Reproducing the Don Smith Effect

[tswift]

Yes, by all means anyone who has something to contribute to the discussion, please do so. My experiments are continuing but I have had no more obviously anomalous behavior in any of the configurations I have tried. Rectifying the output to DC and using a load resistance makes it easy to compute the power and obvious that the there is no power gain, or at least no measurable power gain. Even though I was getting up to around 4 watts of power going into the output transformer, after the transformer I can't get more than about 1 watt. I have tried both the 30 turn and 13 turn secondaries and power wise both seem about the same.

This is a problem I have heard mentioned by John Bedini, overunity is only obvious if the efficiency is also high enough. If you have a device with 5% efficiency and COP 10, then you will still only have 50% of the input power and cannot loop the device. So far all the simple ways that Zelina mentioned for generating cold electricity have very low efficiency. Charging a capacitor near a Tesla coil (or Slayer exciter) by means of an avramenko to ground certainly works, but very very slowly and at very low efficiency. I got out my Tesla primary and secondary and hooked them up to a Slayer circuit and built a small receiver circuit with a cap and two avramenko diodes. I started off with a 10K load resistor on it and even right at the top of the secondary I could only measure about 0.4 mW of power. It's hard to do anything useful with less than a milliwatt. I also built a small Joule thief circuit next to the cap/avramenko receiver to run the output through a transformer to see if there is any COP>1 gain. The Joule thief lights up the LED's if the arrangement is close enough to the resonating secondary but it doesn't seem like there is any obvious power gain. I have included a picture of this "COP meter" circuit, note that a wire to ground between the two avramenko diodes is necessary for operation.

[tswift]

What I have on the bench right now is the same primary and secondary configured as a mini Tesla coil using the PVM12 as a high voltage power supply. The primary capacitance is 10 nF, the resonant frequency about 770 KHz. With 4 turns in the primary, the match is pretty good and I can get 2 inch plus sparks from this rig with the PVM12 running at 12 volts and the spark gap as wide as the PVM12 can jump. Here I have it charging my weaker battery, supposedly with cold electricity. Since my earlier successful test of the self-charging seems to have had something to do with the batteries starting off conditioned on a pulse charger, I am trying to further condition them before going back to that test. The top battery in the picture is positioned close to the Tesla coil and toward the hot end of the coil to receive a large voltage swing from the E-field around the Tesla coil. The battery terminals are connected to ground through two Avramenko diodes, so each oscillation of the secondary will pump electrons from ground through the diodes and charge the battery slowly. Very slowly; as I mentioned before, this arrangement is very inefficient. But the point is to charge the battery while in a surrounding E-field, which should give the electricity going into the battery a cold/radiant/scalar component. At least this is my understanding so far of how the whole "cold electricity" phenomenon works, but obviously until there is a reproducible experiment showing this then it's just conjecture.

[tswift]

Actually, my little "COP meter" idea is proving useful. All it really does for now is demonstrate using an oscillating E-field to charge a cap using an Avramenko diode pair, something that should be familiar to every researcher here. But it's still useful for seeing the relative magnitude and spatial extent of the E-field around oscillating circuits. Surprisingly, the much higher voltage of the real spark-excited Tesla coil produces much lower power coupled into the cap than the much lower voltage of the CW Slayer exciter configuration. When using the PVM12 to charge the battery, the meter also lights up slightly, but the PVM12 is operating at roughly 1/10 the frequency of the Tesla/Slayer, so correspondingly less charge pumping is happening. The best configuration (at least as far as charging rate) appears to be the Slayer exciter.

My eventual idea is to make a real COP meter capable of measuring cold electricity. With a given input current, it should be possible to compare the power dissipated in a load resistor against the power dissipated in a load resistor after going through a 1:1 transformer stage. This should directly measure the cold component (if any) in the incoming current. The real challenge so far appears to be the very low absolute power level I'm dealing with. If there is less than a milliwatt of received power, just the energy to switch a MOSFET gate (for instance) to invert the current from DC to square wave AC to go through a transformer is already more than enough to swamp the measurement of the incident power. It would probably make the most sense to have the switching logic externally powered so as not to be a load and alter the measurement. Much more design and experimentation will be needed on this before it's actually useful.

[med.3012]

The best way to go further in the resonance induction system is to think in low level environment, how to cancel Lenz's effect ? how the electrons move and what we can do according every stage ?

i think there's a lots of thing we are still missing ...

[tswift]

Another day, another experimental setup....

This experiment goes back to test an idea from Zelina long ago, talking about using a pair of bucking coils as sender and receiver. I used the 30-turn coils I had on hand and wound another pair to match them, taking care to get the exact length. With the coils in parallel and a 100 nF cap across, the resonant frequency of each pair was 145 KHz +-1 KHz. Using the two scope channels with one measuring voltage across the primary/sending coil pair and the other measuring current through a small 10 ohm current sense resistor, it is possible to measure the input power. As can be seen from the scope shots, the voltage across the sending coils was 5.52 Vrms and the current was 0.044 Arms, with the voltage and current almost completely in phase (cos theta of 1, close enough). So the input power was approximately 0.25 watt. Measuring voltage across the sending and receiving coils at the same time reveals an interesting phenomenon: the voltage on the receiving coils is approximately twice the voltage on the sending coils. Does this represent a power gain?

What is happening here appears to be normal resonant power coupling. Using the bucking coil configuration doesn't really make it work any different than a single 1/4 wave resonator. It's still a near field phenomenon and as near as I can tell no overunity is happening in this configuration. I tried a range of values of load resistance on the receiving coils and in no case was the power even close to 0.25 watt; more like 0.05 watt. The unloaded voltage gain on the receiving coils is just a case of resonant rise in a high Q tuned circuit. The sending coils are loaded by the power being transferred to the secondary and thus has a lower Q, so less resonant rise. At first it almost looks as if there is a power gain but I was unable to find any evidence of such in this configuration. The signal source is a function generator and the tuning is quite sharp due to the high Q. Also, the coils have to be aligned and touching for maximum power transfer. As the separation increases the induced voltage in the receiving coils falls off quite quickly.

[wayne.ct]

Frequency of operation

Perhaps you can explain the scope shots? In particular, the setup independent variables and dependent variables? In pic #1, there are frequencies of 133.330 kHz and 144.5 kHz. In pic #2, the frequencies are 133.331 kHz and 144.6 kHz. In pic #3, the frequencies are 133.357 kHz and 144.7 kHz. Which is coming from the PVM12 and which is measured on the secondary? Also, are the primary and secondary coaxial or parallel? The picture shows two coils side by side.

[tswift]

Sure, I'll explain as best I can. First off, the scope's frequency detection logic isn't always 100% accurate. The only method to be sure it's measuring what you think it's measuring is to actually count divisions on the screen and do the math, or use the scope's cursor features to do it for you. It's fond of measuring harmonics on non-sinusoidal waveforms, for instance. Also, the two displays on the screen are apparently derived from two different algorithms. Since the one on the bottom line is displayed right next to the trigger settings, I assume it's the frequency of triggering which might or might not be directly correlated to the waveform. I looked at the pictures and I think you might have had a typo when writing your post: on the first one the bottom line shows 144.330, not 133.330. It's quite close to the value displayed above in the measurement block. There is also some jitter and the measurement bounces around a bit within a very small range, but it freezes the display when you save a waveform so just looking at the GIF doesn't really convey this.

Also, in this case I'm not using the PVM12 at all, this is a strictly low voltage setup being driven just by the miniscule power of the function generator itself (20V P-P at 50 Ohms output impedance). One of those comments from Zelina long ago was that "effects can also be seen using a function generator", the idea being that you can tune the whole thing at low voltages and then change to spark-excited for much higher voltages and power levels.

For the experimental setup, I take all my pictures at maximum camera resolution (10 MP) but the forum software considerably shrinks them. I understand this is necessary to save space but it means if you zoom in some of the detail will be lost. Each "coil" in the picture is actually two coils, wound on the same 1.5" PVC. Coaxial. The winding sense is opposite between the halves, so one is wound CW and the other is wound CCW. I try to make sure the length of the wire in each coil piece is identically equal, some people have claimed that wire length or wire weight makes a difference in these experiments. To achieve this, I first wind the estimated length around a slightly smaller piece of PVC (1.25" nominal). Due to the springiness of the copper wire, it doesn't sit tightly against the PVC, but when you take it off the 1.25" PVC and slide it onto the 1.5" it makes a nice snug fit and grips it like a spring. I then use a piece of extruded aluminum L-channel as a marking guide and a permanent marker to make a line down the length of the coil, so there is a stripe at every turn. Then it's simple to use wire cutters to cut off an exactly equal number of turns. Even with small manufacturing variations in the PVC tube and the wire itself, it should be possible to match the coils within about 1/8" - 1/16" (roughly 2-3 mm for our metric friends). Again, some people working on the Don Smith device have claimed that this makes a difference, so I want to reproduce it accurately.

In scope shot #1, the blue trace is the voltage across the first set of coils, the coils being driven by the function generator. The midpoint of the paired coils is grounded, the outer ends are shorted together, and there is a 100 nF capacitor across the coils. The function generator is driving the outside parts of the coils. At resonance it behaves like a 1/2 wave resonator: two 1/4 wave resonators back to back, with a voltage node in the middle. The yellow trace is the voltage across a 10 ohm resistor I have placed between the function generator black lead and the coil center leads. It's acting as a current sense shunt. Looking at the measurement block, it's all in yellow so you can tell the scope is giving you data about the yellow channel, channel 1. Since one end of the resistor is grounded (through the function generator), the end we're measuring with the scope probe will show the current through the resistor by Ohm's law: V=IR. Since R is 10 ohms, the current in amps will be 1/10 of the indicated value in volts. The scope shows 440 mV RMS (1.22V P-P), so this is 44 mA RMS.

In scope shot #2, the probes are in the same places and all I have done is switched the scope so the measurement data shows info from channel 2, which is why it's all in blue instead of yellow. It's showing the voltage at the outside ends of the paired coils. The function generator is capable of 20V P-P open circuit, so you can see it's being loaded down slightly and is only putting 15.4V P-P across the coils, which is 5.52V RMS by the scope's arithmetic. So now we have voltage and current (and we can see that they are in phase in this case), so we can compute power. Since part of the voltage drop is happening across the current sense resistor itself, this needs to be subtracted out for an accurate computation, which I didn't do the first time. So this gives (5.52V - 0.44V) * 0.044 A = 0.22W, to two decimal places. The scope is supposedly accurate to 3% of full scale on each range but I don't have a good way to check it for calibration.

In scope shot #3, I have moved the yellow probe to the outside end of the other set of paired coils, the receiving coils. These are being driven strictly by induction from being next to the driven coils. The voltage is 41.4V P-P, which is 13.8V RMS. A voltage gain of about 2.5x, which is why it's easy to think at first that there is a power gain happening. But if you attach a load resistance across the receiving coils the voltage drops greatly and if you use the resulting measured voltage to computer power across the resistor (P=V^2/R) then it becomes obvious that the power is much less than that going into the driven coils. The voltage across the driven coils (the blue trace) also changes slightly depending on the load resistance, but not by very much. Certainly not by enough to alter the conclusion that there is no power gain happening.

I spent years experimenting before I finally got a digital scope. If you don't have one, it is an indispensable tool. Now they are within the price range of most hobbyist experimenters and anyone reading this who considers themselves a serious researcher should not be without one.

[tswift]

In the last few days I have made several changes to my experimental setup. One of the main obstacles to getting the Don Smith device to work seems to be how to get enough watts out at the output to run anything useful. A 12 volt output would be convenient for running an inverter, but it's not practical to use the main resonant transformer to achieve this large a step-down ratio. If the turns on the secondary are very few then the power transfer is very poor. Most of my experimenting has been with a 158-turn primary and a 13-turn (paired) secondary, for a nominal step-down ratio of approximately 12:1. Using the ferrite toroids in the core helps to achieve a high coupling ratio between the coils but the real step down ratio is hard to determine since I don't have a scope probe that can safely read 4000V+. But to step down 4000V to 12V, a total ratio on the order of 300:1 is required, so even after the 12:1 step down, another 25:1 ratio is required. The simplest way to do this seems to be with a second transformer, and Zilano talked about this approach as well as Don had some schematics showing this configuration. Building a second air-core resonant transformer and trying to match it with the first one seems like quite a task, and best avoided altogether. Since the voltages are lower at this point, it seems reasonable to try a conventional transformer design.

So, I took a toroid core I had on hand and did a little transformer design, trying to compute a conservative volts per turn figure for the 67.5 KHz resonant frequency. For this core (a Micrometals T300-52 powdered iron core) it turned out to be approximately 1.33 volts per turn. I have used this core for a number of previous experiments and I have made a bobbin to fit it for easy toroid winding. I used 26 AWG enameled aluminum wire scavenged from a microwave oven transformer secondary to wind it, this gave right around 300 turns in a single closely spaced layer. I was measuring around 400V peaks on the waveform coming from the secondary of the resonant step-down transformer, so the spark gap is set for about 5000V. I hooked up the scope probe with a single turn secondary and measured 1.4V peaks, good enough. I was conservative with the design calculations to avoid core saturation at this frequency and voltage. I then wound a 10 turn secondary on the transformer to give approximately 14V when rectified, 13V after accounting for a 1-volt drop in the rectifying diodes.

To test the power output, I used one of my STTH1210 high speed 1000V diodes on the output, and from the diode into a cap bank to smooth the output. For this test I was not looking for any overunity effects, just trying to get enough power throughput from the test arrangement to run an inverter at the output. I used a 100 ohm load resistor across the cap bank. At 12V, a 100 ohm load is 1.44W and 0.12A. The smallest 12V inverter I have draws about 0.2A at idle, for comparison.

Here's where things get interesting. The power was absolutely minimal. Like it vanished. I was reading about 2V across the 100 ohm load resistor, which is 0.04W. Where did the power go? At the time I did the test I was confused by this extreme lack of power, especially since one of my other test modifications was to actually get a real 10KV NST and a variac. The NST (actually an OBIT, but close enough) is rated at 240VA and should have lots more grunt than the power available from the little PVM12. I also changed the primary side cap to 50 nF instead of 10 nF. At a 5 KV spark gap setting, this is 0.625J per spark, and with the sparks happening synchronously with the peaks of the AC waveform (if you set the variac carefully) then this is 60 sparks per second, representing 37.5W of input power to the primary. Even with the PVM12 I was able to get a measured 5W of power received at the secondary, at an efficiency of perhaps 30%.

This happened yesterday, so I have had time to think at length about this result. The most obvious explanation is that core saturation is happening in the toroid and hysteresis loss is soaking up a great deal of the power, heating the core. But if this were happening it would distort the AC waveform shown at the low-voltage output of the toroid secondary, and I didn't observe that. The core would also heat up noticeably with 5-10W of power being dissipated, and I should have felt the core to see if it was warm but I didn't at the time. The other thing is that the power being absorbed is large, nearly total. Almost 40W into the resonant transformer, so let's assume perhaps 10W out, coming into the toroid. From 10W to 0.04W is a 99.6% loss. Even with significant core loss, that shouldn't be possible.

One other possibility is that I actually have the Don Smith effect going, but in reverse. Don mentioned in his last document that polarity matters and "positive dies a heat death but negative runs cool". My assumption/theory is that the output transformer in the Don Smith device acts as a scalar interferometer in endothermic mode (as described by Bearden), so it should be equally possible, with a reversal of polarity of the scalar characteristic, for it to operate in exothermic mode. Instead of energy synthesis, it would be doing energy destruction. Electrical energy would appear to vanish. Heat should also be evolved in the process, but the heat would be similar whether it was just core loss or the scalar effect at work.

In any case, it is possible that I very nearly have a working Don Smith device and all that is necessary is to figure out what's backwards. Is it the diode on the NST? The winding sense of the coils? There are a number of possible combinations to work through. If it will start doing energy synthesis instead of destruction and operate endothermically, then the transformer should show a power gain instead of a power loss. This is all speculation, because I wasn't expecting to see any overunity effects at this stage anyway. The real overunity should happen because the cap bank is close to the resonant coil and picks up scalar energy with every current pulse, then the inverter on the 12V output after it should run overunity with a large power gain. The energy going into the toroid should, according to my understanding, just be conventional electricity. However, the paired step-down secondary coils themselves have inter-turn capacitance, and this capacitance could pick up some scalar characteristic from being in the displacement field around the ringing primary. If this is true it would lend credence to those whose theories of the Don Smith device hold that the power gain (or at least SOME of the power gain) happens in the L1/L2 coils. I haven't shared this view until now, but experimental verification is the final arbiter.

As usual, more experiments are needed.

[tswift]

I have spent some time following up on an idea I first posted in my discussion thread. This is essentially a souped up joule thief or pulse charger, with a step-down output section. At the beginning of this thread, I briefly had a positive result for a self-running device that recharged its own battery, but I had been zapping the heck out of the battery in a previous test by dumping my 10 nF cap bank directly through a spark gap into it. I was surprised this didn't seem to damage the battery, but it definitely put some radiant charge into it, as evidenced by the subsequent test where it self-ran at least briefly. So, since that approach definitely seems to work, it's back to zapping. It's the same mechanism at work whether it's an inductive flyback spike or an oscillating tesla coil: the battery or cap picks up a scalar copy of the energy surrounding it when the current pulse arrives. It doesn't even necessarily require high voltages, certainly many joule thief circuits produce unusual effects at moderately low voltages.

But for the best results, I think it's reasonable to assume that the higher the voltage the more the radiant effects will be. So here's the circuit under test: first, a timing section consisting of a 1/2 of a 556 timer set for about 60 Hz with a close to 50% duty cycle. Second, an inverter consisting of the other half of the 556, which is being used to give a two-phase clock signal. Third, another 555 driven by the master clock oscillator, but configured as a one-shot (monostable) to give a pulse output of about 20 microseconds. Fourth, another 555 driven from the inverter stage, configured as a one-shot to give a pulse output of about 100 microseconds.

The purpose of the timer section is to generate pulses to drive the gates of two MOSFETs, one for the step-up section and one for the step-down section. I used the MOSFETs I had at hand in the parts bin, one of them is an IRFP460 (500V 20A) and the other is a different brand spec'd for 900V and 12A. The values aren't really critical as long as the voltage rating is high enough. The first MOSFET charges the flyback inductor with current while its gate is on (20 microseconds). The flyback inductor in this case was a toroid choke scavenged from a computer power supply and measures 83 microhenries on my meter. So since V=L*(di/dt), with a 12V supply this choke will charge to 2.89A in 20 microseconds. Since P=i^2*L, this represents 0.7 mJ of energy for each pulse. When the MOSFET cuts off, this changes to a voltage spike which will charge the capacitance of the circuit. After some calculation and experimentation, I arrived at a capacitance of 235 nF since I had two 470 nF capacitors handy and I put them in series. This gives voltage spikes of about 120V peak.

The cap is charged with a fairly fast rise time by the flyback current pulse from the inductor, and indeed this can be seen by putting the scope probe in the air next to the case of the cap. This is where the radiant effect comes from; the fast-changing voltage changes the "ambient" voltage around the cap so that the entering charge takes on a radiant characteristic. So now that the charge is in there, we have to get it out again. Half a cycle of the master clock later, the other MOSFET gates on and shorts the cap to ground through the output transformer. Again, I used what was handy, it's a 3-inch powdered iron toroid with about 200 turns on the primary and 10 turns on the secondary. Obviously the turns ratio needs to be adjusted to give an approximately 12V output for the circuit to be self-running, but just for initial testing I left it as it was.

The resulting output pulse can be seen on the scope across a load resistor. I tried load resistances from 100 ohms all the way down to 1 ohm in the brief time I had to examine it, and the output did indeed seem insensitive to the load, like you would expect if Lenz's law was being canceled or at least reduced. More testing is needed, but this configuration looks promising. Even higher voltage spikes would presumably produce more radiant effect, but if I increased the on-time of the first MOSFET to give more energy in the inductor, the timing circuits started glitching from the magnitude of the voltage and current transients. I'll work on a schematic later, but at least this is the theory of operation.

[tswift]

Just an update on the previous post, I spent some more bench time tinkering with variations of this arrangement. I wound a 48-turn secondary on the big output toroid and once I did that it was unmistakably not getting a power gain or canceling Lenz's law significantly. I went back to using a battery as the load instead of a cap, since that removes the problem of getting the output back to 12V. Tinkering with the coil and timing parameters, I finally was able to get spikes up to about 120V at the battery. The rise time was very short, under 100 ns. I tried using the bigger powdered iron toroid for more inductance to give larger spikes, but I never could get any combination to give bigger spikes than about 120V at the battery. If the stored energy in the inductor is too large, it either gives the timing circuit fits and causes glitches (even with large bypassing caps at the IC's), or the ringing of the coil from the spike is enough to cause the MOSFET gate voltage to bounce back into the ON region (even though the 555 output transistors should be holding the gate OFF). Due to the Miller effect, a large dv/dt at the MOSFET drain (or transistor collector, for bipolar) can couple through the gate capacitance. This imposes a practical limit on the size of voltage spikes one can generate this way with a MOSFET. Perhaps there are some fancier circuits that can overcome this: instead of being held at 0V when off (or close, perhaps 0.1V) by the 555, adding a negative rail to the circuit and a transistor the MOSFET gate could be held to perhaps -10V, giving a lot more margin against the ringing effect gating the MOSFET back on. I'm still trying simple circuits so I didn't attempt anything this fancy.

As those spikes feeding the battery get bigger, the radiant component of the charge entering the battery should grow. When I was zapping the battery directly with the spark gap it could have been 3KV spikes. The big open question is, how much radiant component is necessary to achieve self-running? Looking at the sort-of Bedini machine variant in chapter 21 of Patrick's e-book, the spikes going into the battery in this case are certainly less than 400V, since the 13009 transistor used is only rated that much. Just on the off chance that 120V spikes cause enough radiant input, I let the battery charge for a few minutes and then hooked up a small inverter to it and attempted looping, powering my 12V bench power supply from the inverter. It ran in this configuration for some minutes, but the battery voltage was drooping the whole time and clearly it wasn't self-powering.

What I learned from this exercise are that MOSFET's may not be the best choice for these types of pulse charging circuits. I like the ease of drive circuitry and fast switching, but the large transients cause issues in the circuit. To get bigger pulses means more inductor energy being switched, and even though big bipolar transistors aren't normally known for being fast they may be a better choice for this type of circuit. Certainly the Bedini-style drive with the bifilar coil and the joule thief at low voltages are examples of simple circuits that work great. I'm essentially trying to take the machine described by Patrick in chapter 21 and replace the mechanical parts with all solid-state. Also, using a ferrite core toroid means a lot less wire for the energy storage inductor, and therefore lower expense. Big coils with lots of copper are not cheap to build.

[tswift]

Well, after a brief and unproductive detour into pulse chargers, I am back to the standard Don Smith configuration to follow up on something I observed in a previous test. I am using the Zelina step-down configuration, where the L1 coil is excited by a spark gap and the L2 coil has fewer turns for a lower voltage instead of a higher voltage. This has the distinct advantage of putting the output voltage in a much more reasonable range of several hundred volts rather than several thousand volts. My L1 coil, in combination with a 10 nF cap bank, will resonate at 67.5 KHz when spark excited (this is with the center of the coil form filled with ferrite toroids to increase the inductance, otherwise the frequency is about double). When checking the voltage on the case of the metal-case storage cap (driven by the L2 coil through a diode), I noticed that in addition to the voltage swings at the 67.5 KHz frequency, there is a very high-frequency noise burst right at the instant of the spark, before the first coil oscillation. I zoomed in on the time axis of the scope and discovered that this wasn't noise but a fairly sinusoidal oscillation at about 1.33 MHz. With the ferrites out of the core, the frequency goes up to almost exactly double this, around 2.7 MHz. A little arithmetic is enough to demonstrate that this is most likely due to the quarter-wave self resonance of the coil alone, without the capacitor. I calculated that given the diameter of my coil form (1-1/4" nominal PVC pipe) and 158 turns of wire plus a little extra for the leads, I had about 21.7 meters of wire on the L1 coil. Using the speed of light, the quarter wave resonant frequency for this would be 3.45 MHz. As Eric Dollard has shown in his research on oscillating current transformers, the longitudinal wave propagates down the transformer at a speed which depends on a number of factors related to the geometry of the coil and wire, but can be greater or lesser than the speed of light. In this case it appears to be (2.7/3.45) 78% of the speed of light.

Looking at the voltage on the scope trace, with the probe attached to the non-grounded case of the storage cap positioned close to the hot end of the coil, it's clear that this coil self-oscillation frequency due to the wire length actually drives much more voltage onto the cap than the 67.5 KHz oscillation does. Why? I assume it's because capacitive reactance is inversely proportional to frequency. A frequency on the order of 100 times higher will couple 100 times more effectively to nearby objects. Perhaps there really is something to the theory about the Don Smith device that says the wire length must be in small whole number proportions? Don said this was important, but reading the Zilano notes it didn't seem to be, just matching the LC resonance of the two coils was supposedly sufficient. If the wire length in the secondary is 1/4 that of the primary, then the self-resonant frequency of the secondary should be 4 times higher than the primary as well. Two octaves up, an overtone, which should be excited by the impulsive discharge from the spark. Complicating the matter is the fact that, as Eric Dollard has pointed out, the self-resonant frequency depends on other factors than just the wire length.

However, this is easy enough to test, right? So since my L1 coil is 21.7 meters, that means I need 5.4 meters in the L2 coil. With the length constrained, the number of turns will be determined by the diameter of the coil. I want a significant step-down ratio, so that means using a larger coil form. I had handy a piece of 4" PVC drain pipe that I have used for previous experiments. The true outside diameter is just a little over 4 inches, so I measured it and calculated about 17.5 turns to get the required length. For good measure I wound 18 so that I can cut it down a little if my numbers are off.

For this test, I omitted the cap bank altogether. Depending on how you read the Zilano notes, this is one of the possible spark gap positions. Supposedly if you get the wire lengths in the correct relationship "you won't need caps", and I think I'm beginning to understand why this is. So what I have in the test is the PVM12 zapping the L1 coil through a spark gap set fairly closely, maybe 1 mm or even slightly less. So the spark frequency is essentially just the PVM12 operating frequency, and the spark is very nearly silent. The other end of the L1 coil is grounded, and I have removed the ferrites from the core so it's just air core.

I have the scope channel 1 (yellow trace) measuring the voltage induced across the L2 coil, and the scope channel 2 (blue trace) probe in the air near the top (hot end) of the coil, not touching anything, to measure the ambient voltage.

And you know what I found? It's very interesting. First, the ring frequency of the L2 coil is not quadruple that of the L1 coil, it's double. Pretty nearly exactly double in fact, as can be seen from the two traces where the peaks and valleys occur. It's a little hard to see because the two waveforms are interacting with each other, the L1 on the L2 coil and vice versa, so some of the lower-frequency wave is superimposed on the higher one and also the opposite. But to me it looks like they are in pretty good resonance. This is with no tuning caps, just solely based on the wire length. It's also interesting in that the aspect ratio of the secondary is completely different, so there's no obvious reason why they would end up in exact resonance. Is this really one of those secrets to the Don Smith device?

Now it will be necessary to put the rest of the circuit on it, the rectifier and the storage cap, and a provision to spark an output transformer from the storage cap. I only have the one L2 coil, not a matched pair of bucking coils wound CW/CCW, but I suspect that this really isn't necessary to make the device work. Some of Don's devices only had single coils.

Just for good measure, I also tested the coil that was already wound on the coil form, which had the same diameter but only 10 turns. This is shown in the last scope picture. The L2 frequency is higher and not in resonance. The induced voltage in the L2 coil is much lower. The resonance really seems to matter.

[ewizard]

@wayne.ct - yes, the PVM12 is from amazing1.com. It's a handy little device, mainly because it's tunable. Inside it has an SG3524 chip driving two MOSFETs in a push-pull configuration into the HV transformer. Also seems pretty robust, at least my experimenting has not damaged this one so far. snip <

I've had one of thoose PVM12 for years but unfortunately it died a year or so ago when running it to a plasma ball. I've been intending to try finding the problem - not sure if it's the HV coil or circuit. So they can be damaged - I never ran it for longer than a minute or so.