Turquoise Energy Ltd. News #39
Victoria BC
Copyright 2010 Craig Carmichael - May 2nd 2011

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

Month In Brief
  * Hybridizing cars: Two Electric Hubcap motors, with planetary gear sets? (having nylon planet gears?)
  * Turquoise Battery Project: First Successful rechargeable salt electrolyte battery. Mn-Ni, 1.5 V.
  * Successful end to the NiMH dry cell car batteries as a research project. (But I'll be getting lots of them for EV use!)
  * Much paperwork/writing
  * Independent franchises business model?

Electric Hubcap System

  * Motor Configuration
  * Shortest Motor
  * Smaller Diameter, Too - new body parts molds
  * Hall Sensor Circuit Board
  * A3938 V2 Motor Controller Circuit Board
  * Smaller 'production version' controller chassies: fewer components and smaller controller plate allows smaller wiring box.

Electric Weel Motor Project (Electric Wheel Motor... Rim Motor...)
  * More metal parts have been cut, a better design emerges based on improved Hubcap motor design. But other projects take precedence.

Planetary Gears Project
  * No mechanical torque converter yet!
  * Instead: hybridize cars with Hubcap motors and planetary gears.
  * Photos of gears and car installation.
  * Lubrication: Oil Drip...
  * ...or Nylon Planet Gears! Nylon gears make non-oil-filled transmissions practical.
  * Can't find nylon gears... maybe just great gobs of grease and the metal gears?

Torque Converter Project

  * New design idea: a variable ratio "planetary gear"?

Turquoise Battery Project
  * Osmium - conductivity
  * "Low" oxygen overvoltage may be the key cause of high self discharge!
  * Refrigerator test, freezer test: self discharge is much lower at freezing temperature.
  * In salt battery, nickel hydroxide forms naturally from nickel metal -- no need to make it.
  * Rare earth hydroxides (La(OH)3) are used to increase oxygen overvoltage. (Hence: my original monel-lanthanum hydroxide mix is probably great stuff! But other rare earths are probably better. -- Samarium or Neodymium?)
  * Designs for next cell (complete design less sealed case).
  * Cell made: self discharge still high.
  * Reversed the polarity: IT WORKS! 1.5 V rechargeable cell holds charge.

Newsletters Index/Highlights:

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/

April in Brief

The essential components to electrify a car seem to be coming together at last:
Hubcap motor with planetary gear, Motor controller (finned box),
Forward-Off-Reverse switch and gas pedal potentiometer,
and a few kilowatt-hours of batteries. (The NiMH dry cells shown are about 1 KWH total.)
(Final version of motor will be an inch thinner and 1/2 inch smaller diameter; machine-drilled holes.)

Hybridizing Cars with Electric Hubcap Motor(s) + Planetary Gear(s)

   After 7 continuous months of motor improvements, I think now one could look far and wide and not find as good a motor as the Electric Hubcap. Despite the time taken by motor developments, there are working motor controllers and better ones to come, and a choice of cheap lead-acid or NiMH batteries (or lithium). But I haven't been successful yet at creating the mechanical torque converter to couple the motor to the wheel.

   In order that hybridizing cars not be further delayed by this one laggard item, I've decided to try gearing the motors to the wheels with planetary gears. The speed reduction will be a compromise. Mid speeds will be great, but there'll be less torque to start the car moving and high motor revs at higher speeds. It appears it will probably be possible to gear a small car for adequate low speed starting torque while allowing driving up to about 80 Km/Hr with two motors/gears, on left and right wheels. Using a single motor would definitely mean low top speeds - city driving only.

   I ended up with a 2.8 to 1 Chrysler planetary gear which I installed in last month's motor. I mounted it on the car and tried it on the last day of the month. Disappointingly, it didn't move the car - the usual sort of "almost but not quite" on my lumpy driveway and "might just have moved on level pavement" - but I discovered the motor controller was only giving it 55 amps instead of 127, so it had less than half torque. I fear raising the limit in the controller will simply blow the controller (and it's the only one working at the moment), so the need to get the A3938 controllers working properly is pressing. If I could find more planetary gears with different ratios, I'd love to try them out. At 10 to 1 the car would doubtless have moved and had great acceleration... up to the motor's top speed at 20-25 Km/hr. At 5 to 1 it probably would have worked (except maybe on steep hills), up to about 40-50 Km/hr.
   With two motors at the 2.8 to 1 gear ratio and full amps, the car would have five times the torque of just one direct connected motor. Is that enough for the streets? Fair chance... I guess I'll find out when everything's ready!

NiMH Project Ends, a Success!

   The NiMH Dry Cell Battery Project is completed as an experimental research project. The cells work so well as car batteries and to run heavy loads that there seems to be nothing more to do but make batteries and use them or sell them. I can't say how long they'll last, but it should be a very long time - maybe 15 to 25 years as a car battery. I accidentally left my running lights on for a couple of hours recently and the car started like nothing had happened. I will of course be buying more to use for the electric motor projects. The first three 'spare' 12 volters are already saving my bad back from the stress of carrying 40-50 pound lead-acid batteries out to my car for on-car motor tests.
   Using them for car batteries not only saves gas and makes your car a bit greener, it's also the way to increase NiMH dry cell sales volumes, which is likely to further reduce their already decreasing prices. I was e-mailed a 'coupon' this month from all-battery.com as a repeat customer for 15% off 'D' cells, which brings them down to the lowest price yet: about 5 $US list each - 417 $/KWH. So I put 80 more on my credit card - if I can't sell NiMH car batteries I'll certainly use them myself as EV batteries. (Indeed I plan to solder some more batteries together in the near future, including a couple of 6 volt ones to get 42 volts.) 80 D cells is almost a KWH, about 1/6th of my minimum target of 6 KWH for 50+ Km of local electric travel range. With D cells, the weight of 6 KWH would be close to 200 pounds; with AA 140 pounds; with lead-acid about 600 pounds.

Turquoise Battery Project: a successful battery at last!

   Better than nickel-metal hydride batteries would be batteries with a shorter name that are also more economical and hold more energy by weight.

   Two things of late have been between me and successful salt electrolyte battery development. The first is trying to get enough conductivity out of my graphite and cell structures and formulations for high currents. Ten times better would be a big improvement and closer to typical levels. Still, that by itself shouldn't prevent making batteries that at least work, even if the current capacity is low.
   The second and more perplexing problem was high self discharge - minutes and hours instead of days and weeks. I've done various things to raise hydrogen and oxygen overvoltage levels to match my higher energy reactants, but this month's cell, made on the 25th, discharged itself as rapidly as the previous ones.
   Well, the ingredients in both electrodes could potentially work in either direction. I reversed the charges on the morning of the 30th. By evening, it was holding -1.1 volts, and by the next morning, lo and behold, it was working, the "backwards" charge holding for hours at around 1.5 volts! I expect I have crated a Mn-Ni cell instead of Ni-Mn, with the Mn charging to KMnO4 at +1 volt and the Ni to metal at -.5 volts.

   This appears to be a working, economical, rechargeable cell with good energy density that should be very long lived. Going by ingredients, it might cost twice as much as a standard single use dry cell... and those are currently going for mere pennies in all the stores.
   But I have a few more things to try to get workable 2 volt cells (33% more energy, all else being equal) before settling for 1.5 volts.
   I'm still thinking about better ways of making the cells, though. One can't spend a whole day or two to make a cell for anything but the most limited and valuable uses. Homemade is great, but electric transport really demands they be rolling off assembly lines into boxes and shipped off in truckfulls. A good export commodity for BC and Canada!


   With improvements and new developments coming fast and furious for some months, I not only have trouble doing a fraction of what needs developing, but the documentation has fallen far behind, and without that, others can't readily make use of my work. With all the changes, and now a motor kit available to order, the whole motor building manual was hopelessly out of date and needed rewriting, and the improved format web site was still waiting for me to do several things. Even the web intro/abstract to the Electric Hubcap was two years old and contained major outdated or superseded concepts and had to be rewritten. Then I got a phone call telling me I'd done some things wrong on my income tax/SR & ED claim last month, so I had that to deal with as well. By then it was getting close to time for this newsletter.
   So from the middle of the month, I spent a lot of time typing at the computer, while I also waited for Waterforce to get around to cutting my steel parts for the geared hubcap motor (he did), and for planetary gear suppliers and makers to get around to answering my queries (they didn't).
   A new motor controller manual is also needed, but it'll have to await a working new motor controller, which is probably close now. But the design had more problems than I suspected. This too shows the problems with poor documentation: I've had design problems leading to serious delays trying to make a motor controller from the A3938 chip partly because the chip itself is poorly documented: sketchy datasheets and no proper application notes.

A Business Model?

   As I try to figure out how best to spread the Electric Hubcap technology as it becomes available, one idea that comes to mind is to do independent franchises. Here at the center, materials and parts would be purchased and made in bulk, and wholesaled to the franchised dealers and installers. Jigs and molds could also be supplied for local making of many or most of the components. The dealers then would assemble the equipment, sell and install it per installer and customer needs. An insurance pool that everyone would pay into could perhaps cover liability so that no one would risk their all to participate.

Electric Hubcap Motor System

The motor profile with flat plate rotor.
Finally, 1/2 inch is to be trimmed from the diameter - less 'lip' outside the coils and rotor.

Pressing a drying PP-epoxy ring harder to try to squeeze out dry spots.
Afterwards I started backing the flexing PE plastic mold pieces with plywood,
and then bought 1" thick to replace these 1/2" pieces.

Motor Configuration

   Just after putting out the March newsletter, I realized that putting the stator side bearing on the outer ring instead of the inner would improve the layout:

1. It would allow turning around the SDS coupling and decreasing the motor width another .5" inches to 3.5".
2. Turning the SDS coupling around also simplifies alignment of the rotor - the bolt heads are then on the right side to adjust easily in situ.
3. It eliminates the need to drill some vent holes in the inner ring.
4. It eliminates the need for a special cover over the center hole in the outer ring.
5. It makes the bearing more accessible for lubrication.

The down sides to the change are small:

1. The shaft is 1-1/2" longer.
2. A spacer about an inch long will be required on the shaft.

However, the longer shaft is still only five to six inches, and the extra length will improve resistance to sideways shaft twisting forces.

Shortest Motor

   With flat plate rotors instead of brake disk rotors, and bearings on the outside, the motor is about as short as it can get without going to extremes or changing the electromagnetic design. The main body is 3-1/2" thick, still by 11-3/4" to 12" diameter. The rotor compartment is just 1.5" across, and the rotor with magnets is almost an inch thick. What's left is useful for cooling air flow.
   The coils are an inch, the magnets are 1/2", the gap between them is over 1/2", and the rotor plate is almost 3/8", making up 2-1/2". Then there's about 3/8" air space behind the rotor and two 3/8" end ring plates, adding another inch, total 3-1/2". The bearings and their holders, and the heads of the bolts that hold it all together, stick out a little farther.

   I was very ready to slap a "final version" motor together near the start of the month, but as of the 15th, Waterforce had only cut my two big Weel motor pieces and not the five Hubcap motor pieces including a flat rotor. Oh well... no shortage of other things to do! When I've checked the fit and made sure the pieces don't need any changes, I'll get 1/2 dozen of everything and afterwards not be waiting on him until six motors are done.

   Late in the month I gave an order for a couple of rotors and plates to Victoria Waterjet in Langford, too. Langford is increasingly hard to get to from Esquimalt during business hours, but I might as well try them out too; they're there! Why have all the eggs in one slow-moving basket?

Smaller Diameter, too

   I was having trouble with epoxy sticking to the PVC outer shells when casting the body parts. The reason the  motors were the specific diameter they were was that I could use the 11-3/4" turquoise PVC culvert pipe for the outside of the mold. At the same time, epoxy would ooze out the gap between the UHMW-PE "butcher block" ring and the PVC shell, and also the PE was a little thin, and I had to back it with plywood. On top of those things, I wanted to have three molds so I could make all three body rings in one gooey epoxying session instead of three separate sessions per motor. This is now the most labourious part of making a motor.
   For all these reasons, I bought some 1" thick ultra high molecular weight polyethylene (UHMW-PE) 'butcher block' to make new molds from.

   Looking at a motor, I decided it was about 1/2" bigger around than it really needed to be. Not much, but I planned the new molds to make 11-1/4" diameter parts instead of 11-3/4". The mold would be routed out of the 1" pieces on the CNC drill/router as a 5/8" deep "pan" with no holes or cracks in the bottom. It would still need a piece of plywood reinforcing the bottom. This proved an advantage in that the mold pieces were now 12" x 12" instead of 12.5 x 12.5": the heavy PE is very expensive and Industrial Plastics only sells whole square feet regardless. The extra material for the extra 1/2" x 1/2" thus quadrupled the price for a single mold piece, and doubled it for two or three pieces (and only by cutting some clever angles out of the larger pieces). Now I'll get four pieces out of the piece I bought that would have only made two.
   I changed the center holes from 2-3/4" to 3". The 4" center hole in the stator inner piece is unchanged. But the pan would have a 2" "post stub" in the middle, and the desired taller 3" or 4" diameter UHMW-PE post would have a 2" hollow in the base and fit onto that. That gives the tall post of the desired height for stuffing the PP fabric around, and it can be pulled out afterwards, providing a spot at the center to get a chisel under the molded piece and pry it out of the pan.
   The one remaining problem is that I have no tube to put around the outside to stuff the PP fabric into, except perhaps to cut down a piece(s) of the PVC pipe to make it a bit smaller. But perhaps when the mold top ring is pushed all the way down into the pan, the PVC ring can come off and be cleaned off separately, before the epoxy hardens. A PE piece would be better.

   I have enough pieces molded now for four of the slightly larger diameter motors. The future ones will be a touch smaller and a touch lighter.

Bigger Bearings & Shafts

   The axle of the motor with the first planetary gear to connect to a car wheel had to be 1-1/16" on the gear end to match the inside diameter of the sun gear. On the 28th I took a 1" trailer axle, cut it to length, and turned down the fat end on the lathe to that diameter, to fit the gear and a 1-1/16" I.D. trailer wheel bearing.
   Owing to the forces involved plus the car wheel bumping around, the bearing and shaft on the gear side would take considerable loads, but since they were rated for the wheels of 2000 pound trailers, surely they were big enough for the job?
   But there was a hitch. The sun gear of the planetary gear wouldn't fit through the bearing race or the hole for it in the end plate of the motor. Nor could it be put on later, as the bolt holes weren't accessible once it was installed. That meant the gear had to go onto the axle first, slip the axle in from the outside, and then put the bearing and the magnet rotor on. That in turn made it difficult (though not impossible) to install and adjust the magnet rotor.

   The sun gear just fits through a 1.375" bearing race. So I'll probably continue to use the Pacific Trailers stub axles, which are 1" at one end, and just over 1.375" at the other; easy to turn down to that diameter for the bearing and a 1.375" SDS bushing. An inch at the end gets turned down to 1-1/16" for the gear. The SDS bushings to hold the rotor on are available for all these shaft sizes. The larger bearing size means revising the metal end plate inner diameters on the motor as well. (now that I thought I had everything set up!)
   The larger diameter won't add much weight to the motor. The weight where the SDS bushing is - 1.5" of the 5" shaft - remains the same since the larger axle diameter is compensated by a larger hole in the bushing - it's all metal one way or the other. An inch at the gear end needs to be turned to the diameter for the gear regardless, and the stator end (about 2.5" long) remains 1" for the 1" bearing.
   I'll use up the five 1" trailer axles I've already bought before switching to uniform starting diameter shaft.

Hall Sensor Board

   I completed the layout of the Magnet sensor board on April first or second. It puts the Hall sensors in just the right places and angles on one circuit board, simply screwed onto the inner stator ring. When I got the boards back, I found I had neglected to check the A1203 pinouts: they were Vcc,Gnd,Out, not Gnd,Vcc,Out, so I had the power and ground backwards. As the board was so simple, I simply connected the ground wire to Vcc and the Vcc wire to ground, and crossed a couple of resistors for the motor temperature sensor. You wouldn't even notice! Of course I corrected the layout for the next batch.

A3938 V2 Motor Controller

   Tristan was designing a new motor controller board and I left it with him. He sent the finished layout and it was very good except 1/2 inch wider, 2" instead of 1.5". This was understandable as there were more components on the new board - filter resistors and capacitors, test point pins, and a 5 volt power supply for external peripherals. But it meant the chassis boxes would have to be 1/2 inch thicker. Being a fussy person about such things, I decided I would try my hand at shrinking it down again. I left out a couple of things that we'd agreed to put on because I was having trouble routing the lines, and then I added a 100uF filter capacitor and an additional .1uF in the battery power line. As I think about it, although there's a dozen capacitors on the busses at the mosfets, they go to "Vs" rather than to the main ground, and it was foolish neglecting the PCB filters from this spikey line in version one. However, it doesn't appear to be the explanation for the failures of the version 1 boards.

   I also changed the header connectors to three, each with a clearly defined function: there's now one 6 pin to the motor for hall sensors and temperature sensor, an 8-pin for the basic controls, and a 9-pin for statuses and instrumentation. All have 5 volts power and ground pins.
   The controls header has the connections to the 'gas' pedal potentiometer, on-off, forward-reverse, brake pedal switch and brake pedal potentiometer. The last two of these are for regenerative or dynamic braking, but they aren't used in the first version. ("Brake pedal switch" should be connected to ground, as it is active high and has a pull-up.)
   The status header includes 5 volts power, the 3 magnet sensors (from which may be derived RPM, speed, distance traveled, direction and illegal sensor states/unplugged detection), battery voltage divided by 10, motor current (1mV/A), motor temperature, and the A3938's 'fault' pin status. You can hook anything to these signals from no connection at all to a voltmeter to a multifunction microcontroller run display with status and diagnostic system giving MPH, mileage, state of charge, remaining driving range, overheat warnings, etc.
   With a combined system that connects to both the control inputs and the status plug, one could introduce computerized overcontrol of the system, without compromising the controller's critical hardwired current limiting system, coil timing et al. Such a system would probably make regenerative braking simple.

   After starting in on making these motor controllers in 2008, I found brushless motor controllers were in fact available - mostly smaller, but a few models seemed big enough. I had one running by then, but more than once I've wondered if designing and making them myself is a good use of my time and resources, or if I should just be buying controllers. But I ran into to a store clerk who said he had had been part of several EV conversions using kits, and he said that although they were costly, the controllers would fail after a while. And evidently the well known line of controllers in the kits is made by hand - "3 a month" when he was involved. Apparently they're not all they might be, and hard to make as well. And then of course there's the "false pride" factor: I wanted to make my own controllers because my BCIT diploma is in electronics and I've created many successful electronic designs... though almost none for 20 years, and they were mostly digital and computer interface circuits.
   My prototypes have blown up a lot of mosfets and chips and have been a continuing source of trouble. But the troubles have been and are being weeded out one at a time as they rear their ugly heads, and with each improvement they become more robust and reliable. The MC33033 version seems quite reliable now, though I have the one in the car supplying reduced currents to the motor for fear of losing it. Assuming the final A3938 (or A3932) Turquoise controllers work reliably, it seems possible they'll work better than whatever else is available. That said, Canadian Electric Vehicles says the controllers they're using now work well unless they're blown up during installation. I'm pleased that it's a "hard wired" circuit rather than a microcontroller programmed design. It won't have any timing glitches or software bugs to cause unexpected troubles or failures on the road, which seems to be a feature of at least some of the available controllers.

   I completed the A3938 version 2 circuit board layout on the 5th and e-mailed it, along with the magnet sensor board layout, to AP Circuits on the 6th. I also sent the magnet sensors board to someone in Vancouver who has a nice CNC machine good for circuit boards. Unfortunately, my CNC drill router is probably too big and clunky for any sort of circuit boards, even simple ones. (Then again... what would happen if I put a dremmel tool on it with a small router bit? Hmm...)
   The finished boards arrived on the 11th. I put together a controller the next day.

   I also decided that I'd shrink down the wiring box that the controller is a part of, from 10" long to 9", and from 6" wide to 4" or 5". As it's progressed, the components in that box have dropped to just the main circuit breaker and a relay to turn the system on and off from the car key. Other than that, it clamps the incoming wires, has a big ground central bolt, and encloses everything. It no longer needs to be very big, and space in a car is limited.

   I didn't get a chance to test the controller until the 20th. It ran a motor for a brief time and then burned out. I recalculated the blanking time capacitor and found I'd been using one about 20 times too large. That's probably the trouble, although I can't think why that would burn out the controller rather than mosfets. I replaced the A3938, but those tiny surface mount chips are just SO hard to deal with.
   That one burned out almost at once too. This time, with the reduced Ct capacitor size and the same Rt, the frequency was 4x higher than the max, and probably that did it. It was burned out in a different way than the previous ones: only the oscillator quit running properly. I couldn't understand why Allegro didn't put out app notes for the A3938, showing typical component values with calculations and reasons therefor. All these costly  prototyping failures for silly, easily avoided reasons is a ______. On the one hand, it's up to me, the designer, to double check everything, but on the other, seeing a couple of sets of "typical values" from the chip maker would doubtless have raised some red flags where mine were notably different. If I wasn't so persistent - and if I hadn't had some "pure luck" initial success showing the "gas pedal like" operation, now months back - I might well have given up and gone to some other chip from another manufacturer.
   I decided to look again at allegro.com. I had originally stumbled across the A3938 and designed the controller around it. But at Allegro.com I recently discovered there's also an A3932 - an almost identical chip with a couple of small differences (main difference: it's half the price!), doubtless the first of the 'series'. Linked to the A3932, but not to the A3938 (thanks a lot!), was an application note for designing around the chip! It is obviously applicable to both chips, but it wasn't caught in any searches for the A3938 part number.
   I wrote and the web manager says he's linked it now. According to the web manager, the person who wrote that paper said it was written for a lecture and wasn't a substitute for real app notes or better datasheets. (Sounds like he wasn't very impressed with Allegro's documentation either!)

   I did glean a a couple of vital details from that paper that weren't mentioned in the A3938 datasheet: the chip was intended to drive "up to 100 nanoCoulombs and beyond" of MOSFET gate charge. Since the IRFP3206 MOSFET is typically 115 nC and there are two in parallel (240 nC typical, 270 max), it becomes apparent why the A3938 gate drive signals often seemed to be individually failing: the drives are underpowered.
   Furthermore, the app notes reminds that the gate signal lines should be kept as short as possible, and (I probably knew this but had forgotten), that the gate resistors should be as close as possible to the gates.

   I increased the gate resistors from 15 to 27 ohms. That will make for slower switching slew rates and more heat in the MOSFETs, but should cut the currents the drives have to supply virtually in half. And I shrunk down the mosfet array, with one inch between transistors on the heat sink bars instead of 1.5", which very considerably shortened the gate wires to the outermost ones. The controller plate is only 6" long now instead of 9". The motor power wire terminal blocks had to be moved out in front of it, on an extended base flange, since there's now no room at the side.
   Another change after I thought I had everything all worked out, this time to the motor controller & chassis dimensions!

New "shrunk" 6 inch wide motor controller layout with short gate wires.
Four MOSFETs are under the (unfinished) PCB where there were three.
Wood block shows where heavy terminal block will sit.
The smaller size will do nothing good for heat dissipation from the MOSFETs!

   In all the planetary gear work, I haven't yet finished the revised controller, but I have good hopes it'll finally work. The last motor/planetary gear test on the car with the old MC33033 controller shows it's badly needed.

Battery Voltage: 36-42 nominal?

   Originally, I wired the three coils of each phase with 60 turns of #14 wire, intending the motors to run from 120 volts DC. Then realizing that was creating a needless electrical hazard (not to mention needing 10 batteries for any 'full voltage' test), I put the coils in parallel instead of in series, so they would use three times the current at one third the voltage - 40 volts. (Now I'm wiring 20-21 turns of #11 per coils and putting them in series again. Paralleling these would make them good for 12-14 volt operation.)
   Three 12 volts batteries gives 36 volts - close enough, I thought. Also I think even 48 volts nominal is starting to get a little high - four batteries can be 54-58 volts when charging. It's getting close to being a shock hazard, and the motor controller is only rated for 60 volts absolute max.
   Now, making my own nickel-metal hydride batteries from dry cells (and sometime my own chemistry batteries), I see no reason not to make whatever voltage I please. I think I'll add in a six volter and try out 42 nominal volts. That could also be achieved with 7 golf cart (6 volt lead-acid... with sodium sulfate) batteries. With the same current, there's about another 750 watts available if the motor - or the controller - doesn't get too hot. And with the efficiencies hinted at by the low idle currents I've been seeing, I don't think they will, at least not around here.
   The mosfets - and the A3938 motor controller - are only rated for 60 volts, so counting spikes and noise and charging voltages up to about 50 volts (PbPb or NiMH), 42 volts is probably about as high as one should go with my controllers so as not to push the 60 volt 'absolute maximum' semiconductor specs and risk seeing it go up in smoke. (I did use 48 volts in one test with the MC33033 based controller with no adverse result.)

The Electric Weel Motor Project

   The next pieces, the ring for the magnets and the stator center, were cut at Waterforce.ca in Sydney by mid-month. I started to realize that the best way to put them together would be the same way I'm now doing the smaller motors: an enclosed motor with the bearings in the end bells, and the giant magnet rotor attached by an SDS coupling to the 2 inch shaft. If the Hubcap is perhaps more of a "cake" motor, the Weel would take on a real "pancake" shape, 28" diameter and still just 4" thick.

   However, the new motor controllers are still not working. Not only are they badly needed for the Hubcap motors, it's almost pointless to build this motor and not have the three controllers it will need, so until those are working properly and there are enough made, they have priority.

Two-piece 26" rotor (magnet positions outlined by waterjet!),
and stator center piece, with bearings and shaft.
The motor diameter is almost identical to the wheel/tire O.D. of the truck it's to be installed in,
so it ain't a Hubcap motor to go on the wheel!

Planetary Gear Project

   It seems people want the motors for vehicles (no surprise), but I still don't have a working torque converter. With gears, a Hubcap motor at about 4:1 gear ratio might acceptably drive a car up to 50-60 Km/Hr for city driving. Highway driving would have to be done on gas, with the motors decoupled from the wheels to prevent over-revving them. One would have to stop and disconnect before hitting the highway.
   Of course, my car did move with the direct drive 1 to 1 ratio. Could I be overestimating the amount of torque multiplication that will be required? A 2 to 1 ratio would keep the motor revs to 2000 at 100 Km/Hr. 3 to 1 - more likely to work - might allow near highway speeds, even 80 Km/Hr (2400 RPM).
   I finally decided to look into planetary gears as an interim measure. I really didn't want to do them. They are inferior to a variable torque converter because:

1. The fixed speed reduction ratio is a compromise. It works fine at mid speeds, but it doesn't give very good torque for starting the car moving from a stop, and the motor revs faster and faster linearly with speed until it hits its maximum RPM, which will define the top vehicle speed. With the torque converter the torque and speed are optimized for any given speed and driving condition, including outstanding start-up torque and moderate RPMs on the highway.
   Theoretically, one could put a two or three speeds transmission on the wheel to solve the basic problems, but in practice I think it will be hard enough just to mount one gear with a fixed ratio.

2. The lubrication requirements are high. In gas car transmissions, they sit in a bath of oil, but mounted out on the car wheel, it's a problem.
   At first I had decided to have a continuous oil drip lubricating them, with a manual shut-off. In summer, cheap vegetable oil is an eco-friendly way to go. Corn oil seems to stay liquid in the fridge almost down to freezing. Now I've decided to try just lots of grease first. I have two identical planetary gears: if the first one in grease breaks down, I'll try the oil.
   The gears turn at absolute speed depending on vehicle speed. The torque converter was hopefully to only turn at a relative speed - the speed difference between the motor and the wheel, which is substantially slower. Consequently, wear and lubrication requirements are much less. But it may be (those designs having failed to perform well so far) that I try a torque converter design that has sort of stator that doesn't turn, then it will be more like the gear - also easier to design.

BUT planetary gears do WORK! Cars can start to be hybridized!

   Any port in a storm! Losses should be low - Having no oil bath eliminates most of the fluid friction. If and when I get a working torque converter, it can replace the gear. That automatically and continuously variable rate of between around 10 or 15 to 1 and about 1.5 to 1 would be better, allowing operation of any smaller car with one motor using the least amount of electricity.

   I was looking into planetary gears at a place on the web, but the guy was slow answering e-mails and it dragged on without me getting to place an order. The ones I first inquired about turned out to be $475 - almost the price of my whole motor kit! I thought the shaft sizes seemed good, but the torque rating was actually about 3 times what was needed, so I asked about smaller models. Only two suitable looking models were under $400, at $215.
   But earlier on the same day that I finally got those prices, I had the idea to drop in on an auto transmission shop, taking with me two pieces (of three) of a broken planetary gear I'd picked up from curiosity in 2006 when I was doing the wave power stuff, just to show what I was after.
   The first shop referred me to a second shop. The guy there immediately identified the gear as a Chrysler gear for a V6 engine. Wow! He wasn't sure of the torque rating, but guessed around 150 foot-pounds. That seemed to be well over the requirement. Perfect! I asked where one might buy such things. "We have them." Wow, pay dirt! He showed me the other pieces that went with what I had, and got on the phone and said he could get me all the drive parts for a working, used transmission - TWO planetary gears plus everything that went with them - for $109, and have it there the next morning. Wow, pay dirt again!
   He didn't know the reduction ratio, but said that each planetary gear could do three ratios. If the inner part turned with the outer casing held stationary it was one ratio, but if the inner part was held stationary and the outer case turned (and was the output), it was a different ratio. Once I had that digested, I forgot to ask about  the third ratio, but later I realized it would be with the inside shaft held stationary and the inner and outer housings turning.

   I picked up the gear set the next day (14th) and started figuring out what would go where. Selecting the more suitable looking of the two planetaries - the same one as the sample I'd already had - by day's end, I had cut the sun gear in half with a zip disk (the two ends went to the opposite planetaries) and figured out how to mount it on the motor shaft, grinding slots for two 1/4" holding bolts that would screw into the axle. The inside diameter, however, was 1-1/16", so the 1" axle shafts were too small. I would turn down a trailer axle I'd bought earlier that was 1.4" on the fat end.
   The outer case with its inside ring gear would be affixed to the motor.
   The planets part - the output - I'd have to weld or bolt a plate to to put the wheel drive pins into. Ironically, this meant that the broken part of my original sample gear isn't needed - I had everything I needed all along except the sun gear... and to know what to ask for. I bought another sun gear and had two complete identical gear sets: one for each motor on both rear wheels.
   The gear ratio turned out to be about 2.8 to 1. That might be enough. If not, with two motors, that will be the equivalent of 5.6 times the torque of a single motor (which did after all move the car, if only barely, in the October 2008 test). Earlier I was estimating 7x would be a good figure, but hopefully 5.6 is good enough. (A test at the end of the month suggested one motor with a 2.8 ratio isn't good enough.) If it is, it is very helpful that with 2.8 to 1 gears the car's top speed can likely be around 80 Km/Hr rather than 50, 60 or 70, revving the motor up to about 2240 RPM. That gets it essentially onto the highway. And if perchance 4 to 1 or less works, two motors at 2 to 1 would give full 100+ Km/Hr speed.
   But I'm wondering if maybe one "double" Hubcap motor (20" diameter, 10 KW) wouldn't be better: it would have 4x the torque of the smaller motor even without a gear, and two would have 8x. But at that point we're up to 20 KW of motor, and they'd have a rough time with a flat tire, the bottom end hitting the road.


Wheel Plate turns wheel by pushing on the lug nuts. (1/4" mild steel.)

This was the first try at mounting the outer ring gear on a flange.
The black piece inside is slippery delrin/acetal plastic for the planet gear assembly to rub on.
(Yes, I had a hard time cutting the inside circle.)

Ring gear mounted on motor end ring.
Motor shaft turns sun gear (center one), 1.062" I.D., so needs 1-1/16th" axle turned on lathe.
I switched flanges to this one, cut by CNC waterjet.

The planet gear carrier in place. Note the threaded holes I made near the outsides to hold the wheel thrust plate.
Another piece of delrin is enclosed for sun gear to rub against.

Completed outer assembly april 27th.
Ring gear is stationary, sun gear on motor shaft turns planet gear assembly, ratio 2.8 to 1.
Planet gear assembly is connected to lug nuts on wheel by plate, and tentatively held against motor (since nothing else is holding it) by a spring from the plate to the wheel. (Might be a good place for some more delrin between ring gear and outer face of carrier to retain carrier - and seal it up a bit.)
Note the very short distance from the motor to the car wheel - it won't stick out as far as the rear view mirror.

The lug nuts and bolts already take the torque of the car engine and brakes...
I'm less sure about these 1/4 inch bolts holding the plate onto the planet gear carrier.

The sun gear on the axle,
and 'dust cover' on the motor, April 29th.

Fitting the motor & gear to the car,
thrust plate at lug nuts.

Marking up for fitting motor spring mountings.
Sticks out farther than rear view mirror - one inch thinner motor will be an asset.
(Truck for Electric Weel motor project is behind.)

Spinning up the motor with 2.8 to 1 gear on 30th. Jacked up, it accelerated quickly.
On the ground, the car only twitched - but then it was only getting 1/2 current from the motor controller.

Lubrication: Oil Drip... Lots of Grease... or Nylon Planet Gears?

   Transmission gears usually sit in the enclosed transmission in an oil bath. If this one needs an oil drip for lubrication, that's okay for testing and demonstrations, and for real zealots, but one can just see it en masse: car gears seizing up daily on busy streets because the oil ran out or was left turned off. Oil slicks causing accidents on the road. Shortages of vegetable oils at the grocery.

   But once I got started, it dawned on me that gears can be made of nylon (or other plastic such as acetal/delrin - there's even a "nyla-steel" made for doing gears), and nylon gears don't require lubrication or are "self lubricating". It is preferred that nylon gears mesh with metal gears. In the planetary gear as configured for this job, only the planet gears spin around fixed shafts, and the other gears contact only with the planet gears. So if one could replace the metal planet gears with nylon ones, it would be perfect, and constantly dripping oil wouldn't be required. An occasional greasing would do fine.
   Then it dawned on me that helical gears might be made in standard sizes, and one might actually be able to find the right gears, ready-made of nylon somewhere, if one knew what to ask for and where to look. Perhaps even a planet gear carrier complete with the nylon gears, ready to just pop in.
   I went on the web, found some plastic gear places, and made a few e-mail inquiries. They went unanswered, or the answer wasn't helpful.

   I also found out that gears are (or can be) made on something called a "gear hobber". I saw a picture, and some ideas for a cheap, single purpose "gear hobber" entered my mind. I could use one of the metal gears themselves, using the teeth as detents to rotate and line up the nylon piece to just the right angle and place to duplicate the teeth. Then mount a dremel tool or an angle grinder on a stand with a pivot to make the cuts. Sort of like a fancy key cutter that follows the shape of the original key.

   But that would be a last resort! If the metal gears will work okay just with grease, they'll be fine, and I should forget about plastic gears. Otherwise, I can use it even with oil dripping for testing and demos for some months before I run out of patience looking for a source of ready-made nylon planetary gears - or at least individual planet gears.

   Either just greasing the gears, or else plastic planet gears, will make the whole idea truly practical. Otherwise, turning oil drips on and off before and after driving... and forgetting to do so... will be unpopular and unreliable.

Torque Converter Project

   Doing the planetary gear work has made me think perhaps I should change tack with the torque converter design: If it were to have a stationary component, like the stationary outer case of the planetary gears (or the stator of a fluid torque converter), it could be made as a sort of variable ratio planetary gear. It would then be a positive action drive rather than the oscillating masses design.
   For me, the biggest attraction of the "all live parts" design was that everything would spin together and other than that the force transferring parts would be moving no faster on the highway than when starting out from a stop: That minimizes lubrication requirements. It also seemed to have the most slack construction standards - extreme precision wasn't required, whereas gear must mesh exactly.
   On the other hand, I have not had any success making the statorless converter yet, and any converter that works is better than fussing around with planetary gears, with their fixed ratios forcing use of bigger (or more) motors with more torque.

   The torque ratios of a design that I'm now thinking of now might range from perhaps 6-10 to 1, to 2-3 to 1. That's less dramatic range than an oscillating masses converter, but it meets the requirements. One thing that makes it seem more feasible now is that I know that Waterforce could cut straight gears (though not helical) with exact teeth to any specifications I supply - I've seen some made at his shop.

   In this design, the planet gears would mesh with the outside ring as usual. But they would have an "X" or "*" on top of each, eg four or five arms. The motor side has drive pins that hit the inner arm of the "X" as they go by instead of a sun gear. The drive pins are pulled towards the axle by a spring. In their innermost position, they just miss the tip of an arm of an "X". At low speed, the pin on the slightly extended spring will reach an "X" only if the arm is pointed almost straight in, so it will only contact some of the arms. As the motor turns faster and faster, the drive pins move outwards, and hit more towards the center of the "X", hitting all the "X"es and turning them farther as they go by. The more planet gear "X"es that are hit, and the farther they are turned each time, the lower the drive ratio.
If I up-scale the size, say to the full 12" instead of 4", and use big nylon planet gears, it might not need more than greasing once in a while.

   It seems likely I'll want to refine that idea some before attempting to building anything. In the meantime, I'm giving planetary gears a try.

Turquoise Battery Project

Manganese versus Vanadium?

   I bought four carbon electrode rods plus 1/2 a pound of battery grade MnO2 mixed with graphite powder for only $2... two packages of two D cells at the dollar store. It would be more convenient if they just put the materials in small jars, but then they'd probably charge much more!

   I thought I'd look more closely at the electrochemical reactions of a couple of elements. Chromium has a soluble ion form, similar to zinc and cadmium that so shortens the cycle life of cells made with them. So much for chromium! But on inspecting vanadium I found that it had a negatrode reaction that looked similar to manganese (and iron, nickel, copper, etc). In alkali it was no better than iron, which is cheap - but its voltage went up slightly with acidity instead of down, and so in salt solution it was probably just a slightly lower voltage than manganese. In addition, the atomic weight of V is 51 versus 55 for Mn, so it would have a few more amp-hours per kilogram, 1052 versus 975, and the VO formed by discharge in preference to V(OH)2 is also lighter (cell needs less water).
   If V was substituted for Mn, the (probably) slightly lower voltage would make it a little easier to prevent hydrogen gas generation and the overall energy density should be similar.

   Plus, vanadium also seemed to have potentially workable positrode reactions, "vanadiate"(?) similar to the permanganate but moving fewer electrons. Was that worth exploring?

   Then I looked up the price of vanadium oxide. At the pottery supply it seemed to be 20 times the price of manganese, indeed more than any other "top row" metal except cobalt! And then there's the dollar store price of Mn. So much for vanadium!

More Osmium

   After discharging for 24 hours (into a 28 ohm load and ending at .3 volts - the highest figure yet), it seems the battery was taking several hours charging to gradually get back up to about 2 volts, and a whole day at least to be really well charged again. It takes more current at a higher voltage to charge than to discharge, but it was behaving more like a real battery. Was the dopant helping?
   I decided to try another osmium/acetaldehyde/cellophane separator film, this time dipping a paintbush down and stirring to make sure a decent bit of the trace of Os in the acetaldehyde in the test tube was ending up in the film. This time if there wasn't enough osmium, it wouldn't be because the densest of all metals had settled to the bottom of the test tube and I was wicking up liquid from the top.
   If there was further but not "night and day" improvement, more powder could be added to the test tube for the next try. In fact however, no improvement was evident.

   A few days later I added more electrolyte and smeared in some more MnO2+graphite.

Better Conductivity

   In the quest for better conductivity, I've compacted electrodes and compacted graphite sheets. But then I put them together in the battery, where they are only nominally pressed together. It occurs to me that I should be compacting them together to ensure good contact.
   In fact, as I thought about it, I became sure that this was a big part of the answer. But there was no guarantee that the electrode material would stay stuck to the graphite when it was removed from the compactor. I decided I'd try perforating the sheet with lots of little holes, and allow the electrode material to fill the holes. That would give them at least some sort of bond, and increase the surface area of contact between them.
   But it's a bit tricky when I need to torch/sinter the positrode after compaction and the graphite would burn. Perhaps I'll try a little steel bezel to protect the edge of the graphite but not the edge of the electrode.
   If that doesn't work, I could compact the electrode alone, sinter it, then re-compact it with the graphite sheet.

   Another idea would be to texture the graphite sheets. Hopefully even without strong compaction there'd be electrode briquette contact points at high spots all over the sheet. I thought of the textured ABS plastic sheets. I cut one to size and put it into the compactor with a graphite sheet. By gosh, the graphite sheet came out looking just like the plastic! I decided to just reopen the present battery (again, sigh) and try it on the accessible negatrode and see if it made any notable difference to the current and voltage. It seemed to help a little.

< div align="center">

Graphite sheets. L: with punched holes; C: compacted against textured plastic; R: plain cleaned with hexadecane and scotchbrite

Textured sheet with punched holes.
This sheet had been used a couple of times and started to fall apart.
When I made an electrode on it, I didn't put in sufficient graphite powder and the resistance
was too high because of that. It didn't look like it would stand any another try.

   Later I tried simply roughing a graphite sheet surface up with coarse sandpaper. That seemed little or no better than texturing.

Surface of graphite sheet roughed up with #40 sandpaper.
mag: 10x

   I think next time I'll try washing a square of carbon (graphite) fibre with boiling hexadecane (outside this time - it really stinks) and using it as a conductive mat, compacting the electrode around it. That might connect better to the electrode powders, but instead I'm wondering how well it'll connect to the terminal post.

Aha Moment: Could Nickel be charging above oxygen overvoltage?

   On the 15th I thought about the high self discharge. Every cell I make with salt electrolyte seems to have horrible self discharge. I thought the Mn negatrodes must be verging on being too high a voltage, but various measures hadn't fixed it. I tried using iron for a lower voltage negative in case it was my Mn hegatrodes, but even that only helped a little. Was there something intrinsically wrong with using salt as an electrolyte? But the old standard dry cell uses salt, albeit it uses ammonium chloride instead of potassium chloride. What then... just potassium chloride?

   It had to be something... what about the oxygen generation voltage at the positive electrode? We know nickel charges properly in alkaline solution at about +.5 volts. We also know that lead charges properly at +1.7 volts in acid solution. But we also know that these are just under the wire: when the temperature hits about 40ºC the nickel alkaline electrode starts to bubble oxygen, and lead acid cells do so notoriously on any overcharge. The addition of manganese oxide increases oxygen overpotential of nickel electrodes... in alkaline solution.
   We might estimate the oxygen overvoltage at about +1.1 volts for neutral solution, and the nickel reactions as being about 1.0 volts. The Mn should raise the oxygen overvoltage a bit more. These figures and assumptions, along with the fact that nickel works in alkaline solution, led me to assume everything was proportional and nickel would work fine in salt solution. But they aren't really very reassuring on close examination: they're only estimates - the midpoints between alkali and acid. What if the voltage shift is non-linear, and the nickel voltage is a bit higher or the oxygen a bit lower? Or what if the manganese reactions are a little higher voltage than expected, or Mn doesn't raise nickel's oxygen overvoltage in neutral salt solution? The batteries do put out about 2.2 volts rather than 2.0. Or maybe oxygen generation starts a little lower in KCl than in some other salt?

   One thing that is known is that the self discharge of every cell I've made using salt solution has been awful. What if it's because the positrode reaction voltages are just at or even slightly above the oxygen overvoltage instead of just below?

   One simple experiment would be to put a battery in the refrigerator, then (if that didn't work) the freezer, and see if the self discharge slowed markedly. CSA/DREO (and others) only had trouble with oxygen production in their Ni-Cd satellite batteries when the temperature rose above about 40ºC. Maybe troubles in salt would stop at a threshold temperature, perhaps below 5ºC, or -5º?
   I set a table next to the fridge in the lab so the battery could be inside and the charger and DVM outside with just a couple of wires going in. I had been doing ongoing tests, and they would simply be continued and any improvements noted.

   With the cell in the fridge at 2ºC, I figured I was onto something the next morning (16th). The voltage with the charge on was over 2.1 (Ni/Mn-Fe cell), whereas it had never quite hit 2.0 volts previously: it wasn't being dragged down as much. When the charge was removed, the self discharge was still far too high to work. It seemed somewhat slower, but it didn't seem it had crossed any "threshold" temperature where it would simply work fine.

   Next test was in the fridge's freezer at -6.5ºC. Hopefully there was enough salt in the electrolyte to prevent it from freezing. (The NiMHs I've been getting are rated to -10.) After it had charged a while, I turned on the DVM and it was good for a laugh. It read about 2-1/2 volts, and when I took it off charge, it still said over 2 volts... on a 1.6(?) volt cell. This must mean that some trace additive element must have charged up that would have been too high voltage to do so at room temperature - probably the calcium, or maybe the antimony. It seemed to have high self discharge again... but this time from 2+ volts downwards towards where it was supposed to be.
   I put a 10 ohm load on it, and the voltage dropped to .6. Evidently the electrolyte wasn't working very well - probably half frozen. (I wonder how much of my cells' high internal resistance is due to poor electrolyte conductance normally?)
   After a minute or two of load, it stopped going up to 2 volts open circuit and went only to 1.66. The self discharge this time (of the main ingredients) was definitely much lower - it stayed above 1.6 volts for several hours instead of several minutes, and was still 1.45 after 12 hours. The threshold was crossed.

   It seems likely that oxygen overvoltage is the culprit, so the next thing to do was try to figure out how to raise the overvoltage so it would work at higher temperatures.
   I found a paper talking about using higher numbered rare earth elements (in KOH) to raise it, and I started thinking about the lanthanum-monel mix in bean sauce I did in the first year of this project. Nickel oxidizes in positrodes in salt solution (unlike in strong alkali wherein it's inert - the key attraction to producing alkaline chemistries). That was why I couldn't get any batteries to work at all back then -- the "+" terminals corroded away no matter what metal I made them of including nickel. (I didn't try gold, platinum...) About the only higher rare earth element not mentioned was dysprosium, the only one I have some of. Evidently samarium is likely the best.
   The metallic nickel in the monel would thus charge to NiOOH and become the active electrode element, and the lanthanum is fairly likely to raise the oxygen overvoltage enough so it will work properly.
   The copper will also oxidize, to a copper oxide as the hydroxide (Cu(OH)2) is, like zinc hydroxide, evidently not very stable and it will change to CuO or Cu2O. That will add no energy, but both Cu2O and CuO are lower impedance semiconductors, and in solid solution with the NiOOH it should improve conductivity substantially and make the NiOOH more available.
   I could probably add manganese, too, and get better amp-hours, and graphite, and get best conductivity. Manganese is also supposed to raise the oxygen overvoltage of nickel electrodes... at least in alkali.
   I originally thought monel would give the positrode fantastic conductivity since it's normally so corrosion resistant in salt water, but since it in fact oxidizes readily in the battery positrode, that never did work. (...except before I started charging the cell - it did look great then!) Adding traces of cobalt oxide and maybe zinc oxide (I'm not very sold on the zinc, though its conductivity is good), could help conductivity. IIRC there's already cobalt in the mix. Wow: the very stuff I started with in 2008, plus graphite powder (and maybe Mn, Sb, Ca, Zn, Sm oxides to improve it), might actually work great! Did I had good stuff in the first place - workable though not working the way I had envisioned it? In that case the key problems were just the corroding metal terminals and the poor conductivity without graphite powder.

Next Cell

I've put in ALL the ingredients and steps here except making the case. The recipe has become too complex to try to remember everything while working, so that was initially partly to  actually refer to when I made it myself, and then because it had a fair chance of being a workable battery.

   So, the positrode for the next cell would be:

* Carbon rod terminal (salvaged from "D" dry cell).
* Backed by a conductive graphite sheet (boiled in hexadecane and sanded with #40 sandpaper to raise the grain and texture it.)
* 25g of the original mix: monel powder (Ni:Cu alloy), La(OH)3, 1% Co2O3, baked in bean sauce. (Note: next time, I think I'll just torch it after making it, and never mind pre-burning the mix. But I have quite a bit of that mix already made & burned.)
* 10g MnO2/graphite mix - salvaged from dry cells
* 1% Sb4O6 - oops, a big clump went in. I spooned some out. - .6g? I meant to use .35 g.
* 2g Sunlight and sufficient water for stiff paste.
* 16g more graphite powder. This was determined by tamping down the wetted powder with the pestle and testing resistance. I added more graphite 3 times. When it was down to around 10 ohms I stopped.
* a bit more water.

It got a lot fluffier with all the graphite powder, and 25g of the mix, about half of it, looked like a good amount.

* Compacted it (25g) - with the roughened graphite sheet. Hopefully they are well connected together.

The electrode was about 3 - 3.5mm thick.

* painted on a thin layer of Ca(OH)2 to raise oxygen overvoltage.
* A second painting: of zirconium silicate to block any stray MnO4- ions.

* Dried it in the oven at 110ºc for an hour, then torched it just about 10 seconds to sinter and harden. (If it's not dry, it will explode from steam pressure when it's torched.)

   The separator would be:

* A piece of cellophane painted with acetaldehyde that's been doped with osmium powder.
* A piece of Arches watercolour paper. This wraps up the sides to form a little paper "tray" for the negatrode, to ensure there's no contact between the electrode materials.

   The negatrode would be:

* Salvaged D cell carbon rod terminal
* Graphite sheet as above (sometime I may get brave and try metal structures again in the negatrode -- should work [Cu2O(s) + H2O  <=>  2 Cu(s) + 2 OH [@ −0.360v], so in the -1v negatrode it should stay 'charged' to metal] -- but not this time.)

* 1% Sb4O6 (.2 g) - .25 g. I put it in first this time so I could pour back or dump out any excess.
* 25 g of salvaged MnO2/graphite mix. There was a lump that looked like (was) 1/2 a D cell. It was very well compacted and read about 100 ohms. I decided to try and get it lower.
* 2g Sunlight + sufficient water for stiff paste
* 10 g additional graphite powder. Even after that, mostly the readings were in the 100s of ohms. It was in fact too damp to tamp down very hard, and the meter leeds sunk in easily. Some water oozed out of the compactor.

* 25 g of this mix was compacted with its graphite sheet. It came out about 4mm thick.

* Eggwhite painted on (absorbs in), baked in oven. (Although, the eggwhite was already freeze-dried - it had been in the freezer for months and all the moisture was gone.)

   The electrolyte:

The usual KCl plus some sodium borate (borax) in distilled water.

   This cell seemed so promising I spent a day on the 25th to make it.

   The graphite sheets came apart from both electrodes. Oh well!, maybe punch some holes in the graphite next time - or try the carbon fibre. I pushed them firmly together in the cell.


   The cell started off at -.32 volts... reverse charged. In effect, I'd made it backwards.
   The "+" side was mostly monel, the nickel metal being a negative electrode form. That would have to charge to Ni(OH)2, the normal positive discharged form, and then to NiOOH (or maybe NiO2?), the charged form. The MnO2 on the "+" side is a "+" charge, but with a lower voltage than the KMnO4 in the charged cell. The "-" side, on the other hand, was MnO2, again an "overdischarged" form for a negatrode. It would first have to charge to the regular discharged form, Mn(OH)2, and then to the charged metallic form.
   This small reverse voltage drove over 100mA through a 1 ohm resistor. After 5 minutes it was over 60mA, and after 20 minutes, it was still 50mA. It seemed like a good sign! Two hours later, it was still putting out 16, and it recovered to -100mV in a few minutes. Then I started charging it, in the right direction.
   It occurred to me a while later that both sides of the battery would be using up water during the initial "precharge":
MnO2 -> Mn(OH)2 on the "-" side and
Ni -> Ni(OH)2, Cu -> Cu2O or CuO on the "+".

   That meant that after a while I would have to unscrew all the screws, open it and add water - and how many times? ugh!

   I added water twice, and it charged over a period of about four days. It still had very high self discharge.

A Working Battery... by charging it in reverse!

   Well... it had started out with a negative voltage, and it put out 100mA from just a -.3 V starting level. What would it do if I charged it backwards? Although the main ingredients could theoretically work that way around, this seemed contrary to reason since all the additives to increase overvoltages were arranged the other way.

   However, I tried it... and it worked! After some time charging it was at 1.1 volts. I took the leeds off, and it soon stopped discharging at about 1 volt. This probably meant it was charged to MnO4 at +.5 volts and Ni metal at -.5 volts. Overnight it got up to 1.65 volts, and when charge was removed, it was still 1.5 volts after an hour. In fact, it would hold almost 1.5 volts for several hours. When given a load for a minute or two, it would drop in voltage but when the load was removed it would recover to pretty much where it had started from. The current capacity is low with an internal resistance of evidently about 3 ohms, but it's a battery!

Voltages with loads (once relatively stable):
no load:                1.5 V
51 ohms,  28 mA:  1.4 V
11 ohms, 100 mA:  1.1 V
1 ohm,    400 mA:     .4 V

   In May some amp-hours tests will be made - and some calculations of what they ought to be to compare with.

   So what was the chemistry? Obviously the positive side (formerly negative) was manganese - there was no other major ingredient in it. It had a choice of charging to MnO2 at +.5 volts (standard dry cell level), or KMnO4 at +1 volt. The negative side had the monel (nickel:copper) and manganese. The copper would turn metallic quickly - it doesn't have much potential. The nickel would charge to perhaps -.5 volts. The manganese would charge to around -1.3 volts.
   The most likely choice, given that it stops right about 1.5 volts, is that the + side is KMnO4 and the minus is Ni, and that the Mn in the negative won't hold a charge because its voltage is definitely too high without eggwhite.

   That would mean that in previous cells it was really the manganese in the negatrode at -1.3 volts that was causing the self discharge, as initially suspected. On the other hand, the cell with a lower voltage iron negatrode also self discharged, albeit somewhat more slowly. I attributed the slower discharge to the lower voltage of the cell overall, but with the culprit 'obviously' not being the negative side. Now I suspect it was both negatives - the manganese with its high voltage, and the iron too, an element which, though it doesn't corrode in alkali (hence nickel-iron alkaline batteries), doesn't fare well in salt water -- as is well known.

   So for the moment I'm assuming that in the positive side, manganese works at +1 volts as KMnO4 -- at least with added eggwhite, and nickel works in the negative after all -- at least as constituted, from monel with lanthanum hydroxide in bean sauce, flamed, and with added manganese oxide or hydroxide and graphite powder. The best form to supply manganese+ in might then actually be as KMnO4, while adding the monel as metal powder on the minus side. Thus the battery would be made as a pre-charged cell.

   Even if nothing else fundamentally improves this, it could still make the most economical, long life rechargeable batteries. They wouldn't be radically different from standard dry cells, but the energy of charging the manganese from dioxide to permanganate is much higher: the reaction voltage is +1 volt instead of +.5, and discharging to dioxide it moves three electrons per molecule instead of one, so the amp hours per kilogram will be greater. It's possible well made cells could hit 200 watt-hours per kilogram even at a nominal 1.3 or 1.4 volt discharge voltage.

   A potential close to 2 volts may well be attainable, though the list of suitable negatrode elements is looking pretty small. There may well be some way to get the manganese negatrode working - it's evidently out by just a small amount. This would be the best solution, and if I can think of any ideas, I'll try them.
   Otherwise, I'll probably try that vanadium (~-1V) after all. It looks like (all else being equal) it should be about .3 volts lower than the manganese negative, which, with eggwhite added, is likely to make it just enough less that it will stay charged. That should make 2 volt open circuit cells, or about 1.8 V nominal in use rather than 1.2, 1.3 or 1.4.
   I phoned the pottery supply and while vanadium oxide (V2O5) costs much more than manganese oxide (especially from old or new dry cells!), it's probably not more than nickel oxide (NiO) after all, 10$ for a quarter pound, though I haven't compared on a metals content basis or checked other sources or quantities. The store did, however, put their one bag of it aside for me.

Victoria BC