Talk:Cold fusion/Skeptical argument/Were the excess heat results ever shown to be artifact?/Input Electrical Power Model

Modeling the electrical properties of a CF cell

The question asked on the page above was:

Is electrical input power directly measured with precision watt-meters, or is it computed from a (possibly simplified) model that leaves out some significant fraction of the true input power? Or are these computed measurements sufficiently accurate to not be a source of error in estimating excess heat?

See w:Wattmeter#Digital. I'm not sure there is such a thing as a "precision watt-meter," rather there are meters which do measure voltage and current simultaneously and calculate the power from that. This is a huge can of worms. I can say what's actually done.

Cell voltage and cell current are monitored, often with high accuracy. Routinely, this data is captured with some sample rate. For long runs, lower sample rates might be used, but the sample rate is sensibly high enough for sampling error to not cause misinterpretation of the data. Experimenters have been cautioned against inadequate recording of data, because a known source of excess power finding would be some anomaly in voltage or current with a prosaic cause, which would normally show up by a detailed examination of both voltage and current. Voltage and current are interesting for other reasons, they are part of how loading is measured (cathode resistance is also often measured simultaneously and continuously, because cathode resistance will vary with loading, thought that's a complex subject of its own, and cell voltage, with a constant-current supply, also varies with loading and palladium nanostructure -- which changes!).

To be absolutely sure that some complex problem with, say, AC noise, isn't causing error, some examination of the power should be done with a high-speed digital storage oscilloscope, and, 'twere it up to me, I'd do this routinely during the experiment, recording the data for later analysis. In this case, power would be calculated from high-sample rate records of current and voltage, and would not be confused by complex waveforms.

It's possible to make a wattmeter that doesn't depend on calculation, but doing this is unlikely to give sufficient accuracy. Hey, I know how to do it! Put the load inside a calorimeter.

And this is exactly what's done. The experimenters run controls. The simplest control is just a resistor, dissipating known power, but, of course, that won't produce the AC noise that is a concern. So they also run hydrogen controls (which will have bubble noise) and they run dead cells, i.e., experimental cells with palladium that, for unknown reasons, don't produce the excess heat effect. When resistors and hydrogen cells and dead cells all agree, they are pretty confident they have accurate power from their voltage and current measurements. Don't you think that's reasonable?

Note that this is exactly what Zhang found. With no excess heat, his measurements were -- no excess heat, well within calorimeter noise. He also found only a little excess heat, sometimes, not a lot, in "active" palladium, but that's a completely separate issue, he's using an idiosyncratic approach where most variations, routinely, kill the effect. --Abd 19:41, 8 January 2011

• The peak-to-peak voltage excursions depend on amplitude of the abrupt changes in the ohmic resistance caused by the random bubbling. For the very simple model that I worked out below, I assumed the ohmic resistance changed abruptly to a new value which persisted just long enough for the power supply to slew to the correct voltage. Then I let the ohmic resistance change abruptly in the opposite direction, so as to produce a perfectly symmetrical triangle wave. In other words, I treated the ohmic resistance as a square wave around the mean value, with a duty cycle just long enough to let the voltage slew do its job. This gives us a baseline view of the fraction of AC power going into a cell with ohmic perturbations yielding a 2 V peak-to-peak AC noise voltage. Obviously, the experimenters need to carefully measure the AC input power by reliable methods, since we see that it can account for as much as 5% of the total electrical power under the conditions I assumed in my sample calculation below. —Caprice 16:57, 9 January 2011 (UTC)
• My calculations below follow the assumption of instantaneous change in ohmic resistance. In the calculation that considers slewing as relevant, I likewise assumed the same: as soon as the slew reaches the set current, the resistance changes again.
• But "perfectly symmetrical triangle wave" in what? Not current! When the ohmic resistance changes, the current changes immediately, and only then slews back toward the set value. This is not a triangle wave, nor is it a sawtooth (It has abrupt transitions in both the positive and negative directions, alternating, I don't know the name for this wave pattern, it's sawtooth in alternating directions.)
• Is it voltage slew or current slew? Further, "fraction of AC power" assumes that total power is the sum of DC power and AC power. I'm not convinced. (This could be elementary AC power analysis, and, remember, I claim utterly no expertise here.) Powers do not necessarily add linearly. How about showing us?
• If the experimenters look at the current signal with an oscilloscope, while bubble noise is active, and see that the noise is only a small fraction of the voltage, they can and will neglect the effect of this noise on calculation of power input. Do come up with a 5% error, Caprice has assumed a very large noise value, so that the effect of the quadratic relationship of current to power significantly deviates from the linear relationship.
• In the end, the proof that input power is being measured accurately is found with control and "dead" experiments, where bubbling effects would be quite the same. This would be a consistent error, indeed. Too consistent! --Abd 23:08, 9 January 2011 (UTC)
• "Fraction of AC power" assumes that total power is the sum of DC power and AC power. I'm not convinced. (This could be elementary AC power analysis, and, remember, I claim utterly no expertise here.) Powers do not necessarily add linearly.
It's not elementary. You need w:Fourier Analysis (and w:Fourier's Theorem) to understand it. And you need to understand real power and reactive power. This is at least a semester or two of mathematics and electrical engineering to explain. There is no way you can learn that in these pages. —Caprice 23:40, 9 January 2011 (UTC)
Well, I'm relieved. But didn't you say, just below, that simple calculus would suffice?
Barry, you skipped over the calculation, below. Are you suggesting that we should just accept it, because you know enough to wave the magic words "fourier analysis"? I describe the real current waveform, under your stated assumptions, above. The power will not be the product of two triangle waves. But, beyond this, the critical question is the dR/R ratio. You have it as large, I think that very unlikely, and the proof of all this is in experiment. In particular, hydrogen and dead cell controls, which also would have the same bubble noise. Why would bubble noise produce RMS power, apparent excess heat, only with active cathodes? Why would only these active cathodes produce helium? How would this effect, limited in scope, produce apparent excess heat far higher than 5%? And, at the same time, escape the notice of experts working in the field? Who do have oscilloscopes, you know. --Abd 04:46, 10 January 2011 (UTC)
• If you are not satisfied with simplifying assumption that make the calculus tractable (and why should you), then you can always go back to first principles and solve for the power by Maxwell's Equations. You can also eliminate the problem of the slew rate of the Kepco power supply by using a Van de Graaf generator as your constant current source. Now, I'm quite sure the mathematical complexities of doing that are well beyond both of our pay grades, and we can both return to preposterously dimwitted flights of fancy to escape the difficulties of solving Maxwell's Equations for a Faradaic RLC circuit operating with steady-state DC current plus gas discharge riding on top of ohmic white noise up to 1 GHz. —Caprice 11:20, 10 January 2011 (UTC)

AC power into a sinusoidally varying resistance

Another way to appreciate AC power is to take a very simple example. Assume a perfect constant voltage source, V, working into a sinusoidally varying resistor, R + r sin ωt, where r << R. Let α = r/R. You can integrate the power over one cycle of the sinusoid to get PAC = ½fα²PDC, where PDC = V²/R, and f = ω/2π is the frequency of the sinusoid in Hertz. Even for α << 1, you can get substantial AC power if the frequency of the sinusoidally varying resistor is high enough. When f = 2/α², there is as much AC power as DC power. So even very small sinusoidal fluctuations in R can draw significant AC power, if the fluctuations are occurring at a high enough frequency. —Caprice 15:34, 10 January 2011 (UTC)

• Correction. ScienceApologist spotted an error in my derivation. There should not be the frequency factor, f, in the model. It's just PAC = ½α²PDC, independent of frequency. —Caprice 20:38, 10 January 2011 (UTC)
Mmmm... From the equations, as r/R approaches zero, i.e., no resistance noise, the point at which "there is as much AC power as DC power," goes rapidly to infinity. The power itself is shown as going to zero, as the resistance ratio approaches zero, by the square of the ratio. Yes, "high enough frequency," if the equations are correct, is important. In real life, noise power at very high frequencies can be neglected except under very special situations. Note that very fast resistance variations would be absorbed (the current supplied by) parallel capacitance, they would not be seen at the power supply at all. (I think circuit inductance has a similar effect, the current is maintained without additional power supply current.) The power supply also has very large capacitance on the output, but it gets complicated in constant current mode. In other words, if R declines in a step dR, of very short duration, no change in voltage and current from the supply are seen, it's entirely handled by local capacitance, perhaps from the bubble capacitance you asserted elsewhere. There is no effect on input power. --Abd 18:49, 10 January 2011 (UTC)
• Above some frequency, f ≈ 1/RC, the capacitive impedance goes to zero, so the above formula for AC power applies to frequencies below 1/RC, where C is the capacitance looking into the cell. Above that frequency of perturbation, the cell capacitance dominates the input impedance of the cell. Caprice 19:12, 10 January 2011 (UTC)
• Again, there is no frequency factor, f, in the corrected formula, so the above comment doesn't apply. PAC = ½α²PDC, independent of the frequency of perturbations in a sinusoidally varying resistive load. —Caprice 20:38, 10 January 2011 (UTC)
• Wow! Edit summary was: AC power into a sinusoidally varying resistance: Correction. ScienceApologist spotted an error in my derivation. There should not be the frequency factor, f, in the model. It's just PAC = ½α²PDC.)
• You did what I tried and failed to do. I invited SA to participate here, he rejected it as a waste of his time. So now he's actually helped. Way cool! My intuition was correct, that there was something quite fishy about the formula and the results you were deriving from it. Much better now, thanks for correcting it. We get by with a little help from our friends, eh? --Abd 21:21, 10 January 2011 (UTC)
• I gave him something useful to do that wasn't a total waste of his time. I gave him a precise mathematical model to review. I asked him to check my derivation. He did that without hesitation, and he immediately found the algebraic error, where I had plugged in 2π radians for the upper limit of the integral, rather than the period, T, in seconds. (That's why I ended up with watts/sec instead of watts as my result.) He carried out the step that I had skipped, which is to routinely follow up with a dimensional analysis check to make sure the mathematical results at the bottom have the correct dimensions. Unlike me, SA is not going to waste his time staring at a blizzard of words in your patented process of textual wallucinations. —Caprice 15:01, 11 January 2011 (UTC)
• He wouldn't have to read that "blizzard of words." He could, for example, look at Cold fusion, and maybe at Cold fusion/Skeptical arguments, and make sure that the pages are complete and accurate. He could add additional arguments, if he has some relevant ones. He could look at Cold fusion/Excess heat correlated with helium/Sources and see if any important sources are missing. He could start his own seminar either on Cold fusion or on some other topic in physics, astrophysics, or whatever. Sure, he's impatient with discussion, because he believes he already knows and understands this subject. He might, however, have something to learn. After all, he's not really studied this field. --Abd 16:20, 11 January 2011 (UTC)
• Right. But the crucial information is level of AC noise at the cell input, and, in particular, the integrated product of cell voltage and current. Trivial to measure, and it would be, as you've stated elsewhere, somewhat similar from cell to cell, under similar conditions of gas evolution. I'd expect hydrogen noise to be smaller than deuterium noise, if I understand the mechanism of the noise formation correctly. But not radically different. Smaller bubbles would be formed, and my first expectation is that the bubbles would be roughly half the volume, which is still 3/4 the diameter, so the effect of each bubble would be almost the same. I'd expect tapping the cell to produce a rise in input power for a moment, until the bubbles were replaced. This could be slow enough to be captured by the power measurement, otherwise it could produce a burst of "excess heat" Such bursts are sometimes seen, but they don't necessarily make a significant contribution to overall excess power.
• Nevertheless, Barry, do you think we should ask for the records from the seismometer that they, of course, if they were competent researchers, trying like heck to refute the null hypothesis, they'd have had running attached to the cell? Why didn't they publish this data? What are they trying to cover up? And, remember, we need that dietary information about the researchers, to rule out telekinetic power, as well as complete records from their psychotherapists, and, hey, have they purchased palladium futures? You can never be too careful. --Abd 20:00, 10 January 2011 (UTC)

Consideration of power supply AC characteristics

All the CF cells used by researchers doing precision calorimetry operate with a constant-current DC power supply.

The advantage of using a constant current DC power supply is that the experimenter can control the DC drive current, which is the faradaic current that is either charging the electrodes or dissociating the electrolyte.

In all the models for input electrical power that I've looked at, the electrical input power is modeled as pure DC power, with no AC component. That is, the constant-current DC power supply is treated as an ideal constant current source, with no AC perturbations around the constant DC current.

However, all real constant current power supplies have a characteristic slew rate which specifies how fast they respond when the load impedance changes abruptly. The Kepco BOP 20-20M 400-watt power supply used by McKubre has a slew rate of 1.25 A/μsec in the constant-current mode and 1.0 V/μsec when operated in the constant-voltage mode.

The slew can be modeled as a linear ramp whenever the resistive load changes abruptly from one value to another. Instead of an ideal square wave (instantaneous adjustment), the rise and fall of the square wave is really a ramp, with slope given by the slew rate.

One can model AC power by using sawtooth or triangle waveforms, which are easy enough to integrate with simple calculus. Depending on the fluctuations in the resistive load, there will be corresponding fluctuations in the voltage which ramp up and down at the slew rate to maintain constant current. One can get an idea of how much AC power is going into the resistive load by computing the AC power of a simple triangle wave with a peak-to-peak voltage.

I did this for two examples, corresponding to a pair of experiments in McKubre's EPRI paper. McKubre doesn't say what the peak-to-peak voltage excursions are when his cells are bubbling, so I assumed a 2 V peak-to-peak around his reported DC values, to make the math easy to do. For the case of 2 A going into a nominal 2.5 Ω resistive load at 5 V for a DC power of 10 W, a 2-V peak-to-peak triangle wave riding on top of the DC comes out as 0.5 W of AC, for a total of 10.5 W electrical power. For the case of 7 A going into 6/7 Ω resistive load at 6 V for a DC power of 42 W, a 2-V peak-to-peak triangle wave riding on top of the DC comes out as 1.8 W of AC power, for a total of 43.8 W.

Thus a voltage excursion of the order of magnitude of 2 V peak-to-peak works out to about 4-5% of the electrical power being AC at a frequency related to the slew rate of the power supply.

The CBS News film crew that accompanied correspondent Scott Pelley to McKubre's lab at SRI may have missed an opportunity to measure the peak-to-peak AC (audio) power going into McKubre's cell. All they had to do was slap one of their audio VU meters across the terminals of McKubre's cells to see if there was any AC (audio) power that McKubre was leaving out of his calculations. comment added to resource page by Caprice, 16:10, 9 January 2011

I don't believe it's true that all CF researchers use constant-current power supplies. Note that Earth Tech, attempting to replicate Zhang, used a constant-voltage power supply. The idea that the Earth Tech replication failure was due to this is interesting. Little doesn't seem to have considered the possibility. The Zhang approach would have vigorous bubbling compared to other approaches, so there would be substantially higher bubble noise. However, the figure of 2 V p-p noise is very high compared to what I'd expect. Bubbles will increase the resistance of a section of electrolyte, the effect of a bubble on current at the surface would be reduced because the bubble will not prevent current from "reaching" the section "underneath" the bubble, it merely lengthens the path, except for actual points of contact with the surface, and the surface itself is really a complex layer that will be partially conductive. (I think! the microstructure of that layer is not necessarily known, and may vary considerably from experiment to experiment.)
Note that gas-loaded cells use no power supply at all.
It's not true that the power source is modeled as pure DC. Rather, McKubre considers the noise, but claims that it will average out. Whether or not it does will depend on the exact characteristics of the noise, it's not enough to say that AC noise voltage exists. Some AC noise would average out, some would produce a change in power, it depends on details that Caprice has not considered. The problem is that power is proportional to the square of the current, so a positive dI existing for the same time as a negative dI of the same value will result in increased power over the average value of the two.
I am not experienced with power considerations in circuits like this. Almost all my work with power supplies has been constant-voltage, where the supply is designed for very good regulation of voltage, and noise is heavily damped by large capacitance on the output. With a constant-voltage supply, resistance noise will affect the current, linearly, and thus the power, linearly. Average power will equal average current times the fixed voltage.
So let's look at what happens with a constant current supply when cell resistance increases by dR. The claim (by McKubre) is that resistance noise is random, so positive dR will be equal to negative dR, averaged out. Let's assume this circuit: the resistance of the cell is switched between R-dR and R+dR, rapidly. R is the average resistance. Caprice has claimed (elsewhere) that the slew rate is irrelevant, but that's can't be true, because if the slew rate is zero, or very low with respect to the switching rate, so that by the time the supply starts to react to decreased current from +dR, resistance has switched in the opposite direction. A possible effect here must require a high slew rate, high enough to produce a significant effect before the resistance changes again. For a first pass at this, I will look at infinite slew rate, i.e., the supply adjusts immediately, and resistance noise produces no current noise, only voltage noise.
• Vs is the steady state voltage if resistance were R.
• Is is the set current. With infinite slew, instantaneous current is equal to Is at all times.
• Ps is the steady state power if resistance were R. Vs is the steady state voltage.
• When R goes to R+dR, voltage goes to Is*(R+dR). Power goes to Is^2(R+dR).
• When R goes to R-dR, voltage goes to Is*(R-dR). Power goes to Is^2(R-dR).
• If the time of -dR is equal to the time of +dR, then average power is Is^2*R. R can be inferred from the average voltage.
• Resistance noise doesn't produce a net effect on average power, and a calorimeter will see only average power.
• There is AC power created by the resistance noise. However, this power cannot simply be summed with the average DC power to create a larger average power. Rather, the AC power is a component of the DC power.
Caprice was not explicit as to how he calculated the total power. In addition, there is a problem with slew rate: slew rate specifications are minimum values for the range of operating conditions. Actual supply design may slew at a maximum voltage rate, and the current rate is a consequence of this. As a result, a ramp may not be linear. But let's assume constant current slew rate, with Caprice.
So, we have a preliminary conclusion here: if there is an effect on average power, slew rate and other specific response characteristics must be involved in the calculation, because there is no effect with zero slew rate and no effect with infinite slew rate.
So now let's look at slew. Let's assume that the slew rate causes a triangular variation in current, rather than the square wave that the instantaneous slew I assumed for the above would produce. Let's assume that, miraculously, the slew hits the set current just as the resistance flips. Otherwise the current variation gets truncated, the set current will not be reached; but will suddenly flip with the next resistance change.
With this, the instantaneous effect of +dR is to reduce current, reducing power. If we assume a pure ramp, current will then slew positively to the set current, which will then be at an increased power.
• Before the positive change in resistance, resistance was at R - dR. Current was at the set value Is. Power was at Is^2*(R-dR). Voltage was at Is*(R-dR).
• When the resistance flips to R+dR, current drops to Is*(R-dR)/(R+dR). Power is then Is^2*(R-dR))^2*/(R+dR)).
• Voltage then ramps to the value needed to restore Is. Voltage will then be Is*(R+dR). The voltage swing for this transition will be Is*(R+dR) - Is*(R-dr) or Is*(2*dR).
• Power increases to Is*^2*(R+dR). This is not a linear ramp. The AC component of the power is not a sawtooth.
• When the resistance flips to R-dR, instantaneous current increases to Is*(R+dR)/(R-dR). Power is then Is*^2(R+dR)^2/(R-dr).
• Current then slews down to Is to restore the initial condition.
• Voltage swings from Is(R-dR) to Is(R+dR), with a sawtooth waveform.
• Current swings from Is^2(R-dR)^2/(R+dR) to Is^2(R+dR/(R-dR). This is also a sawtooth.
• Power swings from Is^2(R-dR)^2/(R+dR) to Is^2(R+dR)^2/(R-dR). This is not a linear sawtooth at all.
• If R>>dR, however, which should be the case, the power swings approach Is^2*R(1 +/- dR). Average power will be Is*2*R.
• Substantial deviation from this requires that dR be substantial compared to R.
Caprice assumes a huge variation, inconsistent with the experimental variations. And even with this, he gets a difference between the steady state assumption and his calculated power of only 5%. Absolutely, for some experiments, a 5% error in the estimation of input power would be enough to explain the reported excess power. However, 5% is at the lower end of what is considered significant power in recent work, and the 5% calculation requires a ridiculously large variation in cathode resistance caused by bubble noise. I'd be surprised if the noise was as high as a tenth of that. Realize that for the noise to be large, there would have to be few bubbles taking up a lot of area, but bubble size is drastically limited, and bubble persistence (and thus size) is limited strongly by circulation in the cells.
What is the average power without the simplifying assumption of R >> dR? I don't have time to do the math, I've already spent too much time today on this. The power waveform is the product of the voltage and current sawtooth waveforms, it should be integrated to find true average power, which would then be expressed in terms of the set current, R, and dR. --Abd 18:33, 9 January 2011 (UTC)
• Resistance noise doesn't produce a net effect on average power, and a calorimeter will see only average power.
This conclusion is incorrect, but I don't know how to explain to you why it's wrong. Even if you assume an ideal constant current source, resistance noise will still inject an AC power transient. To do the math correctly, you either have to go back to first principles, or you have to take the limit as the duration of the transient goes to zero. Taking the limit is easy if you work from the linear ramp model, since the transient energy is the integral of the transient (AC) power, integrated over the duration of the transient. If you take the limit of this integral as the duration of the transient goes to zero, you get the AC (static) noise power, which does not vanish, even in the limit. An easier way to do this is to use Dirac δ-functions for the time derivative of R(t). It's a bit of a cheat, with respect to doing the calculus the hard way, but Dirac's short-cut method works OK here. You'll get the energy pulse for a single step function in R(t), and then you can multiply that by how many instantaneous step-function perturbations in R(t) occur per second, to get back to AC noise power in watts. Compare to the Johnson-Nyquist model for deriving the white noise of a resistive medium. No matter what the underlying mechanism is for generating static, shot noise, or white noise, there is a corresponding amount of noise power to reckon. —Caprice 13:39, 10 January 2011 (UTC)
• The statement is taken out of context, which was a particular assumption, a perfect constant current supply, that adjusts instantly. Resistance noise then introduces no current noise, only voltage noise, and the voltage noise varies directly with the resistance noise, as does the power. There is noise power, all right, it was not claimed that there was not. But no longer does the power vary according to a square law, but directly. Average power then equals average voltage times the set current.
• That there is noise power becomes irrelevant. The frequency of the noise becomes irrelevant. (High frequency noise, of some kind, might possibly affect the voltage measurements, if it interacted with the sampling frequency in some way. But random noise would not do this.)
• Pedagogically, introducing higher mathematics by reference here is incomplete. Perhaps we should define the cold fusion resources as accessible to people without higher math, just as, say, Feynman's little book on quantum electrodynamics is so accessible. If necessary, we can cover higher math on subpages, when needed. It is not needed here, not yet, anyway. And I'd want it reviewed by experts. Indeed, I want all of our work here to be so reviewed, eventually, if they will be so kind. Which means that my mistakes will be totally exposed, eh? To my mind, that would be spectacular! There is no way to learn so quickly as by making mistakes. If one is willing.
• Technically, this page, from its name, is about the input power model, and thus consideration of these details is appropriate, but the context is "skeptical argument," so the ultimate goal is to understand skeptical arguments. And, there, more than the model is involved. Various models predict differing results, and, as it happens, the research data is, so far, quite able to distinguish between them. I.e., no matter how much math Barry does, his model will predict certain results from an assumption of such-and-such noise power of such-and-such a kind. And if those results are not seen, but the contrary, the math is quite likely incorrect, or the model is incorrect, even if we don't understand where or why. Basic science, Barry. Don't leave home without it. One of the things that basic science allows us to do is see through the noise generated by pseudo-experts or even real ones who've overlooked something.
• Now this idea that there can be DC power of X and AC power of Y, and that the total power is not X+Y is counter-intuitive, I'd say. The resolution in this case is that X includes Y, Y is a component of X. Basically, if we consider true DC power of A, and Y is the average value of the AC power, then X = A + Y. This is only true under certain specific conditions, which obtain when the current is constant.
• Normally, voltage and current in AC are varying together. Constant current simplifies this. Only the voltage varies.
• As noted, however, a real constant current supply responds imperfectly, with a limited slew rate. For simplicity, we are assuming fixed current slew.
• Discussion of noise gets confused, again, when it is asserted that noise will be up to the GHz region. Yes, perhaps, but what is the source of the noise power? The only source of power of significance here, aside from anomalous heat, is the power supply. In the case of bubble noise, Barry purports to have shown that there is AC power that does actually increase total power above average power, as determined from instantaneous voltage and current measurements, and I'm tentatively accepting that without having verified his calculations in detail. However, the source of this increased power is the work that the power supply does maintaining current. The noise introduced by this is frequency-limited, as to any significant components, by the slew rate.
• Unfortunately, McKubre describes his measurement protocol this way:
The cell and heater currents were each measured as a voltage dropped across a calibrated, series resistor. Voltages were measured using a Keithley 195A 5-1/2 digit digital multimeter with 0.01% DC volt accuracy and 0.015% resistance accuracy. Resolution was 1 ppm (Ω) and 10 ppm (DCV). Each 5-1/2 digit measurement was averaged 32 times before being recorded. Resistance standards were calibrated periodically against NIST traceable standards, using NIST traceable calibration instruments yielding an accuracy of ~ 0.1%.
• It looks from this like the voltage and implied current measurements were averaged before being recorded, there is no mention of first multiplying them together, and the samples would be taken at different times, not together. I'd have to assume from this that instantaneous power was not determined from the individual measurements, then averaged, which would address the noise problem, as long as the acquisition time for each measurement was short enough, and both values were simultaneously acquired. (Probably not possible with that meter, and not even with two of them, unless the sampling is synchronized. Electronically, that's quite possible, but practically, not so easy. I think my DSO would do it, it's dual-channel, but I'd need to look at details. If it alternates sampling, using a single high-speed ADC, to save cost, not so great. Two are needed for best power accuracy. However, at 1 GS, and low slew for the power supply, which is the case, 500 MS, alternating, is probably close enough. The averaging ("32 times") I'd read as averaging the voltage measurements. So McKubre's approach would not see certain kinds of noise power, indeed. However, I still think it highly likely that he already knew that significant noise, of the kind that would cause a problem, was absent, and the control experiments demonstrate that he was correct about this.
• I'm not sure there is much more value in beating this dead horse; but possibly noise power is a factor in some approaches, we should keep it in mind. --Abd 19:04, 10 January 2011 (UTC)
• McKubre considers the noise, but claims that it will average out.
Yes, he claims that, based on the assumption that he has an ideal (perfect) constant current power supply. But he doesn't. They don't exist. That's just a convenient simplifying model. Most of the time that simplification is fine. But in this case it's a fatally flawed assumption. In fact, the faster the slew rate of the constant-current power supply, the more AC power it will inject. And McKubre is using one of the Kepco models with the fastest slew rate in their line. This is an aspect of their input power model that they simply don't address in the EPRI report. I stared at that paragraph for 30 minutes, trying to understand why McKubre could assume that multiplying average voltage by constant current was a sensible model. It was that word, "sensible" that had me puzzled. First of all, it didn't make sense to me, and second of all, he wasn't measuring ("sensing") the instantaneous AC power at the slew rate of his power supply. Note that the linear ramp model applies to the voltage slew and current slew. Their instantaneous product is quadratic which is where the overlooked AC power comes from. —Caprice 19:04, 9 January 2011 (UTC)
• Caprice assumes a huge variation, inconsistent with the experimental variations.
In the first example, (2A, 5V, 10W), the mean resistance is 2.5 Ω and switches 0.5 Ω up and down, from 2 Ω to 3 Ω and back. In the second example (7A, 6V, 42 W), the mean resistance is 6/7 Ω and switches 1/7 Ω up and down, from 5/7 Ω to 1 Ω and back. That's a 16-20% variation in the ohmic resistance. —Caprice 19:16, 9 January 2011 (UTC)
• Caprice has claimed (elsewhere) that the slew rate is irrelevant, but that's can't be true, because if the slew rate is zero, or very low with respect to the switching rate, so that by the time the supply starts to react to decreased current from +dR, resistance has switched in the opposite direction.
If the slew rate is zero, then it's not a constant current power supply. If the slew rate is slow, then it's an average current power supply. If you multiply average current by average voltage, you still get (average) DC power, and you are sill leaving out the AC power from the (now slower) swings about the average values. —Caprice 19:35, 9 January 2011 (UTC)
So silly. The point is that slew rate must affect the nature of an error here. "Zero" means "not fast enough to respond within the period involved, and a constant-current supply would still be a constant current supply if it, say, responded within a second, for example. This is a standard heuristic practice, to look at extremes to understand how factors affect results. What I pointed out is that the claim is preposterous, so then Caprice quibbles about the name of the supply. Suppose the supply had a dial labelled "slew rate." You can turn it all the way from some high value that means very rapid response, down to a low value that means it will take seconds or minutes to adjust. This is a "constant current supply," the practical value of such a supply is that one sets current under operating conditions, not voltage. And does the error depend on that slew rate setting? Obviously it must, if there is any error at all, because at the extreme settings, there is no error! Yet Caprice's calculation obviously does not consider slew rate, so, my conclusion: it's bogus.
That doesn't mean that he's wrong. We'll have to look closer. I'm not seeing clean calculations with all assumptions being explicit.
The error Caprice is alleging doesn't come from voltage noise alone. That, alone, would produce no error, even though there is certainly AC power. The voltage noise is produced by the power supply adjusting voltage to compensate for changed resistance. The extreme is a perfect power supply which maintains constant current, absolutely. Infinite slew rate. There is no current noise. The power noise and voltage noise are identical, because power is then simply voltage (varying) times a constant current, and averaging works just fine.
In the other direction, where the slew rate is slow compared to the noise frequency, the voltage doesn't vary at all (except long-term, with long-term shifts in resistance, not bubble noise). Bubble noise, again, produces only current noise, not voltage noise. The product is, again, linear with resistance change, there is no quadratic effect.
It is only with some intermediate value of slew rate that we see simultaneous changes in voltage and current, and the current waveform, in particular, is not a sawtooth as Barry stated, because the current immediately shifts with a change in resistance. Only if the resistance changes as a sawtooth would current change similarly, under some conditions. Power is reduced for part of the cycle, and increased for the next part. It's obvious that the affect on average power depends on noise frequency as it relates to slew rate. Caprice "calculated" the power, but did not state how he did it. It requires integration of the function I gave, if I got that function right.
The effect Caprice is asserting would affect hydrogen controls and dead cells quite the same as active deuterium cells. It would consistently produce this ("rms power" effect), as long as bubbling was similar. The cells McKubre was working with simply did not behave like that. This is entirely aside from the rather obvious likelihood that McKubre actually looked at the signals with a scope. If he saw a 2V peak-to-peak signal riding on top of a 5 V average level, he'd be very concerned. Wouldn't you?
Yeah, I'd be happier with some statement of the level of noise involved. But he's an electrochemist, and, my guess, he had lots of experience with electrolytic cells long before CF came on the scene. He'd know the voltage behavior. And so would any other electrochemist, he might not even think to state it.
Perhaps to encourage you to start looking at helium, I'll point to Miles (1993) mentions bubble noise, page 12 of the PDF. He considers that this and other problems limit his calorimetric accuracy to +/- 40 mW or 2%. This was quite early work. --Abd 22:49, 9 January 2011 (UTC)
Miles writes, "The major errors in our calorimetric measurements are probably fluctuations in the room temperature and fluctuations in the cell voltage due to gas bubble effects. These error sources limit our accuracy to about +0.04 W or +2%." That's all he says. He doesn't say what he means by "gas bubble effects." Does he give an analysis or model explaining how he reckoned the contributions from "fluctuations in the cell voltage due to gas bubble effects?" Caprice 20:52, 10 January 2011 (UTC)
Let's see if we can accumulate a few more Totally Stupid Questions before I ask him, okay? Isn't it obvious what he means, in round terms? Miles surely knows the voltage from gas bubbles under his conditions -- and the effect of bubbles would vary greatly with exact experimental configuration, I'd assume -- and he therefore limits his accuracy to what he states. And it doesn't matter, because he's hunting for bigger game, not proof of excess heat, but correlation between excess heat and helium. For that purpose, and given how loose his helium data is, in those early experiments (it's simply given as powers of ten), niggling details like absolute accuracy of calorimetry aren't as important as in other work. You want to criticize calorimetry in relation to helium, you will need to look at the later, more careful work, where helium is measured to -- hold your breath! -- one or two-place accuracy, maybe a little more in some cases. The good news is that helium is correlated only to excess heat, not to absolute power. A ten percent error in excess heat (and helium!) might be a one percent error in absolute power....
Seriously, the question isn't totally stupid, and it might be worth asking (particularly to know the level of noise observed). I do want to accumulate a few of these, though, before standing up to be seen with them. --Abd 21:32, 10 January 2011 (UTC)
• The question is "How does he model the contribution from the bubbling?" He gives a number, for some particular cell, operated over some range of drive conditions, but that doesn't tell me how he modeled it. Where does he give the model or method of analysis? —Caprice 21:41, 10 January 2011 (UTC)
• It's a stupid question, on the other hand, because it's moot, as far as his purposes are concerned. I don't see that he gives the model, and I don't see that anyone asked him. But maybe someone did, somewhere. I'd assume that he saw some noise. And he's aware that room temperature could affect his calorimetry by shifting calibration. So instead of nailing it down and eliminating it, and because it wasn't enough to torpedo his purpose, according to his extremely careful calculations that he wrote on a napkin at a local cafe, or in the cafeteria, remembering only the result, he just made the statement about errors due to room temperature and bubble noise. Experimentally efficient. Real world science, and real-world publication deficiencies, I see it all the time. But the question can be asked. --Abd 22:34, 10 January 2011 (UTC)
• The error from random fluctuations in ohmic load appear to be irresolvable, save for coming up with an upper bound for them. My calculation is one way to find an upper bound. From this discussion thread, it appears that accurately measuring the noise power is beyond the scope of measurement technology, because the frequency of the noise spectrum can extend up to 1 GHz. That means it has to be modeled statistically to get the best upper bound. I have no idea how that might be done. My upper bound is a very simple model that doesn't depend on the slew rate, only on the peak-to-peak voltage excursions. But I'd like to know what would be seen with an audio frequency VU meter across the terminals of the cell. —Caprice 23:11, 9 January 2011 (UTC)
• Therefore, Barry, as McKubre hinted, it is impossibly complex to rule out all error sources. Except with independent confirmations. Yeah, I'd like to see that VU meter reading, but that, too, would have limitations, it's only designed, probably, to handle noise within the audio region, not much beyond it. There are simpler ways. And, guess what? McKubre may be using one of them. He samples the voltage and current waveforms. If his meter is fast enough, and if the noise is random, as he notes, and if he does not average the voltage and current, but instead they are simultaneously sampled and multiplied together before being averaged, there you go. The noise power will be captured. This, in fact, can be done with most DSOs nowadays, and the sample rate could be 1 GS/second easily, my little Rigol scope can do it. (Ah, what I'd have given for that scope twenty years ago! And it only cost about \$400.)
• But, even better than sheer force in capturing every last smidgen of possible error is independent confirmation. Basically, control experiments. This is part of the design of the Scientific Method. Let us pray. Scientific Method. Scientific Method. Scientific Method. Om!
• There are two forms of control that cover the noise problem. The first is using the same setup and the same current with hydrogen controls. In fact, they put the hydrogen cell, in the McKubre work we were looking at, in series with the deuterium cell. Experimentally, there is only one variable: hydrogen vs. deuterium. While there will be differences between hydrogen and deuterium, perhaps due to more bouyant bubbles. which would lead to smaller bubbles, the difference should be quantitative, not qualitative, i.e., if there is an effect in hiding power in the bubble noise and the power supply response to it, it should show up with hydrogen as well as deuterium. It doesn't, i.e., these experiments are not showing excess heat (in McKubre's work, and don't be confused by other work showing some level of heat with hydrogen. That is generally at far lower levels, requiring extremely sensitive calorimetry, which is difficult, obviously.)
• The second control is handed us, fortuitously, by the striking variability of palladium behavior. Some samples of palladium do not show excess heat, under the same conditions, apparently, as other samples. (It's likely that this is related to nanostructure of the palladium, part of this is understood, but not all.)
• So a set of experiments of the type we've been looking at, we now realize, will show variable behavior. The levels of excess heat found cannot be predicted, other than statistically. While this was extraordinarily frustrating, and understandably led to a suspicion of confirmation bias, "dead cells" hand us a handy control, where we may have set up conditions where the only variable is this unknown condition of the palladium, and thus we can study something like prosaic heat aside from excess heat.
• In one of the experiments we looked at, the pair was, I think, P13 and P14. One was hydrogen and one was deuterium. In addition, the experiment consisted of the palladium being loaded and maintained at high loading for long periods. Then current excursions were applied at various times. In both cells, then, there would be bubbling and thus the noise in question. At a certain current density, sometimes, but not always, excess heat would appear in the deuterium cell. Never in the hydrogen cell, and quite often, even usually, not in the deuterium cell either. Same sample of palladium. Sometimes heat and sometimes not. The source is an "unknown reaction." Now, what are the characteristics of this?
• If the source of excess heat were RMS noise in the power supply, due to the supply working to compensate for bubble noise, this would show up in both cells. And it would also show up in each current excursion. So the RMS noise hypothesis is falsified by the data. Isn't that how it is supposed to work?
• The dead cells and hydrogen controls and the inactive periods confirm that the calorimetry is accurate, even with regard to a whole universe of unexamined assumptions. This is the power of control experiments!
• This leaves one major possible prosaic source for excess heat; if there is unexpected deuterium-oxygen recombination, unexpected heat would appear in these current excursions. This is a quite separate issue and will need study in detail. This is where Shanahan set up his tent; complicating it by also asserting a calibration constant shift. (This is like a lawyer who says, "My client has an alibi, he was elsewhere, but he was justified in shooting the fellow." It's legally acceptable, and there could indeed be more than one source of error, combining.)
• Basically, while it's been shown that power supply/bubble noise could be a factor, the independent confirmations that zero excess heat is normally observed, show that the noise issue isn't a real one. Understanding this better probably requires more information than is provided in the publications, as I've frequently pointed out.
• I'm collecting questions to ask the experts; I'll eventually ask them on the CMNS mailing list, instead of bothering individual researchers. Storms reads that list and has often replied to me there, or off list. (He invited me to join it, it's only by invitation. He doesn't manage the list, but he does have some credibility in the field, eh?) --Abd 15:09, 10 January 2011
There are a lot of variables that affect the resistance fluctuations as a function of the rate of evolution of gases on the cathode. As you note, deuterium gas is twice as dense as Hydrogen, so the buoyant forces dislodging the bubbles will be less than for hydrogen. The smoothness of the surface of the electrode also is a factor. A highly polished surface won't have as many crags to keep the bubbles in place as a rougher surface. What they need to do, for any individual cell, is plot the AC power (aka "excess heat") as a function of the rate of evolution of the bubbling gases, for each species of gas, and for electrodes of different degrees of surface smoothness and plumb angle. I reckon they will find some generally reproducible results for correlating AC power to bubbling rates for different species of gas. —Caprice 17:32, 10 January 2011 (UTC)
We already have data limiting the amount of such noise. I've described it. Bubble noise might be subject to many subtle variations, and the surface of a palladium cathode under electrolysis is highly complex, both chemically and morphologically. simpler would be to vibrate the cathode rapidly enough to shake bubbles free quickly, before they can grow and cover more surface. Flow across the cathode would also do this, but would cause more mixing of gases from anode and cathode, perhaps, unless the flow was confined in some way. This would then substantially reduce bubble noise.
But the experimental data shows, clearly, that this is not necessary. There is no significant hidden AC power under bubbling conditions, that's shown independently from examination of the current and voltage themselves, it is shown by a method of measuring error that integrates all power: calorimetry. The calorimetry is independent from the power supply noise, that noise only affects measurement of presumed input power. Calorimetry under conditions of no excess heat, or presumed lack of excess heat, shows the same power as the measurement technique used by McKubre, including calorimetry during conditions of high bubbling, therefore the calorimetry and the power supply measurements confirm each other, setting upper bounds on the noise voltage. Nice try, no cigar.
This has become a pattern here: look at part of the data, then find some theory that could explain it. Proposed this as the reason for part of the results, present it as if all the work is impeached by this theory not having been refuted. And then when it's pointed out that the full results do refute the theory, move to the next proposed questionable detail. Repeat.
Nothing wrong with the process of investigating error sources, including preposterous ones (once all the facts are known). Indeed, this can be highly useful. The problem is confident assertion of error in the work of highly competent researchers, as if finding some cockamamie theory that they didn't note and refute in their publications, is proof that they are foolish and deluded. --Abd 17:53, 10 January 2011 (UTC)

Limit scope of input power criticism

First of all, not all CF experiments use the same power supply arrangements. Criticism should be specific, of specific experimental approaches, not shotgun, assuming that all experiments are done under the same conditions.

Further, McKubre is aware of possible criticism due to rsm noise. See [1].

Constant current or slowly ramped conditions were used in all cases so as to minimize the potential for unmeasured contributions to the input power. Commonly, experiments were performed electrically in series to test the effects of different variables, e.g., D20 compared to H20.

Note: not all experiments were done with constant current supply. Note that what McKubre is often looking for is comparison between D2O and H2O, or some other variable. If cells are in series, the same current is flowing through each cell. Resistance noise will decrease, because of the effect of a larger sample of random bubble variations.

The power input to the calorimeter by the electrochemical current was considered to be the product of that current and the voltage at the isothermal boundary. Under experimental conditions, this input power changed owing to voltage or resistance

variations in the cell, or at times when the current was ramped. This change had two undesirable consequences. A change in input power changed the cell temperature so that the electrochemical conditions were no longer under control. A change in the temperature also moved the calorimeter from its steady state as the calorimeter contents took up or released heat. To minimize these effects, we used a compensation heater to correct for changes in electrochemical power so that the sum of the heater and electrochemical power input to the calorimeter was held constant. A computer-controlled power supply was used to drive the compensation heater element operated in galvanostatic mode to avoid possible unmeasured rootmean-square (rms) heat input. This heater was also used for calorimeter calibration, in which the input power was measured as the product of the heater current and voltage at the isothermal boundary.

Now, Barry, read the above paragraph and see if you can claim that McKubre was ignorant of rms heat input as a possibility. Yes, I can think of questions to ask, as can you. If the RMS noise is fast compared to the ability of the compensation heater supply to adjust voltage, an error could still exist. However, a critical factor here is the level of RMS noise in the overall cell power. If it is low, the maximum possible effect is low. Above, you assumed 2 V peak-to-peak on top of 5 V. That's huge! The voltage swings from 3 V to 7 V, or +/- 40%. I'd be astonished if bubble noise could produce that kind of variation. I don't think that McKubre was really thinking about bubble noise in writing the above. He was concerned about other variations. I've seen no data on the magnitude of bubble noise, if it were this large, I'd think it would have been widely covered. A larger effect, that McKubre would have in mind, would be variations in cell resistance due to changes in cathode structure, and slower rms noise.

Note that bubble noise would be reduced by placing a large capacitance across the interface layer. Your capacitance effect would reduce bubble noise, to the extent it is a real effect. (A large capacitance could be set up using additional electrodes, say mesh, with a capacitor being placed across the cathode and an additional mesh electrode a short distance to the cathode and anode, each. As a side benefit, a mesh electrode like that, properly constructed, would reduce recombination.) Probably simpler to run a compensation heater to keep cell temperature, under no-excess heat conditions, more stable and make that heater be fast-response. I'd think of dumping the constant current supply entirely, and make the setup be constant-power-input, through the primary supply in voltage mode and a compensating supply with voltage controlled to a value to maintain constant power. Fast compensation, with high-bandwidth power amplifier.

(Remember that a constant-current supply has a power amplifier, with the necessary slew rate. The additional supply need not have any higher slew rate. The RMS power you assert would be supplied by the constant-current supply electronics, the additional supply would compensate.)

But I doubt that it's worth the experimental effort. Remember, the background here is that excess heat has been independently confirmed through the measurement of helium. So all this is crossing the t's and dotting the i's. Experimentally, researchers are now interested in increasing the reliability and quantity of excess heat, so unless they introduce some new error, they are safe.

You imagine that they are still in the business of proving (to whom?) that excess heat is real, independently from helium and a mass of other evidence, such as peak power considerations. Remember, if McKubre knows that the noise signal is small, which he could tell by a glance at an oscilloscope under operating conditions, he does not need to consider it. As I recall, I have heard McKubre state that he's looked at the signal. Uh, wouldn't you? --Abd 21:35, 9 January 2011 (UTC)

He wasn't ignorant of it. He expressly explained in his EPRI report why he was leaving the term for AC power out of his model. He left it out because he was assuming an ideal constant current power supply. And lookie here. This same issue came up five years ago. —Caprice 22:15, 9 January 2011 (UTC)
No, he wasn't assuming "ideal," he was explaining "sensible." And that's dicta, in fact. His analysis is quite adequate unless noise power is quite large, which is precisely what you assumed, still only coming up with 5% error. If you did the math correctly, which I have no confidence in, from your prior calculation gaffes, and lack of vision of the forest for the trees.
As to that old discussion, there are some sophisticated arguments there. Shanahan is pretty much the King of the Skeptics on cold fusion, he's got more convincing arguments per kilobyte than anyone. However, no cigar. Helium whacked his position long ago. There is lots of work on deuterium-oxygen recombination that shows it simply isn't happening (usually) at the levels he needs for his explanations. And helium confirms that! Shanahan needs, for his theory to stand, even aside from helium, some unknown effect causing recombination, and the theory applies not at all to sealed cells with recombiners and total accounting for all the deuterium.
This is the bottom line: Shanahan raises many plausible "alternative explanations," plausible if one doesn't look at all the experimental evidence and what is known to those skilled in the fields involved. He does not test any of them. Sometimes work has later been done by some experimenters that does address some of these objections. But mostly his objections are considered preposterous by those working in the field. He satisfies himself, but hardly anyone else.
Still, many people will think like him and like you. I see this all the time. I just responded to someone who had written, in a comment to the recent CBS special on cold fusion, something like "Why haven't they looked for helium? Since they haven't, this is probably all wrong." The person seemed to think that measuring helium was trivial, but he didn't realize that helium is actually difficult to measure accurately when you have D2 around, the mass is almost the same. Mass spectrometers than can resolve them are *expensive.* Miles depended on process that scrubs the D2, and that's messy and inaccurate, which is why Miles' data was so crude, it was good enough to establish "order of magnitude" information about heat/helium. And the person simply assumed that nobody had found helium because the special he watched didn't mention it. Wrong. Very wrong. He didn't look at the literature.
I noticed this lacuna myself. I don't know why, but many reviewers of cold fusion, supporting it, have neglected heat/helium, and the original paper that Storms wrote, that led to a request that he write, instead, a review of the entire field, was about heat/helium. We had been discussing it, and the strange neglect of what is, to me, obviously the strongest evidence for "fusion." So Storms wrote the paper. He may have been thinking about it already, and I don't know it for a fact that the paper was a response to our discussions. But people in the field certainly know about it, as well as anyone who reads Storms (2007) with any care; however, helium was not necessarily what convinced them *personally.* When they saw their calorimeter suddenly start showing lotso heat, they woke up. Many of them had been skeptical but had decided to try it. And then they were hooked. For Mizuno, it wasn't even a calorimeter, it was a cell that just kept on pumping out heat, for days, after the power was turned off. Actually it may have been weeks. (This was a closed cell, and pressure may have been pretty high.) Far more energy was released than any known chemical storage mechanism could explain. Recombination simply isn't even close.)
But isolated incidents like that have not been reproducible, so they don't convince others. Heat/helium is reproducible, quite so. --Abd 23:52, 9 January 2011 (UTC)
• If you want to do religious dogma, then do religious dogma. There is no point trying to do science with someone who insists on doing religion. —Caprice 23:56, 9 January 2011 (UTC)
• Barry, the religion here is yours. I mention reproducible experiment, reproduced by many, reported widely, with no contrary experimental evidence -- heat/helium fits with all the early work, the "replication failures," that found no heat and no helium -- and you talk "religious dogma."
• Suit yourself. Consistently, you are looking only at quibbles, minor points, applicable to nothing, or to some fraction of the work at best. That's how people who are attempting to maintain a belief argue. You are totally welcome to disregard what is being published in the peer-reviewed journals, at increasing rate, in favor of your own shallow arguments. You have no obligation to look at any of this, and no particular need to be interested, unless you are truly interested in what is behind the "difference in conclusions" that you claimed to be pursuing. What is the evidence considered by the two groups, the "skeptics" and the "believers"? Is the difference one of method or one of different evidence being examined? Are there any neutral judges?
• I can point to w:Robert Duncan (physicist) who certainly was neutral or skeptical but who is no longer skeptical about the overall field. Duncan was retained by CBS, as a reputable physicist recommended to them to come up with a neutral report. He was quite surprised. And I've never seen Duncan even comment on heat/helium, which is the single reproducible experiment with predictable outcome that everyone claimed to be seeking. It was done more than fifteen years ago. And the obvious conclusion from that reproducible experiment is "fusion."
• My conclusion and that of many others is that there is a large contingent of people who imagine that they are scientists but who do not understand the scientific method, because they require "models" in order to accept experimental data that appears to contradict existing models. That's not science, that's religion, belief in the conclusions of the past, rejecting contrary evidence. --Abd 00:53, 10 January 2011 (UTC)
I confess that I require any model to be consistent with Maxwell's Equations. That's my religion, and I'm sticking to it, come hell, electrolyzed high water, or lightning bolt discharges from a Van de Graaf generator. —Caprice 11:20, 10 January 2011 (UTC)
Dancing away from the point, you are. There is no inconsistency with Maxwell's Equations here. There are, as you can show and sometimes have shown, some aspects of the calorimetry that are not fully documented, and it's theoretically possible that there are therefore some unexamined error sources. But, because of independent confirmations, it's unlikely.
You just did, in fact, display religious devotion to Maxwell's Equations, chanting the name as if it, by itself, had the power to ward off the Unknown Nonsense, like a magic formula. There are quite a number of models that have been proposed here; none of them violate Maxwells's Equations in any way that has been shown.
You were asked to document your calculations of power from a speculative noise type. You didn't do it. You just gave a (somewhat vague) description of the waveforms. When asked about summing power, you chanted "Fourier." In your latest comments, you may have described the current waveform more accurately, but not well enough to be clear that this is what you meant, and I remain unclear as to what you actually based your power calculations on. I will probably do the calculus myself, but not today. At my age, it's like trying to put together the pieces of a jigsaw puzzle that were scattered about the house years ago. Now, where did that differential go? I know it's around here somewhere.... It may have been over forty years since I needed to integrate an equation.
We are, in practical terms, the same age. There may be a difference. I know I've got some marbles missing, and allow for it. One of the techniques of allowing for it is to discuss things, so that I'm not depending only on what Rosanne called "our own unsteady will power." I know the taste of my foot in my mouth, usually, and have learned that chanting "Oops!" and "Sorry!" helps with extracting it.
A triangle wave with constant absolute slope (except at the transition points, of course), has a very definite shape. The voltage waveform is a triangle wave, but the current waveform is more complex, and if you assumed a triangle wave current, you'd get incorrect results. I have not claimed that your power calculation is incorrect, merely that you have not shown how you obtained it. You may decline, if you wish, or you may chant "Maxwell's Equations" as much as you like. Freedom of Religion, you know.
I'm designing the seminar process here to allow people to take breaks for their religious devotions, to chat and kibbitz, and generally behave more like a real human community, yet still build learning resources. I believe this is being useful, and apparently you have agreed with that. When students at Cal Tech called each other "dorks" and "warm bodies," nobody called the campus police, they just carried on with normal adolescent or early adult rivalry and bluster -- and learning. What I'd add to that today would be training in dispute resolution that can transcend personal attachments, but we are not likely to eliminate what is probably quite instinctive. At least for males. --Abd 14:38, 10 January 2011 (UTC)
• You'll have two or three days to come up with your own independent models or calculations. Once you've done that, we can compare notes. Perhaps ScienceApologist will even have some more contributions to these analytical models. I'll be in Jacksonville giving a talk to the American Association of Physics Teachers. Perhaps I'll have time to schmooze with them about some of the issues that have arisen in my conversations with you and Ed Storms. Caprice 22:51, 10 January 2011 (UTC)

Confirmation of input power from calorimetry

Above is a consideration of a possible source of error in analysis of calorimetry, noise in input power, which can cause an error in estimation of integrated power through the simple multiplication of average voltage by average current.

It's important to realize that this isn't about the calorimetry itself, but about how calorimetry is interpreted. The calorimeter will show heat flow, limited by its own error sources, but unmeasured input electrical power will still show up in evolved heat.

Voltage and current noise in a constant-current power supply, caused by the response of the supply to changes in cell resistance due to bubbling, may create a component of the input power that will be missed by determining input power through measurement of average voltage and average current. The exact nature of this component will depend on various factors.

Researchers have not been explicit, in published papers, about power supply noise. If the noise is large, significant errors could be introduced, but a calculation with 2V of peak-to-peak noise riding on 5V DC, 40% voltage noise, produced an estimate of power error of only 5%. If noise levels were that large, it can be assumed that it would have been noticed and explicitly considered. However, the actual value of observed noise is unknown to those writing here at this time. Noise at the high levels necessary to throw off input power estimation by enough to impeach the study results would be immediately visible on examination of the power supply voltage with an oscilloscope.

Nevertheless, the following can be observed: it was work from Michael McKubre that was the basis for beginning this discussion. McKubre, in the subject report, covered many experimental runs, and runs were done with hydrogen and deuterium cells, and in one case in particular, cited here, the cells were electrically in series. No significant excess heat was reported in hydrogen cells. In addition, current excursions were applied to test the effect of current density on excess heat. These are under conditions where the cell cathode is already highly loaded, loading is maintained otherwise with a trickle current. Most current excursions, though they would introduce the same bubbling as with other current excursions, and thus the same speculative AC noise power, did not produce any apparent excess heat. Rather, the calorimeter normally, outside episodes of excess heat, showed no heat beyond that expected from input power.

Therefore power supply power noise is not the source of the anomalous excess heat, and the input electrical model is confirmed by actual calorimetry, which is independent.

Additional possible artifacts will be explored as new seminars listed under Cold fusion/Skeptical argument/Were the excess heat results ever shown to be artifact?. --Abd 18:58, 11 January 2011 (UTC)

Continued examination of the "AC power" artifact proposal

Caprice wrote the following on a user talk page:

There is still burst noise even if you model the voltage with perfect square waves. The math is a tad trickier. You can do it with Dirac delta-functions. But if you already find it too daunting to do it with triangle waves, it's gonna be even more daunting when those triangle waves collapse into Dirac delta-functions. What's easier is to note that the noise power is not a function of the slew rate. It's much easier to do the math for constant voltage (zero slew) than for constant current (infinitely fast slew), but the noise power comes out the same either way, because it's not a function of the slew rate. —Caprice 22:37, 17 January 2011 (UTC)

The article cited on burst noise has nothing to do with the bubble noise that we have been discussing. If the statement regarding independence of noise power from slew rate were true, the math would be very, very simple. Further, Caprice, in continued discussion in various places, including his blog, continues to assert that the neglect of noise power impeaches all cold fusion calorimetry, ignoring all the evidence from controls that shows that the calorimetry is accurate for purpose.

Let's go over this. The context has been lost; in the quote above, Caprice asserts that there is still noise, but the question is not whether or not there is noise -- there is noise -- but whether or not this noise will affect the estimation of input power through the average voltage measurements that McKubre uses, multiplied by current. The case most relevant is that the slew rate is high.

High compared to what? To the resistance slew. Let's look at what causes bubble noise. A bubble forms, it grows from very small, but becomes larger with time, perhaps as gas escapes from a small crack in the palladium. That process produces no rapid change; resistance will slowly increase as the bubble obscures part of the current path through the electrolyte. However, when the bubble is released from the electrode, water will rush in to replace it, thus lowering resistance more rapidly than it rose. However, the time scale for this would still be milliseconds, not microseconds. Therefore the bubble noise results from relatively slow resistance changes, taking place over milliseconds.

McKubre assumed that the power supply would respond rapidly enough that current noise would be very low. I see no reason to challenge this, though I'd certainly like to see characterization of the voltage and current noise. Current noise is assumed to be absent, and almost certainly is very, very low. This is a high-quality constant current power supply.

Voltage noise, then, with constant current, will track the resistance. Since the resistance noise is random, it will vary as much above the average resistance as below it. This was exactly McKubre's rationale: that current would be constant (a "scalar"), and thus average voltage times current would accurately provide average input power. If the slew rate were low enough that the power supply could not maintain constant current during transitions, the matter would be different. That was the point of McKubre's "sensibly designed" comment, about the power supply.

That this is essentially correct is shown, within the limits of calorimetry noise, by the match between input power and evolved heat, with "dead cells," where the bubble noise would be the same, or at least very close. --Abd 03:33, 18 January 2011 (UTC)

• Try computing the energy from burst noise during the rise or fall time of the voltage. Do this for any slew rate you like, down to femto-second speeds if you like. Note that the energy from the burst noise only appears during the instant (or very brief interval) when the very fast voltage transition occurs, to track the very fast step in the ohmic resistance. Do you wish to conclude that this fixed amount of energy, which is the same regardless of the slew rate, somehow magically disappears if you imagine the voltage step to occur instantaneously? —Caprice 11:31, 18 January 2011 (UTC)
• There is no w:burst noise. The problem is due to an assumption of a "very fast step in the ohmic resistance." Rather, the cell resistance has slow transitions, well within the capacity of the power supply to track them with constant current. McKubre stated his analysis in the paper, Barry simply disregarded it, because he did not understand the nature of the bubble noise. An electrochemist would be intimately familiar with it, I suspect. From long experience with electrolysis, he would know that "constant current" means constant current, with a well-designed power supply used for electrolysis.
• "Slow" means that the noise is limited by the gross physical nature of the bubbles. It is not electronic noise, like burst noise. In order to create a resistance change, a bubble has to grow (slow! you can watch them grow, I understand) and be released (relatively fast, but how fast does a bubble rise? Again, you can see it).
• Barry lost context. There is noise, but it is (as long as the supply slew limit is not exceeded) purely voltage noise, no current noise. Hook a hi-fi to the current shunt, assuming it's at the ground end of the circuit, and you will hear nothing. You might as well hook it to ground, that's a very solid DC voltage. Hook a hi-fi to the power supply voltage, you will hear the bubble noise, the voltage varies directly with the resistance.
• Because it is only voltage noise, the average power equals the average voltage times the set current. If there were significant current noise, we'd definitely have a problem. There is not.
• And then, even if Barry's analysis were cogent, and the assumed noise power not unreasonable, there is the experimental fact that McKubre's calculations of input power match the calorimetry, during periods of no calculated excess heat, even though the current was sometimes high and thus the bubbling would be the same. In other words, the theory was wrong (because a preposterous experimental condition was assumed) and the theory was falsified (by actual experimental observation.) Yet the theorist persisted in the theory, because the theorist has an agenda, and that agenda is not to falsify his own hypotheses, as he grandly proclaimed the cold fusion researchers should be doing, in order to not fool themselves, but rather to find the assumed error in the work of others. That's a formula for fooling oneself.
• Merely asking about the noise power, and investigating it, not a problem. But for some time now, Barry has been going around proclaiming that he found the error in McKubre's work, and extending this to the work of all cold fusion researchers, attempting to toss out all of that huge corpus of work, because he didn't see anyone factoring for the "noise power." Yet he never responded to the simple analysis, given above, that if the slew rate is high enough, there is no noise in the current, the whole function of a constant current supply is to eliminate that noise. If there is no noise in the current, the power noise is only voltage noise, and average power, in the presence of low-frequency random voltage noise, may be measured by measuring average voltage according to the scheme McKubre set up and used, and multiplying by the constant current.
• Now, Barry, what "emotions" are you experiencing (referring to your stated research objective)? --Abd 13:29, 18 January 2011 (UTC)
• Model the fluctuating resistance waveform any way you like. Model it as a sawtooth, or model it as a sinusoid. Assume any slew rate you like. Now compute the integral over one complete cycle. Go ahead, Abd. It's not that hard to do the math. This is well within sophomore level AC circuit theory. As to my emotions, you can read about them here. —Caprice 13:51, 18 January 2011 (UTC)
• Good thing you didn't write those posts here, Barry, you, how shall we say it, made false statements, many of them, when you should know that they were false, including statements about me and my history. Call it reckless disregard of the truth. If you wonder why people react negatively to you, this is it. Because you present the people there with a false understanding of the situation, you will get answers back that provide you little information about the real situation.
• Now, to the point. I have no skill at representing equations in MediaWiki.
• I will assume triangle wave resistance variation, and a slew rate of 1% resistance per millisecond. The produces no sweat in the power supply, the current will be held constant, and the voltage will track the resistance change. I.e., +1% resistance will produce a 1% change in voltage, so that current remains constant. (To be clear, the resistance variation is not a sawtooth, there are no rapid changes in resistance that a sawtooth would represent; rather the resistance slews positively for period t, and then slews negatively for the same period, completing one cycle.)
• There will be (there must be) some delay in the correction of voltage, but this delay is well under a microsecond, under the experimental conditions, and given how far the supply is operating from its margins, the current noise will be very, very low. We may treat the current as constant, as McKubre states, under the experimental conditions. That is not a generic truth, for if resistance noise is high-frequency and of sufficient magnitude (so that the required voltage slew to create constant current exceeds the supply slew rate), current would not be constant. There must also be some error in current constancy, but it would require very large variations in resistance to encounter that as a significant factor. There are high-bandwidth amplifiers operating in the power supply to maintain constant current.
• The voltage noise is a triangle wave also, the voltage being, at every point, I*(R+dr). For resistance excursion from R-r to R+r, the power slews from I^2*(R-r) to I^2*(R+r). I^2 is constant, so the integral of this power is I^2*R*t, where t is a multiple of the cycle time.
• Voltage in the McKubre experiment was measured by a method that produces clean average voltage, to high accuracy. Average voltage, under the triangle wave assumption, is I*R. Calculated input power is I^2*R, and it integrates to the same value as the integral of the DC plus noise power.
• I have not calculated the noise power for a triangle wave, because, under the experimental conditions, it's irrelevant. That noise power is not additive to the average DC power. Rather, there is a baseline DC power, the *minimum* DC power, and the noise power (which is entirely voltage noise) adds to it, bringing the power up to the average DC power.
• I have made approximations in the above. However, Caprice allowed the assumption of effectively infinite slew rate, which allows this. Nevertheless, there must be some delay in slew, i.e., there must be some error term in the current, but this term is effectively clamped to a very low level by the characteristics of the power supply. My guess is that it would take fairly sensitive instrumentation to measure it. Yeah, put a high-impedance, high-bandwidth, high-gain preamplifier on the current shunt, you might hear some noise. But not if you, as was suggested, simply hook up a hi-fi. That shunt would look, to a hi-fi, as a solid ground, noise-free.
• There is a more complex calculation that could be done. I'm not about to do it, and I haven't seen that Barry has done it. He's simply assumed the result, from inaccurate models, and many comments show that. And he continues to ignore the experimental evidence that shows that, if AC noise power is a problem, it's below significance in the experimental conditions described, but he continues to insist on his theory as an explanation of the experimental results. All the while condemning the researchers for not explicitly ruling this out, when McKubre did, in fact, rule it out, he just didn't spell it out for a novice with no experience in the field. Who wasn't paying SRI consulting fees.
• This is, in fact, a common phenomenon among pseudo-skeptics regarding cold fusion. This whole discussion stands as a good example of how that works, in practice. Error is assumed, then vigorous attempts are made to identify the error, and anything that seems, transiently, plausible, is asserted as if it were a fact. The scientific method, as actually used to advance the frontiers of science, has been tossed in the trash. It doesn't matter how many researchers come up with the same conclusions, it doesn't matter how strong the experimental evidence is, because it is assumed to be impossible, therefore there must be some error, and the pseudo-skeptic will continue inventing artifacts, of increasing implausibility, until the cows come home. And then again the next morning. At some point, that kind of pseudo-skepticism collapses, when a paradigm-shifting discovery finally has enough confirmation and, in this case, the die-hard skeptics themselves retire or otherwise fade away, and new researchers simply continue to investigate.
• There are already a number of hot-fusion physicists working on cold fusion, there always were some. These are not ignorant "believers," they are people who were convinced by the evidence sufficiently to bet their careers on it, in some cases. Barry has picked the wrong time to hitch his bicycle to cold fusion pseudo-skepticism, it's dead, as far as publication under peer review is concerned. If he's made such a major discovery, with his "misting" and "noise power" speculations, let him try to get it published under peer review. I'm hoping for it, but I doubt that his speculations would survive any knowledgeable peer review. Rather, we are blessed with them here, and can use them for educational purpose. I've certainly learned a lot! Among other things, I'm now quite satisfied with the use of good constant current power supplies in cold fusion experiments. That's relevant to my own work, I'm not using a high-performance constant current supply, just an ordinary current regulator, as in the Galileo project protocol. So, from this discussion, I'll look at the actual noise power. (But I'm not doing serious calorimetry.) I'll publish what I find! --Abd 15:56, 18 January 2011 (UTC)

Discussion of Burst Noise in Cold Fusion Cells

moved from User talk:John Bessa. --Abd 16:47, 18 January 2011 (UTC)

This is an entirely unrelated discussion about burst noise in electrolytic cells used in cold fusion experiments and has nothing to do with HiFis. It doesn't belong here. comment added by John Bessa in temporary placement in HiFi. --Abd 16:47, 18 January 2011 (UTC)

You would quite possibly hear nothing, even if there was an AC noise power problem. That's because the slew rate of the power supply allows correction of the voltage, full excursion, probably within less than a microsecond, so the problem noise will be above 1 MHz. The amp wouldn't respond to it at all. Connecting the power supply voltage (through a capacitor, probably) to a hi-fi would not be the way to demonstrate this noise, but it would easily be done with any normal oscilloscope, and with an ordinary DSO (I have a Rigol 50 MHz, two-channel 1GS/sec scope that was very cheap), one could measure noise power way above the frequencies of interest. And I consider it a practical certainty that McKubre has looked at his power with a 'scope, but in the event that he didn't, other experimental data shows that the AC noise power problem is not visible; the calorimetry would show it with control cells and conditions of similar bubble noise.
Barry's responses are quite useful, though, because it is allowing the observation of the amazingly persistent skepticism in this field, Barry is arguing, as I've seen many do, against what is now routinely accepted at peer-reviewed mainstream journals, with no contrary publication of equal quality. He is showing the position behind the brain-dead Wikipedia article, that presents what has been called the "scientific fiasco of the century" (i.e., 20th century), as if it were a minor error, long ago rejected, still rejected, with no significant research results. To see the current true mainstream position ("mainstream" in this case means "among those who know and understand the evidence"), see the abstract at [2] and read the article if you are sufficiently interested. Naturwissenschaften is a century-old journal, Einstein published in it, and it's Springer-Verlag's "flagship multidisciplinary journal." The review was featured on the first page of the October issue, prominently labeled "REVIEW." The editors know what they are doing, and Springer is not about to bet the farm on lunacy.
And then ask why a peer-reviewed secondary source, confirming many other peer-reviewed secondary sources over the last five years, is as if it does not exist, as far as the Wikipedia article is concerned. That's a problem whether cold fusion is real or not.... but it's a Wikipedia problem, I gave up working on it, because it became just way too hard, I'd have to appeal to ArbComm, etc.
I notice another recent review Springer has published, [3]. Barry has hitched his bicycle to a falling star.
As to those falling wires, the engineer in me expects that this amp might not be working! My condolences about the rare tubes. Hey, I have some of those in me, too! --Abd 16:44, 17 January 2011 (UTC)
• Well, there you have it, John. You can take static from Abd, or you can discuss the theoretical models of static with me. Incidentally, the name of this kind of noise is burst noise (also called impulse noise or popcorn noise). And it doesn't depend on the rise time (slew rate) of the power supply. Study of it dates back to the days of the telegraph, but it was studied extensively when it turned up in early solid state devices. The most common cause of burst noise is the random trapping and release of charge carriers at the interface between two conducting media. But I digress. With respect to your work on empathy models, do you have any guidance for how to express empathy with someone who is being cocksure, surly, and churlish? What is the recommended way to engage empathetically with someone who is exhibiting hubris? —Caprice 17:04, 17 January 2011 (UTC)
If what I'm writing is "static," the volume control is readily available. John, I'll stop all response here -- or control it as you wish, at a simple request. As to the question about hubris, perhaps Barry could tell us! How should we respond to him? As for myself, hubris is always a possible problem. I do tend to imagine that I understand a subject when I've studied it for a few years. Sometimes I do, sometimes I make mistakes, and the mistakes are how I continue to learn.
Now, as to the source of the noise in question, though I think that John was more concerned about static in his Hi-fi amplifier, the source of noise power in the situation we've been studying is the work done by the power supply to maintain constant current. That is controlled by "slew rate," i.e., the maximum change in voltage that the supply will create, per unit time. The noise is created by bubbles formed on the surface of the electrodes in an electrolytic cell, which reduce the conduction path, thus increasing the resistance, reducing current, and the bubbling thus introduces a random noise, within certain frequencies. My sense is that the frequency of this noise is not very high, because individual bubbles grow slowly, and when they break free and rise, they do so relatively slowly (i.e., we would be talking about a shift in resistance that takes place over milliseconds, not microseconds.) In other words, this would not present a challenge to a power supply with a slew rate over a volt per microsecond (total voltage might be 5 - 10 V.) But we have no data on the actual characteristics of this noise. Barry is speculating, and the speculation predicts results that don't show up with controls. Barry's claim that the noise in this case does not depend on the slew rate of the supply is blatantly incorrect, he's never responded to the point that infinite slew rate would not show the power estimation error he claims, nor would very slow slew rate produce the effect. He's oversimplified the situation; we are talking, not about the existence of noise -- it exists -- but about how additional power might be hidden in noise, not detected by a specific measurement protocol. --Abd 18:54, 17 January 2011 (UTC)
• There is still burst noise even if you model the voltage with perfect square waves. The math is a tad trickier. You can do it with Dirac delta-functions. But if you already find it too daunting to do it with triangle waves, it's gonna be even more daunting when those triangle waves collapse into Dirac delta-functions. What's easier is to note that the noise power is not a function of the slew rate. It's much easier to do the math for constant voltage (zero slew) than for constant current (infinitely fast slew), but the noise power comes out the same either way, because it's not a function of the slew rate. —Caprice 22:37, 17 January 2011 (UTC)
• Ugh. Pseudoscientific complexification of something that is relatively simple. I've responded with [4]. By the way, the power supply voltage would have noise in it, but probably in the kHz range, it would be a hiss, maybe a little rumbly. No current noise, i.e., if you were to listen to the voltage across the current shunt that was used to measure current, you'd not hear anything. DC. --Abd 03:47, 18 January 2011 (UTC)
• You've made your prediction, from your theory. Now do the experiment. Tap the voltage across the shunt resistor and feed it into a high-impedance pre-amp, and on into your Hi-Fi. Do you hear popcorn noise or bacon frying noise that correlates with the bubbling of the electrodes? Can you see also see the impulse noise spikes on your scope? —Caprice 11:43, 18 January 2011 (UTC)
• I certainly will look when I have the opportunity (using an oscilloscope, which would also see noise above the audible range). But this described experiment has certainly been done (with a scope) by many workers in the field. I'll ask them. --Abd 13:44, 18 January 2011 (UTC)

Error in reading power supply specifications.

We have been reading the slew rate specification for the power supply as if it were the transient response specificiation. In fact, it appears, the slew rate is the rate at which the power supply will alter its controlled voltage or current in response to a change in the programmed value. See the Kepco glossary for "slew rate." This explains how I became confused on the issue of current vs. voltage control and the specification. I was thinking of transient response from the beginning, and assumed that slew referred to it, as did Barry, apparently. I do suspect that the voltage slew to correct to constant current is similar to the voltage slew for change to a new voltage, but it's not necessarily the same.

In any case, the slew spec for current mode is then in terms of current rate of change. However, with slow noise, there is very low current change. Barry has postulated, in some places, step resistance changes, i.e., instant changes in resistance, producing an instant drop or increase in current, until the voltage changes to return the current to the set value. However, resistance from bubbles will not change instantaneously, it will be relatively slow, because it's a physical movement of bubbles.

The whole point of using a constant current supply is to have constant current; a properly designed supply, which is what McKubre refers to, will have low current noise when operating with modest changes in load.

Trying to find more on the specifications, The Kepco 20-20M is rated for recovery from a step load within 75 microseconds. The manual is at [5]. There is a figure in the glossary cited above for "load effect" which I don't seem to be able to see. Javascript. However, I was able to access the figure directly at [6]. The BOP dynamic specs are at [7]. The step load specs are probably for recovery from something like a full-range change in load. This is a +/-20V, +/-20A supply. --Abd 03:05, 19 January 2011 (UTC)

• It doesn't matter how you model the resistance fluctuations. The math is easy if you model it as a step, but you can model it as a sawtooth or even a sinusoid and it doesn't change the result very much. That's because, when you do the Fourier Transform (or construct the Taylor Series), almost all the noise power is concentrated in the first non-zero harmonic (or the quadratic term of the Taylor Series). The exact shape hardly matters. —Caprice 02:49, 24 January 2011 (UTC)
Barry, step modeling will create an abrupt transition in resistance, precisely the condition that will create current noise with a real power supply. I now understand the error you have been making.
If we assume that the power supply has infinite slew, i.e., can adjust immediately to load variations, we can consider the current as constant.
So we could, for example, model the resistance as having a sinusoidal variation, and let the resistance variation be such as to make a 2 V RMS signal appear across the resistance, given the constant current, let that current be 1 amp. Notice, the current is DC. AC current is zero.
Now, let the average voltage be 5 volts. That is, the 2 V RMS voltage is added to 5 VDC. A 2 V RMS voltage varies, peak to peak, from -2.8 volts to + 2.8 volts. The average RMS voltage (of the signal created by the resistance variation) is zero. Thus the average sum of the DC voltage and the AC voltage is 5 volts. The voltage varies from 2.2 volts to 7.8 volts. Very noisy, eh? What is the total power?
It's 5 watts. The power varies from 2.2 watts to 7.8 watts, instantaneously, average 5 watts.
What would McKubre measure for this? He takes many voltage samples and averages them together to produce a periodic average voltage, that is recorded. The sampling is regular, but the resistance changes are random, so the average voltage measured will approach the true average voltage of the signal, integrated over the time period. The total power for the period over which the voltage is averaged is equal to the average voltage times the constant current. Nevertheless, there is AC noise power. It's not missing! It is simply part of the overall power.
You are calculating noise power and assuming that it is additional to what McKubre finds. It is not, constant current is a special case. The noise power causes the instantaneous power to increase and to decrease, but because the current is constant, the average voltage covers all of that. If the current varied directly with the voltage, which is much more normal, this would be different, because increased voltage normally means increased current, so there is a squaring of the variation, which doesn't then average linearly.
You have been assuming that if there is AC noise, as seen in the voltage, there will be, therefore, an error from using average voltage. That is only true if the current is not constant. Get it? --Abd 03:36, 24 January 2011 (UTC)
• Tell Mike McKubre to do this experiment: Find an older style desk telephone with a carbon microphone. This would be one of those traditional desk telephones with the round mouthpiece that was the standard model until the 1990s. Take out the carbon microphone and drive it with a constant current. Put the scope trace across the carbon mike and also across the 1-Ω series resistor to measure the current. Tap the mike to get a fluctuation in the resistance of the carbon grains. What audio signal shows up on the two traces? —Caprice 04:14, 24 January 2011 (UTC)
• Barry, this would show very little of value. That signal, however, is an audio signal, it's in audio frequencies, and has, I'm pretty confident, very little high-frequency noise as part of it. The current would be constant. I doubt very much that there would be anything above 20 KHz. There might be nothing that high, even. The noise signal would be entirely voltage noise. What I have as experimental report from those who have looked at the current with an oscilloscope is that the cell current is, in fact, clean, noise-free. Do you have one shred of evidence to the contrary? And do you have any theoretical reason to expect bubble noise to be like the burst noise that you have sometimes called it? --Abd 05:57, 24 January 2011 (UTC)
You can do the experiment yourself. A carbon button microphone from a Model 500 telephone handset has a resistance of about 4 Ω. Drive it with a 12 V power supply in series with a 100 Ω resistor. For all intents and purposes you will be driving it with a nearly constant current. Put Channel A of your scope trace across the carbon button microphone and Channel B across the 100 Ω resistor. Tap on the microphone with a pencil and look for the signal traces on both channels. —Caprice 00:12, 25 January 2011 (UTC)
Experiments without a stated hypothesis being tested before the experiment may be interesting, but are not conclusive about practically anything, they won't be properly designed and objectively observed. What is being tested by this experiment? What result do you expect that you think would be unexpected for me? Why should I waste my time on this? And why shouldn't I drive the thing with a constant current, rather than a "nearly constant current?" I have a constant current power supply, after all. And the whole discussion here is about the behavior of a constant current power supply, not about carbon microphones and noise power in circuits where current is not held constant. --Abd 18:41, 25 January 2011 (UTC)
• Have you done the experiment yet? Have you written up your analysis and results and submitted them for peer review? —Caprice 19:19, 25 January 2011 (UTC)

Looking at a scope display of bubble noise?

Elsewhere, Barry, you mentioned looking at a 'scope display of bubble noise. The context was that this was for McKubre. Where is this image?

I will point out that if I'm correct about confirming what McKubre wrote about the "sensibly designed" constant current supply, there would be no visible current noise from bubbling, only voltage noise, down below about 10 KHz. I rather doubt that McKubre (or Naudin, if this was really about Naudin) would be showing us a display of the current, rather, normally, they would show voltage.

There is voltage noise introduces, there is no question about that, though there is a question about the magnitude of it. If there is no current noise, i.e., the current noise is below significance, then estimating power by measuring average voltage and multiplying it by a set current will be accurate as to total input power of the cell.

Elsewhere you referred to multiplying two triangle waves together; but there is only the voltage triangle wave (for a triangle wave resistance variation.) The current is constant, as long as the load is not changing faster than the capacity of the supply to track it.

The proof would be in the current as seen on an oscilloscope. I find it practically impossible to imagine that the researchers never looked at the current! After all, McKubre states that the current is constant. How would he know? Sure, he might have assumed it. Him and hundreds of other researchers who have done a similar experiment. And nobody ever looked? These people talk to each other, you know. If there was a problem here it would be known. But I'll ask.

McKubre actually measures the current, as I recall, in the same way as the voltage. He doesn't depend on the power supply setting. He takes the individual measurements and averages them; those individual measurements are fairly fast samples, we could check the specs for the meter he's using. If current were noisy, those samples would have high variance, it would stand out like a sore thumb. His raw data would show the problem, but that data isn't published, if it was kept at all. If the current data is stable, it's awfully boring! --Abd 03:17, 19 January 2011 (UTC)

• The scope screen shot is here. Very likely, the red trace is the voltage across the 1-Ω series resistor that's used to measure the current, and the multi-colored trace is the voltage across the terminals of the cell. If that's correct, then the AC noise power can estimated by multiplying the fluctuation of the red (current) trace by the fluctuation of the multi-colored (voltage) trace. If that's correct, then the AC noise power in that scope screen shot would be about 0.25 mW AC, on top of 30 mW of DC, or about 0.8% of the total electrical power. —Caprice 03:17, 24 January 2011 (UTC)
Not "very likely" at all. I agree, the multicolored trace is probably the voltage across the cell, probably, though I'm suspicious about something, and I'll tell you why below. However, the red trace isn't the current, for sure. This is a very low sweep rate display, if you don't know why I say that, ask. The red trace is either calculated average power, or it is calorimetry data. They wouldn't show current in a display like this, regularly, because it's a stable value, very dull. If that's the output of a constant current supply, it's truly a lousy one! I have a LabView display from an operating CF cell, and it shows voltage and current. The voltage is captured at low rate, and it looks sort of like the red trace. The current is flat, straight, no visible noise at all. It shifts rapidly from one level to another when there is an adjustment. And that's what the people I asked told me they see. In other words, Barry, faced with real bubble noise, which would be intrinsically low frequency noise, compared to the capacity of a decent constant current power supply, that supply supplies, miracle of miracles, constant current. McKubre covered all this by saying that the current was constant. You, quite simply, didn't believe him, and then accused him of making a sophomoric error.
Now, why am I suspicious of the "multicolored trace"? Well, they might have been using SuperWave stimulation. After all, McKubre did replicate the Energetics Technologies work. And SuperWave is extraordinarily complex, it's designed to be that way, to cause the system to be constantly shifting as to equilibrium. This is how ET gets stronger results. Could this be a problem with measuring power? Sure, I'd be much more inclined to believe it. But ... that problem is easily handled with controls.
I'm suspicious in that other indications I have are that bubble noise isn't as large a factor as you have suspected, i.e., the voltage noise is not so much. But maybe.
Controls are of various kinds; relevant to the discussion of bubble noise, there would be hydrogen controls, dead cell controls, and platinum cathode controls. In the latter two cases, there is the same bubbling, but no excess heat, thus verifying that the input power estimation is correct for the operating conditions. Hydrogen controls could have some variation due to the difference in buoyancy, as you've noted, but they would not be entirely different.
You stated 1 ohm shunt resistor. Did you find that figure somewhere? --Abd 03:58, 24 January 2011 (UTC)
Note: my speculation about this being SuperWave was correct. That display gives us no information about normal bubble noise, period. It's simply showing the SuperWave voltage pattern, at very slow sweep rate. --Abd 19:06, 25 January 2011 (UTC)
• What we've been calling fluctuations in ohmic resistance from the formation of bubbles is a very old topic in electrolytic cells. Over a century ago, a man named Arthur Wehnelt constructed a device known as the Wehnelt Electrolytic Interrupter. An Electrolytic Interrupter is an electrical current interrupter consisting of a cell containing two electrodes in an electrolytic solution in which bubbles, formed at frequent intervals by application of current to one of the electrodes, continually interrupt the passage of current. Rob Duncan and Richard Garwin have pointed to the Electrolytic Interrupter as a model for the source of AC power in McKubre's cells. McKubre told Duncan that the "excess heat" effect is "improved by surface roughness on the cathode." Increasing the surface roughness would help to keep the bubbles in place longer before they grow big enough to cast off from their moorings. Writing about the Wehnelt Electrolytic Interrupter in 1910, Karl Taylor Compton observed, a "strident thumping noise, due to the explosive interruptions" which "were almost deafening." Of course, Wehnelt's device was intentionally designed to maximize the circuit interruption effect. The bubbling in McKubre's cells doesn't produce quite so violent an effect as Compton noted in Wehnelt's cell. But just how much noise does it produce? Curious scientists want to know. —Caprice 06:06, 25 January 2011 (UTC)
• "Quite as violent," a drastic understatement, but, my guess, you want to imply similarity. Duncan and Garwin have pointed to the Electrolytic Interruptor? Where? Barry, you keep ignoring a number of central issues:
• The calculation of input power is confirmed by the calorimetry, in numerous ways. It's accurate, by experiment and measurement of actual power, using a method totally immune to bubble noise, which cannot, with closed cells, change the actual heat evolved.
• Electrochemists do electrolysis all the time. They know what happens. They know about bubble noise. They know how to address it, what is necessary and not necessary. What the electrochemists tell me is that they have measured input power, from the power supply, in many different ways, and since all these ways come up with the same answer, they simply use what is simplest: a constant current power supply with recording of average voltage. Mostly they just use standard data acquisition equipment, now, with LabView, one researcher is using precise information that the power supply itself, computer controlled, reports over the instrument bus. But they have used high-speed oscilloscopes to look at the actual current with high bandwidth, and the noise you are assuming, as to current noise, simply is not there.
• This is not at all surprising. Constant current power supplies can handle the noise generated by low-speed physical movement of bubbles. There is voltage noise, for sure. But power, if current is constant, may be accurately estimated by taking many voltage measurements and averaging them and multiplying them by the constant current. If the current varies, all bets are off. It gets complicated.
• The scope trace that you thought was cell voltage was, probably. There was no current display there. What I found out was that this experiment was SuperWave stimulation, very complex waveform. On my suggestion that specific noise figures be given, McKubre has promised to check with ENEA, which also confirmed the Energetics Technologies work (SuperWave), on this. However, from McKubre's paper on the ET confirmation, published in the American Chemical Society Low Energy Nuclear Reactions Sourcebook (2008), the SuperWave signal seems limited to 100 Hz as the maximum frequency component. Very well within the capacity of the power supply to keep constant current.
• That scope display, taken off of the CBS Sixty Minutes video, gave you no information relevant to your point, you totally misinterpreted it, but you have claimed in many places that this shows how they have ignored "AC noise." That was your fantasy, based on your profound ignorance of the experimental conditions in this work, yet ... you keep asserting it. --Abd 19:04, 25 January 2011 (UTC)

The confirmation of McKubre's power input estimation

Calorimetry is used for direct power measurement, it is immune to problems with AC noise. Since McKubre was using calorimetry, with various cells under various conditions, and most conditions don't produce any apparent excess power, the question is then if cells, other than the ones showing "excess power," were operated under conditions where Kort's theory of significant AC power noise would predict an "excess power" error due to unmeasured noise.

And the answer is, yes, they were operated that way. McKubre's work involved loading cells and maintaining loading for a very long time with a "trickle charge." Then current excursions were created, where current was ramped up in a series of steps. Most of the time nothing happened, with hydrogen and deuterium cells. But sometimes there was excess heat.

(Yes, Barry, "excess heat" means, precisely, "unexpected heat," or "anomalous heat." It does not mean "fusion heat," though that might possibly be the source of it. In Cold fusion work, it means the heat that the experimenter is unable to explain through prosaic process. One person's excess heat might later be explained as artifact. There are CF reports where excess heat is marginal, and a small percentage of input power. Generally, though, those results aren't reported as "success." Sometimes they are reported as being of interest in some way, and good research reports all the results, not just "success." On the other hand, sometimes in CF research there has been a great deal of work with no result at all. When, suddenly, what seemed to be a totally dead cell starts pumping out heat, it's pretty exciting, and it just might be reported at the next ICCF if they were doing something new and unusual. And it might or might not be reproducible. That's cold fusion research. Very frustrating at the same time as very exciting. Eventually, cell reliability became quite good. Miles was hitting over 60% by the early 1990s. Miles. One of the supposed "negative replicators," per the 1989 DoE report, because they didn't return his phone call (before their report was issued) that he'd started seeing results. One more of those CF stories....)

In any case, most of the time that current is ramped up, there is no excess heat. Bubbling would be just as high or almost as high. The difference between an excess heat period and a dead period is blatant.

I'm sure that when you see a graph like some we've looked at, showing the stepped power and the excess heat that jumped up with it, in the deuterium cell, but practically no change in the hydrogen cell in series with it, you didn't realize that most of the time that this current excursion was applied, nothing happened in either cell. You thought that this was a "typical" result. Aha! you thought, it's the bubble noise from the high current. And you explain away the hydrogen lack of "excess heat" with some other reason. So set aside the hydrogen control. Just look at the deuterium cells, and realize that this current excursion wasn't the only one, just one -- atypical -- where the excess heat effect showed up. Some deuterium cells didn't produce any of those effects, but they were still bubbling, quite the same.

(Yes, there will be a small difference. Excess heat is fairly well correlated with loading, so there might be more bubbling, the higher the loading. But at this point in one of these experiments, there is a lot of bubbling with little increased loading. F-P experiments are pretty inefficient.)

The confirmation of McKubre's input power estimation is the calorimetry in the periods where no anomalous heat appears. This demonstrates that noise power is not significantly outside the calorimetry error. What you have asserted is a hypothesis, that there might be significant power. The actual experimental results show, no, there isn't. Yet, elsewhere, even though this has been pointed out, you are announcing that you have found the "artifact" that explains CF results (this and misting, a separate issue, though connected in that misting would also increase, if there is any mist emitted from the cell, with increased bubbling).

Since you pointed out the comments here, I'll say that your comments on [8] are shameful. --Abd 03:43, 19 January 2011 (UTC)

• The hypothesis that the AC noise power contributes about 1% to 5% of the total input electrical power is consistent with both the analytical models and the screen shots of the scope trace, assuming we are reading them correctly, as outlined above. —Caprice 03:23, 24 January 2011 (UTC)
Recalculate, assuming constant current. You were not reading correctly re current. I've confirmed with researchers, I'll provide specifics later, there is no visible current noise. Faced with bubble noise, a constant current power supply delivers constant current. There is AC noise power, all right, but it doesn't damage the estimation of input power as a product of average voltage and the constant current.
Further, the hypothesis that there is significant unaccounted-for input power due to bubble noise is falsified by the behavior of the cells in McKubre's P13/P14 report from EPRI, which we've been looking at. There are periods of maximum bubbling with no excess power in either the hydrogen cell or the deuterium cell (which are in series), and, while the graph of excess power in the deuterium cell in the paper, for the third of three current excursions, shows tracking of the current, and thus, presumably of bubbling, the hydrogen control only shows some low level possible tracking, down in the calorimeter noise, the error bars still cover zero excess heat, and even that increase only appears at maximum current, and doesn't appear with the current step before that. Thus excess heat doesn't correlate well with current, as would be expected if bubble noise were the source of excess heat through power supply artifact. The calorimetry confirms, instead, that the estimation of input power is accurate within calorimetry noise. Excess heat is a chaotic phenomenon that was, at that time, relatively unpredictable. Sometimes it was seen and sometimes it was not, but these cells would always be bubbling similarly at high current under the conditions described (which is that high loading had already been attained; if loading variations were behind the appearance of excess heat, it would only be as a minor threshold effect, bubbling would not be radically different.) --Abd 04:09, 24 January 2011 (UTC)
• If you assume constant current, then you either have to treat the noise pulses as traveling waves (as is done in telephony), or you have to use Dirac delta-functions when computing differentials with the Calculus. Either way, it has to come out the same, because the mathematical models have to agree with natural phenomena as governed by the (discoverable) laws of nature. If you discard the correct models for the laws of nature, and replace them by flawed models, they won't yield reliable analytical results or predications. The notion that bubbles disrupt the flow of current in an electrolytic cell has been known for well over a century. Before the invention of vacuum tubes, bubbling in electrolytic cells was exploited to devise valves that could rapidly turn the current flow on and off (so as to drive an induction coil to produce high voltages). Note that over a century ago, Karl Taylor Compton reported that the electrolyte becomes warm in these cells as the current flow is interrupted by the formation and release of the bubbles. —Caprice 13:51, 25 January 2011 (UTC)
• This is complexification in order to hide ignorance. Barry has allowed, previously, the assumption of "zero response time" to variations in resistance. That means that current is constant. Apparently he is now challenging that current is constant, which, then requires finite response time, and the introduction of current noise.
• In the devices mentioned, current is actually interrupted by bubbles, i.e., blocked. That's a huge variation in resistance, and no constant current supply could keep current constant under those conditions. Yes. I assume constant current, which means that I assume that the resistance noise is not fast enough in transition time to cause significant variations in current. From the supply specs and an understanding of what bubble noise must be (slow, by comparison with the speed of the supply to respond), I do assume the same as the researchers assumed, including the (reasonable) skeptic Britz. Britz has now agreed to re-examine this in the sense of determining the actual noise figure, instead of simply looking at conditions and observing that it is "below significance." He correctly noted, in response to my question, that he had not calculated the actual noise expected, but he also confirmed the reports of others as to oscilloscope observation. The noise he claims to be the cause of excess heat, which means it must be very significant, isn't observed. By anyone. Skeptic or "believer."
• Given how much has been published in this field, and given how obvious this possible error is, to those without the electrochemistry experience, one would think this would have been specifically raised in skeptical literature. It hasn't been, to any notable extent. It's just part of the general "buzz" that assumes that there must be something wrong, but nobody has actually shown it, and it would be simple to show. If it were this easy to replicate the excess heat results, it would have been done long ago, and then the researcher would have debunked these results by showing that the source of the excess heat was power supply noise, by simply measuring it. That's how real science would address an error like this.
• Fake science sits in a chair and makes up one theory after another, but without assuming any responsibility for making predictions from the alternate theory and verifying or falsifying them. Barry's theory of power supply noise is easily falsified, by many different already-recorded observations made plain in my comments, but, in fact, had he been doing real science, he'd have looked for this evidence himself, instead of vigorously arguing against it with every argument yhe can imagine, no matter how bogus, every piece of dramatic polemic he can think of, like "Electrolytic Interruptors."
• To others looking at this, without knowledge of the field, it may look reasonable what Barry claims, sometimes. Let me point out that the views I'm expressing match what is being published, over the last five years, in mainstream peer-reviewed journals. I'm explaining how the science is currently understood in this field, by those who really count in scientific publication: the peer-reviewers at mainstream journals. To the unwary, it can look like I'm pushing some fringe view, because what I'm explaining was a fringe view, say ten years ago. The situation shifted, and that understanding, given that there was no dramatic demonstration, just an overwhelming accumulation of often-complex evidence, hasn't penetrated to most "ordinary scientists," i.e., those who have not followed the literature.
• To agree on a detail, it is obvious that there must be *some* error in maintaining current. That, however, doesn't make it significant. The issue is how fast the resistance transitions occur, if it's power we are concerned with.
• Barry has raised many issues connected with this, often making some true statement that is applied in such a way as to produce quite misleading impressions. For example, Barry has commonly complained that I haven't done the exact math to determine noise power, assuming some waveform in the resistance. I haven't delved into that complexity because it is moot, for more than one reason: a constant current supply actually does hold current constant under these conditions, "constant" meaning "within the necessary accuracy for an experiment" -- well within it, in this case -- and the experimental calorimetry confirms the power estimation assuming constant current. That is more than adequate, especially since we are reviewing work that is almost twenty years old! The basic results, the reported phenomena, have been massively confirmed by others. That's how science is supposed to work. --Abd 19:35, 25 January 2011 (UTC)
• You can't repeals the laws of nature with a blizzard of words. Without being aware of the prior literature, I postulated two obvious errors in the energy budget model — treating entrained mist as evaporated water and ignoring AC noise power. As you now know, this was not an original discovery on my part. Kirk Shanahan and others raised the issue of entrained mist back in 2005. And the study of the effect of bubbling in the electrolyte dates back more than a century. —Caprice 20:17, 25 January 2011 (UTC)
• You cannot own the laws of nature with tendentious ignorance. Those are speculative errors in the model, and the evidence is clear that they are not real errors. You've built a house of cards, a fantasy, now, with no support from any experts, with bad analysis and misreading of evidence and total failure to attempt to falsify your own hypotheses, it's obvious. Suit yourself, it's all visible, and the record is clear, when it's reviewed. --Abd 03:43, 26 January 2011 (UTC)

Dieter Britz evaluation of Moulton power model theory

In response to a question I raised on the CMNS mailing list, and from direct correspondence between Barry Kort (Moulton) and Dieter Britz, Britz wrote an evaluation of the input power model, examining the suggestion that significant bubble noise would affect the power calculation.

http://www.dieterbritz.dk/coldfusion/powercalc.pdf

I previously showed that, even if, in theory, there could be power estimation error, the behavior of CF cells in calorimetry and the comparison of active cells with inactive controls -- all of which would have the same bubble noise or similar -- demonstrated that this was not an actual effect large enough to have a major impact on excess heat calculations. Further, CF researchers, when I asked, claimed to have confirmed their power calculations with methods that Kort had previously said would discover the error: one reported using a high-bandwidth wattmeter, and high-speed data acquisition so that instantaneous power could be calculated from dual measurements of voltage and current. Since all methods agreed on the input power -- as would be expected from the analysis Britz made, which only duplicated in detail what I had done on a napkin, so to speak -- cold fusion researchers settled on the easy method of collecting average voltage data and multiplying by set current.

If there is artifact in these experiments, it is not from this method of calculating input electrical power. --Abd 17:26, 25 February 2011 (UTC)