Karen Nakamura: January 2010 Archives

DealExtreme is one of my favorite online stores. It's a distributor of inexpensive electronic gadgets based in China. I'm always finding something new there. The latest treasure is this little-but-very-bright bare LED:

DealExtreme lists it as a 10 watt LED (SKU 5876). Unbelievably it's just under $12 with shipping included!

Looking at the die shows that it is 9 discrete high-powered white LEDs in a single package. DealExtreme is bad about specs, but the comments in the DX forum seem to suggest that 700 mA at 12 volts is a reasonable spec for this LED. This would yield 8.4 watts.

(I'm wondering though if it isn't 3 x 350 mA @ 3.5 serial LEDs in a 3 parallel strings, which would be 1050 mA @ 10.5 volts. But for now, I'll run it at 700 mA).

DealExtreme lists it as 500-600 lumens @ 6500K color temperature.

As with most LEDs, you need a good current regulated driver circuit since you can't just run these things off a resistor. I decided that the easiest and simplest driver would be one based off the amazingly versatile LM317 chip.

As before, these sites have good javascript based circuit diagrams for calculating LED driver circuitry:

• http://diyaudioprojects.com/Technical/Voltage-Regulator
• http://www.reuk.co.uk/LM317-Current-Calculator.htm

Plugging my values (700 mA) into them yielded the need for a 1.8 ohm resistor with my LM317. Here's the schematic that I designed around those figures (courtesy of ExpressPCH):

Bodged together and plugged into a li-ion pack from my model helicopter and voila, an amazing amount of light. I'm thinking of using it on the headlight of my Piaggio (which currently uses a 3-watt LED) or to replace the bulb on my old 15-watt Niterider headlight, which has seen happier days.

(More photos and photometric testing after the jump)

The weather was finally nice enough to commute to work this week on my Piaggio Boxer EV with Prius NiMH batteries.

Here's the data from my CycleAnalyst:

	Run #1 1.18 (To)	Run #2 1.18 (From)	Run #3 1.19 (To)	Run #5 1.20 (To)
Distance	5.2 km	4.85 km	4.21 km	4.19 km
Efficiency	45.1 Wh/km	39.2 Wh/km	42.2 Wh/km	47.6 Wh/km
Energy used	234.23 Wh	189.87 Wh	176.92 Wh	199.03 Wh
Charge Used	5.90 Ah	4.28 Ah	4.01 Ah	5.01 Ah
Max Amps	106 A	101 A	91 A	101 A
Average Speed	20.6 km/h	23.3 km/h	24.8 km/h	24.1 km/h
Max Speed	36.6 km/h	40.3 km/h	39.0 km/h	36.6 km/h
Starting voltage	---	49.9	50.2	47.2
Ending voltage	43.8v	45.9v	---	44.2v
Run time	15 min	12:30	10:10	10:24

My commute is slightly uphill on the way to work and downhill on the way back, which accounts for the difference in energy efficiency going to and from work.

The bike feels much lighter than with the SLAs and faster too (even with the gear reduction) so I have to say it's an unqualified success. I just hope I can get good life out of these batteries.

After Run #4, I was in a rush and so I put the charger on and went to a talk and then came back. About 3 hours had passed and the charger had over charged the batteries. It actually wedged the battery holder apart. So I'm worried now that my batteries will be weakened -- even though most of the bulging has subsided.

I reinforced the battery holder this morning and we'll see how it holds up.

I just bought a cute little folding bike. At first I thought it was a Brompton but a little more research shows that it's most probably an Italian bike, my best guess is the Amica built by the Carnielli bike company. I'm not positive since someone repainted the bike a horrendous orange.

Found a great article that describes how to make a LED voltmeter using a chip designed specifically for that, the LM3914. http://www.evconvert.com/article/led-bargraph-battery-monitor

I'm working on a bicycle assist motor project. I thought about using a hub motor or chain drive but given the economy, I'm going with a friction drive. Friction drive is cheap, has few little moving parts to go wrong, and is cheap. I think I can make it using parts almost entirely found around the shop -- with the exception of the friction roller.

Now what I like about friction drive is if you use a roller with a one-way bearing and take advantage of some physics, the motor can release from the wheel entirely when freewheeling, so the bicycle remains entirely pedal-able on its own.

I decided to go with rollers from the defunct EV Warrior project. They're available on the surplus market, have one way bearings, and are nicely knurled. Other people are making their own friction rollers from BMX wheel hub extensions, but they don't have one-way bearings.

Here are some dimensions almost entirely for my own benefit.

Part	Inch	mm
Shaft OD	0.500"	12.70mm
Shaft ID	0.315" (a tad over 5/16")	8.00
Shaft Width	3.016"	76.61
Roller Width	2.375" 2 3/8"	60.35
Roller OD	1.275" ~1 1/4"	32.38
Key Notch Width	0.130"	3.32
Key Notch Depth	0.411"	10.4

On this cold New Year's Day, I've been thinking about the problem that NiMH batteries like to self-discharge. I'd love to trickle charge them with a low current but my current NiMH charger seems like 1) overkill; 2) liable to slip into the wrong charge mode and boil them dry; 3) too big of a hammer for this little nail.

So I thought of the 1.5 watt 12 volt solar chargers I had bought for my car back pre-Prius. These would be perfect except for the voltage. I think they are actually around 15 or 16 volts nominal, but my packs are 43.2 volts. So I need to boost the voltage....

Scrounging around the web, the best solution appears to be the LT1070 chip from Linear Technologies (www.linear.com). It requires minimal external components and comes with a through-hole TO-220 package for us non-SMD people.

So... how to design the right circuit..

This is my back of the napkin calculations using the design notes and should be taken with a huge grain of salt. Do not trust my calculations!

Vin = 14 volts
Vout = 48 volts

R1 = R2 * ( 48v / 1.244 - 1) = 46.606 k ohms
R2 = 1.24 k ohms

Duty Cycle = (48v - 14v) / 14v = 70.8%

L = (14V * ( 48V - 14V)) / (0.5 A * 40 kHz * 48 v)

L = 476 / 960,000

L = 495.8 uH

C1 = arbitrarily 100 uF with a low ESR
C2 = (48v * 1A) / (40kHz * (14V + 48V) * (0.33 Vpp))
C2 = 48 w / 818,400
C2 = 58 uF

Here are some of the magazines that I've been reading recently in my quest to beef up my robotics / electronics skills (disclaimer: Amazon referral codes embedded):

• Nuts & Volts -- one of my favorites as it has projects from beginning to advanced
• SERVO Magazine -- very useful nuts and bolts robot construction

• Robot -- more glossy and less informative than Servo

• MAKE: Technology on Your Time - fun projects but often not enough detail. expensive
  

Amazon is also offering $5 off eligible subscriptions until Jan 31st, so now is a good time to bite!

Happy new year! A new decade!

I've been playing with LEDs for my EV and robotics projects. It doesn't seem to make sense to use incandescent bulbs in an EV build -- it'd ruin the whole concept of going green.

LEDs are tricky to deal with though, especially the high-output "star" type LEDs that are emerging. Rather than voltage regulation, you have to regulate the amount of current that goes through them. This isn't fixed, because as an LED heats up, its resistance goes down (unlike an incandescent filament whose resistance goes up as it heats up, thus self-regulating). If it gets too hot, it goes into thermal runaway and you soon have what ledophiles call a Dark Emitting Diode (DED) -- dead, get it?.

So, you need some form of a current regulating system. For small 3mm or 5mm LEDs, people just use a fixed resistor since the current demand is rather small, around 20 ma. This limits the maximum current that can go through -- but it also limits the max brightness because you have to put in a safety factor and you can't easily adjust for fluctuating voltage.

Here's a good javascript calculator for series/parallel LED resistors:

http://led.linear1.org/led.wiz

The problem with 5mm LEDs is that the clear plastic casing limits the amount of heat that the LED can output (and yes, LEDs do produce waste heat, although not as much as incandescent lights). Heat control is one of the main factors affecting the output of LEDs and the reason why manufacturers went to the star configuration, which allows you to directly back the LED with a heatsink -- which lets the LED current jump from 20 mA to 350 mA with a concomitant light output.

---

Now, if you want to use a high output Luxeon or Cree star, you also have to current regulate as mentioned before. The typical white high-output LED takes 350 ma with a forward voltage drop of 3.5 volts. With a light output of 120 lumens, this is good enough for a moped, scooter or bicycle headlight.

Cree even has a high-power star that consists of four of their 120 lumen LEDs mounted on a single die. This produces 480 lumens, although you'll need to regulate four x 350 ma. There are other stars that bundle 2 x 120 or 3 x 120 lumen LEDs. More than enough to blind you -- or for a car or motorbike headlamp.

For high power LEDs, the LM317 seems a good choice for current regulating at a low cost. Here's a good javascript calculator for that:

http://diyaudioprojects.com/Technical/Voltage-Regulator/

and some more info on why current regulation is necessary:

http://users.telenet.be/davshomepage/current-source.htm