Is joule thief circuit gets overunity?

[picowatt]

TK, Lawrence, et.al,

Because I now realize I was using odd assignments for channel 1 (A1=CH1 probe, A3=CH1 gnd) and even assignments for channel 2 (A2= CH2 probe, A4=CH2 gnd), I am reposting the following with corrections.  It proably made very little sense as it was!

The main point here was to assist in visualizing the correct polarity of the voltage measured at the input CSR when the battery is discharging.  Past posts I read here pointed to a bit of confusion regarding this.

TK,

Referring to the schematic in your post 469 a few pages back, if you draw a short circuit between the A2-A3 point (scope grounds) and the A1 point (batt+), you will have maximum current flow thru the input side CSR.  Assuming for the moment that the battery (and the short circuit) has an infinitely low internal resistance, the measured voltage at A1 (batt+) would be zero volts, and at A4 (batt- and CSR), the voltage would be -1.5V (assuming a 1.5 volt battery).   

Current flow would threfore be 1.5 amps and the actual battery voltage would be (A1-A4), which is 0-(-1.5), or 1.5volts.

In the recent captures, the negative going ripple on A1 is due to Vdrop across both Rint and the CSR when current is being drawn from the battery.  Any negative voltage observed at A4 represents current drawn from the battery.

Therefore, the negative going dips/ripples on the A1 voltage are due to Q1 turning on and loading the battery with the toroid, which  produces current flow thru the CSR and causes A4 to be a negaive voltage (base current is also drawn to a lesser degree).  During the Q1 on time, the voltage across the LED is effectively Vce(sat), so the LED is turned off because the voltage at B1 is below Vled(on) during this time. (this assumes the schematic as drawn in your post 469 without a second battery in series with the LED).

When Q1 turns off, the voltage at B1 rises above Vbatt as the energy stored in the toroid discharges in series with Vbatt.  The voltage at B1 is clamped at the LED voltage as the LED turns on briefly (the B1 LED voltage must also be corrected by subtracting the output CSR Vdrop similar to the A1 Vbatt correction).

As the waveforms contain some fairly high frequency components/harmonics, all of the caveats regarding wiring inductance relating to accurate voltage/current measurements must also be considered.

That's my take on it...



Hopefully the above post will now make sense.

PW

[picowatt]

TK,

I looked for, but could not find, the core specs and winding info for the toroid.

What did you use?

PW

[TinselKoala]

I am puzzled that wih only an analog oscilloscope, how can one do the spreadsheet analysis?  Without the spreadsheet, how can one do the multipliction and averaging of thousands of sample points?

Well, in the first place you don't really need thousands of sample points, if the sample points are selected intelligently instead of by rote, and in the second place..... how do you think people did these things before DSOs and easy-to-use spreadsheets were developed?  How, for example, was the first digital oscilloscope designed and tested?

One way that has been used in the past with great accuracy is to make an image of the scope trace on paper, then you cut out the waveforms carefully with scissors and you weigh the pieces of paper on an analytical balance.

Another way is to use, instead of an oscilloscope, an integrating power meter like a Clarke-Hess 2330.

Using a completely manual method on my scope displays of some traces from a different, but quite comparable, electrical OU claim, I was able to come to within a few percent of the "theoretical" values obtained from displaying the same traces on a 4CH DSO that did the necessary computations _of many thousands_ of sample points in real-time. Of course the manual process is painstaking and takes a good amount of time, like about 2 hours of sustained concentration to process a single screenshot.

The image below will give you some idea of one stage of the process as I implemented it. Please don't laugh, this really does work "close enough for government purposes" , and the issue of spikes and sampling errors is addressed by doing the process at the correct time scale resolution.

There are other ways, too, working from screen imagery of scopetraces. For example many graphics programs allow you to select or define an area, and then they will return the pixel count of that area, a very effective way of performing a graphical "integration".

[TinselKoala]

TK,

I looked for, but could not find, the core specs and winding info for the toroid.

What did you use?

PW

It's a green-painted toroid from a defunct PC PSU. It has a 10 turn primary of 1.08 mm enameled wire that was already on it, and about 55 turns of 0.70 mm enameled wire secondary that I wound onto it. The bare toroid is 20 mm in diameter.

And I also should point out that the PCB JT has a metal-can 2n2222 transistor in it, not a TO-92 2n2222a or even a metal can 2n2222a. However I am happy (more or less) to pull the one and put the other in, I have all three types in stock. But there are other transistors that will work even "better" if we want to get down to comparing equipment... so to speak. 2n2369a works well, MPSA18 is great.

[TinselKoala]

OK, ok, so I put a genuine TO-18 Motorola 2n2222a in there, along with trying a few other flavors. (The old one was a Motorola 2n2222). The plastic MPS2222a really surprised me. It didn't give a big brilliant light, but at less than 2ma PEAK current draw from the battery it will run for a long long time. The 2n2222a seems the best of the 2222 variants that I've tried but it's hard to see any difference with the 2n2222 by the same manufacturer at these frequencies.

I found a data sheet that shows the differences in ratings between the 2n2222 and the "a" variant.