George Egely's Magic Wand

    I left home for Budapest at 6 am on Thursday, got home near midnight on Friday. George and his colleagues made me very welcome, and on Thursday we spent a lot of time in the lab, where I collected a machine to test independently in my lab. George and Co ran some tests, mostly to show me how to operate it. It will take a few weeks for me to gather together the things I need to run a test here. They have built 5 of the latest iteration of the device, and others have gone to David Nagel, Guido Parchi, and Andras Kovacs.


    The photos below are from the test run in George's lab near Budapest.


    This is the general layout. The DC PSU is a fairly scary set-up, but they run the project on a shoestring basically, nobody gets a salary. The red variac feeds the secondary of a microwave transformer visible to its right. This is fitted with a diode bridge rectifier and a couple of chunky smoothing capacitors. Normally they stick to a maximum input voltage below 3kV DC, at a few mA. This reduces the rate of electrode erosion.



    A close-up of the PSU.



    This is the input waveform as created from the high-voltage DC input by the onboard LRC 'tank' circuit.



    This is the general set-up before starting the run, There is a 'spare' electrode assembly on the bench next to it. The electrode gap is adjustable, and when the machine is running you can clearly see sparking across the gap. The two analogue mechanical dial thermometers are coupled to resistors embedded in insulation directly beneath them. The one on the left is input, on the right output. The tubing is connected to the vacuum pump and hydrogen generator.



    The pressure gauge - this is the H2 pressure. (Not D2)


    More soon.


    BTW- this is a lab workhorse device, mine is slightly more elegant Just slightly.

    I have a question. I am putting together a high-voltage PSU, which will provide DC at 2-3 kV and a fe mA to the primary circuit of the Egely device. Like this:-



    I have two 7.5 KV 0.1mF capacitors. Are they (one or both, series or parallel) going to provide a reasonable amount of smoothing, or should I try something else.?

    your voltage and current specs imply an effective load of about 1megohm. With C= 0.1 microfarad, time constant is RC = 0.1 seconds, so it should provide fair smoothing at 50 or 60 Hz. Using both capacitors in parallel would help, or you could just save one for future use.

    Quote from Alan Smith

    I have two 7.5 KV 0.1mF capacitors. Are they (one or both, series or parallel) going to provide a reasonable amount of smoothing, or should I try something else.?

    It also depends on the output impedance of the HV AC generator and the level of "smoothness" that is required. But in general something around 0.1uF should be fine.

    I know George has been using some sort of series inductor for smoothing (I'm now much less worried about that, since discovering the feed to the "device" is DC and not pulsed).


    If the post capacitor ripple still looks a bit much, on the oscilloscope, maybe running a few turns of the output wire around a lump of ferrite (or soft iron) might help. Whether you'd need an extra series diode, like George also uses, I don't know.

    Quote from Frogfall

    If the post capacitor ripple still looks a bit much, on the oscilloscope, maybe running a few turns of the output wire around a lump of ferrite (or soft iron) might help. Whether you'd need an extra series diode, like George also uses, I don't know.

    As a decades ago radio ham I'm familiar with inductors, but it's a good idea so thanks for the reminder. There are some classy iron-cored toroids in the 'maybe useful' box, also dig out a radio to look for RF emissions, which I suspect will be plentiful and broadband


    I have a mid-way review of another project this week, so that's currently keeping me busy re-writing data into civil service speak, also waiting for a bunch of 470k resistors to build a resistor chain to feed the input-side scope and multimeter with something less lethal than 3kV. All good fun of course.

    Quote from Frogfall

    I know George has been using some sort of series inductor for smoothing (I'm now much less worried about that, since discovering the feed to the "device" is DC and not pulsed).


    If the post capacitor ripple still looks a bit much, on the oscilloscope, maybe running a few turns of the output wire around a lump of ferrite (or soft iron) might help. Whether you'd need an extra series diode, like George also uses, I don't know.

    There is from me no concern about the 50Hz ripple on the supply. Nor for any HF ripple.


    The issue is that any measuring device connected to this system anywhere is going to experience HF interference from the spark - as seen on the scope trace. This will be enough to induce an ac voltage on any unscreened lead or circuit.


    Now, as I understand it, input power can safely be measured via two DC quantities (the voltage is stable enough for Vdc * Idc ~ P to work). However DC measurements from a DMM should not be assumed accurate when its leads have large amounts of HF interference. They may be accurate - or there may be a DC offset introduces in the internal amplifier circuits if these pick up the HF.

    Which is why I view DC values measured from scope - even though they will be less accurate then the DMM, as more certain in this case. You can do both, and they should be the same. But if not you cannot assume the DMM is more accurate than the scope even though its spec says it should be so.


    if there is too much HF stuff on the scope you can connect the probe via a 1cm wire threaded through a ferrite bead which with the scope input capacitance X1 will knock out the HF stuff.


    Technically, a large HF amplitude ~ 5kV on the PSU output could pull more power from the PSU than the DC measurements suggest. However I don't see this as a worry because such a large HF amplitude is not feasible from RFI.

    I'd like to highlight just how important Frogfall 's thoughts are about heat dissipation on two resistors, one before, one after the reactor cell being irrelevant information about reactor cell efficiency/COP. I wouldn't go as far as stating "it is impossible to use these two values to infer any kind of cell efficiency", because there is one (and one only) scenario where it is possible in this exact circuit, even Dr. Egely was very careful to make a note on this during ICCF24 presentation:


    So Dr. Egely's calorimetry and COP statements are valid only if the capacitor have been fully charged and discharged in my opinion.


    Unfortunately I did not see data which suggests this criteria was honored. In the presentation capacitor was shown to be only partially charged and discharged:


    Expanding further on this topic, I don't believe it's even possible to apply t≥5τ criteria with the electrical setup presented, because:

    >>1. there is no timing mechanism which guarantees t≥5τ neither at charge, nor discharge, only charging current and ark discharge breakdown voltage seems to provide a timing;

    >>2. charging current (rate of voltage rise) and capacitor discharge threshold level cannot be set to match t≥5τ criteria, arguments:

    >>>>2.a. at t≥5τ charge current should be near zero, voltage on capacitor should be near exact the input voltage;
    >>>>2.b. discharge trigger level might not be constant, because it's governed by an arc discharge device with gas pressure and composition that can drift and with electrode characteristics that can drift;

    >>>>2.c. in order to make a suitable timing, input power supply voltage level should exactly match the breakdown voltage of gas discharge device at every discharge event (which is practically not likely), then one might need a pure luck at every event to force the discharge to happen, and not to wait perhaps minutes, hours, or even to infinity.

    >>3. capacitor discharge by voltage breakdown in gas (ark) usually stops before reaching 0V, in other words to keep the ionized gas in ionized state, so it can carry current down to near 0V on the capacitor, there is need to keep the excitation up of the gas, which usually is not there anymore at low voltages. But I have hopes in this regard, maybe the nuclear events provide this ionization energy.


    As arguments on why the presented calorimetry does not work if the capacitor is not fully charged and fully discharged, in other words t≥5τ criteria is not met:

    - consider the charge resistor being similar to a current monitoring shunt resistor, and the output resistor to be a load resistor. Similarities can be drawn with any other power converter using a shunt to monitor a current and a final load resistor to dump all useful power: small power is dissipated on the shunt resistor by design (why waste energy at some monitoring component) and most of the power (for ex. >95% in a converter having good efficiency) gets dissipated on the load resistor.

    Ex.: a converter of 100W drawn at input can have 1W dissipated on a shunt resistor and 95W dissipated at output load resistor (other 4W are dissipated on other internal components such as MOSFETs, tranformer/inductors, capacitors). Does this mean the converter COP is 95W/1W = 95 by looking at heat on two resistors? Over-unity of 94? Wow! Absolutely not! The converter is COP = 0.95, 95% efficient. :thumbup:

    - I've made a quick spice simulation to get some crude estimate on powers for 2 resistors in a circuit similar to Dr. Egely's. I intended to replicate parcial charge and discharge of capacitor based on the limited info I had from ICCF video.

    Charge a capacitor to almost 3kV, discharge to almost 1.5kV, with repetition rate of aprox. 2ms.

    The beauty of a spice simulation that it accurately simulates and stores many datapoints (voltages, currents) of which we can easily apply some simple numerical maths: multiply voltage (V) and current (A) on desired components obtain power (W), average instantaneous power or integrate it over time to obtain energy (J). For example on first resitor:


    LTspice simulation file attached ("CapacitorDischarge.zip/CapacitorDischarge.asc").

    Here are the results:


    The "shunt" resistor before the capacitor dissipates p_r1 ~= 0.52W

    The "load" resistor after the capacitor dissipates p_r2 ~= 1.37W

    Apparent efficiency by comparing the two powers/energies is COP ~= 2.6, which is of course falsely assumed in this case, because t≥5τ criteria is not met

    Real efficiency is by comparing input power to sum of power at both resistors: COP ~= 1 --> this is that we should strive to achieve: measure input power in DC domain reliably and integrate to obtain input energy, then place the whole unit (all converters, resistors, ark discharge devices, everything) in a heat calorimeter and measure output energy. What do you think?

    Thank you - a lot to think about there. The problems of measuring input are much discussed at the Egely lab. Attached is a picture I took of the blackboard in the lab during a discussion about this, which was (of course) in Hungarian.


    Its good that some people are running simulations ( Tibi.fusion  SM6FIE ) - but we do need to ensure that we don't introduce even more confusion.


    If I had the skills to use the graphical software, I would have a go too (I'm not an electronics engineer)


    As I mentioned further up this thread, I was more concerned about the twin resistor calorimetry technique when I thought the device was being driven by a primary "pumping" oscillator (**). But as R1, C1 & the spark gap are acting as their own oscillator, I am less concerned. After all, even if C1 only partially discharges (and videos show that the discharge level does indeed vary when the input voltage is only marginally above the spark initiation voltage) - that means that an equally partial energy pulse drives the spark.



    I'm afraid I don't know how this circuit detail is meant to model the spark gap Tibi.fusion - maybe you can explain the logic in more detail, thanks.


    (n.b. (**) It appears that some of the Moray machines used some kind of pumping oscillator to drive or trigger the "tubes" - quite possibly using crude forerunners of the Tunnel / Gunn diode (initially his Swedish Stone, and then later a semiconductor "detector" of his own devising). He once claimed that his machine would still run without the peculiar detector, but its oscillations would be very uneven.)

    Egely also mentions the importance of an inductor and a diode in series on the input-side because of "explosions" (= energy generation) in the cell.

    Maybe not that important for the simulation, but to protect the PSU.

    By the way a nice Spice.


    If we think about the COP-calculation and assume that the PSU is perfoming well (efficiency e.g. ~ 90%), then the input average energy

    to the cell could be calculated by Ec_in = Ein - ER1. Here:

    Ec_in = the actual energy send to the cell

    Ein = the measured energy to the PSU

    ER1 = the measured resistor energy with a calorimeter


    Here it doesn't matter how the capacitor C1 voltage actually varies. The only interesting thing that matters is how much energy is transferred from

    the capacitor to the cell, but this way it can be measured without a direct measurement.


    The output energy ERL is also measured with a calorimeter.

    The final COP would then be simply ERL / Ec_in
    This requires of course the input power to the PSU measurement, but that is standard measrument.

    At least a this way it should be possible to get a rough estimate of the COP.

    If the real COP would be e.g > 2, then even this method should give some interesting results.

    To look (as ever) on the bright side, even if the COP of this device is only 2, the low cost and simplicity suggests that it would be economically feasible to daisy-chain them, with the first single system feeding say 3 others, and those in turn feeding power to another 9. A bright smiley thought for Sunday :) .