Joule Thief 101

[TinselKoala]

@MH: Without my entering the debate too deeply.... Just what is wrong with the idea of equalizing the impedances of the capacitor and battery by using a variable resistance in series with the capacitor? I'm not asking whether or not this is actually necessary to make the test valid (I don't think it is necessary), I'm asking you to explain what is wrong with the idea. After all, the necessary added resistance would only be a few ohms at most. Approximating the change in battery impedance by using a fixed resistor at the average value would probably be "close enough" as they say.


Meanwhile, back at the Real Data Dude Ranch.... here is the graphed raw data from my testing of last night.  Which circuit is more efficient? Let the happy reader decide.

The total light output over the time of the test can be seen in the Lux-Seconds curves as the integration (total area) of the curves. The voltage vs. light output efficiency can be derived from the Lux-Voltage curves.

[MileHigh]

@MH: Without my entering the debate too deeply.... Just what is wrong with the idea of equalizing the impedances of the capacitor and battery by using a variable resistance in series with the capacitor? I'm not asking whether or not this is actually necessary to make the test valid (I don't think it is necessary), I'm asking you to explain what is wrong with the idea. After all, the necessary added resistance would only be a few ohms at most. Approximating the change in battery impedance by using a fixed resistor at the average value would probably be "close enough" as they say.

Let's see if Brad can figure that out first, it's his pearl of wisdom.

"'Eye sea,' said the blind man."

[TinselKoala]

MH: You may recall Mark Dansie's earlier comment about testing a circuit that was optimized for the particular power cell they are using. It turned out that this optimization included taking into account the output impedance of the power cell. One test that was performed involved substituting a capacitor charged to the nominal voltage of the power cell. This resulted in the immediate failure (smoke release) of the MarkE-designed circuit's DC-DC converter chip, due to the ability of the capacitor to deliver a lot more current at the design voltage than the original power cell could. It turns out that a proper test with a capacitor would use a series resistance (albeit very low, like 0.1 to 1.0 ohms) to avoid blowing the converter chip.

The moral of the story is that one may not neglect the output impedance of the power source arbitrarily. Now, whether or not the battery's output impedance actually matters very much in the present testing of the JT circuit efficiency.... that is another question. After all, the power dissipation of a couple of ohms at say 40 mA is... shall we say.... not very great.
I²R= .040 x .040 x 2 = a bit over 3 milliWatts.

Personally I should think that a real test of the circuit efficiencies should use as low a supply impedance as possible. Maybe PW and .99 and others might like to weigh in on this topic.
After all, many times in testing various circuits we have seen on this forum, the suggestion has been made that capacitors should be used instead of batteries. But also, we have seen cases where the battery type and condition made a big difference in performance. So perhaps the battery indeed should be considered a critical component of the JT.

Opinions? Discussions?


Meanwhile one last graph from the raw data: Voltage vs. Duration
(Note the little bump in the Blue data: this is probably due to my delay or error in recording that one data point.)

[tinman]

Brad, my lightbox was built when I was doing some efficiency tests for another project, to compare with results from some other researchers. This is why I used the 18 inch distance from LED to sensor. The box is 5 inches square by 18 inches long (inside dimensions) and is painted inside with gloss white paint. The Extech LT300 lightmeter sensor dome is in the center of the back end wall. The front end wall is removable for access, and has a small breadboard attached to its inner surface for mounting LEDs etc.  This puts the LED about 1/2 inch from the actual surface of the front wall. The actual distance from the LED under test and the sensor inside the dome is 18 inches, as the sensor itself is recessed a bit in its housing.

I'm using one LumiLED Luxeon 3535L surface-mount LED (part number MXA7-PW57-H001) in the center of the front wall breadboard, soldered to some pins to fit the breadboard. These LEDs have a nice wide dispersion angle and no focussing lens. They are superefficent, but are generally meant to be supplied with 100 or 200  mA  DC max (depending on exact part number). At that drive current they are blindingly brilliant (but need a heatsink), but even at 20 mA drive they are amazingly bright. I've attached the full data sheet for these LEDs below. They are great, you should order some!


   

Thanks for the info TK,i will begin construction of a light box so as it matches your light box dimensions.
I will call into jaycar on the way home tonight,and grab a small bread board,as that sounds a great way to change out  different light globe's/LEDs

Has nobody drawn graphs from my raw data yet?

Sorry TK,i havnt had time,as i have been working on my setup.
MH seems to have much spare time 


Brad

[poynt99]

Personally I should think that a real test of the circuit efficiencies should use as low a supply impedance as possible. Maybe PW and .99 and others might like to weigh in on this topic.

I think I've essentially weighed in on this already with the challenge I posted. The power source was to be a bench supply set to 1.5V. It's the only way to establish a known fixed input condition. The Ro of most bench supplies should be low enough such that 100mW of power draw isn't going to drop the voltage much nor cause ripple on the output.