Hi Professor,

I bought some components i was missing and was able to replicate your circuit. It now shows waveforms similar to yours - input current and power oscillating around zero.

Using a 10mF capacitor instead of a battery, i let the circuit run and the capacitor discharged from 2.48V to 1.50V in 104s which i believe gives an average input power of 37uW.
For this setup i used a 2N2222 transistor, Cb = 151pF, Rb = 1MOhm, Ro = 0. The led is red, very bright and needs very little current to light up. It was very visible as shown in the picture. The other picture shows part of a cycle of the input current. The frequency is around 10.6Khz made of very short pulses. I didn't measure the real capacitance of the 10mF capacitor which may have a higher value.

If i try the same setup but change the transistor with the 2N3904, the led is dimmer but perfectly visible and the capacitor discharges in 296 seconds which i believe corresponds to an average power of 13.18uW.

I'm not sure we can conclude much from these results and it's very difficult to compare results from different experimenters since there is no way to measure the led brightness. If i use the diode i get similar discharge times (a little longer).

I also tried running the circuit from the capacitor and charge a second 10mF capacitor via a schottky diode. I let the source capacitor discharge to 1.50V and at that point disconnected the charging capacitor. Calculating Ein and Eout, n is around 0.42.

Regards,
Jaime


EDIT: The coil has 20-21 turns in the primary and secondary with what appears to be a normal ferrite toroidal core. The inductance is unkown.

good morning Steven


i've been running my SJ1 variant tests from a single  (depleted) AAA NiMH, and the voltage has been around the 1.12-1.15V range

just for fun i constructed another unit to add to the comparative results you're getting back from folks

the circuit is a minor mod to one of my earlier tests with a variant of your SJ1 circuit (see schematic below for this test circuit)

C1 0.022uF
D1 1N5187 (schottky)
Q1 2N3906
T1 approx 50:50:100; 0.45mm wire; tri-toroid (ferrite)
C2 1000uF (nominal)
L1 approx 0.5mH
LED1 6mm(?) HiBrite (visible, but not bright)

i'm using a tertiary winding to decouple the AC o/p from the DC operating conditions of the oscillator


this circuit takes 433 seconds (7min 13sec) to discharge a nominal 1000uF cap from 2.24V to 1.5V

C2 1000uF (nominal)
2.24V => 2.509mJ
1.50V => 1.125mJ
         -------
    Ein: 1.384mJ

Pav: 1.384/433 = 3.2uW


this is all just ballbark at the moment, obviously - to be more accurate, the 1000uF cap would need measuring

also, the pulse repetition frequency is outside your stated conditions, so this circuit doesn't qualify for your competition

it starts at approx 28Hz, with no visible flicker, and the frequency increases as the supply voltage falls

the pulse 'burst' is approx 15uS long


since i have the o/p DC decoupled with the tertiary wind, i can now revert back to the NPN config of your original SJ1, so i'll be able compare efficiency between the two

[EDIT:  i'll also try with a 1N4148 (& then a schottky) replacing the LED, to see the effect on discharge time]


on the subject of the lower frequency limit for your stated conditions, i feel that in general, any frequency of LED drive which the eye perceives as 'continuous' should be counted in - this would be one of the 'design limits' for a commercial lighting product

i agree that care would be needed, in these tests, to ensure isolation from utility & broadcast sources of 'ambient' energy,  but it's not difficult to discount these other sources of input by using a metal case, or taking measurements with the device in a microwave oven

in fact, with wi-fi, bluetooth and cordless phone signals around the home this sort of shielding will likely be advisable anyway, in which case the lower frequency limit does not need to start at 200Hz

just my 2c


good idea to start a competition, looking forward to seeing how this develops!


greetings
np


http://docsfreelunch.blogspot.com

   One of the problems with this test is that the frequency will change in the devices as the batteries or charge cap drop in charge.  So their ideal running voltage and frequency sweet spot will be vary, and be lost.  Especially when using a weak 1000 uf cap. The special quality of these low microamp circuits will be overlooked. 
   Question:  if these circuits are really running on micro amps, why are they discharging their source battery or cap as quickly as they are?
   I've got 4 BwJt running every day 24/7, they all have the same trans, cap 103, and pots, and no resistors.  They all work very different since they have different coils on them.  The leds are also all different. All of them drain the battery in a day or two.  There are even difference in using the same type of transistor, etz...
   If all the test are showing that the capacitor or batteries are being discharged, and are not maintaining their original voltage, this shows that there is no real or special efficiency with this set up. Just another Jt draining its source.  What it the point???  On the other hand if you had a unit that does NOT get discharged at all, but instead charges up to a higher value (like Koolers do), that would interest me.

Dr. Jones' Contest Conditions:
Required is a JT-type circuit, with at least one transistor and one bifilar (or higher)-wound toroid and it must light an LED to observable brightness in a lighted room.   The final test is conducted with a 1N4148 replacing the LED, and the device placed in a microwave (OFF) or other Faraday cage to exclude ambient energy sources. (The winning device cannot be poaching from the local electrical grid; operation at 50-60Hz is excluded.)   The input power is to be measured by the capacitor/time method:
Ecap = 1/2 C V**2
so
Pinput = 1/2 C (Vstart**2 - Vstop**2)/time

Where Vstart is approx. 2.55V and Vstop is 1.5V.   
The circuit with the LOWEST Pinput wins, and the award amount is currently:
$100 - ($Pinput in uW/10).

(18 June 2011)

Of course, this method for Pout neglects the power dissipated in the LED (and other components)...