Skip to content

Untested and speculative ideas


This page is here to draw attention to some of the more speculative ideas I’ve contemplated over the years, but for various reasons I was unable to test them in a conclusive way. Some of them may not even be seriously testable without millions of dollars of investment and years of consulting with experts, but I still found them interesting in theory.


Use of near-field communication for power electronics control

After success of my simple wireless power systems in providing power to various loads, it’s natural to start exploring the possibility of sending information as well as power over the same channel. While this is already practiced extensively in applications like RFID, I’ve never seen such means of communication being utilized for a task such as providing gate drive to power semiconductors. It’s easy to imagine why –  radio communication has never been considered being reliable enough for this task, where a single bit error could cause destruction of the switching devices. And in most power electronics, there are many sources of radiated and conducted noise, and other interference that may compromise weak radio signals.

However, near-field communication combats this problem by transmitting information with many orders of magnitude higher energy than typical radio, and there is a chance it may indeed work for this application. The simplest form may be on-off keying based modulation, utilized with simple wireless power scheme where the receiver output voltage directly drives a power device like IGBT or SCR. More sophisticated variants may include duplex signalling, where every drive module would send a status-reporting bitstream back to the controller.

This concept could be best utilized in applications where other forms of gate drive become difficult, such as high voltage semiconductor switch series stacks. Guard rings, which are generally utilized in such modules to prevent corona formation, could easily double as good “antennas” for wireless power and nearfield communication. Usually, systems like this required extensive fiber-optic communications to control and monitor every device in the stack, and interruption of a single fiber link could have grave results. In these cases, Near-field communication may offer advantages both in lower cost and increased reliability compared to existing technologies.


A while ago, Steve Conner posted about an interesting article that described radical improvement in properties of MOSFET devices when cooled to cryogenic temperatures by liquid nitrogen:

Last year I’ve managed to do a test on my own, by cooling some newer poly-silicon gate MOSFET’s in liquid nitrogen. I was only able to do simple DC load tests here, without any transient or switching tests – but the results were still quite impressive, showing around 10x drop in ON-resistance on all devices tested; however some devices needed higher gate voltages to turn fully on. Obviously, the next step would be to do switching tests and construct a practical example of an inverter running at cryogenic temperatures, in order to fully asses the value of such an approach. A cost-effective mixed-gas Joule-Thompson cryo-cooler like described here could be used in conjunction with an inverter, replacing the tipycal forced air or fluid cooling system. While I’m currently unable to fully test this idea myself, I would be very happy to research it in future once I get more resources available.

High frequency SSTC burner igniter

For many years I’ve had trouble, over and over, with stove burner igniters not doing their job properly, sparking and sparking but not igniting the gas as they should. The traditional igniters (those built into stoves) are generally based on  iron cored HV transformers very similar to a car ignition coil. A capacitor that is slowly charged from current-limited, recitfied mains discharges periodically (few times a second) by a SIDAC trigger device into the transformer primary, very much like a spark gap Tesla coil. This produces a spark pulsing a few times a second on the output, which usually jumps towards a grounded object like burner head. The small, short lasting spark needs to occur in exactly right place and time where the gas-air mixture is correct for the ignition, and the practice has shown that small amounts of dirt either on the ceramic spark pillar, or the burner head itself can shift the spark or gas flow away from ideal circumstances, making the ignition difficult.

For a long time I’ve had an idea that would not only solve these problems and make ignition more reliable, but also reduce the overall cost of the system and reduce the radio frequency interference produced.

In order to ignite gas more reliably, we ideally want a more voluminous plume of plasma, that also lasts a lot longer than a typical igniter spark. An ideal way to produce something like this would be a small SSTC operating in MHz range, and producing a small plasma plume atop a sharp breakout point. The high frequency output would not need to connect towards a grounded surface: it could be directed simply by a sharp breakout point tip, making it much less sensitive to dirt and electrode wear.

Unlike 50Hz transformers and pulsed igniters, the output would also be completely electric shock-safe due to insensitivity of nerves to HF currents – of course, it still wouldn’t be burn-safe and care should be taken not to touch it!

A SSTC operating in MHz range would produce narrowband interference around it’s resonant frequency, which should be easier to filter than the low frequency interference caused by pulsed igniters.

And finally, a MHz range SSTC would need very little copper, and no iron at all in comparison to bulky iron cored transformers, as well as no potting (Tesla coils can stay air cored) making them less expensive to construct.

Of course, some problems are still left to be solved – such as minimization of the driver circuit to reduce it’s cost. The coil would only need to produce an undemanding 10 mm or so long plasma plume, but to keep the cost down it would probably have to run from mains (no step-down transformer) which complicates the driver circuit. The output could be pulsed (for example 500ms ON, 2s OFF) to reduce heatsinking requirements.

This is the kind of igniter I would love to see in future stoves, water heaters and burners!

Somewhat ludicrous ideas

Electrolasers and QCW Tesla coils

The idea of ionizing air with a laser beam and using the path to conduct high voltage has been around for a long time, but all I’ve ever seen were a couple of pictures and videos without a clear description of setups used. Generally, the idea seemed to utilize an extremely high power pulsed UV beam from a frequency-tripled Q switched Nd:Yag or a similar laser, to provide the ionized path for a high voltage source that can discharge rapidly, like a Van De Graaf generator.

While I’ve found little exact info, it seems that the required laser power is rather massive and unavailable to most hobbyist experimenters, along with exotic nonlinear frequency-tripling optics.

Then I saw this video featuring volumetric displays in air, made using a focused, pulsed IR laser beams:

The ionization in this case is caused by extreme heating of the air at the point of beam focus, rather than from ionizing UV radiation.

Immediately after seeing this, I thought of QCW Tesla coils and their slowly propagating sword stremers, which could perhaps be guided in air using series of ionized dots in air. I have no idea whether it would work: the voxes size may simply turn out too small to have any effect on the spark. But it’s definitely something to try, should ever a QCW coil and a big scary laser meet in a same place one day!

Lofstrom loops

While not an idea of my own, this alternative idea of a space launch system has sparked a lot of my curiosity.  The concept is explained on these pages. While it’s obviously a gargantuan project, which may need billionic investments to even determine if it’s remotely possible – which naturally made it receive little interest from serious scientific community – I was fascinated by the mere idea of having a magnetically levitated, hypervelocity rotor moving in a vacuum sheath at many km/s, and all sorts of physics and control theory unknowns brought by it. I would find it very fascinating to do at least some theoretical research or simulation about this subject – of course, I would first need to spend quite a bit of time consulting an expert in magnetic levitation before even attempting something like that. I proposed that a similar device, though running completely on ground in a circular loop and at much lower velocities, could be used as a method of energy storage for grid load leveling, a much more probable first application of this interesting concept. But all this is pretty well in dreamware category until I get a lot smarter than I’m now!

Big Railguns

Ever since I’ve seen US Navy’s Rail Cannon that is being developed for military purposes, I started imagining some more peaceful applications, such as launching satellites into space. Despite numerous big problems faced by all electromagnetic space-launch concepts, I could quite clearly conclude that rail guns are still the most promising technology. Launching a typical sized satellite or a manned spacecraft by electromagnetic accelerators are difficult merely because of gigantic power requirements, which may be on scale of electricity consumption of a large country, sustained for several minutes, making them extremely speculative so far. But, I’ve contemplated something that may be much more easily attainable with current technology: an experimental rail-gun could launch a soda-can sized microsatellite into orbit, as a proof-of concept experiment that could later be supersized.  The naval rail-gun has already launched heavier projectiles at velocities approaching orbital, and a land-based rail-gun intended for space launching would not have the same length constraints as a ship-based weapon: this means it could be made hundreds of meters or even a kilometer long, greatly reducing acceleration, current and power requirements for the launch. Perhaps this would make it possible to use a large rotating machine for the power supply, instead of a gigantic capacitor bank, which would further reduce the cost. The generator could perhaps even be hacked from one or several commercially available power plant generators, removing the need to designing a special compulsator from scratch. I’d even dare to project that a team of capable scientists would be able to construct such a project on a budget of a few million and demonstrate it successfully with some luck!

The most difficult part of the railgun, which I’m primarily concerned with, would be the current collectors for the projectile. While I’ve built a toy-sized rail gun model, my experience from working on it may mean absolutely nothing for a massive, space launching system. I’m pretty sure that brush design becomes exceedingly difficult at hypersonic velocities, and some innovation in this field would certainly be of great benefit for the idea. I’m currently investigating the possibility of using liquid or low-melting point metal alloys as a means of enhancing the high-current projectile to rail contact. However, some serious research in fluid dynamics would be required to make such an idea work at many km/s.

Some of fantasy articles to read up:

Neural networks and FPGA’s

In recent years I’ve started to become increasingly interested into alternative computing methods, particularly with the concept of neural networks. While the field is still rather unknown to me, I found it fascinating because it underlies the still least understood computing device known – the human brain. Artificial neural networks, despite their simplification in contrast to natural ones, are still a little developed branch of research, and I believe a lot could be done for their improvement.

I was particularly fascinated by the concept of using evolutionary algorithms to develop neural networks (word “learning” many not be exactly appropriate here) that converge over many generations towards performing a particular task better and better. I really got excited by this idea when I saw a simple practical example of this idea in action, shown on the following video:

It would be amazing to apply the same concept to real-world problems, such as recognition of text and speech. Moreover, one other thing I would love to investigate is how could artificial neural networks be implemented on high density programmable logic devices like FPGA’s, to attain orders of magnitude greater speeds than possible with computers. It would be exciting to look for an optimal combination of series operation and parallelism, as well as explore the possible benefit of reprogrammability of FPGA’s to directly benefit the concept of evolving neural networks – for example, each new generation could be created by flashing the ROM with new synaptic weights, over the data of previous generation.

Brain-computer interfaces

For a long time I was curious about possibilities of interfacing electric signals from human nervous system to outside world, have been reading about current achievements and consulting some people around it. So far, the technology seems rather primitive and there are many hurdles to overcome. It would be ideal if we could take every axon of a nerve like a wire and stick it into an amplifier input, but that’s easier said than done. Axons are tiny and delicate structures, which are also actually a protrusion of a single cell – and the distal part tends to die if it’s separated from it’s parent neuron. In central nervous system, permanent scarring tends to occur on both sides preventing almost any regeneration!

Hence the most of the ideas are based on stimulating neurons, such as retinal ganglions which is the area I’ve been investigating lately. Actually, primitive visual retinal prosthesis have already been implanted into some patients, though they provide only basic functionality with like 10×10 pixels of resolution!

I was wondering if this could be made any better, and after consulting an expert in VLSI design there doesn’t seem to be an electrical reason why arrays of many millions of electrodes could be used instead. However, I’d need to consult someone with good knowledge in neuroscience and physiology before I can take the idea with any degree of seriousness.

My idea was as follows:

  • A monolithic integrated circuit would be created with millions of electrodes that could be multiplexed to the signal source by analog multiplexers.
  • The chip would best be implanted over fovea in order to have the best effect, from inside of the eye so it contacts the ganglion cells directly; I have no idea about difficulty of this procedure, but I feel steps would have to be taken to prevent the hard IC with sharp edges from damaging delicate tissue.
  • Before attempting to display any image, the chip would perform some sort of imaging of the retina surface. At first I thought of optical imaging, but then I figured out that it would be much simpler, as well as provide better resolution, to simply do conductivity mapping using high frequency currents that don’t trigger the nerves. Hopefully this would create an image where individual ganglions could be identified, and only electrodes near each ganglion would be programmed towards it; I think this approach could greatly reduce signal to noise ratio of the prosthesis.
  • Finally, one would need to completely figure out the human vision encoding system before attempting to send any signal; I don’t know much about this either, apart from the fact that the intermediate neurons in human retina perform a sort of edge detection which would now have to be done by computer.
  • The implanted device would be as simple as possible, and powered by wireless power; information processing would be done by external computer and a camera.

Enough ranting for now! In any case, I’d love to discuss what of this is possible, and what not some day.

Teleoperated robots

A while ago I’ve ran into this interesting video of a robotic “ditch-witch” that used robotic exo-skeleton style arms that are guided by the operator to drive a pair of large and powerful robotic arms, and provide feedback to the operator which allows very precise and fluid control. As it can be seen from the video, human brain still makes a superior control system for applications like this!

After seeing that, I naturally couldn’t resist not to think – “Wouldn’t it be extremely cool to do the same for the legs as well?” – indeed, the same company has already created a sort of a full body robotic exoskeleton, and I don’t see why would it be impossible to use a similar setup to operate a fully humanoid robot, and allow all sorts of complex aspects of human locomotion like running, jumping and even acrobatics to be exercised by a robot, controlled by a human pilot. Providing enough feedback in order for the brain to fully accept the robot’s body as it’s own may be demanding – the whole feedback exoskeleton may need to be suspended in a gimballed mechanism in order to provide orientation/balance feedback to the operater. The target robot would also likely need to have a realistically articulated spine, which may also turn out to be the hardest part to provide feedback for, despite it’s crucial  for human balance.

The target may not need to be a real robot at all: it could be a virtual model as well, providing means of all sorts of virtual reality simulations.

Leave a Comment

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: