,
    wound with palladium wire (Fig. 2). This is the positive electrode. The negative electrode is a nickel mesh which is
    fitted against the inside wall of the reactor, and connected to ground (Fig. 3).
    The older method is described in detail elsewhere [1]. It includes several rounds of cleaning the electrodes by
    heating and evacuation to remove impurities, followed by weeks or months of glow discharge to erode the center
    palladium electrode and sputter palladium onto the nickel mesh.
    The highest power observed with the older method was 480 W output with 248 W input, or 232 W excess (Fig. 4).


    Note that reference [1] here is in fact to paper A which does detail the results, but describes them as obtained using 20kg reactors.

  • Noise is a disturbance signal which overlaps a basic signal, reducing its intelligibility

    The intelligibility of Ascoli turbulence or Ascoli foam is very low.


    THHnew invents a fictional single time constant with his eyeball and backup blather

    postulating noise


    but nowhere gets down to mathematical modelling.


    What are the error bounds on THHnew's 7000 second time constant.

    which fit in 95% of the data?


    Heating phase 7000+- 3000?

    Cooling phase 7000+_ 2000?


    I am just trying to make THH assertions mathematically intelligible.

    There is no point with Ascoli...the noise expert.

  • yes - I know that. It is just that the 2019 paper, as I quoted before, says something different.

    THH has to clarify that with Jed.


    Cruciform 2017

    "

    The reactor is made of SUS 316. Its volume is 2740 cm3 and its weight is 20.3 kg. "


    Preliminary calculations made before based on the 120W spreadsheets

    are consistent with the reactor mass being ~20 kg.

  • yes - I know that. It is just that the 2019 paper, as I quoted before, says something different.


    Dear THH,

    may I suggest you to limit the frequency of your replies to RB?


    It is clear that his interventions aim to boycott the discussion and to make the findings on the main deficiencies of Mizuno's tests less evident, by filling the pages of this thread with posts on specious arguments of detail.


    People who read this test and have a real interest in finding out about the reality of Mizuno's results can easily eliminate the RB disturbing noise by ignoring his posts. If you answer them, they must skip twice as many posts, having to select among your posts those that deal with really important topics.


    I hope you agree with me that at the moment the priority is to get from JedRothwell an answer to the question of the difference between the "Input power" columns of the two 120 W spreadsheets (1).


    (1) Error bounds for Mizuno R19 results

  • I think you are confused about data modelling. Trying to model the data too accurately

    Using the standard Excel trend line modelling

    for the heating and cooling curves

    of the 120W excess heat experiment

    does not seem to provide strong evidence for a single time constant for reactor thermal output.


    The heating curve :

    The software refuses to model an exponential function!

    The heating data appears to not suit any single time constant.

    A polynomial fit of order 2 gives an R coefficient of 0.93.

    The cooling curve:

    Polynomial fitting is not easy.

    At best R=0.97 Order 6


    Exponential fitting is possible but the time constant =33000 far away from THHs eyeball figure of 7000s.

    R = 0.79


    {y values are the delta T in degrees C,,

    x values are the time in seconds

    from the start of heating /cooling.}


    The contention that the thermal output

    of the active reactor can be modelled

    as a large 20 kg ss mass

    with constant temperature input during

    heating is not supported by the data.

    If this were so then good exponential fits

    would be possible


    The temperature input from reactions occurring

    between 20g of nickel mesh and 3 mg of deuterium is anything but constant,

    but fluctuates over a huge range and is indeed

    sometimes negative when endothermic processes

    predominate over exothermic processes.

    This fluctuation increases with external heating and suddenly reduces when this is ceased at timemark

    20,500. This is difficult to explain.


  • Agreed: even my patience has its limits.

  • even my patience has its limits.

    Apparently THH's eyeball has its limits.


    The 7000s time constant that THHnew eyeballs and proposes to model the 2017 Mizuno 120W reactor has very weak support.


    There is better support for no time constant

    during heating and a long time constant of 33000 s on cooling.


    A simplistic modelling of the reactor/mesh as a passive thermal reservoir doesn't cut it in Excel software.


    Perhaps THHnew has proprietary software that can justify this modelling?


    The data shows that the reactor/mesh cannot be modelled with any confidence as a passive thermal objectwith first order exponential character.

    An alternative explanation for the reactor is that energy is not being stored in less dense thermal energy but in other much more dense forms.

    Suggestions from LENR theory include magnetic states of heavier (Z>19) nuclei and the high energy states of hydrogen.

  • List of all the output power curves in the JCMNS Vol.25 (Nov.2017)


    The 120 W tests, that have been extensively reviewed in recent weeks, were reported for the first time in an article published in November 2017, in Volume 25 of JCMNS (1). In addition to the tests at 80, 120, and 248 W of nominal input, whose output power curves were reported in Figure 27 and 28, the JCMNS article contains the output power curves of other 3 tests.


    The following table lists all these curves and the respective tests in the same order as the figures:

    N.

    JCMNS 25 Nov.2017

    Date

    Reactor

    Test

    Pin (nom)

    Pout

    Time constant

    Electric heating

    1

    Fig.18

    ?

    Control

    Calibration

    100 W

    -

    n00 s

    External (d)

    2

    Fig.21

    May 10, 2017 (a)

    Active

    During treatment

    "

    120 W

    n00 s

    External (d)

    3

    Fig.24

    ?

    Active

    After treatment

    "

    180 W

    n00 s

    External (d)

    4

    Fig.27

    ?

    Control

    Calibration

    80 W

    -

    n00 s

    External

    5

    Fig.28

    ?

    Active

    Excess heat

    "

    ca 125 W (c)

    n000 s

    Internal

    6

    Fig.27

    May 20, 2016 (b)

    Control

    Calibration

    120 W

    -

    n00 s

    External

    7

    Fig.28

    May 19, 2016 (b)

    Active

    Excess heat

    "

    ca 210 W (c)

    n000 s

    Internal

    8

    Fig.27

    ?

    Control

    Calibration

    248 W

    -

    n00 s

    External

    9

    Fig.28

    ?

    Active

    Excess heat

    "

    480 W

    n000 s

    Internal

    (a) On the graph

    (b) In the spreadsheet

    (c) Esteemed from the graph

    (d) Presumed on the basis of the time constant


    The list above shows 2 groups of tests:

    (A) – Includes 3 tests (from 1 to 3) performed at 1 nominal input power of 100 W;

    (B) – Includes 6 tests (from 4 to 9) performed at 3 nominal input powers: 80, 120, and 248 W.


    From the 120 W spreadsheets, we know that the group (B) tests were run in May 2016. It's not clear when the groups (A) tests were performed. In the JCMNS article they have been described as first, but Fig.21 reports the date of May 2017. In any case, the article describes only one type of reactor body, that is the 20 kg cruciform body, so we can assume that all the tests were carried out using 2 identical reactors – one control and one active reactor – placed side by side inside the air flow calorimeter.


    The control reactor was used to run 4 calibration tests (N. 1, 4, 6, and 8). In these tests, the power output matched the nominal input power and all the corresponding curves show a time constant of several hundred seconds.


    The remaining 5 tests (N. 2, 3, 5, 7, and 9) were performed with the so called "active" reactor, ie the reactor whose internal Ni mesh was deemed able to produce an excess heat from LENR. The reported values of the output powers measured during these last tests exhibit comparable levels of excess heat (from tens to hundreds watts) and power gains (from 30 to 100%). However the output power of the two tests of group (A) (N. 2 and 3) behave differently from the three tests of group (B) (N. 5, 7, and 9). The difference is represented by the time constants. The time constant of group (A) tests is the same of the calibration tests, that is a few hundreds of seconds, while the time constant of group (B) active tests is 10x longer.


    This difference in the time constants of the so called "excess heat" tests of the groups (A) and (B) is very significant. As finally admitted by JR (2), the longer time constant of group (B) active tests reveals that the heat source was totally internal to the reactor body. It follows that the shorter time constant of group (A) active tests reveals that the heat source was totally external to the reactor body. But the Ni mesh is internal to the reactor, therefore it may not have been the source of the alleged excess heat claimed for the group (A) tests. This fact leaves a big measurements mistake as the only plausible explanation for the differences between the input and output powers in the so called "excess heat" tests reported at figures 21 and 24 of the referenced JCMNS article.


    (1) https://www.lenr-canr.org/acrobat/MizunoTpreprintob.pdf

    (2) Mizuno Airflow Calorimetry

  • The time constant is so interesting.

    Thanks for Ascoli in highlighting this important phenomenon.

    I was wondering why the heater decay in the

    calibration/inactive reactor was astray.


    I believe that the 'time constant' is dependent on two things


    1. the thermal inertia affects due to the heat transfer kinetics (convection/radiation) and the massxCp

    ( mainly in the( 20 kg x0.5J/gC) reactor and 15? kg x1.5 J/gC of acrylic calorimeter walls..not the heater)


    2, the LENR inertia effects of the dense energy storage and dense energy transfer kinetics

    in the 23 g of nickel mesh/3mg deuterium


    I believe that the time constant difference is IMPORTANT.

    The much longer time constant with the active reactor is mainly due to the LENR inertia effects


    A check of the nonheater on time constant for the 120W( active versus inactive)

    reveals that the two ' time constants ' as measured by linear rates are similar


    Active Reactor: 0.00031C/s

    InactiveReactor:0.00028C/s


    There are provisos...including the error in some RTDs may be systematic

    and ~ 0.1 degrees (0.1 degree=1.34 Watts)..

    however this should not affect the gradients.


    The rate of heat loss can be approximated by the decay in the

    RTD temperature ,T(oR) .just after the reactor and before the blower.

    There is no actual reactor T data available.

    The 'time constant' if there exists a single one....

    can be calculated from the decay rates.. using a linear Taylor's series approx. for e-x

    since for small time intervals the value of x is small, IMHO .

    For the active reactor 33000 s appears to make a ~800s time interval small.

    This of course needs verification with further experimentation

  • There is more than one time constant in the Mizono experiment and the interactions between them are not easily solved without simulation tools. The same differential equations for heat flow are the ones used in EE for RC circuits. I have created a simulation based on this and used a SPICE simulation to create the waveforms. This is the circuit, where:

    Cr = Heat capacity of the reactor

    Cm = Heat capacity of the mesh

    Rhr = Thermal resistance from heater to reactor

    Rh = Thermal resistance of the heating element

    Rra = Thermal R from reactor to ambient (fan speed dependent)

    Rrm - Thermal R from reactor to mesh

    Voltage is analogous to temperature.

    The simulations go from 0 - 70 ms, to correspond to 0-70 Ks in the experiment.


    Waveforms below alternate between those from the Mizono spreadsheet and SPICE sim of this circuit. I am using 4 columns from the Mizono spreadsheet - Time, Air In Temp (for ambient), Air Out Temp, and Heater Temp. I used Reactor temp = Air Out Temp. That would not be quite right, but probably just a constant offset. Plots show delta Temp from ambient.



    The reactor temp rises more quickly than it falls because winding the heater around the reactor conducts the heat with thermal resistance that is low compared to convection to ambient air to cool it off later.


    The long tail of the blue simulation is from the extra heat storage in the mesh after the heater is turned off. It takes a long time to make its way out to the reactor and then to ambient, likely because much of the mesh is in the vacuum and has low thermal conductivity to the reactor body.


    I chose values to make the waveforms look as similar as possible. There is no published information on heat capacities and thermal resistances in the experiment. I could get times and max Temperature (Voltage) from the spreadsheet, but had to iterate to find other values to make things fit. Someone else might be able to derive some values based on other published data.


    The simulation does not prove or disprove the existence of excess heat. It just shows that you cannot tell either way based on the shapes of the curves. The long tail might look like heat after death, but the small value is within the noise of other measurements in the same trace.


    With the values shown, the simulations show that you may need to wait a really long time to get accurate calibration constants.


    A great experiment would be to run it once with He, then run the same reactor with D keeping other parameters the same. Then differences could easily be subtracted out to show any net excess.

  • The long tail of the blue simulation is from the extra heat storage in the mesh after the heater is turned off

    Great work Robert:)

    Four questions if I may be so bold as to ask

    1. Does the SPICE sim give a regression value saying how good the fit is on the heating phase , and on the cooling phase?

    2. Is it possible to include the thermal inertia of the insulated Perspex walls.. maybe the mass is 11000 g or more? Cw= 1.5

    3. Can the voltage on the vertical axis be made to correspond to energy content in Joules?

    4. What is the expected maximum temperature of the Nickel? if


    a)The mass of nickel is 23g Cn =0.44...the mass of the reactor is 20300g Cr=0.5

    b)The mass of nickel is 23g Cn =0.44...the mass of the reactor is doubled 40600g Cr=0.5..( this is a simplistic attempt to look at the effect of including the calorimeter wall)?

  • Great work Robert

    Four questions if I may be so bold as to ask

    1. Does the SPICE sim give a regression value saying how good the fit is on the heating phase , and on the cooling phase?


    No. SPICE does not do that directly. There are ways of varying parameters or adding comparisons to algorithmicly generated waveforms, but that would take quite a bit of work.

    2. Is it possible to include the thermal inertia of the insulated Perspex walls.. maybe the mass is 11000 g or more? Cw= 1.5


    Yes, more elements could be added easily if enough information could be gleaned from the experiment.

    3. Can the voltage on the vertical axis be made to correspond to energy content in Joules?


    Yes, that could be done. Not sure how much time I have to continue this though.

    4. What is the expected maximum temperature of the Nickel?


    The Nickel should eventually reach the same temperature as the housing if the circuit is generally right. There is a thermal path (Rrm) to heat it from the housing, and no thermal path to cool the mesh (no resistor to ground). But that assumes that the heater stays on long enough. If the mesh heat capacity is large (large Cm) and the vacuum does a good job of insulating the mesh (large Rrm). The time constant to reach equilibrium is very large.

  • Thanks Robert.

    More time is probably not worth it for just one experiment

    But your indication of at least two time constants.. one fast and one slow , is valuable.

    The fast and slow time constants definitely explain the curves much better than a single constant

    Of course adding even more constants would give closer approximations.


    My interpretation of your two modelled curves is that the blue fast one corresponds to the reactor plus walls, mass =20300+11000? g

    The orange slow one corresponds to the tiny 23 g mesh, where LENR takes place.


    Of course this would yield really high temperatures for the 23g mesh if the energy is stored as thermal energy...it might vaporise

    My contention is that the energy is not stored thermally..

    perhaps in the ~500 ev D-D form or in isotopic magnetic states.( these range from 1kEv way up over 1000keV)


    Of course other interpretations are possible.. further experiments to sort out the contributors are needed...

    eg increasing the mesh amount relative to reactor.

  • The reactor temp rises more quickly than it falls because winding the heater around the reactor conducts the heat with thermal resistance that is low compared to convection to ambient air to cool it off later.


    The long tail of the blue simulation is from the extra heat storage in the mesh after the heater is turned off. It takes a long time to make its way out to the reactor and then to ambient, likely because much of the mesh is in the vacuum and has low thermal conductivity to the reactor body.


    I chose values to make the waveforms look as similar as possible. There is no published information on heat capacities and thermal resistances in the experiment. I could get times and max Temperature (Voltage) from the spreadsheet, but had to iterate to find other values to make things fit. Someone else might be able to derive some values based on other published data.


    Robert:


    (1)


    I broadly agree with the Spice simulation. I'd point out that the total heat content in the mesh (300g Ni) compares with the total heat content in the SS reactor (20kg?) SS as:


    (Mmess / Mreact) * (SHCmesh / SHCreact) =


    0.3 / 20 * (0.44/ 0.5) = 0.0132


    Therefore:


    Rrm / Rra = 1/0.0132 => Rrm = 25/0.0132 ~ 1900 (ignoring the 10% correction from Rhr + Rh which is in parallel with RRa)


    You can adjust the relevant capacitor Cm to give whatever time constant you think is correct. The key thing is that the mesh has very little affect on the casing temperature, because it is much lower mass, so its time constant does not matter. If we reckon the mesh has some non-linear temperature dependent significant reaction, we need a much more complex model to deal with this.



    (2) Your Spice simulation is still dominated by a single time constant ( Rra * Cr). That is realistic. If you don't believe me set Rrm to infinity (remove it) and simulate again, there will be a less than 5% difference. Although the atual difference as above is even less than this.


    (3) The actual waveform is not quite the same as the exponential - but it is pretty similar. The differences can be investigated. (see also point (4))


    (4) If the external heater is cooled by the air stream directly, it will give a step (both up and down) which is much faster than the main reactor time constant, because the heater is much much smaller in mass than the reactor. There is some evidence of this from the waveform, but at an amplitude less than 20% of the main amplitude. That is a problem with this solution to the different time constants. It does not work!


    I therefore admit I was wrong before - this model does NOT explain the different time constant. However the mesh heating up / cooling much more slowly than the reactor body also does not explain it. Why not?


    (5) If the mesh is loosely coupled to the reactor thermally, with a longer time constant than the reactor body, and contributes 50% or so of the total heat getting to the reactor, we can see that it will heat itself up to a temperature higher than the reactor. In fact, it will exhibit hysteresis in which it is either on - and hotter than the reactor, or off - and colder than the reactor.


    This argument goes as follows:


    • Mesh heat capacity is 0.0132 that of reactor.
    • To get a time constant between mesh and reactor higher than the reactor time constant we need a very thermal resistance between reactor and mesh
    • In that case, if the mesh generate significant power, it will get much too hot and melt.


    We can quantify this. Suppose the reactor case temperature in the active 120W in 250W out test is 200C above ambient (the result will not depend much on what we choose for this, it could be 100C. or 400C.


    • The reactor cooling is therefore roughly 1W/C
    • To get a mesh time constant the same as the reactor case time constant (and BTW we need it to be 10X longer for this explanation to work) we need 13mW/C thermal resistance from mesh to reactor case.
    • Then the temperature rise between mesh and case for 100W from mesh is 7000C (100/ 0.132).


    These numbers are so extreme that you can see it just does not make sense for the mesh heat/cool time constant to be anything like as low as the reactor/ambient time constant.


    We need to find some other explanation for the very long time constant of the active run when compared with the control run.


    I think this is a serious problem with these results, because they don't make sense. Whatever temperature-dependent reaction we suppose in the mesh - they still don't make sense.


    Sorry that I did not consider this aspect before!


    Anyone got any ideas how to mend this?

  • Modelling the "active mesh" hypothesis in Spice.


    You would need a nonlinear negative resistance to model the mesh, where the (negative) resistance is large at 0 voltage, and decreases exponentially as the voltage is increased. It would, as the calculation above shows, lead to hysteresis, unless the mesh is very closely connected to the reactor (Rrm not larger than Rra).

  • Redux (revised)


    Ascoli's useful observations on the 2016 data (that is the 100% excess power 2017 published results that preceded the R19 and R20 results):


    • The input power is measured differently between the control and active run spreadsheets. In the control case with a mains power analyser. In the active case with figures computed from V*I. This is not evidenced on the spreadsheets where the power column is used and filled with V*I or other measurements which by inspection would come from a power analyser.
    • The resistance of the heating element for control and active runs differs by a factor of 2.
    • The dynamics of the control data are 10X faster than the active data
    • The mass of the reactor used for these results is inconsistenctly stated as 50kg old-style (from the 2019 paper) and 20kg - seemingly new-style though not called that (from the 2017 paper!). It would be good to have some confirmation of which reactor was used, and also which methodology. the old-style reactors were measured individually by the calorimeter. the new-style reactors used two reactors, control and active, together at the same time.


    On investigating the dynamics issue we find that the active data has an internal heater, The control data has an external heater. This explains the dynamics if we assume that the external heater does not make good thermal contact with the reactor and therefore heats up the reactor casing much less than the internal heater. It also explains the different resistance. There remains no explanation for the different input power measurement between active and control runs, which is unfortunate, although not in itself a problem. The exact methodology of these results (old-style or new-style) has still not been clarified; the two papers contradict each other.


    There is nothing here that necessarily invalidates these results. However, the poor methodology (very different systems used for active and control runs) and poor documentation - e.g. the difference in power measurement is not made explicit on the spreadsheets - is unfortunate and makes it more difficult to accept extraordinary results as real rather than some mistake caused by poor methodology and record-keeping.


    As one example. It would normally be obvious from reactor temperature whether control power out was more or less than active power out (at least at the 2x level reported). In this case that cannot help, because the active reactor, with external heater, would get much hotter even with the same power as the control and no mesh inside.


    It should also be possible to do something similar (compare dynamics) with the R19 results given detailed spreadsheet data.


    On the positive side: we can see that looking at the dynamics of these runs, together with the reactor case temperature, allows an independent measurement of the total (output) power, as an excellent cross-check on the output calorimetry that should be accurate +/- 20%.


    Sentence in bold is wrong. we have no explanation for the 10X slower active run, because if e.g. 9/10th of the heater power was cooled directly by air then 90% of the output air temperature change would switch quickly with the heater. We observe 10% - 20% at most does this. Unless the heater heat capacity is 1/10th of the reactor heat capacity, so that both time constants switch together and they are not distinguishable. But that is impossible, the heater is much much smaller than the reactor, I think? Maybe a metal frame round the heater massing 1kg or more could do this, and mean that the two time constants were roughly the same?

  • The input power is measured differently between the control and active run spreadsheets. In the control case with a mains power analyser. In the active case with figures computed from V*I.


    That is completely wrong. Where do you and Ascoli come up with this stuff? The spreadsheet is generated by the multichannel A/D HP gadget, and it only records V*I on fixed channels. The analyzer has its own memory and it can be dumped into a computer, but it is not part of the automatic data collection system. Perhaps it could be interfaced, but it is not in these studies. It is used to confirm the readings, as is the AC meter between the wall and the power supply.


    When you turn on the Variac to set the power, you look at the analyzer screen. It is easier to read than the computer. The two always agree.


    The analyzer was purchased for the plasma discharge experiments, which have rapidly changing input power.


    The HP gadget measures electric power on the same channels (corresponding to spreadsheet columns) with the same wires during calibrations and active runs. The wires are unplugged from one machine and plugged into the other. There are two other channels connecting to the blower power, and several other channels are connected to the pressure gauge and various other things.