Unconventional electrolysis

Quote from magicsound

Unfortunately, I haven't yet figured out how to get a functioning link to the modified circuit for posting

It's through the File > Export as Link... menu item:

Perhaps a further improvement to the simulation could also be adding the inherent inductance of the spark gap's discharge channel. I found about this on a few papers, like for example:

• https://www.researchgate.net/p…tching_UWB_HPM_generation
• https://www.researchgate.net/p…a_Submillimeter_Spark-Gap

I haven't worked out the calculations yet, but an inductance in the order of 0.3-1.0 nH/mm is quoted and used.

0.2 Ohm for the inductor ESR seems low. I think it's higher, but I do not have the proper instrumentation for measuring it accurately.

Quote from Robert Horst

LTspice is a little harder to learn, but gives much better simulation of real components. That gets more important as you use MOSFETs and other active components,or need to simulate the core saturation of inductors. It also allows much more complexity in stimulating the circuit with different waveforms. I run it in a Windows10 VM on a Mac.

I definitely should check that out - I haven't yet as of today also due to testing with the new power supply and associated preparations.

Quote from Robert Horst

Regarding your power source, for either a power supply or battery you may want to add a resettable fuse (also known as a PTC or positive thermal coefficient device). Here is one that might be a good choice: https://www.digikey.com/produc…500/RHEF1500HF-ND/5029799

I didn't know about these resettable fuses as discrete components, thanks for the information. I was considering getting a high current circuit breaker intended for car audio systems, like this one:

magicsound

That's unfortunate. It's likely that any form of Javascript blocking could cause some problems to that web application.

On a different note, earlier today I disassembled the electrodes and made a few photos.

Just after removing them from the jar

After reaching suitable conditions for producing arc discharges the gap is mostly clear, with wet patches/clumps of conductive metal/metal-oxide particles remaining. Gap width was larger than it should have been.

The condition of the electrodes at the interface, before drying them.

Electrodes after drying and wiping with paper towel, but not washing. The anode is the one on the bottom of the photos. It used to be the cathode; for this test I swapped polarity to even out wear. It appears it has gained quite a few new craters and damage compared to last time.

To continue these tests without the emission of excessive amounts of toxic fumes from spacer material, I have ordered some inexpensive mica sheets normally used as insulating shims for transistor heat sinks and I should have a bunch within a few days. No testing until then. The material should be mostly inert to acidic environments and will not burn. Since these sheets are 0.09 mm thick it should be possible to stack them to obtain the desired thickness.

Some properties of mica: http://www.icrmica.com/icrmica_mica_introduction.html

I thought of getting thicker sheets normally intended for replacing waveguide covers in microwave ovens to cover the backside of the electrodes, ultimately obtaining a kind of "sandwich" as in the associated diagram below, but I will pass for now.

I also got a current probe; although it will be no oscilloscope it should allow a better understanding of the operating parameters especially in case this actually works for producing anomalous effects, although in the end it's for the most part a variation of the Russian-type "water explosion" experiments (including Alexander Parkhomov's recent "Woodpecker" device), so on this basis they shouldn't be totally ruled off.

If magicsound decides to go forward with attempting to replicate a similar cell to verify any RF behavior as previously planned, the results should help shaping what I will do next. For what it's worth, since that post I found that:

• The polarity of the electrodes (which in my case have a slightly different shape and surface wear) does not seem to be critical
• A different (much nicer) power supply and cabling produced similar results, at least acoustically speaking
• The electrode portion not immersed in water might be a contributing factor to the acoustic noise produced
• After the discharges (and the noise) start occurring the gap appears to be for the most part clear of hard to remove deposition products
  • It looks like it's mostly produced by electrically conductive particles forming a short circuit
  • It could mean that a thick oxide layer will not help here
• When I obtained that only very slight amounts of 10% HCl would immediately start a reaction, I also previously aerated the solution with ambient air
  • I later wrote that excessive gap width was the main cause for a a higher need for larger amounts of HCl solution in later tests, but it probably was not the only factor as I didn't previously add air
  • Could be important information for replicating the same exact effects quickly, although I don't think it's crucial
• A Nd magnet does not seem to be necessary to start the noise-inducing reaction
  
  • I didn't use one last time - I only later realized that I forgot adding one

Today I received the current probe and the mica spacers. Some observations after some more testing.

• I reused the previous materials without altering anything else, except for the spacers; everything appears to have started working more or less as previously within a couple minutes or less.
• The mica spacers worked: no strong bad smell by burning materials was produced for the most part.
  • However, for mostly hassle-free operation I had to increase the gap on the bottom of the electrodes from roughly 0.4mm to 0.6mm (from 4 to 6 spacers).
• The current probe (which also works as a multimeter) revealed several features of the system.
  • The coil has a lower resistance than I assumed, but that was to be expected in retrospect: 0.3 Ohm.
    • Even this value might not be very much accurate, however.
  • When a "welding-type" fault occurs where a conduction pathway would not be blown apart, currents up to 36A can be passed through the coil, which heats up quickly together with the electrodes.
    • Surprisingly, even under this much current being passed, which is close to the limit allowed by the PSU of 40A, line voltage would still be about 11.3V.
  • Under the intended mode of operation current varies within the 4.5-7.5A range, with occasional spikes either up or down.
• Contrarily to what I expected, the hissing noise appears to be occurring mostly on the electrode gap portion outside of the water.
  • This could indicate that total immersion in the electrolyte-metal particle solution might not be desired, but it's to be tested more in depth.
  • I found this by lifting the electrodes from the solution while they were hissing.
• I found that the orientation of the AM Radio appears to be somewhat critical in order to "listen" the same hissing noise there as it is acoustically.
  • The previous finding this time has been replicated: I could hear the sound there too.
• Subjectively speaking, water loss seems much quicker when the repetitive discharges occur rather than when a "welding-type" fault occurs, even though power into the cell is significantly lower.

A photo of the testing arrangement

Jar hissing. This could be heard on the AM Radio too. Eventually I turned off the power supply due to excessive foam production. I think this video might also be showing that such foaming is preventing the hissing noise in a way or another. (direct link).

https://streamable.com/z0beq

More of the same, minus the foaming (direct link).

https://streamable.com/7nmq2

Wire from the coil moving under spontaneous intermittent operation (direct link).

https://streamable.com/v1sal

Attached a recording of the noise with my cell phone in "audio recorder" mode.

Sound 43.zip

EDIT: a couple graphs. Voltage is across electrodes. Measurements have been manually spot-sampled and are not continuous, even if at first the connecting line might suggest otherwise. No significant change in Geiger readings observed.

Today I tried a kind of controversial test. It's not going to prove anything due to the large error margins and issues with entrainment, etc but I wanted to have a sort of measurement of how much heat was produced with the equipment I had at disposal (which is pretty minimal).

I weighted the jar with the electrodes before and after an experimental run. Apparently in total 18.85g of water have been evaporated or dissociated (measured before and after the run).

Using a video and OCR'ing the data (a rather painful process) of DC voltage and current measured across the electrodes**, I've calculated that roughly 100 kJ of energy have been put into the cell, which was not insulated (as usual).

(**EDIT 2018-12-15: important clarification: while I wrote "across the electrodes" I actually performed measurements across the terminals connected to the electrodes and the coil. This means that when the electrodes completely short-circuited most of the heating went into the coil as it has a non-negligible DC resistance which can be roughly quantified in 0.3 Ohm. This should be consistent with the observation of the "welding-type" continuous short-circuit events apparently not contributing much to cell heating, as I later note in the comment)

(Link to the simplified circuit simulation. Note that in the actual circuit power supply polarity was the opposite)

If water just got evaporated it would have taken 2.257 kJ/g = 42.54 kJ in total, meaning that more than half the energy got lost into the environment. However electrolysis also was ongoing to some extent. As far as I aware of it should take a minimum of 237 kJ/mol to dissociate water, or about 13.16 kJ/g of water.

Overall I had the feeling that water evaporated relatively quickly, but I can't rule out various artifacts going on for this sort of test, or other effects such as the combustion of metal particles (from the electrodes), etc.

A more accurate test at least on the input energy side would probably require a power analyzer or at least measuring the current drawn by the power supply at the AC outlet. This power supply is supposed to have a power factor of .99, but at low loads it is relatively inefficient.

I tried to upload a video of the instrumentation in the usual place, but it only allows a maximum length of 10 minutes.

EDIT: due to significant foam production during this test I had to turn power on-off regularly. Furthermore the electrodes would often not operate as intended and a kind of continuous "welding" reaction would occur instead of the desired intermittent operation. The former is where current draw would be the highest. Foam/bubble and vapor production however occurred for the most part during intermittent operation (the one which causes hissing acoustic noise and AM radio noise).

(Data partially sanitized from spikes that have not actually occurred due to OCR errors)

EDIT: for those really interested, here is the video (split into 3-parts) of the instrumentation:

https://streamable.com/df6ez

Part1

Part2

Part3

In the data the total resistance of the jar+coil was naively calculated as V/I, but one has to consider that due to the error margins in particular of the current probe, values calculated when power is not applied are not going to be very reliable.

By filtering out resistance values calculated under such conditions (for example when voltage is lower than 10V and current lower than 0.15A), I obtain that the minimum total resistance measured across the circuit is about 0.293 Ohm (obtained during sustained short-circuit conditions), which is very close to the 0.3 Ohm initially assumed for the coil. A fair assumption could be that the coil has a DC resistance of 0.28 Ohm. This seems in agreement with the current probe I recently purchased (which also works as a multimeter), which reported 0.3 Ohm with its probing terminals.

At this point I tried going further and calculate power and ultimately cumulative energy to the coil alone, assuming that its DC resistance is constant with load and temperature. Once I obtained this, I subtracted these values to the total power and energy previously calculated to obtain those at the jar, and this is the result:

It turns out that about 70 kJ might have actually gone into the electrode-containing jar, out of the total 100 kJ of the coil+jar. This makes the previous results more interesting, although now of course the main issue becomes that of water loss, which might be possibly dominated by entrainment of water droplets into the steam and gases produced. Visually speaking, roughly 3/4 of the jar's water content (with the jar nominally containing 25 ml) was lost, which seems consistent with the weight difference measured.

Attached filtered/cleaned data used for the above graph, which however still contains many brief spikes caused by OCR errors.

Robert Horst

For what it's worth, the coil at the end of the test was only slightly warm to the touch. It's made of a stranded tin-coated copper wire with still its insulation in place, and dissipates heat slowly from what I've seen in other tests. I might upgrade this at some point in the future with a properly made coil, if it will prove to be useful.

My current meter is a rebranded version of this (see attached manual):

https://www.tacklifetools.com/product/product/index/id/29

It's supposed to integrate measurements over 4000 counts and read frequencies up to 100 kHz, although its accuracy over noisy DC waveforms is unknown.

I'm aware that the margin of errors could be large for several reasons here, regardless of the old wet/dry steam debacle. However, if there's truly excess heat, it should be large enough to be measured by probing current at the power outlet. That still won't be perfect, but it should be better.

Will read the paper in detail later - I was thinking that in my case the acoustic noise (associated with radio noise of similar quality) could be mainly due to electric discharges occurring at a high rate within the unimmersed portion of the electrodes, causing the combustion of hydrogen and oxygen in the presence of traces of a fine metal particle-containing acidic slurry.

Quote from Robert Horst

Regarding the noise, I came across an interesting paper with many measurements of acoustic standing waves during electrolysis. See:

Kenneth E.Tempelmeyer, Electrolysis Bubble Noise in Small-Scale Tests of a Seawater MHD Thruster, 1990.

https://apps.dtic.mil/dtic/tr/fulltext/u2/a227548.pdf

After skimming quickly through the paper, there might be a certain degree of similarity in some of the spectral plots, but I have little idea if my testing conditions can be comparable to those experiments. Unlike them I also recorded sound through the air under uncontrolled conditions and with a microphone which has a far from linear response. The colored spectrum here is from earlier testing; the other from fig.15 of the paper.

On a partially related note, something I didn't fully consider which the paper highlights in the introduction is that with the electrolysis of seawater (or, my guess, chlorine-containing water) hydrogen can be generated at both the anode and cathode. I have to stress that only trace amounts of chlorine are present in the aqueous solution I have been using for my tests, so it's not like they pose immediate chemical hazards. I would say that during operation only a faint poolwater odor can be generally noticed.

However since the underside of the electrode not immersed in water but exposed to the jar opening look kind of oxidized (unlike the immersed portions which look for the most part black, just like the aqueous solution) perhaps I should pay a bit more care.

EDIT: on an even less related note, I found that altering the contrast and gamma and desaturating webcam images directly from the source might help with OCR, which in retrospective makes sense. Next time I'll try this way. Unfortunately the display of the clamp meter is not as readable as that of the cheap digital multimeter I have used for voltage readings.

magicsound

Nice to see that you were serious with it.

For the sake of avoiding related frustrations it's probably worth mentioning a few practical problems I've encountered so far which you might experience as well:

1) The electrodes can get rather hot, either simply by the electrical power applied or hydrogen-oxygen (or other) combustion and start degrading the retention system. Lately I have used standard (nylon?) plastic zip ties but I think I might have to upgrade this for prolonged/repeated testing. For short (in the order of a few hours) testing sessions they seem fine.

2) A more important note is that depending on several variables including mainly the amount and size of particles suspended in the solution and depositing on the surfaces of the electrode gap (which seems to be governed by current density and how acidic the solution is), at 12V a 0.2mm spacing could be too low and cause the power supply to occasionally fail in providing a sufficient amount of power to destroy the short-circuiting path(s) continuously forming. When this happens an unrecoverable short-circuit will occur and the electrodes will have to be separated or cleared manually in order to continue. During such event most of the heat will also go into the coil, so it will have to be monitored (not necessarily with a thermocouple), but there will be acoustic and visual indications of when this happens before it gets damaged.

A larger spacing mitigates this issue. Last time I used 0.7mm but I still had occasional problems with it. Alternatively, a higher supply voltage (i.e. more power) could possibly provide short-term help, but I haven't had the opportunity to try a power supply that could bring it.

3) Another issue which might or might not be a problem in your case is that a black foam composed of conductive, ferromagnetic fine metal particles can get produced during the test. Last time I had to interrupt it several times due to it. Depending on the dimensions of the container and level of the aqueous solution, it might cause inconvenient spilling outside of it. The bubbles formed might also contain dangerous concentrations of O2-H2, so beware (although in this experiment series I haven't had larger explosions occurring from it yet).

magicsound

I don't think a mesh cover would affect negatively the experiment, so it's probably worth trying. In earlier tests I used a piece of Al foil to reduce the amount of spatter, but in the latest one for some reason it turned out to not be sufficient anymore.

A potentially important note is that in the past few testing sessions I have kept reusing the same metal-particle containing solution (stored in two small jars). I haven't added any more HCl since last time I started with clean distilled water. This could have been a related factor.

On a loosely related note, in the last testing session of which I previously posted data, I found that a ferromagnetic load/core on the coil appeared to make the reaction at the jar quicker to occur. However I made only a single attempt at trying this, so it would need further investigation. If you can find a large ferromagnetic steel or iron bolt to place into your coil you could check if this will turn out to be the case as well in your planned experiment.

(here was first without a core, then with a core - in my case composed of several ferromagnetic tools. EDIT: note that I've used an improvised ferromagnetic core all along, so the test was checking out if removing it would make the reaction weaker - which it apparently did)

While waiting for magicsound's results, I wanted to make another test. Unfortunately one of the zip ties I used to hold the electrodes together broke and a quick fix I attempted didn't restore the electrodes to the previous tight fit. The electrode gap got significantly wider than usual and probably mainly because of this I haven't been able to reproduce the hissing noise as in the past few tests.

Right now I'm struggling with OCR'ing current data from the current clamp, which doesn't have a very readable display.

I tried to measure voltage directly across the electrodes, and kind of expecting high voltage spikes below 1000V or so I made a voltage divider to avoid damaging the multimeter. 2 MOhm on across both ends and 0.4 MOhm across the last resistor, for a 5x voltage reduction. Unfortunately I didn't manage to observe any spike justifying its usage, but that's also probably because the reaction didn't proceed as usual.

Since I used the 200V scale on the multimeter with the measured values divided by 5 this also means that measurements for the low voltages observed haven't been very precise.

The syringe was used for periodically introducing distilled water as the solution got evaporated or electrolyzed, but I forgot to weight the jar before and after the test. The water level was gauged by eye, so this was not a very accurate test either. While the total amount of water introduced in the end could give some indications about the heat produced, since electric measurements are quite rough this will probably not be as useful as initially expected.

I will post some graphs when the OCR process has finished, but this time I'm not trusting my own data very much.

After a lenghty OCR process I came up with a graph, but it's very noisy and probably unusable. The original video data is kind of of human-readable but I'd need a better method to process it. The main issue is that the lighting wasn't constant during the test. EDIT: uploaded a slightly better/less noisy version.

The last refill at the end of the experiment wasn't accurately measured but it included at least two completely filled syringes of water (5+2ml) and something more than that. Input power/energy is calculated across the electrodes/jar, with no contribution from the coil.

I would need to come with a better method for normalizing webcam images, gamma correcting and processing them with some sort of morphology filter. Since all of these filters have to be applied the 130+k frames of the video and are CPU-intensive, it takes a long time and it doesn't always work anyway because the the instruments' displays don't update instantaneously and often show "half-updated" values which will have to be discarded. I haven't investigated in detail yet, but many of the spikes visible in the graph above might be from such half-updated values.

For processing I'm using ImageMagick and a script to perform the batch-processing.

Some useful related pages:

http://www.imagemagick.org/Usage/color_mods/

http://www.imagemagick.org/Usage/morphology/

http://www.imagemagick.org/Usage/

Here is one of the worst case examples (showing valid figures) that has to be processed:

(voltage 1/5 of actual value)

In retrospect I should probably have not attempted to color-correct webcam images at the source and only attempt this as a post-processing operation. Using constant artificial light would have been better as well.

I came up with something that could work using ImageMagick I previously mentioned with this multi-line command (graph updated in the previous comment):

convert \
-crop "881x225+66+63" -normalize -level 80%,92%,0.1 -shear -11x0 \

-draw "rectangle 0,0 40,225" \
-draw "rectangle 200,0 263,225" \
-draw "rectangle 427,0 490,225" \
-draw "rectangle 655,0 718,225" \
-draw "rectangle 883,0 922,225" \
-draw "rectangle 78,25 158,100" \
-draw "rectangle 310,25 380,100" \
-draw "rectangle 533,25 613,100" \
-draw "rectangle 764,25 844,100" \
-draw "rectangle 78,125 158,200" \
-draw "rectangle 310,125 380,200" \
-draw "rectangle 533,125 613,200" \
-draw "rectangle 764,125 844,200" \
-fill white \
input.jpg output.jpg

The top portion of the following image:

Becomes this:

This is probably still not perfect because what now looks like an "8" was likely a "9" or a "4" (due to the slow-changing display some digits might be in an inconsistent state), but overall there seem to be less errors and spikes compared to the previous method.

The main point of interest with the above command is that I fill empty spaces in the image with pure white boxes, taking advantage of the fact that the display has fixed width segmented digits. In this way I can get rid of unwanted noise which disturbs the OCR process.

To optically recognize this I use ssocr

https://www.unix-ag.uni-kl.de/~auerswal/ssocr/

Using the following parameters:

-C --charset=digits -d -1 crop 40 0 844 225