i used a 1n4148 diode and i tried a 10uf, a 1uf and a .1uf cap. not getting the results i expected. im very interested in figuring out how to acuratly measure the output. if nothing else this is a great learning experience.

A generally accepted way of measuring the LED current is with a current sensing resistor (CSR). You already know that the voltage will be about 3 to 4 volts across the LED. To sense the current, we put a 1 ohm resistor in series with the cathode (flat spot) of the LED and the negative or ground. For each millivolt on the DMM (across the resistor), there is a milliamp of current flowing through the resistor. If you measure 20 millivolts, then there is 20 milliamps LED current.

Some people say this is not accurate - the meter may be misreading the voltage. But even so, it gives you a relative indication; so if you make a change and find that the meter has increased from 10 millivolts for example to 15 millivolts, then the LED is getting half again as much current. The reading depends on the DMM, but I've found that the expensive Fluke DMMs measure about the same as the $4.00 DMMs from Harbor Freight.

 

A generally accepted way of measuring the LED current is with a current sensing resistor (CSR). You already know that the voltage will be about 3 to 4 volts across the LED. To sense the current, we put a 1 ohm resistor in series with the cathode (flat spot) of the LED and the negative or ground. For each millivolt on the DMM (across the resistor), there is a milliamp of current flowing through the resistor. If you measure 20 millivolts, then there is 20 milliamps LED current.

Some people say this is not accurate - the meter may be misreading the voltage. But even so, it gives you a relative indication; so if you make a change and find that the meter has increased from 10 millivolts for example to 15 millivolts, then the LED is getting half again as much current. The reading depends on the DMM, but I've found that the expensive Fluke DMMs measure about the same as the $4.00 DMMs from Harbor Freight.

This may hold true with a constant DC source voltage/current circuit, but is not so reliable with the blocking oscillator configuration where pulses of current are sent to the LED.

 

The circuit may not work to charge a battery. Once a charge builds on the battery needing recharge, it would have an EMF which opposes the charging and charging would stop. Somebody correct me if I've got it wrong.

Most cheap NiMH chargers just continuously push C/10 current through the cell.. C is the capacity in mA-H, so if a cell is 1000 mA-H, then the charger just keeps pushing 100 mA through it. Supposedly this isn't going to hurt the cell

Any charger that does not monitor the voltage can do the same thing. The cell gets a bit warm, but I think the added heat shortens its life.

 

You are probably using a DVM which is measuring a value that is closer to the average DC voltage?

If you want to try something neat, disconnect the cathode of the LED and connect it to the positive power rail. The LED will still light.

hgmjr

But then the coil has to overcome the LED's forward voltage _and_ the battery voltage combined. With the LED connected across the transistor (normal config), the battery voltage does not oppose the coil's V, it aids the coil's V. in this

Also, in the oddball LED config, the voltage across the feedback winding is the same (but opposite polarity) as the primary winding. The primary winding has to overcome the LED voltage and battery voltage, which adds up to at least 4.5V. This is getting dangerously close to most transistors' maximum emitter to base voltage, which is usually 5 volts. If this is exceeded, it can damage the transistor. The damage is a permanent loss of current gain, which can make the circuit start to act unstable, or quit working when the battery voltage drops. The transistor has been 'used up'. Reminds me of this story.

 

This may hold true with a constant DC source voltage/current circuit, but is not so reliable with the blocking oscillator configuration where pulses of current are sent to the LED.

I'm puzzled about what 'reliable' is supposed to mean above. I can make a measurement across the CSR (current sensing resistor) today, and I can come back and make that measurement tomorrow, next week, or next year, and i can rely on those readings being the same as the first reading.

I thought about this and revisited my Supercharged Joule Thief circuit and changed the way I used to make the measurements so that there would be no question as to the validity of the meter's readings. My measurements were not inconsistent with previous ones.

 

Well, I tried your "Supercharged JT" circuits using LTSpice, and I'm sad to say the initial results were not very promising.

I simulated "Figure 2", the upper schematic, and initially got 18.9% efficiency with a 1nF cap; that improved slightly to 22.3% when it was changed to 680pF.

Then I simulated Big Clive's Joule Thief:
http://www.emanator.demon.co.uk/bigclive/joule.htm
and 89.95% efficiency was indicated; which is better than what I thought it would be.

Then I tried simulating your "improved JT" in Figure 1, and it came in at a disappointing 16.5% efficiency. That circuit is essentially the same as the original, except:
Transistor changed from BC337-25 to BC338
C1 changed from 47uF to 22uF
R1 decreased from 1.5k to 820 Ohms.

The Fair-Rite toroids you used would have around 306uH inductance per winding. I didn't simulate the resistance of the wiring, or the parasitic capacitance of the toroidal transformer, as I'm not sure what it would be offhand. Also, the coupling between primary and secondary was ideal, along with the inductor being lossless and immune to saturation.

I'm not sure how you're calculating your efficiency, but it's Watts_Out/Watts_In * 100. The only source in Figure 2 is the one 1.5v battery, and the only load is the LED.
Figure 1 has two sources. I didn't bother simulating the coin cell portion, as it would have minimal impact on the results.

 

But then the coil has to overcome the LED's forward voltage _and_ the battery voltage combined. With the LED connected across the transistor (normal config), the battery voltage does not oppose the coil's V, it aids the coil's V. in this

Also, in the oddball LED config, the voltage across the feedback winding is the same (but opposite polarity) as the primary winding. The primary winding has to overcome the LED voltage and battery voltage, which adds up to at least 4.5V. This is getting dangerously close to most transistors' maximum emitter to base voltage, which is usually 5 volts. If this is exceeded, it can damage the transistor. The damage is a permanent loss of current gain, which can make the circuit start to act unstable, or quit working when the battery voltage drops. The transistor has been 'used up'. Reminds me of this story.

You may want to give it a try before you dismiss it as unworkable. I have made at least six joule thief circuits in this configuration. They work quite well. I have not had a single transistor fail and they are still working efficiently after a couple of years of operation. The flyback voltage developed across the coil easily powers the LED. The LED forms an effective clamp that keeps the voltage across the primary of the 1:1 transformer from developing a voltage on the secondary sufficient to do any damage due to excessive reverse voltage across the base-emitter junction.

I get close to a week of operation from almost every exhausted battery that I use on it. They make very effective night lights.

Maybe we can entice sgtwookie into simulating a JT in the configuration I have suggested.

hgmjr

 

Well, the reverse-bias condition isn't included in standard transistor models, as it would dramatically increase the simulation run-time.

I believe that Ron_H has written some models that do simulate the reverse bias breakdown (using Zeners, I think) but I don't happen to have copies of them, and I'm dubious about the ability to model what's actually happening with that kind of non-conventional utilization.

 

You may want to give it a try before you dismiss it as unworkable. I have made at least six joule thief circuits in this configuration. They work quite well. I have not had a single transistor fail and they are still working efficiently after a couple of years of operation. The flyback voltage developed across the coil easily powers the LED. The LED forms an effective clamp that keeps the voltage across the primary of the 1:1 transformer from developing a voltage on the secondary sufficient to do any damage due to excessive reverse voltage across the base-emitter junction.

I get close to a week of operation from almost every exhausted battery that I use on it. They make very effective night lights.

Maybe we can entice sgtwookie into simulating a JT in the configuration I have suggested.

hgmjr

I didn't dismiss it as unworkable. What I did say is that the LED across the transistor allows the coil to add the battery voltage to its own so that it doesn't have to develop such a high voltage. With the LED across the coil, the coil does not get this benefit. When I put a scope across the LED of the JT, I get a waveform such as the one in the picture in my blog. I read about 4.5 volts across the LED. When the LED is across the coil winding, the voltage developed across the feedback winding is the same as the LED (the windings are 1:1). The negative 4.5V across the feedback winding is getting close to the 5Volts that most transistors are rated as maximum.

When the transistor's emitter to base junction is reverse biased to the point that current flows, the current makes a permanent reduction on the beta or current gain of the transistor. The transistor may still work, but the current gain will never be what it was initially. You can try it for yourself and see. I didn't believe it until I did it to a few transistors and I was surprised that it actually does damage the transistor's gain. Be sure to cut off the leads of the damaged transistors when you're done so they don't get mixed in with the good ones.

 

Well, I tried your "Supercharged JT" circuits using LTSpice, and I'm sad to say the initial results were not very promising.

I simulated "Figure 2", the upper schematic, and initially got 18.9% efficiency with a 1nF cap; that improved slightly to 22.3% when it was changed to 680pF.

Then I simulated Big Clive's Joule Thief:
http://www.emanator.demon.co.uk/bigclive/joule.htm
and 89.95% efficiency was indicated; which is better than what I thought it would be.

Then I tried simulating your "improved JT" in Figure 1, and it came in at a disappointing 16.5% efficiency. That circuit is essentially the same as the original, except:
Transistor changed from BC337-25 to BC338
C1 changed from 47uF to 22uF
R1 decreased from 1.5k to 820 Ohms.

The Fair-Rite toroids you used would have around 306uH inductance per winding. I didn't simulate the resistance of the wiring, or the parasitic capacitance of the toroidal transformer, as I'm not sure what it would be offhand. Also, the coupling between primary and secondary was ideal, along with the inductor being lossless and immune to saturation.

I'm not sure how you're calculating your efficiency, but it's Watts_Out/Watts_In * 100. The only source in Figure 2 is the one 1.5v battery, and the only load is the LED.
Figure 1 has two sources. I didn't bother simulating the coin cell portion, as it would have minimal impact on the results.

I used SwCAD III for a few years, then I upgraded to a new PC and never put it on my new one. I just figured that it was a waste of time when I could grab the soldering iron and tack solder together the real thing in a few minutes and get real results with my freq counter, PS, DMM and LC meter.

The way I calculate efficiency is to measure the current and voltage (typically 1.5V) of the supply, and multiply them to get the input power. Then I assume the LED voltage is about 3.2V for white or blue, and multiply that by the LED current I get from measuring the millivolts across the 1 ohm CSR. I then divide the LED power by the input power.

My calculations are usually about 50 to 60 percent for a conventional JT, and 70 to 90 percent for my Supercharged JT. I welcome anyone to replicate my circuit, and I really hope others do, so they will see that it really is more efficient by a substantial amount compared to the conventional JT.

I really appreciate that you've gone to the trouble to simulate my circuit, but I have no explanation as to why the simulation efficiency figures are so far off. If you put one together on a proto board or even with clip leads, I think you'll confirm my measurements and come up with the same efficiency as I did. Thank you.

 

At the end of my blog, Ronx posted a comment that he independently confirmed what I've said about the higher efficiency of my SJT.

I used SwCAD III for a few years, then I upgraded to a new PC and never put it on my new one. I just figured that it was a waste of time when I could grab the soldering iron and tack solder together the real thing in a few minutes and get real results with my freq counter, PS, DMM and LC meter.

The way I calculate efficiency is to measure the current and voltage (typically 1.5V) of the supply, and multiply them to get the input power. Then I assume the LED voltage is about 3.2V for white or blue, and multiply that by the LED current I get from measuring the millivolts across the 1 ohm CSR. I then divide the LED power by the input power.

My calculations are usually about 50 to 60 percent for a conventional JT, and 70 to 90 percent for my Supercharged JT. I welcome anyone to replicate my circuit, and I really hope others do, so they will see that it really is more efficient by a substantial amount compared to the conventional JT.

I really appreciate that you've gone to the trouble to simulate my circuit, but I have no explanation as to why the simulation efficiency figures are so far off. If you put one together on a proto board or even with clip leads, I think you'll confirm my measurements and come up with the same efficiency as I did. Thank you.

 

At the end of my blog, Ronx posted a comment that he independently confirmed what I've said about the higher efficiency of my SJT.

Do you have an exact parts list for this circuit, down to the ferrite cores ans diameter of wires and number of wingdings...etc.