Turquoise Energy Ltd. News #41
Victoria BC
Copyright 2010 Craig Carmichael - July 1st 2011; Happy Canada Day!

http://www.TurquoiseEnergy.com = http://www.ElectricHubcap.com = http://www.ElectricWeel.com

Month In Brief (summary)
  * Motor controllers: problems (no wonder the car doesn't move!) & solutions
  * AKOO Stuff: Hubcap & Weel motors - e-bike - torque converter design - LED house lighting - battery experiments
  * Wind power: costs hit parity with fossil fuel power costs in 2010
  * Off Topic Editorial Rant: loss of freedoms; US war crimes & Wikileaks; failure of our governing systems to evolve.

Electric Hubcap System
  * Motor Controllers... A3938 Motor Controller: more failures
  * Microcontroller based motor controller: the way to get the exact features you want!
  * MC33033 controller, e-motorbike: more tests, more failures.
  * Back to square 1: IR2130 triple MOSFET driver + 'standard' logic chips like I was using in 2008/9!
  * New IR2133 (with some devious connections!) gets it down to 2 chips, with optimum 'current ramp modulation' control.
  * Motor system thinness; Metal protection: "Cold Galvanizing Compound" spray; First flat magnet rotor.

Electric Weel Motor Project (Electric Wheel Motor... Rim Motor...)
  * Plywood stator base instead of composite mold to speed up protoypte
  * Welded the magnet rotor (finally)

Torque Converter Project
  * Spring-loaded MTC protoype - better configuration idea

LED Lighting Project A little energy efficient home lighting, anyone?
  * 10W, 1000 lumens LED... run at just 4 watts as a 3-AA cell plastic dome light... takes on 3W tungsten, 15W CF and 60W tungsten.
  * 4 W hallway globe light - runs on spare 9 V power adapter plugged into light fixture. (250 lumens? Far brighter than previous 7 W CF.)
  * .9 W "off the shelf" car inside light... lets me read street names on maps after dark.

Turquoise Battery Project
  * Electrode conductivity is now good, but finished cell conductivity is still low.
  * trying other electrolytes

Newsletters Index/Highlights:
http://www.TurquoiseEnergy.com/news/index.html (oops - I moved the index in March and neglected to change the link until now!)

Construction Manuals and information:
Electric Hubcap Motor
- Turquoise Motor Controller
- 36 Volt Electric Fan-Heater
- Nanocrystalline reflective rear electrodes to enhance DSSC Solar Cells
- Simple Spot Welder for battery tabs, connections

- Electric Hubcap Motor Kits, Parts - Build your own ultra-efficient 5 KW motor!
- Sodium Sulfate 4x longevity additive & "worn out" battery renewal.
- NiMH Dry Cell Car Batteries
(please e-mail me to order batteries)
- NiMH Custom Batteries
(EVs, E-Bikes, Scooters, etc. - no extra charge)
- NiMH individual Dry Cells (D - 10 AH, $10 -- AA - 2.5 AH, $2.50)
- Motor Building Workshops

...all at:  http://www.TurquoiseEnergy.com/

June in Brief

   June only had 30 days, but somehow this isn't as brief as expected. It seems like little of note has been accomplished, but the bowling ball is a little farther down the alley towards the several pins at the end.

Motor Controllers

   Motor controllers came to dominate the month. It has gradually dawned on me that they aren't performing well at low speed -- just where lots of torque is needed to start a vehicle moving.
   The variable frequency current control method that I'd found in the A3938 seemed superior to PWM, but my several A3938s all burned out, and I finally gave up on them. (Turns out they were designed only for small motors, <1 HP).
   Unknowingly, the low PWM frequencies I chose for the MC33033 greatly reduced low speed torque. (understandable now, but not mentioned much less explained in any of the literature)

   The first trials with Tristan's electric motorbike, with a MC33033 controller, showed it: the bike just didn't draw enough power to start moving, notwithstanding that it was only a bike with one person on it, and it had a 4 to 1 gear ratio advantage! The car moved across level pavement with direct drive to the wheel in 2008 when I used my own "primitive" design controller. The electric outboard worked fine because the propeller has negligible load at low speeds, so the motor had no trouble getting up to a speed above the low-speed/low-torque effect. I speeded up the PWM to 'recommended' frequency, but then the controllers started failing. After a week of blowing MOSFETs, I finally got a short bike ride on level ground before the next failure.

   I resolved to make a better type of motor controller, based on the A3938 type "current ramp" modulation.
   For a while I thought I'd use a microcontroller for the "brain", with the same power driver circuits and MOSFETs as in the MC33033 controller. I started studying Atmel's nice "SAM7S16" ARM chip and how to apply it to this application. Then I decided that a microcontroller would take too long to get going. And the burnouts seemed to show that better MOSFET interfaces were what was really needed.

   It started to look like the best choice (of a relatively unsatisfactory lot) was the IR2130 triple gate driver and some "old standard" small scale chips - an improved version of what I had made in 2008 before I'd ever heard of single chip brushless motor controllers!  After designing a tricky circuit that got it down to just three chips, it occurred to me to look on irf.com, and I found there's now an IR2133 with a couple of extra features. By a couple more 'workarounds' on this new device I got it down to a two-chip controller with the better 'current ramp' modulation instead of PWM. The month ended with long sessions of circuit and circuit board design, with the board layout completed July 1st.

   If I think my stuff often isn't thought out very well at first and only gradually improves, I can take solace looking at IR's chips: the power MOSFET driver parts seem to be second to none but their chips' logic functions aren't very well considered. With a couple of different pins on the IR2130 it could easily have become a single chip high power brushless motor controller. Instead, the IR2133 has put in one of the functions - in a convoluted sort of way - kept three rather useless inputs, improved one feature, and removed a useful internal timing feature I had planned on using and now must fudge with resistors and capacitors.

Weel Motor

  I continued building the big Weel Motor at the established snail's pace. I cut two plywood circles for the stator rings, to be covered in PP-epoxy rather than molded entirely from it. I tack welded together the two parts of the magnet rotor.

Pieces for the Weel motor.

  I can't assign a very high priority to this project until there are motor controllers to run it with, and, being short of epoxy to do the rings the day I cut them, I moved on to other things.

Torque Converter

   I started building a new torque converter before I realized that a better motor controller had to come first. It had a pivoting wedge piece to hit the drum to make torque hits, but I wasn't entirely confident about the arrangement.
   A mechanically inclined friend came over and talked about centrifugal clutches. The converter couldn't grab on like a clutch, but from that I got the idea of having the wedge spring out centrifugally into a slot in the output drum rim rather than pivot into one. That seems like it should have considerably more force, with both the spring and its own inertial mass going for it, so I'm changing the design and I'm calling it "most likely one to succeed".

Motorbike Project

   Tristan's motorbike project seemed to be stalled. He now had a job in Vancouver and hadn't made it to Victoria in months to work on it. There was a prospect of him coming one weekend, and knowing his time would be limited, I decided to put a few hours into it myself and give it a jump start. I'd swap him his older version Electric Hubcap motor that he'd made for the 'modern' one in the shop, which I fitted onto the bike. Then I put an MC33033 board (working, at least!) into his motor controller. Lastly I turned down a motor axle to make a friction fit for the chain sprocket gear and pounded it on, then I put that shaft into the motor.

Motor (red) & controller (aluminum box) fitted on bike

   Tristan was only able to spend the day of Sunday the 19th, but we got it far enough along to try riding it. We found the aforementioned low low-speed torque wouldn't get it going except downslope.

Tristan riding his "e-Bike" in the yard (slightly downslope).
3 Ni-MH 12 volt batteries were strapped on the back for 36 volts, but the amps being drawn were quite low.

LED Lighting

   In May I ordered three very bright LED "emitters" meant for lighting. They arrived. I made a 3 volt, 4 watt battery operated plastic dome light and a 9 volt, 4 watt hallway globe light that I ran straight off a scrap 9 VDC, 450 mA power adapter, plugged into a light socket screw-in plug adapter in the globe light. 4 watts works out to about 23¢ of electricity per month. The hallway is much brighter than with the 7 watt CF light that was in there before. They were about as bright as 40 or 50 watt regular bulbs and were lovely white color that I'll call "ultra bright moonlight". I decided I'll gradually convert many of my house lights over and save lots of electricity, and have ordered some more. Around 15 watts of LEDs will probably replace a 100 watt bulb. (I've generally been using two 23 or 26 watt CF bulbs - about 50 watts. One 23 watt CF seems considerably dimmer than a 100 watt tungsten bulb to me. So 15 watts is 1/3 of CF watts for equivalent brightness.) At 15 or 20 $ per LED and needing two or three for bright light areas, the initial cost starts to add up - and with no satisfactory 120V LED installation system in place, so does the installation time. There are ready-made screw-in bulbs, which I'll also try out, but from the sounds of things, they won't have the brightness or be as satisfactory as the single emitters. There are also small triple emitters I can try, which a friend is using for his solar panel + NiFe battery DC power house lights. Of course, once installed, LED lights may well last longer than me and I'll never have to buy or change bulbs again.

4 watt LED globe light.
I like the diffuse light from the globe much better than small, intense points of light.

Turquoise Battery

   This month I decided to try Mn & Ni combo again for a "+", with the same nickel "-" as last month's 2 volt vanadium-nickel cell, and with the graphite-epoxy electrode glue as well as the "Diesel Kleen" conductivity raising technique.
   The electrodes seemed highly conductive, but cell conductivity was still poor. I used lots of graphite in the electrodes and in the glue, though I didn't use enough Diesel Kleen on one of the electrodes.
   I got another surprise: the cell was 1.5 volts instead of 2 volts.

I was assuming the voltages were:

Ni/Mn+ ... Mn-  2.3 volts (+1, -1.3)?, but high self discharge from the Mn- side)
V+ ... Ni-  2.1 volts (+1.1, -1)?

So this should have been +1, -1, but the total was apparently only 1.5.

Ni/Mn+ ... Ni-  1.5 volts

   Either the Mn was only charging up to MnO2 (though it was in fact made with KMnO4, the higher charged state), or the nickel "-" side was only -.5 volts... what I thought it should be prior to making the V-Ni cell. Did that mean the vanadium was in fact +1.6 volts and the nickel -.5 ? ... but if so, why did the vanadium seem to have no serious problems with self discharge?

   Near the end of the month, I started experimenting with different electrolytes to reduce internal resistance. Ammonium chloride didn't help... how about organics?
   I tried filling the cell with toluene instead of water, and the case started cracking to pieces. A whole corner fell right off! That was it at least until I have a new case.

Wind Power Cost at parity with fossil fuels in 2010

   Adding to the recent North Carolina report that in 2010 solar power became cheaper (for new plants) than nuclear power, this from Globalspec Newsletters: "According to the World Wind Energy Association (WWEA), average installed wind power costs are now on parity with fossil fuel plants." I don't know how this compares with nuclear, but the report continues: In 2010, China installed 19 GW of wind capacity, as did the rest of the world combined for a total of 38 GW. (Compare with BC's Site C Dam at .9 GW.) Many of the largest wind turbines are installed just offshore and the current plan - or dream - is to go from 120 meter blade diameters (5 MW) to 200 meters (20 MW). Offshore is safer and the noise won't bother people. But I wonder if the requisite number of windplants wouldn't cost less than the Site C dam and occupy less area than that to be flooded?
   I'm for it, but I must comment that if we're going offshore around here, 20 MW windplants will surely be harder to make than small, quiet 20 MW floating wave power units that won't kill flying birds. The wave power units, around Tofino, would also doubtless produce more energy per year.

Off-Topic Editorial Rant that really has no place here

   I noted some very disturbing related things this month.

   More and more we see those we elect to serve us passing laws and bylaws that curb individual rights and freedoms in one issue after another. They wish to try to control and regiment people rather than to empower them to make their own decisions and live their own lives as they choose. It's become so prevalent that things that would have brought storms of protest in past decades now pass without debate, without much comment, and without any consideration of the overall effects -- often just to solve one immediate small, perhaps emotional or popular, current issue. No attempt is made to educate to prevent mistakes -- easier just to make something illegal!
   As just one example, common chemicals I've wanted for battery and solar cell development are unavailable or restricted simply because (and this is just my best guess as to why, as no one seems to have bothered even to explain the rationale to the public) the medical profession thinks a few people may be using them unwisely for health purposes, and wants to be able to dictate their medical usage. But making something illegal or unavailable affects all its potential uses, not just health. Problem substances include among other things element #53, iodine, needed for electrolyte in DSSC solar cells, and potassium permanganate, the best positive battery electrode substance to give cells of the highest energy density (used as a poultice). But there are more and more examples of things restricting freedom to live one's own life besides substances - you can think of a few.

   Now freedom of speech along with freedom of the press appears to be under serious attack in the USA and anywhere they touch. It seems any violent act or smear campaign against any messenger who even might possibly bear unwelcome news is just fine. The great bastion of freedom is moving towards police state as well as bankruptcy.
   First I watched a Swedish documentary on youtube.com, WikiLeaks Rebels. In a featured "leak" was footage from a US military helicopter attacking and killing Reuters journalists on a street in Baghdad from the air (2007?). Then, when Julian Osange, founder of WikiLeaks, couldn't be located (he's been forced to live at 'no fixed address' for his own safety), he was charged with rape of two women - former girlfriends, neither of whom appeared to have initiated the action, wanted to pursue it, or were afraid of him - first to discredit his work by association, and no doubt "hopefully" to get the Swedes to arrest him. This follows another recent multi-million dollar smear campaign I heard of (though perhaps just a rumor?) by American drug companies to discredit Michael Moore, who had done an unflattering documentary on the drug industry. Before that I heard of a carefully prepared TV documentary news report, an exposé on an unsafe Monsanto product being injected into cattle everywhere, which didn't air because Monsanto threatened the network with suing them if they did. (This story too can be found on youtube.) Instead the previously much heralded "hard hitting news" stars who made the documentary ended up sacked. Then in mid June I read a news report wherein a CIA agent said someone from the Bush presidential office had asked him to "dig up dirt - anything" on a university professor who was publishing to the web quite a different view of the war in Iraq to that being put out by the White House, with reports and footage from first-hand sources including Arabic. The agent pointed out that it was illegal for the CIA to spy on American citizens, and reported the incident to his superiors. (Hurrah, integrity lives!) The White House spokesman then proceeded to find another CIA agent, and also made "improper" remarks about the professor without any evidence whatsoever. The professor, surprised, felt he was a pretty small fish to be gone after so hard... what pressure must be on the bigger players? (Those involved were all named in the news article, linked from a Lycos web search page, but I didn't write the names down.)

   The long and the short of it is, I suspect that killing the Reuters journalists (and several bystanders) was no mistake but cold blooded, calculated policy. Consider: independent journalists would be likely to report things "out of line" with Washington's "clean war" image (like that over 100,000 Iraqis were killed) - just like the professor they went after. The pilot first identified "RPG"s. Later he called them "AK-47"s. The camera equipment became not just "looks like weapons", but specific weapons, which surely look little alike, and still less like cameras. And no one on the ground was paying the slightest attention to the prowling helicopter - hardly likely for combatants openly shouldering weapons on the street. Was the pilot that stupid? Then the pilot urgently demanded authority to fire again on one of the journalists who wasn't dead and was painfully dragging himself along the ground down the street. When a good samaritan taking his kids to (?)kindergarten stopped to help and took the wounded man into his van, the van was shot, killing the man as well as the journalist. (The children miraculously survived. They will probably not think of America very fondly.) This would make no sense either: a badly wounded soldier is no threat, and shooting a rescue vehicle is everywhere considered reprehensible. A wounded journalist, on the other hand, if he lives will write a terrible story for public consumption: he is even more dangerous and must be finished off.
    With all that, I strongly suspect the pilots had orders, from the top, to shoot "unauthorized" journalists, and to report cameras as "weapons" in case their transmissions were overheard. The White House "demanded" that WikiLeaks.org return all the vast evidence of their crimes against humanity. (The whistleblower was a brave young intelligence officer who will probably pay a very heavy price.) They got just the unwanted publicity they were trying to stifle by force. The world needs an international police force, not a bullying nation operating in defiance of law and decency and without the sanction of the international community, for its own ends.

   American democracy (not to mention Canadian) is proving unable to evolve as needed to continue its serviceability to a free society - only ineffectual if not actively corrupt people seem able to win election, shutting out those truly fit to lead. Most emphatically, the "illiterate's X" voting system, which has never worked right, needs to be replaced with choice ranking. Australia and New Zealand have it - and probably most of the civilized universe. And Canada needs to let the people elect their PMs and Premiers, just as we elect our mayors - not allow them to be picked by the largest partisan faction ("party") in the legislature.
   The two mistakes give rise to a partisan faction ("party") controlled system where 'the leader' has all the power. Canada in particular comes closest to electing a dictator of any nation. Such "party" systems are open to corruption by rich vested interests. Decisions on public matters are made privately behind closed doors - mainly by the leader of the largest faction - and "rubber stamped" in the deliberative houses where they are supposed to be openly proposed by members, discussed and decided on. In April's federal election, while others had signs supporting individual candidates, I made my own political comments along these lines:

   I thought this could be educational not only for what was said but simply because most North Americans have never even seen a choice ranking ("STV") ballot. Ideas are scary for being unfamiliar when being discussed.

   A photo taken from a back row seat inside the US senate during a recent session showed the 'minority leader' speaking, while two senators ahead were playing cards on their computers, another was on Facebook, and a fourth was doing something else unrelated (texting?), accounting for all the people nearby in the view. My own member of Parliament Dr. Keith Martin - who was one of Parliament's leading movers and speakers - finally threw in the towel in frustration in the last election, saying the system is so dysfunctional no one can accomplish anything. Apparently many if not most of the people we elect to serve us and improve civilization feel just about as powerless as the rest of us.

   Quoting from a greater American statesman than those seen recently: "Those who make peaceful evolution impossible make violent revolution inevitable." - JFK.

   I trust it won't come to that. Change begins with each individual. Individuals are changed by faith in the first source and center of infinity - the Universal Father, trust that the universe is friendly, in recognition of the human family ("We all breathe the same air." - JFK again), and by education. The internet is helping people - presumably including future leaders - learn, by providing open access to previously obscure (or deliberately hidden) documents and materials, and providing many previously unthought-of solutions, of which WikiLeaks and OpenLeaks are just one example type. You can think of more. And the improving communication is bringing our extended human family together, as never before.

Electric Hubcap Motor System

   I got serious about motor controllers this month. After trying to get the A3938 controllers to work early in the month without success, I put together some MC33033 controllers. Tristan came over and we assembled his "electric hubcap" motorbike far enough to try and ride it. It wouldn't move except downslope, and it was only after a week and a dozen blown MOSFETs that I got it to move along level ground... and then another MOSFET blew. So, much of the month was unexpectedly converted to motor controller trials and tribulations, and the MC33033 (as built) proved much more troublesome than expected. I think (maybe) that over about 24 volts the spikes are making the MC33033 'huccup' and leave on for too long MOSFETs that are supposed to be off. It doesn't help that it leaves half of them on regardless of PWM.
   Before the end of the month, I finally realized I could ditch the motor controller chips entirely and use... an LM339 quad voltage comparator (timings, modulations and on/off gating) and a 4070 quad XOR gate (forward/reverse), with 3 'beefier' half-bridge MOSFET gate drivers than the present MC33033 board has. Yes, worked down to essentials, that seemed to be all a brushless motor controller really needs, and it would give the "current ramp" type of modulation I wanted from the A3938. On detailed considerations, the IR2130 triple MOSFET driver seemed like the best choice of drivers, and I could use its internal op-amp (with a couple of seemingly unrelated pins) and drop the LM339. On the IRF.com website I found that's been superseded by a new "IR2133", which, with some creative fanagalling, can do the job with just one other chip. (the quad XOR gate) So after a long run since spring 2009 I've come right around the loop from several general purpose chips to dedicated motor controller chips and back pretty much to my original chips -- but fewer of them, to make a better motor controller.

   The very long subsections on motor controllers written below, something of a 'diary' of events as they unfolded, give 'blow-by-blow' details of all this. Unless you're really into motor controllers, it's probably a good section to skip.

The End of the A3938 Motor Controller?

   The control method of the A3938, ending with a short fixed "off" period after a variable "on" period while the coil current "ramps up" to the set current level, may (possibly) be called "direct torque control" (DTC), "delta modulation", "delta-sigma modulation", or "chopper control". None of them seemed exactly applicable. (See Wikipedia "PWM" article.) I prefer to call it "current ramp control".

   I decided that if I blew the last A3938 chip, I'd go back to the MC33033 controllers until some future time. It developed that the "is being discontinued" notice only applied to the DIP package of the MC33033 (the one I was using), not the chip as a whole. Other packages remain available, so the circuit board would just need a little redesign. I'm glad the DIP version was there while I was troubleshooting the prototypes! Much of the trouble and time spent trying to get the A3938 to work has resulted from the miniscule packages, that are hard to place and solder and can't be socketed.

   I disconnected one pair of transistors from each motor drive output: Vbb, the output node, and the gates. This was intended to reduce the load on the A3938 gate drives from 200+ nano Coulombs to 100+ nC per the recently discovered "app note". In all the crowded wiring I was nervous about doing this, but I checked everything I could think of and taped off bare wires. What could go rong?
   I turned on the power and turned on the switch. I turned the motor a bit by hand and BLAM! Two mosfets and the 30 amp fuse were blown. But as I disassembled things I found out why: the voltage sense from one of the outputs to the board was soldered to the disconnected mosfets, so the voltage sense reading for that output was invalid.
   By this time I had removed the six drivers that were in use, so I reconnected the other six and rewired everything. Evidently the chip somehow wasn't blown in the carnage. However, the motor still wasn't getting power. Evidently it wasn't the gate drive timing delays that were the problem... so now I knew to look elsewhere.

   After getting some shoot-through currents, I had set the dead time between turn OFF of one mosfet and turn ON of another to the max, about 5 microseconds. On the oscilloscope I was seeing voltage spikes under 1 uSec, so I decided to try some compromise value.
   First, to limit potential damage, I put in three 1 ohm resistors in series with the three motor leeds to limit the current, and another 1/2 ohm in series with the main power.
   I tried 100 K ohms, which should reduce dead time to about 2 uSec. That didn't seem to help either. But on the scope I could now (finally) see a steady high level voltage on one of the motor supply wires. How could motor wires be high and low, and yet the motor still drew no current?
   I looked at the plug on the motor. I had made up a special test plug last month with thin (#18) wires  - again to limit destructive current for testing - and my memory is of carefully making sure the pins had clicked home into the sockets, but there it was: a pin not properly inserted, and apparently not making a connection.
   With that fixed, the motor slowly turned. It was working! Very soon I could smell hot resistors. It turned both directions, which meant that by considerable luck I probably had all three phase wires right. I removed the 1/2 ohm resistor from the power and tried again. The motor turned a little faster. My hand came near one of the one ohm resistors, and I could feel the heat, without even touching it. Obviously most of the power was being used heating them up instead of turning the motor.
   I removed the resistors. When I reconnected the power I thought I heard a "snap" noise from the circuit board. The power voltages from the chip indicated it was fried, evidently without even trying to run the motor (I tried anyway, to no avail). No evident rhyme nor reason. That was June 6th.

    That pretty much did it. I really hate to give up on a controller chip from which much was hoped, after so much investment of time and money, and that was running motors pretty well last fall when first tried. I finally remembered why I couldn't even duplicate those results: the initial time when it worked, the motor coils were connected with skinny alligator clip leeds (and so was the power IIRC). Those were something like the current limiting resistors, and I'm not sure it ever worked once wired "properly".
   Since it worked with resistance in the lines and not without it, perhaps I should try putting some resistances or RC filters in series with the output voltage sense inputs to the chip to limit current? (The app note boasts about how the high voltage section was implemented with a minimum of silicon - who cares? That means nothing beside reliable operation. The MC33033 boasts of "industrial ruggedness".)
   Another thing I noticed after I switched back to the MC33033 is that Motorola used an RC filter on the current sense line (which I duly copied) while the same raw, spikey current signal connects straight to the A3938 in Allegro's drawings. I did it their way without giving it any thought, but that line is certainly a potential source of chip-blowing transient voltages. Perhaps all four lines coming back from the motor coils to the A3938 should be filtered.
   Filtering the lines to reduce spikes to the chip might make it work. I may (or may not) get a few more A3938s for the 5 PC boards I already have. Or that might just save the chip and blow MOSFETs instead because of slower voltage sensing times!
   Mid-month I started thinking of using a microcontroller, which can implement similar control plus regenerative braking - and anything else I can think of.

A Revised MC33033 Motor Controller?

   After my last A3938 chip failed I put together a couple more MC33033 controllers just to have some working controllers, and for a while considered trying to "trick" them into doing the A3938's "short fixed off period" with a bit of additional hardware. Then I decided to just raise the PWM frequency a lot.

   The "current ramp" control technique of the A3938 is preferable for car drives. It's got superior low speed torque, lower power dissipation (less heat in the transistors), and a more familiar "gas pedal" control feel. But in the time I've spent troubleshooting I could have put together a few working controllers of the old type. A controller for the shop was certainly needed now, and one for Tristan's bike, the big Weel motor will need three more by itself, and to be able to offer reliable controllers known to work with the Hubcap motors would doubtless be a big asset for selling them. "Optimized" controllers would be even better.

   The big asset to the MC33033 controllers with the 3 separate IRS2003 half bridge MOSFET drivers was that I thought I had them working well.
   The biggest problem with them is that the main control and the current limiting control are separate. When the motor is stopped or turning slowly, the current will rise quickly. When the maximum current is hit, the PWM cycle shuts off to avoid exceeding it. But the cycle is no shorter, so at low speed the motor spends most of each cycle turned off, being supplied a low average current overall. Unless the oscillator frequency is set very high (which brings in needless switching losses and heat in the controller at all times) this limits bottom end torque, right where torque is needed most to get a vehicle to start rolling. This has doubtless been a factor in my torque converter failures, especially the last 'clock escapement' ones. In fact, in 2008 with my original multi-chip controller, the car moved across level pavement, however sluggishly, with the motor tied directly to the wheel.
   That feat hasn't been repeated since making the MC33033 controllers, presumably because of the low low-speed torque. And I suspect commercial controllers use this - "the usual" - PWM type control. This may help explain why some bigger motors or higher gear reductions are being used in electric conversions than seems to me should be necessary.

Tricking the MC33033 to do "current ramp" control?

   But could this unsatisfactory control be modified somehow on the MC33033 chip? The signal that says "the MOTOR ON part of the cycle is terminated" is unfortunately internal only, but this condition is indicated if all three low drive signals are off while running. With a 3 input CMOS (15 volt) NOR gate, a high "MOTOR Power OFF" signal could be generated. But once again, there's no external input to make the desired use of that signal. But it could be tied through a diode and a resistor to the RC (ramp oscillator) pin. The resistor would be much smaller than the regular oscillator pull-up resistor, which would cause the RC pin to rise quickly to the trigger threshold, rapidly (but not instantly) terminating the cycle once the motor had shut off, regardless of whether it shut off by PWM signal or overcurrent signal. The "short fixed OFF time" of the A3938 would become a "short variable OFF time" in the MC33033. The operation would become more akin, and varying the resistor would allow a range of characteristics.
   There's also a similar MC33035 with 24 pins instead of 20, but on checking it, it didn't seem like any of the extra signals would be of much use for implementing the changed operation.

Assembling Controllers

   On the 8th I got out an MC33033 controller board, which, when all wired up turned out to be the blown one (yes, I should have labelled it "blown" back when it blew), then the other one, which was rather hacked and took a while to put into shape. But within 2 or 3 hours I had it running a motor. (46 W at 500, and 93 W at 1000, no-load RPM.) I didn't try anything beyond basic operation.
   In the afternoon I got out a brand new MC33033 board and soldered all the components on. (I had to scrounge a couple off old boards to avoid a shopping trip or (worse) waiting for parts - seems I'm always out of something!) I hooked it to the power drivers for Tristan's Bike (a project much delayed since he's got a job in Port Alberni & Vancouver). By the end of the day it was ready to test and, if good, to run.
   In this one I bent up the "enable" pin (Hah! Try that with surface mount components!) and added a resistor and transistor in series with it to make the external pin "*enable", active low. (That way the motor will be OFF instead of ON if the operator controls plug isn't connected - planned for all future versions.)
   The next morning I tried it out. It worked great. ...from 1 working controller to 3 in a day.
   This time, the same motor seemed to start turning more snappily, and it read about 35 W at 500 RPM and 78 W at 1000 RPM, no-load. Not bad for a 5000+ watt motor, but evidently the power needed had quite a bit to do with the controller driving it. Perhaps it could be even better!

   What was the difference? I found that the first controller had had a 6x larger oscillator capacitor installed, lowering its PWM oscillator frequency proportionately. It shows what a difference can be made: if it had had the "short OFF time" to terminate the cycle after the power was cut, it should have worked as well as the higher frequency, perhaps even better - and indeed it would have been running at a higher frequency under these no-load conditions - but retaining the much reduced switching rates when high speed, high power was being asked for. Even if nothing else, this would reduce the heat in the controller when it was delivering high power. I switched it to the smaller capacitor (which also soon proved to be too big, in the first motorbike test).

Microcontroller Based Motor Controller

   Not to go way off on a tangent, as I thought about it towards the middle of the month, it seemed to me the A3938's problems were mainly with current and voltage spikes. The MC33033 only gets one of the problem signals itself, and specifies that it be filtered. The IRS2003 half bridge drivers seem to handle the rest of the nasty stuff well - it's their only job. So the key to reliable operation isn't really the MC33033, it's the little IRS2003's. At the same time, there were some problems and question marks cropping up in designing circuits to trick the MC33033 into operating a little differently.

   So perhaps it would be preferable to use a microcontroller chip. Then the control scheme, its adjustments and timings, and many other features, could be exactly as desired and tweaked in the software. Including regenerative braking, all implemented exactly as I like it best. Plus, any number of "dumb" statuses and analog signals could be reduced to a 3- or 4-pin serial interface on the controls plug/cable to the dash (eliminating the third header connector, and a second cable to the dash), and the controller could warn of motor overheating (even reduce max. power itself), etc. The IRS2003 driver parts of the PCB, and the heatsink assembly and mosfets, could remain the same.

   Though it's rather vexatious to start from somewhere near scratch (though far better armed with solid experience and knowledge) and with the need to write some exacting control software, I decided this would be the ideal way to go. I selected an ARM RISC chip, the SAM7S16. This might be thought "overkill" in processing power, but it should be easier to program in assembly language than some low performance CPU, and it's still just 6 $.

   This would mean not only a motor controller matched to the Electric Hubcap motors, but a truly superior, "ideal" controller that will squeeze maximum performance out of a motor at all RPMs, and maximum range out of the batteries. I wasn't, however, looking forward to all the study and work.

MC33033 Controller Again (yikes!)

   On yet further reflection, I decided something working was needed soonest, and the only obvious way to get working controllers soon was to use the MC33033, or the A3938 if it could be tweaked into working. The MC33033 at least worked, now, and evidently just needed a circuit board revision because the DIP size chip is no longer available. Raising the frequency would give it more low end torque, "workaround" solution or not.

   On the 20th I raised the bike controller's frequency up about 15x to the "recommended" 25 KHz, and the controller promptly burned out just as the bike wheel started turning. That's why I hate experimenting with the controllers -- if it's working at all, leave well enough alone! Evidently there was still a problem with the controller after all, which only became troublesome at high PWM frequency.

   Just two mosfets were blown and I soon had the controller fixed. But what was the problem?
   Was the filter on the current sense line too well filtered, and sensing of the currents too delayed? I changed the .1 uF filter to .022 uF, and put in a current sense "resistor" (thinner wire) that should cause it to trip off at around 30 amps. It seemed to work well at the high frequency at 24 volts, with currents around 20-25 amps, occasionally 30 and a 40 and once 60 amps reading. That 20-25 amps at 24 volts also seemed to be just enough 'juice' to propel the bike at a walking pace with no one on it. To ride it (besides downslope), higher current or higher voltage was needed.
   But currents went wild at 36 volts. On the bike with the current probe attached and the meter bungied to the handlebars I saw 80, 160 and 140 amps flash by, instead of 20-30. There, then, was the trouble causing the blowouts.
   There was the familiar cracking sound, and then another before I could pull the battery jumper leed off. (No breaker-switch yet.) The ultra-reliable 40 amp fuse remained in perfect shape throughout. (Thanks a lot!) But this one time, the cracking sounds had come from behind, at the batteries, and it turned out nothing was wrong with the controller.
   I put the .1 uF filter capacitor back in, and at 24 volts everything was still fine. I got out the 5 volt battery, and at 29 volts, the currents went wild again. If the cause can be determined and fixed, I think there's controllers that will work acceptably, otherwise not.

   But what was the trouble? Switching shoot-through currents come to mind - they're not amenable to being controlled, and MOSFET turn-off time probably increases with voltage. At the lower PWM frequencies, shoot-through current might be occurring -- but so much less frequently it makes little heat and does no damage. The IRS2003 has a fixed dead time of 1/2 a microsecond. It seems to me I had to set the dead time higher than that on the A3938 to eliminate shoot-through currents. What could be done with a fixed time chip? One technique that's been used is to put an ultra-fast diode in parallel with the gate resistor to speed up turn off. But that effect will be minor. Rechecking its specs, I discovered (rediscovered) that the fixed dead time of the IR2130 triple MOS gate driver that I used in my original controller was 2.5 uSec - 5 times longer! Here, then was a big difference... IF that's what the problem was.

   One thing I thought of to help prevent damage during tests: limit the current by using a single bank of NiMH "D" cells in series, whatever the voltage. They're rated for 30 amps continuous or 50 amps intermittent, and the voltage drops off rapidly above 70 or 80 amps. This is a good current capacity range for the tests, though the 1200 uF of supply filter capacitors will supply more for brief periods.
   I made the battery pack, but it gave me the thought that perhaps the filter capacitors were insufficient. The rapid switching might confuse the current probe? Maybe there were no real very high currents? After all, the 40 and then 30 amp fuses hadn't blown.
   I found some 470uF, 200V caps and put in about 1400uF instead of 1200. (I had to remove some of the 100uFs to fit the 470s.) Blam! More transistors blew. This time they took out the 30 amp fuse. So much for "the currents weren't real."

Causes of the problems?

Problem 1 - A3938 Controller: transient voltage rating

   Most chip inputs don't want to see voltages much above the supply or below ground. I knew special precautions were taken for the motor power signals - IR's half bridges were all "negative transient tolerant". But the A3938's "absolute maximum" for negative voltage coming in from the motor phases was only -5 volts... and I found no mention of spike tolerance. The spikes are 10 to 12 volts. Ouch -- negative 12 volt spikes might well be what was blowing the A3938s! The belatedly found app notes spoke of motors only "up to 1 HP".

Problem 2 - IRS2003 half bridge: too short fixed dead times?

   Possible the .5 uSec fixed dead time is too short - at least with the long switching times enforced by their low gate drive current. One could wish these had the 2.5 uSec of the IR2130, but ALL IR's many 8 pin half bridges have a similar .5 uSec fixed dead time - maddening! Shoot-through currents could be why MOSFETs keep on blowing. At low PWM frequency, the MOSFETs could recover, but at high frequency, they overheat fast. (I wish I had a spare motor for bench tests. That should be the next project. I have the parts and it would only take a day or two.)
   I'll have to go to the 14 pin drivers with resistor programmable dead times.
   Counting the resistor, that's double the connections, 16 wires instead of 8! Or I could go back to the 28 pin IR2130 triple MOS driver. But that chip has separate high and low inputs, and no "output enable". Two extra logic chips are needed to recombine the inputs and to allow modulation, so it saves nothing.

Problem 3 - MC33033: Low drivers stay turned "on".

   The MC33033 only modulates the 'high' side drivers: the 'low' side (as I use them) stays on. That's okay if everything is fine, but it'll cause extra trouble if there's a problem. I started thinking about using an external modulation. If the MC33033 "output enable" was used to accommodate an external modulation signal, as well as gating on/off (with the PWM voltage held high and ignoring the MC33033's PWM), it would turn both high and low drivers on and off. And with external modulation, by LM339 or whatever, it could run the way the A3938's does, with variable current limiting and a short fixed off period.

   Hmm... was the MC33033 even needed for such a system? Some models of half bridges could perhaps be modulated directly. Ah, the vast array of possibilities!

Simplify, simplify, simplify!

   By June 25/26 I thought a big part of the key is in the 3 half-bridge MOS gate drivers.    I'd get the type with one input for high/low select, and another for "shutdown" (SD), AKA "output enable" or "on/off".
   For the simplest brushless motor controller: connect the hall sensor directly to "in" for each phase, and a modulator to "SD" of all three, through an "on/off" switch. The simplest modulator is a 555 timer for a PWM generator. Total chips: one 555 and three half bridges.
   Or instead of PWM, one could use an LM339 quad comparator and use it to do "current ramp" modulation, which compares the current against the reference voltage from the control. (I haven't worked out a specific circuit, but I'm pretty sure the LM339 has it in it somewhere.) Total chips: one LM339 and three half-bridges. This also (inherently) protects the motor and transistors against overcurrent to the motor, like the A3938. This is complete: one comparator does the current sensing, one does the fixed "off" time, and one interfaces the "on/off" switch.

   Those were getting really simple! Unfortunately, I overlooked that there's a band where they are supposed to be "off". More inputs are needed to accommodate that. Again unfortunately, the IR2110 has the required inputs -- but latches the SD input so the drive can be turned off, but it can't be turned on again until the motor has rotated to the next position -- which it won't do as it has no power because SD is latched off. All these clever ways to make chips that don't quite work!

IR2130 (IR2133) Motor Controller - back to 2008!

   I finally decided that the best solution (of several unsatisfactory choices) would be to use the IR2130 triple gate driver (2.5 uSec fixed dead time - 5 times longer) and some "standard" logic and interface chips -- about the same as I'd been using before I ever heard of the MC33033 or other single chip brushless motor controller chips! This time however, I would make it with 'current ramp' control. It seemed it could be done with 4 chips, counting the IR2130.
   Then I looked closely at the IR2130's internal current amp and overcurrent shutdown. It occurred to me that since a triple external gate was already needed anyway, I could 'trick' the overcurrent shutdown into unlatching. Furthermore, the current amp could be used to compare the control ('gas pedal') pot with the current, eliminating the need for the LM339. Also there was a built-in 9 uSec delay turning back on after overcurrent: the "short fixed off period", built right in! Now the IR2130 controller could be made with just 3 chips: the IR2130, a 4011 quad NAND gate and a 4070 quad XOR gate. (If the motor didn't need to reverse, just the IR2130 and the NAND gate - 2 chips!)

   On the 29th I check at IRF.com, and found that the IR2130 has been replaced by a new chip, the IR2133. It had a "FaultClear" input and a "ShutDown" input. Together these allowed elimination of the 4011 quad "NAND" gate (or 4071 quad "AND" gate), reducing the chip count to just one IR2133 and one 4070.

The IR2130 (or else 3 half-bridges) would be needed even for a microcontroller "solution", so that would be 2 chips, and the microcontroller would have many more pins than the 4070. Overall the circuit complexity would be about the same or a bit more complex -- and it would have to be programmed.
   The advanced but non-essential features that the microcontroller could bring can wait. I think simple logic dictates that simple logic is the best way to get controllers working ASAP, and it'll have the main feature: the good current ramp control.

...all assuming I can get them to work well however it's done!

Motor System Thinness

   When I see the motor on the car, I realize that the less it sticks out from the wheel, the more acceptable it'll be. When the first version moved the car with the motor rotor bolted directly to the car wheel, it was quite short. Now, a "regular" configuration motor, a torque converter, and a flexible coupling to the wheel, each add their bit of width. As I measure the pieces of the present motor with the disk brake type magnet rotor and the torque converter rotors, it's about 7". With the flat rotors, that'll reduce to 6".

Top: CNC waterjet cut bearing holders for the ends of the motors, each held to the plate by a thin tab.
Below: new flat plate magnet rotor, in a thin composite motor rotor compartment shell.
(The gray color of the steel rotor plate is polyester powder coating.)

   Some bolt heads stick out. If I switch the ones on the shaft side of both the motor and the torque converter input rotor to flat-head bolts, the torque converter rotor (an inch thick) can be almost touching the motor body, saving about 3/8" - so 5-5/8". I think that'll make it protrude less than than the right side rear view mirror.
   Beyond that, there's making the walls and pieces thinner, but this is probably inadvisable and for small returns.

   Here we see an advantage to my previous design with both shaft bearings on the stator, no inside side wall, and the motor rotor doubling as the torque converter input rotor: it saves around an inch of thickness. But the end of the motor has to be open. The shaft of the enclosed motor is mechanically stiffer with the greater distance between the bearings. A combo configuration remains an option, but is probably best avoided without a finished torque converter design. Once there is one, the idea can be re-examined, again using the magnet rotor as the inner torque converter element.
   A possibility for an enclosed motor-converter would be a concentric axle system: a hollow pipe torque converter output shaft would surround the motor shaft at the wheel end. All the components would be enclosed and only the output shaft (with a wheel lug-nut drive plate attached) would appear outside. The motor casing would be thicker by the width of the converter drum.

   One other possibility for a small reduction is to use thinner supermagnets: 3/8" instead of 1/2". The 1/2" ones are more than strong enough, and the flux gap is also over 1/2". The steel magnet rotor thickness might well be reduced from 5/16" to 1/4". If using 3/8" magnets also reduced the best flux gaps by 1/8", total width savings would be about 5/16" -- and presumably the the supermagnets would cost less (and there'd be some reduction in their potential to do damage or injury during handling). But it may be found that the lesser depth of field of the thinner magnets reduces motor performance. In that case, thinner magnets might be better matched to thinner coils, eg 3/4" instead of 1". That would save even more width, but the whole motor would be less powerful.
   I have high grade (N42) 3/8" magnets and the thinner rotor, and will give it a try when I have time, comparing one motor's performance with the two different rotors installed. (However, so far I haven't even done complete tests on even one motor since improving them! I still need some sort of mechanical motor test load.)

Lower Power "Electric Hubcap" Motors?

   The above thoughts on thinner magnets bring the idea of a lower powered motor of the same design with the same ~95% efficiency, with thinner coils and magnets. For example: 1/2 inch thick coil cores, 1/4" thick supermagnets, for a 2.5 KW motor, still 2000 RPM. Everything including the body parts would be a little thinner (use less PP & epoxy each), making a motor of the same diameter but only perhaps 2.25 to 2.5" thick instead of 3.5".
   These wouldn't be for cars, but eg for bicycles and cordless lawnmowers. For a mower, 2.5 KW is still considerably more power than can be had from a 120 V, 15 A house circuit (120 V * 12.5 A = 1.5 KW). The 1.5 KW mowers are generally underpowered for tall or damp grass. 2.5 KW, with higher efficiency, should make the difference, and the low voltage (and no cord) eliminates shock hazard. (Now, just need those better batteries...)
   Much as I'd like to try out this mower (with some NiMH batteries?), I'd need to order the thinner magnets and coil cores, and I don't know when I'd find time to build it. And it wouldn't cost a whole lot less than the 5 KW version.

   Another option would be to use six 'regular' coils and eight 'regular' magnets, and make a smaller diameter, 3000 RPM, 3.3 KW motor. But then new molds for the PP-epoxy body parts would be needed, so that would take even more setting up.

   Such motors would be pretty simple to develop at this point and might be very useful, but would be just one more sideline from the main project, hybridizing cars.

Metal Protection

   Protecting the steel of the magnet rotors and bearing holders having entered my thoughts, and the powder coating being very nice but too costly, I looked around for a good alternative. I thought perhaps there might be a spray of two-part thermosetting epoxy that could be simply sprayed on, and then set in an oven, essentially working much like the dry powder type.

   Instead, at Industrial Plastics, I found a spray can of "Cold Galvanizing Compound", "97% zinc". Evidently with "ketone peroxide" (MEKP?) in the spray, the zinc powder "bonds with the steel" to form a galvanized coating. As a plus, even if it's scratched, the nearby zinc becomes a 'sacrificial anode' and prevents the steel from rusting, or at least, prevents the rust from spreading underneath the coating. This sounded like a reliable bond and about the best possible protection. The guy at the powder coating place had mentioned zinc as being their primer coat, too.

   I had been impressed by the prep steps for the powder coating. There's degreasing and then cleaning the metal to a shine before coating. I degreased a flat magnet rotor by heating the surface with the hot propane torch. I thought it was pretty clean to start with, but one side in particular seemed to become wet everywhere as it was heated, and this would finally boil off. Not much happened on the other side, though a few small patches changed color a bit.
   I thought it would be a snap to shine up the metal after that with a portable belt sander, but it took two belts (they broke after a while) to do one side. There was some blue-beigey color coated on it that was very difficult to get rid of. Considering it didn't come off with the torching, I'm wondering if it was some sort of metallic coating already, and if perhaps removing it is pointless. I decided that was probably the case and finally sprayed the other "un-silvered" side too, and then a pair of bearing holder pieces, but I plan to ask at a metal shop.
   The zinc finish looked rather rough and not uniform - like it was formed from powder and sprayed by hand - but good.

Two rotors: (L) powder coated polyester, (R) "cold galvanizing compound" (zinc powder) spray can.

   A couple of spots have scratched off on some bearing holder "washers". I'm thinking of trying out some sort of heat treatment to fuse the zinc to the steel better - oven, kiln, torch -- ?

Thinner Motor with Flat Magnet Rotor

   I gave the motor in the shop over to the motorbike project, so I needed a new one to continue the torque converter project. Now that I have the flat magnet rotors, the obvious thing to do was to put a motor together with one, making one of the planned 3.5" thinnest motors that I'd already molded two bodies for. I had all the other parts except one bearing holder set, and assembled motor coils that were handy, removed as a group from an open format motor. This was the "first try" set with the rutile instead of the superior ilmenite. As the sodium silicate was unbaked, I could easily wash the rutile off and redo it with ilmenite. This time I'd try sanding the wire a bit to roughen things up, as the rutile was flaking off badly. Five sets of bearing holders were on order at Victoria Waterjet. It boiled down to putting together the magnet rotor. I did that on the 16th, and the same day the bearing holders - 10 sets - were ready.
    I lightly sanded the magnet side of the polyester powder coated rotor. It was very smooth and I was just slightly nervous about how the epoxy would adhere. I think it would have been fine. (Pictures of this rotor finished and unfinished are above.)
   Then I got into motor controllers again and didn't finish making this motor. It's mostly done and won't take long now. When the rutile coil coatings are replaced with ilmenite, it can be assembled.

Electric Weel Motor

   I wasn't looking forward to making a mold for the 27.5" stator rings for this huge motor, or to the epoxying job to fill that mold. And the 27-1/2" ring wouldn't fit in the oven to rapidly set the epoxy like the smaller ones. Then I had the idea - at least for the prototype - of cutting the rings out of 1/4" plywood, and then simply painting on a couple of layers of polypropylene fabric with epoxy resin to stiffen it. Since it was just 5" wide rings fastened to a metal center, that should be strong enough. I calculated the two rings could just be made from 1/2 a sheet of plywood (4' x 4'). That worked, however the 1/2 sheet of plywood cost only a little less than a full sheet at Rona.
   That just left the rotor end wall. That was going to need more strength, so I might just make it of 1/2" or 3/4" plywood, again PP-epoxy coated. (Although, being behind the magnet rotor rather than in front, it could probably be made of metal without causing drag. Wood or composite will be lighter.)
   I'm not sure this can be made under 100 pounds as I hope: the 27 coils alone weigh almost 25 pounds.

Weel Motor pieces:
- Stator" 18" steel plate with 17" - 27.5" plywood outer ring (a 2nd ring goes over the 27 coils)
- 26" rotor, 2 pieces tack welded together, with magnet positions were layed out by the CNC waterjet.

Torque Converter Project

   I dropped by a recycling place and found a sheet of 1/8" aluminum for the two input rotor pieces, and perhaps for the outer rim of an output rotor. I cut them out with a jig saw, ~11.4" diameter, then fitted them to the SDS coupling. Everything was rough hand work because they were too big for the lathe. (Again I need that pulsejet cutter that I didn't finish making, fitted to the CNC machine!)
   I cut some 5/8" thick pieces to bolt across between the two, making the rotor 7/8" thick. The center holes were sized to fit it onto an SDS bushing, which would compress onto the motor axle.
   I thought the plastic pivoting wedge pieces I'd cut were about the right size for the first try, but that the pivot hole should be a little higher up. On the 15th I redid them, this time in aluminum. I then decided I'd put just one on, and cut one square "slot" in the output drum, and see how that worked.

Wedge with spring on input rotor (shown on single disk of double rotor).
The snakey tempered brass spring is much stiffer than you might expect.
(Say... perhaps it should be redone as a spiral coil - the traditional spring shape.)

   A complication arose when committed the motor in the shop to Tristan's bike project. I was then without one for the converter until I put another one together. Then finding that low low-RPM torque from the motor controllers was a big part of the problem getting torque converters to move the car, I diverted my attention to trying to get better performance out of the controllers, and finally to designing a new controller.

   Say... since I have the motorbike for the moment, perhaps it would be possible to test torque converters on that instead of on the car. The torque output of various converters and adjustments that wouldn't quite move a car might be better evaluated and compared - does 'A' or 'B' climb the steeper hill faster?

   On the 29th a mechanically minded friend dropped in and we talked about the converters. He mentioned centrifugal clutches from washing machines. I got the idea that instead of pivoting sideways as above, the wedge might be hinged out on a sideways 'arm' so it would drop pretty much straight 'down' into the slots in the rim of the drum, perhaps as per this sketch:

Centrifugal wedge torque converter.

   That would probably have better force of torque hits than what I've started making. It could have both a spring and weight. The mass would increase the force of each hit as the motor RPM went up. It would really rumble as the car started moving. At some point of higher speed and low torque, it might well lock the motor and output rotor together for silent higher speed cruising.
   There are some adjustments to ponder: The wedge has to have angles both directions to allow forward and reverse travel. The point of the wedge would start to fall into each slot as it arrived, pushed outwards by the spring and its mass. If the slots were wide enough and the wedge angles too shallow, the wedge would start pressing on the trailing edge as it went in, a retrograde force, but in a narrow enough slot, it wouldn't be falling fast enough yet when the leading edge of the wedge hit the far side, and only propulsive forces would be generated (to the drum rim) as the falling wedge was suddenly forced to rise again. On the other hand, the faster it's dropping when it strikes the slot edge, the more force the hit will have.

LED Lighting Project

   Okay, it's not the most amazing or exciting research project -- but it is brilliant, dazzling! No new inventions, but I'm exploring practical ways to use these devices. Perhaps commercial possibilities lurk therein, or improvements for hybridizing cars.
   In fact, by the time I'd done a couple of lights, I started realizing that what's holding back LED usage for lighting isn't so much the LED cost - that'll come down when usage goes up - as the need for more practical and more attractive lighting products based on them. Few people are going to try to do wiring and to figure out electrical current supplies, heatsinks, and light diffusers and fixtures from scratch, and such bulbs as exist seem to emit nasty, bright pinpricks of light. Especially, they need good diffusers.

   In May I ordered three different bright light emitting diodes (LEDs), for about $10 each. They arrived on June 9th. These "emitters" are literally blindingly bright and can damage your retinas - one reviewer described them as being "like welding arcs". (Somewhat like some of the newer "always glaring" car headlights?)
   I got mine from "DealExtreme.com" (Hong Kong). A solar power/LED lighting enthusiast mentioned another source, "LedSupply.com" (USA). (Any economic Canadian sources?)

   I tried them out clipped to the regulated, current limited power supply, clamped in a vise and facing away from me. It was after dark. I put one up by the window and went outside. I could see the illuminated window area. Then I got far enough back that I could see the LED emitter itself. WOW it was a brilliant point of light, too dazzling to look at even from way out at the street!

   The next day I went to Canadian Tire to find frosted fixtures to put them in. I found some battery powered plastic dome lights for $5 and bought two. You press the front dome to turn it on or off. I mounted the 10 watt, 1000 lumens emitter in one on a piece of aluminum for a heat sink:

Left: LED emitter (DealExtreme.com #51989. XMLAWT-0-1A0-T60-00-0001 ...?) installed on a heat sink cut from a scrap piece of aluminum. The #4-40 machine screws go through (a) slots at the edge of the LED base (b) the aluminum sheet (threaded) and (c) thread into the plastic. Observe battery polarity when soldering the wires.
Right: the unmodified light with frosted tungsten (incandescent) "flashlight" bulb at center.

   For battery power, three NiMH AA cells give more than the required 3.0 to 3.5 volts. So I put in a resistor in place of the fourth cell to limit the current. LEDs need a specific current rather than a specific voltage, and either the voltage drops until the current is okay, or the LED burns out. I chose .5 ohms because it was the lowest value resistor I had. I bent the leeds around so it would make the connections in place of the fourth cell without falling off:

Battery/resistor arrangement. The bottom sides of the aluminum heat sink (center, bright) and the machine screws can be seen.

   All that was left to do after that was turn it on. The voltage across the resistor turned out to be .5 volts, so the LED was drawing one amp. The three cells made a battery of about 4 volts, so the power was 4 watts. (1/2 a watt went into heating the resistor, so the LED was actually only using 3.5 watts - about 1/3 of its max power.) With a smaller resistor, the current would be increased. Since it's supposed to be a 10 watt LED, the limit would be about 3 amps, at which point it should be putting out around 1000 lumens of light - more than a regular 60 watt bulb (~850 lumens). But it's non-linear: at 3.5 watts, it probably puts out 400+ lumens. Here's how it fared at 4 watts (apologies for the overexposed bulbs):

15 Watt compact fluorescent bulb in glass frosted shade versus 4 watt LED light in translucent plastic dome shade. Even at 4 watts, the LED light is so bright that the whole lamp assembly glows - the sides and even the back as well as the front dome. The CF bulb was rather orange. The LED was brilliant white. The translucent plastic does a very good job of diffusing the "much too bright" point source of light.

Left: 60 watt incandescent bulb (850 lumens) in the frosted glass shade
Center: 4 watt LED
Right: The unmodified plastic fixture with the original ~3 watts incandescent bulb. (Yes, it is on!)
Rear: Daylight!

   The dazzling LED is using only slightly more power than the feeble glow of the original tungsten bulb unit.

   Note: The LED is radiating light only 180 degrees, towards the camera, while the incandescent and CF bulbs are sending it all directions, which makes the LED seem brighter by comparison. The LED could be made three times as bright, but I hear they do actually burn out in a few years if run at full power, whereas they'll last for ages at 1/2 or 1/3 power, and they're actually most efficient in that range. At 1/2 power, they give somewhat more than 1/2 as much light as at full rated power. And they won't get very hot.

   Even at 1/3 power, the AA batteries will only last a couple of hours before they need recharging. It seems odd that the makers chose to use 4 AA cells rather than 2 D cells, which would last twice as long. But lucky for this project, since 3 cells are required for the LED.

   This particular light would be a good car/camper dome light, closet light, trouble light, or a light for a shed or garage without electrical wiring, for about $20 counting the batteries.
   Since it's portable, I tried it out in an outdoor shed and in various rooms in the house. Even in rooms with a 100 watt overhead light, this extra 4 watt unit made a substantial contribution to both the quality of the light and the brightness of the room, and in a hallway with 28 watts of CFs, it was about half the light. I like the light. I didn't expect to, but now I want to add it wherever I want bright, working light. Perhaps the cool quality could be described as "ultra-bright moonlight". But there are various flavours to try out.

   My chief conclusion so far is that LED lighting can give you nice light using very little electricity. If it's not bright enough, turn them up; add more; add varying colours - the electricity will still be almost nothing. In the winter certain incandescent bulbs help heat my chilly house, providing some light with the heat on those long winter nights, but I think it would be very nice to switch all the ones where that's not a consideration from tungsten, or from CF when and as these mercury bearing bulbs burn out. (AFAICT, CFs don't really seem to last longer than incandescents.)
   For electric transport, all LED lights would make the range running at night with all the lights on virtually no different from daytime running with them all off.

LED Lighting with 120 VAC Power the Simple Way

   For home use, changing batteries in lights will wear thin quickly as a regular operation, even if they can be recharged indefinitely. In addition, slip-in battery connections aren't very reliable, and in these lights in particular the cells fit too tightly and the springs can't push them together to ensure any connection - after a while the LED light would go bright and dim, flicker, even shut off. It worked more reliably after I chopped some of the offending plastic supports from the battery compartment with sidecutters to loosen things up.

   I wanted to find a simple way to power LED lights from 120 volts AC to replace household lights. For the next experiment I wanted to try various stray power adapters left over from long obsolete equipment, and find emitters that work well at their typical voltages, eg 9 VDC, 12 VDC... (or put, eg, three or four 3-volt emitters in series.) The power adapters could be plugged into those screw-in adapters that turn light sockets into plugs, in order to put plug-in LED lights where lights are supposed to go. (What a devious, twisted plot!)

   I took the 9 volt, 10 watt, 500 lumens white-yellow LED (Great product, DealExtreme #05876 says it's 12 volts - buyer beware!), and found an old 9 VDC, 450mA power adapter. I found the power adapter was putting out 13.8 volts with no load. I used two 1 ohm resistors in series to limit the current, and the LED drew 370 mA. I cut that down to one ohm, and it went up to 420 mA. With no resistor at all it went up to 440mA, and the voltage was about 9.1 volts. That's 4 watts - a little under 1/2 power (good for longevity) -- and the full power capacity of the power adapter. In essence, the tendency of the adapter's voltage to drop with increasing load made it a great current limited LED power supply well matched to the ~9V LED. I have a whole box full of various adapters that people have given me over the years. (even after throwing a bunch out once and giving some away. Once I saw a used computer store with a good selection, too, or perhaps any secondhand store would have some.) Now many of them can be put to productive use, saving the purchase of many LED drivers!

   I had more than one idea about what light to put it in, but a frosted globe light in the upstairs hallway looked like the simplest project - It had lots of room in it.

Bare LED emitter bolted to some scrap aluminum for a heat sink, strapped to 9 V power adapter,
power adapter leeds soldered to emitter's side lugs.

Screw-in light socket power plug adapter (Rona 3 $) --- LED - Power Adapter - Heatsink assembly plugged in.

Finished LED light with globe washed and remounted: very nice hallway light, only 4 watts, for ~14 $.
Around 250 lumens of "super bright moonlight" shines downwards, but reflections from globe light ceiling somewhat. (Unfortunately, with this ceiling.)

If it was never turned off, it would use about 23¢ of electricity per month
(instead of $2.30 for 40 W bulb - 100 W bulb is 6 $/mo, @ 8 ¢/KWH).
(plus however much the transformer is wasting - which might as much as double it. So what?)
Inside, the heatsink gets quite warm; outside, the globe and its metal top remain cold to the touch.

   I did a smaller dome light (Rona, 2 AA cells, 4.5 $) with the third LED and a 4 volt, 300mA power adapter, but it doesn't seem very bright. It needs more current. I also decided I prefer the larger domes - better still, a globe.

   Not all the lights will be this easy, but now I want to convert lots more lights! Kitchen - bathroom - livingroom - shop - garage ... If installations don't get ugly or time consuming, it appears that every dollar invested will pay for itself in electricity saved in under a year of lit time, and light bulb replacement costs thereafter are virtually eliminated. The light is pleasant and they have no mercury vapour or other hazardous wastes.
   I'll start with the lights that are turned on most often. First, I think the 100 watt bulb over the kitchen sink (a 26 W CF wasn't bright enough) should become about four 4 or 5 W LEDs in small plastic domes, spread out more above the counter area. Or else in a globe. Perhaps four 3 volters (like the first light above) in series powered from one doorbell transformer. (Somebody gave me some of those, too.) Then perhaps another transformer and 4 more LEDs, mounted in the existing dome for the central ceiling light of the kitchen.

   Even more ideally, a solar panel or two on the roof, with NiMH, NiFe, or new chemistry batteries to store the electricity, and some low voltage DC wiring for lights. It all pretty much runs for free once installed. Someone tells me Jack Layton lives almost "off the grid" in Toronto. Wow, a politician who actually does cool things!


   I thought that would be it for this month, but I went into Industrial Plastics to buy a second spray can of "cold galvanizing compound" (13$) and more epoxy resin (150$), and found an LED display in their marine department. (Capital Iron has one too.) I found among the lights one made for my car's interior light for 12$. I bought it and put it in. It was just .9 watts, but much brighter than the old incandescent. Gosh, it would seem I can read street names on the map in the car at night again! (I haven't been able to do that for many years.) It said it was good for 10 to 30 volts - all the driving circuits were built in on a tiny PCB. Any that were 3 or 4 watts (the max), were 30$ and up. These were all DC lights for boat and camper... or car. I think there's a separate and poorly served market for nice LED lighting products for household 120 VAC - permanent fixtures and lamps. I ordered another dozen of the 3v, 10w, 1000 lumens LEDs like the one in the first battery dome light. They should arrive some time after the postal workers lockout ends.

Turquoise Battery Project

   It appeared that the one remaining requirement for properly working batteries was better conductivity. Two promising and complementary techniques looked like they'd do the job: compacting electrodes with the Diesel Kleen to increase the electrode's internal conductivity, and gluing the terminal posts to the graphite sheets with mixed epoxy resin and graphite powder to improve the connection to the terminal post. (And these seem to represent the first important advances to this area of salt-electrolyte battery making arts since the 1880s.)
   Another consideration was that the vanadium used in May's battery was a somewhat costly substance, and poisonous. It seemed a mix of nickel and manganese would give about the same voltage and at least similar amp-hours for a substantially lower cost. Manganese is 'dirt cheap', so depending largely on the price of nickel, these batteries could be potentially cheaper than lead-acid.

   All these items were to be tried out in the next cell. But with all the other projects that needed work, and a couple of days' distraction with LED lighting, I didn't manage to get to the batteries until the 18th except to finish putting a case together. On the 18th I made the two electrodes and glued the positive into the bottom with the "grafpoxy". (I made it, I can name it...)
   The "Diesel Kleen"-ed electrodes smelled awful drying in the oven, 80 minutes each. I do wonder exactly what's happening with that. My best guess is that the mix dissolves the graphite powder, because it's good stuff to clean off the counter of carbon/tar (hydrocarbons)/graphite stuff, that usually only scouring with abrasive cleansers will clean. Whatever it does, the electrodes seem to conduct better than without it, and have that metallic sheen, where standard dry cell positive electrodes just look like porous "black sandstone".

One electrode had too much 'diesel kleen' and material oozed out during compaction.
Here is some of that 'metallic looking' carbon-manganese-nickel material compared with coarse 'sandstone'
carbon-managnese mix from a standard MnZn dry cell. The shiny material has lower resistance.

   Cell Internal Components


* Graphite Sheet

* Briquette:
15g monel mix (instead of just Ni(OH)2)
5 g KMnO4
14g graphite powder
1.5g Sunlight dishsoap
.2 g Sb4O6
3.5g Diesel Kleen

* Blended with mortar and pestle. (resistances read around 8-9 ohms)
* Compacted (lost 13.8g - oozed out. Made 3.5-4mm thick trode, 15-20 Ω)
* Painted on calcium hydroxide layer
* Baked in oven 80' at 110ºc
* torched


Microporous cellophane painted with acetaldehyde with osmium dopant
Arches watercolor paper painted with zircon


* Graphite sheet

* Briquette:
20 g monel mix
13 g graphite powder
1.5g Sunlight
.2g Sb4O6
2.9g Diesel Kleen

*****Same processes as above (8-11 Ω, none oozed out, electrode 4.5-5.5mm thick)

   Cell Assembly

Positive Electrode:

Conductive Epoxy: 3.6g epoxy resin + 4.7g graphite powder

1. Spread the conductive epoxy on the bottom of the case (careful not to get any on the sides).
2. put in the + electrode. Slide it around to spread the glue.
3. Insert the terminal post to touch electrode's graphite backing sheet, epoxy filling gaps.
4. Allow to set. Resistances, surface to post, measured 8-15Ω.

Electrode backed with 'grafpoxy' coating & glue for terminal post.

I didn't use enough 'diesel kleen' - only small areas are the silver color and it fell apart.

   In spite of having seemingly very low resistance electrodes this time, the internal resistance of the battery was as high as ever if not worse. It would seem the problem lay elsewere than in conductivity through the electrodes and to the terminal posts. Could the ion flow be poor? Perhaps KCl wasn't as good an electrolyte as I thought?
   I added some ammonium chloride - the same stuff as in a standard dry cell. That didn't seem to do much, though the cell smelled faintly of ammonia next time it was opened.
   Then I thought of organics. Addition of a little methylene chloride didn't seem to do much either. What else might work?

   I tried drying out the electrode and filling the cell with toluene (methyl benzene CH3-C6H5 - a polar molecule(?) solvent) instead of water. As I did up cover screws, the lid cracked in several places and a whole corner of the case simply fell off. Evidently this stuff makes plexiglass very brittle! That was it for battery experiments for this month.

   I must also consider the possibility that the electrolyte simply doesn't penetrate the very fine-textured electrode pores very well.

Victoria BC