Turquoise Energy Ltd. News #42
Victoria BC
Copyright 2011 Craig Carmichael - August 2nd 2011

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

* Advanced "Current Ramp Modulation" Brushless Motor Controller made and tested
* "Battery Sticks": Stacking dry cells in pipes makes using a zillion of them for big (EV) batteries more practical.

Month In Brief (actually pretty brief this time - lots of pictures)

Electric Hubcap System
 * 'Final' version Hubcap motor: "the kit" parts photo, assembled, runs great
 * IR2133 Motor Controller Works Great (Yay!)
 * Advantages of "Current Ramp Modulation" ("CRM") over "Pulse Width Modulation" ("PWM")
 * Finally a great controller (tentatively $499) for a great motor (kit $499!)
 * Regenerative Braking control! (a simple add-on control circuit)
 * Zinc undercoat: spray it on, heat treat it in oven, then spray paint finish coat over it. Forget powder coating!

Nickel - Metal hydride Battery Project (focus: Battery Cases)
 * NiMH Battery Update: bought a pile of D cells, NiMH prices, sold a car battery
 * Stressed solder joints fatigue with road bumps & vibration.
 * New Case Design: Battery Sticks! load NiMH D cells into Plastic Tubes. Fast. No soldering.
 * Motorbike tests, with new controller... and battery sticks.
 * 6V (5 D cells), 10 AH, 1 foot sticks, and 12V (10 D cells), 10 AH, 2 foot sticks: 14-20$ - batteries not included.

Electric Weel Motor Project (Electric Wheel Motor... Rim Motor...)
Torque Converter Project

LED Lighting Project A little energy efficient home lighting, anyone?
 * The kitchen lights: 20 watts replaces 200.
 * LED PVC pipe & mushroom diffuser table lamp (Similar lamps from me: $95)
 * Bedroom: Installed ceiling LED globe or mushroom Lights: Simple - almost trivial!
 * My LED table lamp seems to outshine four Wallmart LED lightbulbs. (YES!)

DSSC Solar Cell Project
 * Overall & Latest Concepts overview... but no actual work.

Turquoise Battery Project (no report)

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. - eg, 10 AH 36V $275)
- NiMH individual Dry Cells (D - 10 AH, $10 -- AA - 2.5 AH, $2.50)
- Motor Building Workshops

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

July in Brief

   July saw valuable new ideas for easily making large dry cell batteries and LED lights, and important electric drive components finally starting to come together.

   I finally assembled an Electric Hubcap motor in the planned 'production kit' form, and ran it with the newly designed and made IR2133 motor controller. The motor ran great, but the controller still had some "glitches".

The parts of the Electric Hubcap motor kit
(body 'ring' parts were drilled & painted shortly after photo, coils were redone with ilmenite)

   The motor/controller set is key to moving cars regardless of everything else, so I then focussed on the controller. It had the very same troubles with spurious very high currents as previous types. Re-reading IR's app notes on driving high power MOSFETs led to a minor change to the power MOSFET wiring... which cured the problem! Since then it's been stable even at the full 42 volts with heavy loading and high stress tests. Evidently nothing more than unshielded MOSFET gate wires 2 or 3 inches too long caused much of the grief and many of the blown MOSFETs and chips over the past couple of years.
   But the new IR2133 controller is the best. With it I was finally able to verify that regen braking is done simply by reversing the motor thrust. This causes it to pump current back into the batteries as it slows down. A simple add-on circuit tied into the brake pedal will activate this process.

   At last I seem to have a fine controller, with the superior "current ramp modulation" and regen braking, to go with the fine motors!

'Production kit' version Electric Hubcap motor

Motor controller with IR2133 as controller chip.
Longer gate wires became twisted pairs for shielding.

   Other things inevitably got neglected. The promising [Mn/Ni]-Ni battery went untouched, as did the torque converter.
   I did do some work on the Weel motor: I put the polypropylene-epoxy skin on the stator parts, painted the metal center, and entered drill co-ordinates for the CNC machine. For 162 hole positions, I finally thought to use a spreadsheet to generate the numbers instead of a hand calculator, making what would have been a very tedious job manageable, and I got the holes drilled.

   And with a dozen bright Cree XM-L-H LED emitters, I did a few more LED house lights. And I've ordered some more LEDs and components - this time a greater variety to try out, and some LED drivers - thinking also of how much fuel LED lights could save in gas cars, and the increase in range (however slight) for electric vehicles. After mid-month, I realized I was turning on LED house lights without even thinking about it. I was already used to the whiteness of the light and it seemed normal. This played some small part in getting my lowest electric bill in decades.

Kitchen Counter Hanging Globe

2 Emitter PVC Pipe Mushroom Diffuser Table Lamp

Bedroom Globe at peak of sloped ceiling - 2 emitters pointing down
Unfortunately, no one seems to make receptacles that fit the electrical boxes used for lights. Instead,
I had to install a light socket and put in a screw-in receptacle for the power adapter, after removing an unsuitable old fixture.

Typical LED light circuit

   I also got some more NiMH D cells, "on sale" at the lowest price yet - 365 US$/KWH list. I made and sold a NiMH car battery ($275), and I now have 250 cells available - 3 KWH - enough for 35-40 Km range with only 100 pounds of weight and for about $1500. Unfortunately, prices (no doubt for all batteries) went up substantially before the end of the month, D cells (in quantity) suddenly going from about $5.80 list to $7.20. (A bit of a shock.) I also note supermagnets have tripled in price and more since February. Seems I got my supplies at the right time, and I wish I'd bought more of everything! These things are made in China. It would seem western society really needs to tear down the barriers erected by corrupt vested interests to making such hi-tech things here, and start learning how to produce our own.

12V NiMH battery in Honda Insight

   I took the battery out of my car and checked it. On the bottom side, some solder joints had come loose. The weight of nearby cells was stressing some of the joins apart on road bumps.
   I'm now trying out an entirely different construction, putting them end to end into tubes such as 5 cells (6 volts) or 10 cells (12 volts), with tube ends that press them together for good contact, and links between tubes. That seems to eliminate all the problems... and the soldering... to make for cheap, reliable car batteries. Any number of small cells are now easily employed to build huge (EV) batteries. And I can just sell cases and let people find their own D cells (or AA or C?) to fill them with.

   On the 29th I tried the controller on the motorbike, with just the first three 12 volt "battery sticks" held on by bungy cords for 36 volts. I barely got a slow ride on level (if not slightly downslope) ground at 40 amps. On August first I tried 6 battery sticks mounted together on a piece of plywood with currents to about 60-80 amps, probably limited mainly by the batteries. The ride was better, but still disappointing. The controller performed great throughout. 9 sticks and currents up to the design limits of the motor and controller might make it work okay, but the 210 pound bike really seems to need a bigger gear reduction to let the motor spin.

   And, I wrote up my latest thoughts on a design of DSSC solar cells using the borosilicate glaze/glass with nanocrystalline titanium oxide that I developed. If I have no time to work on it, perhaps it can at least point directions for future experiments, by me or others. There would be no expensive ingredients in these solar cells, so they could potentially become 'dirt cheap' and solve a lot of clean power generation problems.

Electric Hubcap Motor System

Electric Hubcap motor - kit version, parts, assembly

   I could have put a motor together fastest in one of the three first, larger diameter cases - two of them are ready to use except for paint and some hole drilling. But I wanted to do one in the new production size case -- the first one to be done just as I wanted to sell as kits.

   I made the first flat plate magnet rotor in June. Now I made the outer edge cover of the rotor compartment. I wrapped a couple of winds of 2" PP strapping around the rotor end cover and epoxied it, using the UHMW polyethylene disks from the molds to hold the shape and diameter. After it set, I wrapped and epoxied a couple more winds of 1.5" strapping on the inside. This gave a rotor outside edge almost 1/4" thick, with a flat to set the center ring/cover on and a lip around the edge to hold it centered. (It would have been more ideal if it made the rotor compartment about 1/8" or so wider. But I can't seem to find 1-5/8" and 2-1/8" strapping.)

The $500 Electric Hubcap motor kit: all the parts spread out on the workbench.
(Magnet rotor without magnets is shown - supermagnets and epoxy for the rotor aren't supplied with the kit.)
At the time of this image, the body pieces still needed painting and their holes drilled.
The coils were a reused set, from a recently disassembled motor.

From Top-Left: Magnet sensor board on inner ring, rotor compartment cover with bearing holder & bearing,
1" axle/shaft with SDS taper-lock shaft bushing to hold magnet rotor, plastic bug screen, 2" wide PP strapping to hold magnets, magnet rotor,
some spacers, APP power plug housing, coils & bolts, outer stator ring, bearing holder pieces with spacer, bearing & bearing 'cup'.

   Next I spray painted the PP-epoxy parts with Dolph's polyurethane insulating spray paint (color choice: brick red only - source: Troy Electric motor shop) to make the surfaces heat resistant.

 Then I put the bearing and bearing/planetary gear holding plate on the rotor side, and assembled the axle (the one set up for attaching the sun gear) and rotor.

   Next day (12th) I fired up the CNC machine and drilled the stator holes in both pieces, and in the ones for the three larger diameter motor cases as well.
   I also took the 'used' coils apart and dissolved off the rutile/sodium silicate, which didn't work as well as ilmenite does. I found some of the sodium silicate seemed to have become insoluble all by itself and there were a few sections I couldn't get off. Perhaps if it's left dry long enough, it will all gradually become permanent. Then I spent time sanding the coils, hoping the ilmenite would stick better and not flake off. (It didn't seem to help.)

   On the 16th I spent some hours trying to get washers for the coil bolts. The ones I had were too thin, and when tightened turned dish shaped instead of spreading the force. I had tried before in passing to get some, but Rona and Capital Iron both had only the thin ones. I knew where I wanted to go: Fastener Force. But it was Saturday. I would have had to wait until Monday. I finally found some thicker 'metric' washers at Canadian Tire View Royal. At the end of the day I finally had the motor together.
   The next day I tried to run it out at the car. It ran, but it didn't seem I had the phases connected right. (There are two "wrong" ways out of six where it seems to run but draws way more current going one way than the other, three where it won't run, and one right way. I haven't found a system... yet... to figure them out on a new motor besides trying them out.) While swapping wires around, the controller made some alarming noises a couple of times, and the currents shot way up. Then on one try it blew. Sigh!

Finished Motor, 32 pounds all up.
(finished except for bug screen - must be a better way to fasten it than staples!)

   The new motor ran great with the new IR2133 controller on the 18th, though the controller still had some glitches.
   At about 16 volts, I got no-load readings of 2.2 amps at a steady 500 RPM (35 watts) and 4.7 amps at 1000 (75 watts). These are fantastic figures for a 5 KW motor! As an indicator of how far things have come, the earlier Electric Hubcap motors used 250-500 watts no-load at 1000 RPM.

Recapping the big improvements:
- the PP-epoxy composite stator rings replaced the metal (disk brake rotor) plate, reducing magnetic drag
- the very low loss iron powder coil cores replaced the iron laminate cores, reducing drag & current
- the ilmenite in sodium silicate coil coating creates a low magnetic resistance path that bends 'wasted' lines of magnetic flux around into the core, providing more flux with the same electrical current. This innovation of mine gives it an edge over every other motor.

IR2133 Motor Controller

   After it gradually dawned on me that low low-speed torque from the MC33033 controllers was a big part of the reason the car wouldn't move regardless of mechanical designs, and after all the failures of the A3938 controllers I'd been trying to make since last fall, I designed a new one around the IR2133 chip -- taking the design back almost to 2008 when I actually got the car to move with the IR2130 controller, but with important improvements, especially including the variable frequency 'current ramp modulation' ("CRM", with its implicit overcurrent protection), and needing only two chips instead of four or five.
   The IR2133 can be used as an advanced single chip brushless motor controller. I used certain signals and features in (surely) unexpected ways to achieve that, not as suggested or implied in the datasheets, and along with the missing forward/reverse function - one missing pin requiring a whole second chip to replace - I have little doubt that the maker, International Rectifier, has no inkling that it can be so used.

   The variable frequency CRM does three things that PWM doesn't:
* From motor start and at low RPMs current ramps up quickly and the peak allowed current is rapidly reached. With PWM when the peak current is reached, the rest of the fixed frequency cycle is "shut off" to prevent overcurrent, and there's no more torque until whenever the next PWM cycle begins. With CRM The motor is only shut off for a short fixed off period (lower 10s of microseconds), then the next cycle commences, so (unless the PWM frequency is set high) the average current is much higher.
* As the motor speeds up, it takes longer for the current to rise (owing to back EMF - generated voltage - reducing the effective voltage to the motor.) A fast switching fixed frequency PWM is then wasting energy and making heat in the controller with unnecessary switching. The CRM frequency will naturally drop as the current takes longer to rise to the setpoint.
* Current lags voltage in an inductor (a motor is an inductive load), so using current control instead of voltage control also helps correct the phase angle of the current with respect to the actual positions of the magnets, again improving torque. When the commutation switches as the magnets rotate, the very first modulation stays ON until it brings the new set of coils up to the selected current and torque using the full available voltage. Only a control that anticipates changes of commutation, or magnet sensors advanced from the switching mid-points, can improve on this timing... and such measures are pretty much academic at the Hubcap motor's speeds. CRM is simple and effective.

   Sometimes I've wondered why I was bothering  with my own design when there are commercial controllers out there. I guess it was the lure of potentially coming up with something better - or simpler to make - than what exists. I think the layout is excellent, and the "CRM" will prove better than "PWM", which is what other controllers are probably using.
   Certainly the Electric Hubcap motors have turned out to be exceptional, and people certainly were asking, "Why are you bothering? There's lots of motors around." And a year ago my only real answers were "To get that 'pancake' shape, and an easy to build motor."
   On the other hand, after 3 years until this month I was still struggling to come up with a controller that really worked well at all. Certainly I - an experienced (if somewhat dated in practice) computer and interface circuits designer - didn't quickly understand the rigorous requirements of every detail in multi-kilowatt power switching control circuits. I suppose while I was grasping the basics, I glossed over some finer points in my mind - particularly the need for short gate drive wires, and the need for driver chips to be able to handle heavy voltage spikes in the phase sense lines without damage.

   Near the start of the month I polished up the circuit design and board layout as best I could think of, and e-mailed it off to APCircuits[.com] to get the boards made. Then I ordered the actual IR2133s and a few other parts, and was happy to think of setting it aside for a while. The parts arrived within 24 hours of hitting "purchase" on the Digikey.com web site! The boards arrived the following day. No rest for the wicked! In another day I had put the controller together.

Motor controller with the new IR2133 two-chip controller board

   But I got my respite. I wanted to try it on a motor in the shop under controlled conditions, and for that I needed to put together another motor (above). And I wanted it to be a 'final version' motor. I didn't have it together until the evening of the 16th.
   On the 17th I tried the new motor on the car's MC33033 controller. The motor ran, but while trying to get the phases the right way around, that controller blew, too! Now the only controller that might work (without repair) was the new one.

   In the new controller (new IR2133 PCB replacing the MC33033 PCB) I found four blown mosfets before even starting. No doubt a legacy of the last try on the motorbike. When that was fixed, I found a solder blob short on the circuit board (power to ground), and then a resistor that hadn't been soldered on one end. My eyes aren't what they used to be. I did inspect the board as carefully as I could when I assembled it, using a magnifying glass.

   Before connecting the power leeds to the motor I checked all the voltages and signals I could think of - uncovering a wrong value resistor and a couple of things that needed changing.
   I found that although the datasheets said the 10 to 20 volt IR2133 inputs could handle logic voltages "as low as 2.5 volts", it seemed they couldn't handle any higher than 5 volts! Higher voltages were dragged down to 5.6V. They might have mentioned that! Admittedly there was a clue in the "recommended operating conditions" being 5 volts max, but "absolute max" said 15V. I did wonder why it wasn't 20 or 25V like the rest of the chip. Had I been more cautious, I'd have e-mailed IR tech support for clarification. I designed the whole board around 12 volts. Now it seemed some things would have to be redone and a 5 volt supply added for no other reason than to satisfy this silly 5V logic input voltage limit on a 20 volt chip. For now, I would simply bend up the output pins on the CMOS XOR gate and put in diodes in series to make the gates "open drain", and allow the IR2133 to drag pull-up resistors down to 5.6V. That meant unsoldering the 4070. I had socketed the IR2133 (yay for socketable DIP chips!), but not the 4070, which seemed unlikely to blow. I didn't have the stomach for it that day, so I made an LED light for my bedroom -- something that was highly likely to (and did) work without any special trouble.

   When I did connect the motor, it was through three .47 ohm resistors. Once I had the phases right, it ran! The current ramp modulation seemed to work great, the signals oscillating more or less as they should! After a few more tests, I couldn't see anything amiss or think of anything else to do. I very reluctantly removed the resistors before they got fried. After all the A3938 failures, I could just see the IR2133 now going up in smoke the same way, inevitably taking out a few mosfets with it. It didn't. Hurrah for International Rectifier's great high voltage MOS driver chips!

   After all the troubles, I was cautiously optimistic that this controller would work well and reliably. But 16 volts wasn't 36 or 42, and 22 max amps DC wasn't 127. So I went through a checklist sequence:

- check and decrease 'short fixed off period' with smaller capacitor - it's long
- view the currents on the scope and see what the real instantaneous MAX current is
- try higher voltages: 24, 36, 42 volts
- permit higher currents to the desired max... then (if all is still going well)
- try the motorbike!

- Off period is variable and short enough as is. (I wish there was some way to add a little hysteresis to the switching - that should probably give more uniform results.)
- looked like there were very short transient pulses uniformly of 100 or so amps - these were no doubt transient switching spikes. It looked like currents might be increasing as the control was turned up - or did high currents simply get more frequent? There was lots of hard to figure stuff on the scope, including what looked like some RF oscillation. The oscillation could be seen simply by having the scope probe near the controller with the motor running. Nothing looked really neat and clean... except maybe the motor coil signals. I think a recording/graphing scope would be useful.
- 24 (actually 27) volts: 40 amps, works fine
- 29 (32) volts: 45 amps, works fine
- 36 (40) volts: worked for a while, but there were some very high currents with the control pot turned up even part way. I saw the meter say "150 amps" and then it malfunctioned.

   Here again, the higher voltage caused problems and major current spikes. This time nothing blew up (luck?) - it just quit working. It seemed one(?) phase still worked but not the other(s), and the motor came to a stop. When depowered and repowered at 24 volts, it worked fine again. Seemed suspiciously like CMOS latch-up. Prime suspect: 4070 CMOS XOR gate, even though it shouldn't be getting anything but simple logic voltages. On the other hand, its signals come from the hall sensors in the motor, near the coils. In the last controller design, I had put in some filtering... this one I didn't bother.
   Another possibility was that the power mosfet maximum voltage had been transiently exceeded and caused some sort of latch-up. The IRFP3260s are rated for 60 volts - should I switch to 70 volt rated IRFP3207s? That seemed less likely, though.

   But on going over more of IR's AN978 application note, I became more impressed with the need to keep all the connections short and straight. It said "gate connections 2 inches max" - a figure I'd long since forgotten if I even noticed it on first reading. If it's longer, it's an antenna. But the mosfets need to be spread out on a heatsink for cooling. Even the driver chip itself is 1.6 inches long. This requirement is impossible. But evidently twisted pair gate wires can help for longer runs - the twisted pair gives common mode rejection of induced voltages.
   Evidently also, negative turn-off spikes on the mosfet source leeds (as well as other induced voltages) can cause the gate to become positive compared to the source, spuriously turning the mosfet on at the wrong time.
   Spurious switching would explain much of the trouble I've had. I decided to shorten the gate leeds if possible, and to run twisted pair gate wires as recommended in AN978 to all the gates. Not much I could do about the length of traces on the circuit board, or about having the mosfets spread out for good heat dissipation, which is just as necessary.

- There seemed little need to increase current limiting. If it was to be increased, I would increase it by upping the voltage to the control pot and having higher voltages with more noise immunity, rather than by reducing the 'sense resistor'.
   In fact, I might allow 3mV per amp with a 3 mΩ sense resistor, and set the pot voltage to 381mV for 127 amps instead of 127mV. The noise immunity on the current sensing should be better. But that sense resistor - 4" of #8 nickel-brass wire - would get warm if not hot!

Next Steps...

   Having similar runaway current problems at the same voltages as the MC33033 controller seemed to indicate that the layout was at least a part of the problem. With the high currents being switched, every piece of wire must be viewed as an antenna: an inductor, capacitor and resistor all at once. My MOSFET layout is set up not only to keep the transistors apart for heat dissipation, but to make for short, straight runs on all the high power signals. This is good, but the gate leeds must reach everywhere.

   Evidently the thing to do was to use a twisted pair of wires, and run the gate driver to the gate and also the mosfet 'source' signal to the common return (low side) or to the phase output (hi side). Any spike on the source is coupled to the gate, and external signals are "common mode rejected" - actually coupled to both leeds at once, canceling the effect at the mosfet.
   Consider the spark you get when you touch and remove a 12 volt battery leed from a coil - it can take a notable chunk out of a lead-acid battery terminal post. These are the sort of forces the controller is dealing with continuously, and it's a miracle any solid state controller can work at all! I guess it should be no surprise that some things you wouldn't even think about, like a few inches of skinny wire, turn out to be quite critical.

   I twisted another wire around the longer gate wires including the 1.5" runs between transistor pairs and soldered them at the appropriate points, ignoring only a couple of 1" runs to phase "B".

   Another problem was the motor sitting there and not even drawing current when it was supposed to start turning. Flipping the direction switch and starting the other way usually worked. It seems to be some problem with not getting the floating hi-side supply charged, and I had it in 2008 with the IR2130 too. The one advantage to having the low side transistors stay on when the high side turns off is that the coils are pulled to ground, charging the floating high side supply with each modulation. I've already put in 10KΩ resistors to tug the motor lines towards ground so the floating supply capacitors charge. So a second mod was to try 1uF ceramic bootstrap capacitors instead of .47uF to hold the voltage longer and get the motor to start turning. Once it's turning, commutation brings each phase to ground (low driver on), repeatedly charging its high side supply capacitor. But it has to start turning first...
    In IR's app notes, they talk about using "ultra fast recovery diodes". But then they use some weird numbered diodes that aren't even nearly as fast as a common 1N4148 fast signal diode. That's puzzled me ever since I started, but I finally figured it out: they're selecting high voltage diodes for line voltage motors, and the 1N4148 is only good for 75 or 100 volts. Under about 50 volts, the common 1N4148 is in fact the best choice.

   A third next step was to increase the sense resistor from 1 milliohm to 3mΩ, making the voltage rise at 127 amps .381 volts instead of .127. That provides more noise immunity for the current sense. Thinking a 12" piece of copper wire was really long, I used a piece of #8 nickel-brass (aka "nickel-silver") wire. Measuring the voltage drop with a 1.0 amp current, it turned out a 4" length was .003 ohms. I didn't immediately replace the resistor supplying the speed control pot, so the peak current would be limited for the time being - theoretically - to around 45 amps.


   With these mods, I tried out the motor on the 24th at 17V and everything went well. It started considerably better - it usually started turning with no trouble... unless you turned up the control very gradually and held it back by hand, in which case it would go one step and then cease drawing current on the next one. Then if you turn it off a moment and try again, this step works and it stops at the next one, etc. But the mod has helped. On further trials I found that it seems unstoppable if you just 'tromp' on the gas a bit instead of taking it as slowly as possible. If it proves necessary, I'll increase the capacitance as much as may prove to be necessary for reliable, high-torque starting. I'm not convinced further increase from 1uF is really necessary, but ceramic capacitors up to 2.2uF are under 50¢ at Digikey, so there seems little reason not to order some and use them.

Success - no spurious currents!

   Then I tried it at 24 volts, 29 volts, 36 volts and 42 volts and had no problems. No current readings were seen above about 35 amps. In fact, currents dropped as voltage was increased. At 36 volts I could hardly catch it above the 20s of amps as the motor gained speed so fast; at 42 seldom above the teens. That's a far cry from suddenly seeing 80, 120 or 150 amps 'randomly' flashing by on the meter at the higher voltages. The twisted pair gate wires seem to have done the trick.

Further Tests - Higher Currents, Regenerative Braking

   I increased the control potentiometer voltage to theoretically get 100 amps peak current. If I turned the control right up and then switched the motor on, I saw actual (average) currents up to about 62 amps with the control turned right up. This suddenly applied maximum power at maximum voltage is the scariest test to do. It's where any weak point is most likely to cause everything to blow up. But everything ran smoothly, and the motor accelerated very quickly.

   I also tried reversing the motor from high speed. As I've suspected all along but never been certain of, the current flows backwards - into the batteries - as the motor is forcibly decelerated. Thus, a small external control circuit could enable regenerative braking:
* If RPM indicates insufficient speed (perhaps under about 15-20 Km/H), disable the circuit. (This is to prevent the car from trying to go backwards after stopping.) If the circuit is enabled:
* If "brake" is on (signal from car brake lights), use analog switch to switch from gas pedal pot to brake pedal pot, and reverse the signal to the direction switch in the controller.
   When the car has slowed enough, the RPM sensor will again disable this circuit. Note that if your left foot rides on the brake pedal and presses on it far enough to turn on the brake lights, that will override the gas pedal and you'll be braking regardless of what your right foot is doing.

Note: The MC33033 had an undocumented "feature" - if the motor was turning the "wrong" way, the low side MOSFETs shorted it out to bring it to a quick stop. So The IR2133 is the first controller I've been able to try out regenerative braking on.

   On the 29th, after getting not much of a ride on the motorbike, I reduced the resistor that sets the control potentiometer top voltage, this time from 39KΩ to 30K, which would theoretically limit current to the desired maximum, 130 amps. But that limits the peak current, and the average will be less. At this setting (and 36 volts supply), I saw currents on the meter (amp clamp) up to 75 amps as the motor accelerated. I had to remove the motor from the electronics bench and C-clamp it to a separate table as everything was vibrating off the bench. I reduced it further to 22KΩ. This time I saw one reading of 103 amps and several in the 90s. The motor accelerates so fast that only one high current reading flashes by before currents drop off with speed. I may never be seeing the actual maximum. Then I tried reversing it while it was accelerating at maximum power, full forward straight to full reverse. I wouldn't have dared try this with any previous controller. Everything worked normally - no problems or unexpected effects. (!) I'll try it out on the bike before increasing the current further  - the currents might already be hitting 127 amps, which will become visible under heavier load.
   As anticipated, the current sense resistor got very warm - almost hot in a short time. I may have to cut it from .003 Ω to .002 and drop the sense voltages by 33%. There was little or no heat evident in the power MOSFET transistors, which are around that same resistance but the average load is shared by six (x 2 - high and low sides) and they are mounted on the heatsink.

   It certainly appears I finally have a reliable 5KW motor controller. Theoretically the doubled 120 amp MOSFETs can handle 240 amps continuous, and momentary currents up to 1680 amps(!). I don't plan to go anywhere near those values - a good safety margin is vital, and the motor itself would overheat if driven too hard.

   "Version 2" will have a few changes to reduce gate drive runs on the board to the very minimum feasible lengths. Also since 5 volts is needed anyway, I'll replace the 4070 CMOS XOR gate with a 74ALS86, eliminating any chance of CMOS device latch-up. Other than those points and using a 9 resistor SIP pack to replace several pullup resistors and save some soldering and board space (for the 5V supply), it seems good to go.

Ultrasonic Noise

   Following the main testing sessions, my tinnitus went wild, a loud, high-pitched whistling in my ears being clearly evident especially for a day or two. I had it before recently, too, but was unable to identify the cause... it was probably when I increased the frequency of the MC33033 controller. It will probably take weeks to more or less fade away. (And I've been so careful lately, wearing ear protection at the computer and any source of continuing noise, too, trying to get some peace and quiet back!) The motor coils 'sing' at the modulation frequency increasingly with the current. There's an audible squeal, but it doesn't seem very loud. However, paying close attention, I started to sense at the top of my hearing range that there are very loud harmonics to this 'innocent' squeal, and I'm sure these loud ultrasonic sounds - 15, 20, 30... KHz - are ultrasonic earritants. This is in fact a main reason I chose to use lower PWM frequencies when I didn't realize it made a difference to operation: If there was irritating noise, I wanted to hear it and know it was there. And on a car, a certain amount of sound is actually desirable, as a warning that something is coming. Only during one of the earlier tests did I think to wear hearing protection part way through, since I barely hear anything... until later.

   Fortunately, the motor goes on the outside of the car (dogs beware!), or under the hood. But if a zillion cars are using high frequency controllers switching powerful motors, might it be a bit like car headlights? -- individually okay, but in mass traffic lines, at times one can hardy bear to look towards the glare of beams pointed almost straight in the eyes. Muffling sound is of course possible if it should prove necessary, but it's easier and the need for it is more obvious if that sound can be heard.
   But perhaps that whole concern is best left to a future where people have finally overcome the many obstacles and are finally driving in numbers under electric power! It's probably a good trade because then the rumbling and bellowing combustion engines will at last be following the rest of the dinosaurs to extinction.

Metal Parts Coating: Heat Treating the Zinc and Spray Painting a Finish Coat

   I read a bit more about zinc coatings. It's a good primer that paint adheres well to, but not a good finish. It also seemed pretty soft. I heated the finished parts up in an oven to about 225ºC, about the temperature used when putting it on as a powder coating primer. That hardened it up considerably, though it could still be scraped off with a thumbnail.
   I spray painted some light beige paint over top of this as a finish coat. This seemed to work quite well. It's probably as good as powder coating except for a less perfect looking finish - depending on the spray paint used.

Nickel - Metal hydride Battery Project

   The NiMH Battery Project is back, this time as an exploration of the best ways to make NiMH dry cells into batteries rather than covering electro-chemical aspects. The quality and potential of the chemistry can now be taken for granted, and the problem is we want car & EV batteries and we only have AA or D cells.
   A big leap in battery design concept was made: the elimination of soldering, by stacking cells end to end in tubes with screw-on ends. It simplifies (almost) everything.

Battery Sale - then a big price increase

   All-battery.com had a "Fourth of July" sale on NiMH "D" dry cells, and I ordered 140 more of them - 1680 WH. I now have 250 D cells available - 3 KWH, 95 pounds. That's probably 35 Km or more of Electric Hubcap travel range, and all 250 at the sale price would have been about $1300 to my door. Similar range with lead-acid would weigh about 270-300 pounds, take up much of the luggage space, cost $700 or more, and they would hardly last half as long (if!) even with sodium sulfate added. (Price isn't counting putting heavier springs in the car's rear axle.)
   I also have about .9 KWH in NiMH AA and 4/3AF cells if I choose to employ them, for about 3.9 KWH total. But I'll probably stick to using all identical cells, thought there's no important reason to do so.

   Significant for electric transport, the sale price worked out to only 367 $/KWH, for the first time under $400. (US$ list price before shipping, tax, etc. Lithium types are about 500 $/KWH - straight from Thundersky.com in China to the battery dealer - and up.) I hoped this sale was part of the continuing NiMH price reduction trend and would be repeated and subceeded. (add to spelling dictionary.) Then NiMH dry cell battery packs will become ever more attractive and economic. Lead-acid may gradually become a thing of the past and electric transport more economical to purchase.
   But near the end of the Month they went up substantially, from under 6 US$ list to over 7. This probably applies to all batteries and is part of the price trend of Chinese made products generally.

   For range comparisons: someone reports using around 100 WH/Km in his electric car (a Canadian Electric Vehicles Suzuki Swift conversion, lithium batteries). Another electric car owner (self converted) with lead-acid batteries says he only has about 20 Km range. But it gets him to work... at an auto service shop.

   I made and sold a battery for a Honda Insight hybrid, which had (as usual) a lead-acid 12 volt battery under the hood, that had gone bad. It had to fit in a vertical space, and I ended up making a plexiglass case. (an extra pound of weight, but looks nice!) (I wouldn't be surprised if the unusual shaped lead-acid replacement battery would have cost as much as or more than the NiMH.)

Ni-MH dry cell battery replaces lead-acid in Honda Insight.
Foam spacers can be seen filling in the empty spaces to keep it from rattling around.

   I also made a new frame to hold the hexagon pattern I seem to be employing almost exclusively with D cell batteries, which fits them into a smaller space.

Straight 5 x 6 cells holder for soldering "side 1" (behind)
and hex pattern holder for "side 2" (front).

   For electric transport batteries, I liked the Insight battery's vertical profile. One could place several side by side in a small space. I've looked all over the web for appropriate cases - I can see having to make them. The acrylic is heavier than necessary and considerable work. PP-epoxy?

   All that of course is for new batteries. Cheapest (though laborious and messy) course is still to scrounge old lead-acid batteries (or buy them cheap from a recycler) and renew them with sodium sulfate.

   Another interesting NiMH thing I did was order eight 2 AH, "10C" high rate AA cells. I used them in a new battery pack for the cordless drill. The original packs were 1 AH high rate ("20C?") NiCd cells that would put out as much as 20 amps (though that was pretty much short circuit current). Since NiCd dry cells generally short out after a while and don't last, the originals packs were long since kaput. (To get the above specs, I tested an individual cell that was still working.)
   2 AH NiMHs supplying 10C is also 20 amps. Where the drill had been underpowered with the regular NiMH AA cells I bought at a store some time ago, these new ones gave it back the power it originally had and was capable of. The double amp-hours of the NiMH cells in effect doubles the "C" rating compared to the NiCd cells - something that might not be readily apparent from a specs sheet. And the drill battery was lighter and had considerable empty space in it as the AA cells were smaller than the originals.
   For electric drive, these cells would give you the most power with the least battery on board. But they wouldn't take you very far. For transport, 10 times as much substance at 1C rate with regular cells gives lots of power, plus useful range.

Trouble at the Pass

   Later, on two occasions I had a short dream just as I woke up that my car starter wouldn't turn over. Finally one day I took my own car battery out and had a look. On the bottom side, three of the solder joints had come apart, and two more had weakened and came apart when I pulled it out of its case. The weight of the batteries was stressing some of them on road bumps and they were working loose, and I was lucky it hadn't stranded me somewhere.
   These were soldered quite solidly when I made the battery in February - I tugged and twisted the wires with pliers to make sure they wouldn't come loose - not without redoing the occasional join.
   I repaired the joins and put a piece of 1/2" styrofoam underneath. This should deform and help even out the weight. After the repair, the starter regained some zip that said "better than lead-acid", whereas recently it had only seemed "as good as". The fact that the car continued to start at all with three disconnects vindicates my paralleling all the cells at every second step to support any weaknesses. Perhaps I'll adopt stacking cells vertically as in the Insight battery to reduce the stress - lucky that's the only one I've sold! Perhaps I should do even more, like put in solid plastic supports for every cell.

New Battery Design!

   Then I thought perhaps instead I should change the whole design: put the cells end to end in round plastic tubes (PVC pipes?) and connect them at the ends. Someone had mentioned this being done in hybrid car batteries. I had never considered this to be a viable solution, automatically discarding it without due consideration: one thinks of those unreliable little battery compartments in tape recorders and flashlights, with inductors (coil springs) on one end. But the terminals at one end could screw in and clamp the whole string, to securely tighten the connection points and make the whole thing reliable. The plastic ends would have just enough flex for thermal expansion and contraction. That would eliminate all that soldering. Then, depending on the end connections and pipe mountings, it would be simple to put together batteries. And it would be simple to replace individual cells or all of them.

For tubes with 5 cells - multiples of 6 volts:
* a 20 D cell battery (240 watt-hours) of 6, 12 or 24 volts, 12" x 2.75" x 2.75" (4 tubes of 5).
* a 30 D cell battery (360 W-H) of 6, 12, 18 or 36 volts, 12" x 2.75" x 4" (6 tubes of 5).
* a 40 D cell battery (480 W-H) of 6, 12, 24 or 48 volts, 12" x 4" x 4" or 12" x 2.75" x 5.5" (8 tubes of 5).

For tubes of 10 cells/12 volts, there'd be half the tubes but they'd be two feet long.

   I got excited about this idea and spent much of the next day, the 27th, shopping for appropriate tubes and fittings (instead of trying out the motorbike with the new motor controller). It was only supposed to be one short trip to Rona, but when you're trying to do something new, anything can happen. First I looked all over without being really satisfied with anything. I almost got some plastic electrical conduit pipe at Torbram until I found out the price. I ended up getting 1-1/4" ABS pipe and fixtures at Andrew-Sherret. But when weighed, this seemed rather heavy. It would add the equivalent of 10 extra cells to a 30 cell battery - 33% extra weight. I had seen some lighter PVC pipe at Rona, and I ended up going back and getting a length of it and some PVC fittings. The batteries were an even sloppier fit - oh well. It turned out that this pipe had a bigger I.D. but the same O.D. as the others, so it took the same outside fittings.
   I made up a 5 cell, 6 volt pipe with the PVC fittings, then I realized the ABS fittings were substantially lighter and also fit on this pipe. I made a second pipe, 10 cells, 12 volts, with the PVC pipe but ABS end pieces. The ABS fittings used had a recessed PVC screw-on plug, which made for a recessed electrode that can't easily short to anything. This may be an advantage, or it may just make it hard to connect over to other pipes. In many batteries that's done simply with flat bars of metal with a hole at each end, but here a flat piece won't fit on - unless an extra long terminal bolt (threaded all the way) is used.
   Construction was simple and quick: drill 1/4" holes in the endcaps and put through a 1/4" x 1" stainless steel bolt, sticking out, for a terminal, check the fit and cut the pipe to length, and glue the endcaps on. Slide in the cells and screw on the removable end plug - hand tighten to a reasonable pressure (preferably after the glue is set). They only take 10 minutes to make, and by the next evening I had 7 pipes, and nine boxes of batteries (8 per) were loaded in. AA cell pipes would be about the same with smaller pipe.
   I'll temperature test them in a fridge, a freezer and in a warm oven to ensure the contacts are reliable in any weather with thermal expansion and contraction.
   I made two 17 cell pipes, 20.4 volts and 3-1/2 feet long. 14 of those could mount under the ceiling of the station wagon (pretty much from behind the front seats to the top of the tailgate), to provide my 3 KWH, at 41 volts, for the Electric Hubcap system. Then instead of cargo space, they'd take up only rear headroom that - with my usages - is rarely wanted, and in any event is still adequate, if only just, for average adults.

   The weight penalty is still 20% with the 10 cell pipe and the lighter fittings. The 1-1/4" "irrigation" pipe and the fittings are all a little oversize, and the end pieces perhaps a little overkill in wall thicknesses and strength. There's no 100 PSI of water here. But I looked on the web for a source for slightly smaller tubing closer to the battery diameter and found nothing useful. Well, this pipe is good with appropriate wall thickness. If I find lighter end fittings I'll use them, but the system is servicable as-is. I'll replace my current car battery, and (if all goes well with that for a few weeks), the one I sold, and not have to worry about solder joins working loose on the road.

   So: Rather than making and selling NiMH batteries, I'll simply make and sell D cell battery PIPES, probably with "pipe racks"(?) to mount them. Then people can make their own NiMH car, EV or other batteries. I'll call them "Battery Sticks" - 12V, 6V or custom size. So simple! Inventory will be minimal, and so will shipping costs.

   On the 29th I finally tried out the motorbike with the new controller - and with the three 12V "battery sticks". Currents hit 40+ amps. The cells are rated for 30 amp continuous, 50 intermittent. On the first try I could soon smell overheated plastic. I thought it was my thin jumper wire (a #10 wire that I could yank off for a quick disconnect, there being no breaker/switch), but it was one of the tubes. When I made them I loaded the batteries before I was sure the glue was set, so I didn't tighten the ends. One got left that way, and the poor connections heated the whole pack of batteries and the terminal bolts going through the plastic endcaps. I tightened the screw-on end, and it seemed to go a long ways. That was because it was simply pushing the bolt at the far end right through the softened plastic. On examination, I wondered if PVC would be better than ABS. But then at the other end I found the bolt was half way through the PVC plug as well, so that was little or no better. Conclusion: be sure to tighten them up before use! The enlarged hole now fit a 3/8" bolt, so I put one in.
   I got a bit of a ride on level ground. Then I got out a 6V stick to get 42 volts, but it didn't even seem as good. I brought out another meter (brand new - measures inductance!). The voltage measured 40 instead of around 45, and when I hit the throttle, it dropped to 31 and even 24. In a few minutes of running I had pretty much discharged the batteries with such high currents.
(40A / 10AH = 4C = less than 15 minutes to discharge, and even less owing to the high discharge rate - maybe 10 minutes or less total.)
   In these further tests, none of the sticks seemed to get notably warm despite the heavy currents.
   I should use about 9 sticks to have 3 banks in parallel, or the equivalent, ie the battery packs already made. But the sticks got a good testing at high currents, and a weak point was revealed.

Melted end: the screw-on end wasn't tightened, making poor connecions and heat.

   Another test would be to compare voltages between a soldered battery and a "battery stick" at similar high output current. I tried that, using the "pentagon" headlights panel, and (after measuring voltages at the load instead of directly at the battery terminals the first couple of times) came up with the following representative results:

Soldered Battery (3 banks, 30 AH) Vi=13.74, @ 40.1A after 10 seconds: 12.81V, 1' recovery 13.58V
Battery Stick       (1 banks, 10 AH) Vi=13.74, @ 14.0A after 10 seconds: 12.81V, 1' recovery 13.65V

   It would certainly seem the conductivity through the battery stick is as good as the soldered connections. If there were any poor connections, the 12.8 volts would have been lower, perhaps substantially lower. No telling about durability, but the soldered joins aren't holding out well in moving vehicles, so my money's on the battery sticks. If anyone else would like to buy one or some, I'll happily make them. Batteries not included.

"Battery Sticks" 21 volt, 12 volt and 6 volt, recessed "+" electrodes except on one 6V stick.
A number of possible end fixtures were found.
After making these two - 3-1/2 feet long - I decided to drop the 21 volters.

Higher current motorbike test with 6 tubes

   I got six sticks mounted on a piece of plywood, 36 volts, 20 amp-hours, 60 amps rating. On August 1st I got that mounted on the motorbike. With currents up to 60, 70 or 80 amps instead of 30 or 40, it performed somewhat better, even managing to go slightly uphill. But it was still a disappointing ride. (So I didn't post a video.) I weighed the bike: 210 pounds all up. Add a 150 pound rider and the load is 360 pounds. It did better when I got off and walked beside it, using the throttle so I didn't have to push it uphill.
   Higher currents for further performance improvement will need 3 sets of cells in parallel, 30 amp-hours. The motor coils got warm for once, but the transistors in the controller still felt cool. The .003 ohm current sense wire was warm. The third set of tubes would allow briefly pushing the motor and controller to their 127 amp design limits.
   Or a larger gear reduction (or a torque converter) would improve performance. I wasn't measuring RPM but I expect it never got very high. Twice the revs at the same bike speed would have given it some good oomf.

20 AH, 36V Battery. 30 pounds for 60 cells - 38% extra over the cell weight.
The pipes were plumbing, but the clamps were from the electrical dept.

Best easy way to mount the batteries I could see.

12 volt, 10 amp-hour, D cell Battery Sticks (~2 feet long, 325g, 1950g full)
 1 - 2 | 20 $
 3 - 5 | 17 $
 6 - ? | 15 $

6 volt, 10 amp-hour, D cell Battery Sticks (~1 foot long, 225g, 1045g full)
 1 - 2 | 19 $
 3 - 5 | 16 $
 6 - ? | 14 $

(Now, must get those and the 95 $ LED table lamps onto the catalog on the web site!)

Electric Weel Motor

   On the 21st I applied the polypropylene-epoxy skin to the first plywood stator ring, and the next day to the other one. I must remark they seem pretty flimsy compared to solid molded PP-epoxy composite.

Coating the rings with PP-Epoxy (non-woven PP 'landscaping fabric')

   I had sprayed zinc on the stator center plate in June, now I heat treated it. It barely fit in the oven diagonally. (What will I do for the magnet rotor, which is considerably larger?) Then I spray painted it. I used up most of a can of gloss yellow rust paint that I had on hand. It should be much more rust-proof with the zinc undercoat, and it's a much thicker coating.

   I had marked off one of the stator rings for coil bolt hole drilling, but I finally decided that although a CNC program would take considerably longer to write than simply drilling by hand, it would do a better job, and it would simplify making future Weel motors.
   The waterjet places have software to automatically space points and define lines spread evenly around a circle. My CNC system has nothing, and I couldn't find a CAD program to download that would do it. I could see spending many hours with a calculator doing sines and cosines to figure out 162 X,Y hole positions. But when I went to do it, it suddenly occurred to me to use a spreadsheet. The coil bolt holes were listed in front of me in half an hour. Doing the rest, and then copying and pasting them to the CNC drill sequence programs for the two rings, took longer. It was initially done in a morning, but there were three mistakes in the formulas -- sine instead of cosine or "+" instead of "-"... and each mistake was duplicated 27 times. The corrected numbers all had to be copied, pasted, and rounded off to three decimal places again. But the bulk of the anticipated long hours of tedium were eliminated. I kick myself for not thinking of using a spreadsheet when I did the smaller motors! Within two days I had the stator ring holes drilled. Until now I had considered the CNC machine to be "overkill" in size compared to my needs, but a 30" x 30" bed is none too large for a 28" x 28" piece - I had to shift the center point a couple of times to get it to work!
   Later I drilled and tapped the 27 mounting bolt holes in the metal stator center piece (3/16" steel, 19" diameter). Wow, 27 holes in two separate pieces that all line up! That would have been a painstaking manual job without the CNC machine. I didn't think to do any bolt holes in the plate to mount the motor in a vehicle, or to mount the triple motor controller on it - in fact, I haven't really thought much about arrangements for them.

   Next would seem to be to coat the coils and wire them on. But I've decided to put a second layer of PP-epoxy on the stator rings to stiffen them up, so that comes first.

Torque Converter Project

   In getting the motor controller running, I've done nothing on the torque converter. But here's a little second hand report on somebody else's attempt to improve the drivetrain.

   They are developing a gear transmission especially for EVs, and claiming it'll give 15% efficiency improvement. That's the right direction, and the first time anybody seems to have tried to do much about all the nasty drive train losses or awkward RPMs that sap performance.

http://www.gizmag.com/... [article link]

"the efficiency of electric motors still varies at different speeds - they operate at a peak efficiency of around 90 percent but this can fall to 60-70 percent, particularly at low speed. The question is whether or not it's worth adding a multi-speed transmission to the EV drive-train to optimize efficiency at all speeds. According to Antonov, the answer is definitely yes."

   But is it really so different from other gear transmissions?
Bypassing the drive shaft and turning the wheel from the outside with a mechanical torque converter should still be better in all respects, except for sticking out from the wheel.

LED Lighting Project

Conclusions first...

   I did some more LED house lights - the kitchen ceiling and counter, a table lamp, and my bedroom ceiling. Here are my main conclusions:

* BAD: Screw-in 120 VAC LED light bulbs seem to be an awkward solution for LED lighting. The ones in the stores now seem like toys - very expensive toys - compared to the emitters I'm getting. LED lighting seems to call for new solutions with low voltage DC power. At this point, that appears to mean buying LED emitters and making your own lights. There are some LED emitters and 12 volt DC designs locally at RV and marine places (Industrial Plastics has a good selection), but the real solution for house lighting seems to be to buy them off the web, eg, at ledsupply.com or dealextreme.com, and run them with power adapters.

* GOOD: The best actual tiny LED emitters made for lighting have great quality white light and use next to no power. Diffusers covering all directions are absolutely necessary as these 'point source' lights are so bright they can damage the eyes. Aluminum heatsinks are also required since all the power goes into that tiny emitter and electronics can't run anything like as hot as light bulbs. In the course of doing several lights, I've found a pretty decent 'standard' or 'general' configuration using off the shelf parts.

* POWER: The bright lighting LEDs seem to drop about 2.9 volts, or 8.7 volts for the "triples". The simple and best way to power LED lights from 120 VAC seems to be with DC power adapters. Specifically, 3, 6, 9, and 12 VDC adapters that will supply an amp or more are good for 1, 2, 3, and 4 single emitter LEDs. 9 VDC adapters rated for 350 to 500 mA are good for triple emitter types. Use a small series resistor (eg, 1/2 ohm, 5 watts) to set and limit the current. Keep the current below the adapter's rating, eg, maybe 800 or 900mA for a 1 amp adapter. At and above the rating, the power used from the AC line goes up sharply without supplying much more current to the LEDs, reducing the efficiency and making the adapter run hot. With most power adapters, an extra filter capacitor in the circuit appears to be necessary to minimize 60 Hz flicker/strobe effects.

* LED CURRENT: The emitters are most efficient and run coolest at 1/4 to 1/2 their maximum rating. And they'll probably last longer than you will at those levels. The "2.9 volts, 1000 lumens, 10 watts" emitter will supply considerably more than 500 of its lumens at 5 watts. I'm running them at about an amp, which is only 3 watts but perhaps 400 (?) lumens or better. The 9 volt triple emitters use triple the voltage, 1/3 the current, for similar power and light output. The 'maximum rating' is just the top end of where they won't quickly burn out (given a sufficient heatsink), not a recommended or efficient operating power level.

* LIGHT DIMMER: A 'universal' power adapter with a switch having several voltage settings makes a great 'light dimmer' where several lamp brightnesses can be selected.

* THE PATH AHEAD?: LED lights powered by power adapters can quickly be manufactured and commercialized.
   I'll even do one-off globe or mushroom diffuser LED lights similar to those shown myself, for $95. The lamps are especially simple to use: just buy one and plug in its power adapter. With an external a power adapter, it needs no 120 VAC electrical safety ratings. (Good news for crafts lamp makers!) They still need the heatsink and should have a fuse.

   But retrofit installed LED lighting is almost as easy, done the same way - as long as a power adapter with a visible wire to the light isn't considered too objectionable. The base, a 3" PVC pipe endcap, is simply screwed to the ceiling or wall. The power adapter is plugged into the old light fixture with a screw-in power receptacle, so it runs off the original light switch. There's no wiring, so the light can easily be moved or removed if desired, leaving only a couple of screw holes.

   For new installations, eventually there'll be complete light fixtures with everything - power supply, LEDs, diffuser - all built in. Or perhaps new buildings will start installing low voltage DC wiring for lighting and drop the heavy AC light circuits entirely.

For installed lighting with external light switch, just drop the on-off switch.
For three LEDs in series, use 9V power adapter.
(3V for one LED, 12V for 4)
When checking LED current at resistor, be sure emitters are facing away from you or are covered.

   I should probably mention that I like brighter lighting than most people seem to. As far as I'm concerned (and within reason), the more light the better. I like to heat my office with lights and have both warmth and brightness to ward off SADS on our overcast Pacific northwest coast short winter days and long nights. Elsewhere, and in warm weather, I could save considerable energy with more efficient lighting.

   I ordered a dozen more of the super bright LEDs. When they arrived, I put 3 in series into the kitchen central light, powered from a 9 VDC, 1 amp power adapter, again plugged into a power plug socket that screws into the light bulb socket. A few days later I did 2 in parallel for the light over the sink/counter with a 3 VDC, 2 A adapter, this time hanging a small (6") frosted globe from the ceiling. In both units, I ended up putting .47 ohm, 5 watt resistors in series with the LEDs to limit the current to 1 amp.
   The light is whiter than any other type of lighting. With one LED in a room (4W) I call it "ultra bright moonlight", with somewhat more it might be called "pale" light. In the literal sense it's a cold light - you no that no matter how bright it is or close it you get to it, it has no heat to it at all. My early impressions are that I like it better than tungsten and much better than compact fluorescent (CF). It may be better for preventing SADS, and I suspect it's better for your eyes. Perhaps when it becomes common, fewer people may end up wearing glasses?

Kitchen Ceiling Center Light

   As I expected, the patterned ceiling light glass in the central fixture didn't diffuse the 'point source' LED's light as well as the globe light did. It was good enough, but plain frosted gives much better diffusion.
   I noticed there was a strong strobe effect, too - like old (or new) 60Hz fluorescent tube fixtures. Evidently that power adapter didn't have much of a filter capacitor in it to match its one amp rating!

Kitchen ceiling - center, LEDs mounted in the original fixture.
But using the original fixture seems to be the hard way to manufacture LED lighting.

On: strobe effect interacted with camera to create this odd pattern.

   Someone asked about the efficiency of using power adapters, so this time I checked it out: the AC power going in versus the final DC power out to the LEDs. But my DVM has two current ranges: 0-300mA and 0-10A. When I put it on the high range, it read .16 (occasionally .17) amps. That didn't seem like the best resolution, but when I put it on the low range, it would flash "overflow"... and then instead of reading around 160mA, it read 32mA - 1/6 of the other reading.

120V AC: 160mA (19W) or .032A (3.84W)
9.0 VDC: 1.2 A (10.8W)

   So we can believe almost anything from the transformer using up to almost 9 watts (so, only 56% efficiency), to having it magically deliver the 10.8 watts to the LEDs while only drawing 3.8 watts from the mains (281% efficiency - zero point energy!). For 100% efficiency it would have drawn .09A.
   I decided to check it out with some other current meter before drawing firm conclusions.

   After several hours the heatsink and the power adapter got pretty hot, and I decided to put in a resistor and a couple more bits of aluminum for heatsink. (The fixture would have more room without those silly lightbulb sockets in it!) I used a 5 watt, .47Ω resistor. This time I measured the AC current with an amp clamp on a split-wire extension cord. That gave:

120 VAC .09 or .10 Amps (11 - 12W)
DC: 1.0 amps (9W)

I neglected to measure the DC voltage, but since the power adapter load current was reduced, and because I put in a 4700uF capacitor to stop the flicker (it worked), it would surely have been no less than before - 9 volts. So the resistor didn't reduce the output much, and the efficiency seemed to be over 75% - perhaps partly the meter, and partly because the adapter was now operating within its one amp output rating. That's more like it!

Kitchen Counter Light

   At first I thought I'd try fitting two or three plastic dome lights over the kitchen counter, but the fact that they weren't made for such mounting, plus the need to get the wires from one to the other looked like it would be troublesome. In addition, I thought maybe if the light was lowered, it would light the sink and counter better. This suggested a hanging globe light.

   I found a 6" frosted glass light globe at Canadian Tire, and I got a white PVC plumbing pipe end cap for a cover. I put 3 screws through the PVC to hold the glass on, and a 3" piece of 3/8" lighting fixture pipe for the wires and to hold the works via the center of the aluminum heatsink mount. I drilled 3 extra holes in the end (the top) to let heat out. This globe I hung from the power adapter cord. The power adapter was at the ceiling plug so only the 2 LEDs and their current limiting resistors were inside the globe.

    Kitchen counter hanging 6" globe light
The two LEDs face down.

   This adapter was 3V, 2A. So I put the two LEDs in parallel instead of series, and used a couple of 1.5 ohm resistors (what I had on hand on a Sunday) to limit and balance current between them. This resulted in only .6 amps current, * 3V = 1.8 watts each. The next day I bought a small selection of low value resistors and replaced the 1.5 ohmers with .47 ohms.

120 VAC: .09A (10.8W - amp clamp), 1.1A (13W - DVM)
3.38 V (unit), 2.90 V (an LED), 1.0 A - (6.8 W from adapter: 5.8 W LEDs; 1W resistors)

   Assuming the amp-clamp reading was right (given some weird readings with the DVM), the efficiency was still only 54%. Evidently the higher voltages with LEDs in series reduce the losses. There was at least an extra 1/2W owing to the second resistor, and probably more transformer losses owing to the higher current, with nothing gained by the low voltage.
   The light had some flicker, and I decided to pull it down one more time and add a filter capacitor: if it's going to outlast me, it should at least be done best.

   Inefficiencies and all, now the entire kitchen was about 25 watts for light, with about 15 watts of actual light.
   I couldn't decide if the kitchen was as bright as it was before with two 100 watt bulbs - the difference in color of the light made an exact comparison very much a subjective judgment. Adding a 100W lamp on the kitchen table naturally brightened the room, but it did before, too. This time the color difference with and without it was also quite notable.

Store-bought 120 VAC LED Bulbs

   Canadian Tire had a selection of ready made LED bulbs for line voltage, from night lights to a 10 watt, 450 lumens bulb... for $60. That's probably about as bright as each $9 emitter I'm using, and it looks like their efficiency is under 50%. Somehow spending over $200 to replace a $1, 1700 lumens, 100 watt bulb with 1350 lumens of LEDs in 3 bulbs seems uneconomic to me in spite of the 70 watt electricity savings. The story was much the same at Zellers. They seemed like toys - pricey toys at that.

   Nonetheless, I bought a couple of cheaper ones to compare. One was a dull (60 lumens) decorative "candle" light for $15. Judging by the shadows in a hallway chandelier, it seemed substantially brighter than a whitish 7W CF bulb, but not as bright as a yellowish one.

   The other was a 2.2W 'spotlight' of only 100 lumens for $20 - again not much 'bang for the buck'. I tried it over the kitchen sink. It gave sufficient light over the sink, but the dishrack to the right and the counter to the left were left in relative darkness. My 5.8W globe gave brighter light even on the sink, and this brightness extended in all directions. Furthermore, the spotlight had the worst strobe of anything I've ever seen - I think it must have been turning fully off as the AC crossed zero. It was for that reason unpleasant to wash dishes under. I've heard the filter capacitors dry out and go first, but this bulb didn't even seem to have one. And of course, there was no way to add one to the sealed unit. The heavy glass construction seemed out of place for a solid state, low voltage light that had almost no heat. Why not lightweight plastic?

   Screw-in LED bulbs will probably come down in price and rise in brightness, but I don't think they're a very appropriate format for LED lighting. LEDs are inherently a low voltage, low power DC light. The power adapter approach despite its inefficiencies appears to be the more efficient, and it's more flexible: the power adapters can be replaced, replaced by batteries if required, or by low voltage DC wiring as desired. (I think it would be worth considering putting in low voltage DC house wiring for lighting for new construction.)
   I think in general permanent LED house lighting fixtures should be made complete with appropriate LEDs, rather than considering the LED as an accessory to be screwed into place later by the user. But as I tried things out, I discovered that making globe or mushroom diffuser fixtures to replace existing light fixtures was simple to do -- if one uses power adapters, no wiring is required.

LED Table Lamps

   Doing the globe over the kitchen sink gave me an idea for LED globe or 'mushroom' diffuser table lamps. A freestanding LED lamp would simply have a power adapter plug on the side and a small power switch. Here's a universally applicable idea, capable of infinite variation by designers and crafts lamp makers!

   I found that while the globe fit into the 3" PVC pipe fittings that fit over the pipe, the lip of the mushroom's hole was slightly smaller and fit into the 3" pipe itself. This suggested that the simplest LED lamp structure was a mushroom diffuser, a 3" pipe of the desired length, and a base of some sort. A frosted plastic diffuser could be better - then the lamp would be lightweight and not fragile. But glass ones are what's available, since plastic wouldn't take the heat of incandescent bulbs. The only plastic diffusers I've found so far are the small battery dome light ones.

* Take a length of 3" PVC pipe
* Mount the LED emitter(s) on a heat sink held showing above the top of the pipe
* Drill holes for adapter plug and switch, and wire it all up according to the diagram below
* put in 3 screws 1/4" from the top about 120º apart, and mount the mushroom diffuser covering the LED(s). (WARNING: Don't look at lit LED emitter without diffuser - it's bright "like welding arcs"!)
* put it on a table and plug in the adapter, and turn it on!

    6" mushroom diffuser table lamp with switch
The two LEDs face left and right, hence the darker band in the middle.

   If there's a power failure, or if the light is wanted where there's no plug-in (eg, camping?), replace the external power adapter with a battery pack. (Hmm, I really want a plastic diffuser for camping!)

General LED lamp circuit with 2.9 volt LED emitters.
An installed light fixture doesn't need a switch if it's run from a wall switch.

   I threw the first 'mushroom' lamp together with two emitters and a .27 ohm resistor, and a power adapter plug at the bottom of the PVC pipe. The only suitable 6 volt, 1 amp power adapter handy was my 'universal' one with settings from 3V to 12. It turned out that the different settings made for a great dimmer switch with several brightness levels, each a little brighter than the last!
   It also showed how the power adapter's efficiency drops when its output rating is neared and exceeded. At the "7.5 volt" setting, it drew around .03A (just 3 or 4 watts) from the AC line; at "9 volts", .06A (7 watts); and  at "12 volts", .12A (14 watts - and the adapter gets very hot). At the same time, the LED current was a little under an amp at "9 volts" and a little over an amp with the "12 volts" setting - certainly not double. (I didn't measure at "7.5".) The brightness at only 3-4 watts (1.5 or 2W per emitter) surprised me. You could easily read sitting by it, though 7W was better. The "6 volts" and "4.5 volts" settings dropped the brightness considerably, and the light went out at "3 volts". (In fact, 5.8 volts are needed to overcome the forward drop of two emitters, so the voltage at "4.5 volts" must have exceeded 5.8 with the light (bad pun alert!) load. It's common for unregulated adapters to be considerably higher than they say with no load.)
   Later I added a switch, and a 4700uF, 25V capacitor to reduce 60 Hz flicker. Still later, I found black ABS floor flanges for the pipes that make a stable if not decorous base.

'Installed' LED Light Fixtures the Simple Way: a path to the Future!

   Then it started to occur to me that all that was needed for 'semi-permanent' LED mushroom light fixtures on ceiling or wall was a short length of pipe and the base. The base had screw holes for mounting. But this idea got even simpler!

   Globe light fixtures suddenly started to look ridiculously easy: a 3" PVC pipe endcap, the globe, and (obviously) the light circuit parts. Holes could be drilled through the end of the cap to affix it to the wall or ceiling, one on the side to plug in the power adapter, optionally one for a switch, and a couple more to let heat out. And of course three screws 1/4" from the open end to hold the globe, which would be put on after installation.
   There would be no permanent wiring to add. It would hardly be more work than hanging a picture! To move or remove the fixture, just take the globe off and unscrew the screws holding the base. The only mildly objectionable feature (don't look!) would be the external power adapter and its wire. Where the adapter could be mounted in an existing light fixture with the screw-in receptacle, the same switch on the wall can be used to turn it on and off.
   Essentially, it was my kitchen counter light with the base of the endcap screwed to the ceiling instead of hanging, and the power adapter plugged in instead of hard wired to the light. On the 17th, I put one in my bedroom.

When mounted on the angle of a sloped ceiling, a mushroom diffuser would look more lop-sided than a globe.

Inside the light. (Yes, you can fit the globe over the fanfold aluminum heatsink et al.)

   With a 1 ohm resistor, the current was 1.1 amps, so the LEDs (at 2.9 volts each) were using 6.4 watts - and the resistor 1.2 watts. The AC current was around .09 amps, or about 11 watts power in. This made the brightest light I've had in the bedroom. So the power adapter (6V, 2.1 amps) was using almost 4 watts. However, this same adapter uses a watt or two even without a load, so little better efficiency could be expected from it, except that if its voltage had been a little lower, I could have used a smaller resistor.
   With the 4700uF/25V capacitor I used, no flicker at all was evident.
In fact, this one light continues to glow dimly and then ever more dimly for a couple of minutes after it's shut off. It makes a good night light to see my way into bed after I turn it off!
   Same as the kichen counter, I ran the wire through a plastic clip that I screwed to the light fixture so that if an adapter ever comes unplugged, it won't fall. (in this case onto my bed.)

I mounted the fixture at the next rafter over.

   The same idea could be used with mushroom fixtures. I bought a new 10 foot piece of 3" PVC pipe, and to my chagrin it's slightly smaller than the old one, and the mushroom diffusers don't quite fit in. (maybe if I sand the pipe on the inside?)
   I visited a house where LED lightbulbs from Wallmart were in use in the kitchen. There were four fixtures hanging spread out from the kitchen ceiling. The light, from many small emitters in each "bulb", mostly pointed downwards, though they had better spread than my Canadian Tire one. After you looked towards one, you saw spots in front of your eyes. I put my two-emitter LED mushroom table lamp on the counter. The kitchen seemed much brighter with it than without it, in all directions, and there was no penalty for looking towards it.

   So I'll continue doing it my way with the Cree XM-L-H (2.9V), the nice '500 lumens' 9V ones with no I.D. (DealExtreme.com SKU 5876) or similar emitters, frosted diffusers, and a DC adapter. I'll try the constant current drivers tho - they just might be more efficient than the power adapters & current setting resistors, though less convenient since they have to be wired into the 120 VAC.

Dye Sensitized Solar Cells Project

   I haven't had time to return to the DSSC solar cell project to try and make actual, working solar cells. But I think I should write up my later, though untested, concepts on the subject - before I forget them myself. Some of the planned materials and methods are different from what others have done before, and the ideas are way ahead of the actual implementation. I fear these ideas might go to waste without ever being constructed and tested. Perhaps they might add value to someone's future designs even if I never get back to it myself. So the DSSC Solar Cells project seen mainly in newsletters #28 & 29 is back, if only as a cross section diagram with some explanatory notes, in this issue.

   The first special newer idea was that the borosilicate glaze with nanocrystalline titanium oxide would probably be even more useful as a front cover glass than in a rear reflector. (The first borosilicate nanoparticles were made in Sweden in 2008.) Without having attempted to properly test it, judging by other borosilicate glasses, it should be most transparent in the visible light spectral area (400-700nm) where the energy of sunlight is the strongest, and it should block shorter wavelength UV light, which will prevent deterioration of the internal components and the dye.

Typical borosilicate glass (lower) has very good transparency where the sun's energy is strongest,
but is opaque to the damaging far ultra-violet

   A second idea is to use tartrazine (yellow food dye - grocery store) as a dye, probably in combination with blackberry juice dye, to better absorb the whole spectrum. (The dye currently in favour is an uncommon, pricey substance found at scientific suppliers. If those are the sort of things you try out, they're what you'll find... I just bet more common and cheap dyes can be found that work just as well!)
   Then, a related idea is that the dye(s) can probably be 'set' in the TiO2 to make them permanent. The chemical to do this would be a thiosulfate. Thiosulfates are "used to set dyes in textiles". ...Or maybe here in titanium dioxide.

   A third special idea is to use zinc oxide as the electrolyte instead of iodine and iodide. The zincate ion, so troublesome for using zinc as a rechargeable battery negatrode, can transport the electrons back from the rear electrode to the front.


1. Nanocrystalline borosilicate cover glass
2. antimony-doped tin oxide - conductive electrode collector
3. dyed sintered nanocrystalline titanium dioxide
4. zinc oxide electrolyte space
5. wire grid electrode collector
6. nanocrystalline borosilicate glass rear reflector
7. titanium oxide glaze
8. porcelain bottom cover

Victoria BC