Maker Pro
Maker Pro

Another Joule Thief Thread

Raven Luni

Oct 15, 2011
798
Joined
Oct 15, 2011
Messages
798
Greetings,

I bet alot of people are sick of hearing about this, I've had a read over the previous topics here and the wikipedia stuff etc. but I'd actually like to get an in depth conversation going about the properties of this circuit, how it can be varied for more efficiency etc.

I was looking at my facebook crap the other week and one of my friends who is a scooby diver posted a pic of her new torch. The ensuing conversation with her fellow divers showed that some of these people are paying up to £500 (possibly more) for a simple light. Most of them are LED based and the more expensive ones have things like microprocessor control for temperature monitoring. What they have in common is a burn time measured in hours Of course I thought "I can build a decent light that would run for days with the change in my pocket".

And you can see where I'm going with this: Joule Thief.
I picked out some decent looking power LEDs to test with - 3 Watt cool white - tiny surface mount things packing the punch of a camera flash - currently using 3 in series and they work great my Joule Thief driver at 3V.

What I'd like to know is what properties can be changed for greater efficiency. I'm using a TIP41A with a 33 ohm resistor and a hand wound toroid (12 turns bifilar). I've found that the resistor value sets the lower dropout limit. The higher the resistor, the higher the dropout voltage, the hotter the LEDs and the cooler the transistor and vice versa. 33 ohms is a good tradeoff allowing for a small heat sink and not cooking the LEDs too much.

What effect would changing the transformer ratio have if any? I've seen examples of large secondaries used to power CFLs and such. What effect would adding a shunt regulator have? Would this result in more constant current being available and a possible need to change to a parallel configuration? I've got more questions but these will do for now. I've yet to measure the current being drawn in my circuit - I'll get round to that tonight.
 

BobK

Jan 5, 2010
7,682
Joined
Jan 5, 2010
Messages
7,682
I have built a "standard" Joule Thief, and looked at the waveforms on the scope. The duty cycle is extremely low, I would estimate a few percent from memory, so you are not getting that much light out of the LED. The efficiency is high because there is no resistor in the circuit, so most of the power is going into the LED. In my opinion, to get a really high-powered light out of it, you would have to find a way to increase the duty cycle.

Bob
 

Raven Luni

Oct 15, 2011
798
Joined
Oct 15, 2011
Messages
798
In my opinion, to get a really high-powered light out of it, you would have to find a way to increase the duty cycle.

I was just thinking - the output is connected across the collector and emitter of the transistor so the spikes occur when it is turned off. If it was put on one side of the transistor, would you get the inverse duty cycle?
 

weird_dave

May 9, 2012
36
Joined
May 9, 2012
Messages
36
I was just thinking - the output is connected across the collector and emitter of the transistor so the spikes occur when it is turned off. If it was put on one side of the transistor, would you get the inverse duty cycle?

Not quite, think about the power source voltage and the forward voltage of the LED....

Edit: 3 LEDs in series that is....
 
Last edited:

Raven Luni

Oct 15, 2011
798
Joined
Oct 15, 2011
Messages
798
I realised that not long after - its the inductive kick that causes the voltage spike isnt it
 

BobK

Jan 5, 2010
7,682
Joined
Jan 5, 2010
Messages
7,682
I looked at my breadboarded Joule Thief again, and I was wrong before when I said the duty cycle was really low. I now remember that that was on simultion that I was seeing that. On the actual circuit with a 1.2V battery running a 3.3V LED the duty cycle looked like about 15%. This would translate to a boost factor of 85/15 or about 5.7 which seems like it should be boosting almost 7V, but then there is the Vce on the transistor, which must be eating up somewhere close to half the input voltage.

This indicates that the Joule Thief (at least my implementation) is <50% efficient. Not particularly good for a boost converter, though not bad for the low voltages and currents involved.

Bob
 

Mongrel Shark

Jun 6, 2012
260
Joined
Jun 6, 2012
Messages
260
I've been looking at the efficiency of the Joule Thief too. Got a few videos on you tube with some hd scope footage if your interested...

I've been meaning to do a video showing the base waveform too. Some circuits can show an almost flat line on the base, with a dip into neg when the LED fires. Sometimes equal to the pos spike from collector to emitter Sometimes less than the pos. Almost never for as long a duration.

Try changing your resistor for a pot (say 100k) and looking at different amp draw and voltage drop in the battery, brightness etc.. as you turn the pot. You can also play around with different caps across the resistor (or pot) to get it to work from lower power.

Also look into magnetic amplifiers. You can do cool things with different amounts of turns on each coil.



I've also been meaning to do a torture test and see if a joule thief can last longer at equal brightness than an "On all the time" led. But I'm not sure how many batterys to use on each circuit or what circuit to drive the "Control" LED with..


Let me know if you'd like to see anything specific on the scope.
 

Raven Luni

Oct 15, 2011
798
Joined
Oct 15, 2011
Messages
798
Maybe use a large value capacitor charged at a specific voltage - that way you get a more accurate power level and dont have to worry about battery levels. The only problem is you cant get anything above 1 or 2 farads thats rated for more than a couple of volts.

Setting up the control should just be a matter of making sure the brightness levels are the same. You can use anything photosensitive for that.
 

Raven Luni

Oct 15, 2011
798
Joined
Oct 15, 2011
Messages
798
I've been experimenting with component values and measuring current drawn from the battery. I was particularly interested in testing the claims about this circuit:
LEDTorchCctB.gif
,
which I've found to be greatly exaggerated but with some merit.

With my setup I'm using 3 3W LEDs in series powered from 2 D cells with a TIP41A power transistor.

The 'standard' circuit with a 12:12 transformer (toroid / ferrite) and 33R resistor can light a room enough to read comfortably and the LEDs and transistor are fairly warm. This draws about 125mA from the battery.

The variant in which I used a 16:12 transformer draws about 160mA without a capacitor with no discernable difference in light level, but much less heat (I wonder where the extra energy is going).

What I have noticed at these power levels is that the resistor value significantly affects the LED brighness (I'd say its proportional but no data to back that up). Now we come to the capacitor. I found that using a 10nF capacitor resulted in a slight increase in brightness and the current dropped to about 150mA - far from the 300% claimed in that article but certainly hints at an optimal resonant frequency. Larger values gave dimmer light and less current. I havent tried anything smaller yet (will have to hunt through my box later).

What I would conclude so far is that the circuit can be tuned for optimum efficiency, but wild claims like 300% improvement are for space hippies. I can only guess that in a low powered circuit, light levels are harder to differentiate and / or no proper reference was being observed by the author of that article. They were also using a 2.7K resistor which would have 'lowered the bar' quite a bit.
 

Raven Luni

Oct 15, 2011
798
Joined
Oct 15, 2011
Messages
798
Ah - looks like I had the meter on the wrong setting (it defaults to ac amps instead of dc) when measuring the original circuit. It actually draws alot more than the 16:12 variant - thats why the latter didnt heat up as much and suggests the heat is basically wasted energy. So theres a picture emerging with regards to efficiency. A number of factors have to be just right. I'm going to have to make some proper test pieces. I'll make a new thread for that later.
 

Mongrel Shark

Jun 6, 2012
260
Joined
Jun 6, 2012
Messages
260
Send me a pm when your do. I will be interested to see what you find.


Claims of 300% improvement are not unbelievable to me. But I'm not seeing much advantage to the circuit proposed to do it.

From what I have seen on my scope. You can get full brightness, or even more than full brightness, at more efficiency due to the pulses. You lose a little between pulses, but not always, there are ways to reduce, and perhaps eliminate the leakage. You can go a bit brighter and not seem to harm the led because it has time to cool in it's off times. If I count the time it's on and the time it's off on the scope. I can work out a ratio, Around 5off:1on seems to be best for brightness. I can get a ratio of up to 12:1 but it's dim. if you get less than 3;1 then you risk not saving as much as your leaking. I've also put my signal generator through the coil I have been playing with today, It suggests that a flat top waveform with a freq between 100hz and 5khz will be best. different frequency's work better with some colored leds. There are tons of variables...


As you suggest the the base resistance to brightness ratio is smooth, although not always linear.

I have tried putting a cap across base and emitter since yesterday too, Found much the same as you. Slight drop in amp sometimes. playing around with some of the variables I mention below, the cap can switch the circuit off or increase amp draw while dimming the light. So it's not always advantageous. Only in certain setups.


I have found a few interesting things the last few days. I've just been running 1 led, but I found I can run a few in parallel with little change in amp draw, sometimes it even drop's a little, light from the first led will dim a little with every one added, but I think overall light produced is increasing for a while.

I can also get up to three in series to run. they get proportionately dimmer and draw a little more amps with each one. So 2 in series will produce around half the light as 1, but twice as much of it if that makes sense. So with 3 leds in series you can light 3x the area at a third of the brightness. Approximately.

You can also run a led or group of leds, across the collector and base, with the anode to the collector and the cathode to the base. It will soak up the negative pulse on the base. this can change the way the circuit looks on the scope a lot. seems to reduce the frequency every time. if you do this with a cap across the base resistor the base led will not dim, it will normally keep the collector led bright too. and any flashing mode triggered by bigger caps will not happen. Although you can get the led really dim on a few micro amps with just led across collector and emitter. Adding the extra led could be a good way to save some waste on leaky circuits with little resistance on the base.


If no cap across resistor then the second led on the base can help find the "Sweet spot" of resistance on the base where you get the best brightness to amp ratio. I use a pot as a variable resistor on my bases. As you turn it up you will see it hits a point where the amp draw goes up a LOT, and the led starts to dim as you put less resistance on the pot, sometimes it will stop oscillating and the transistor will stay on, putting the led out. As you turn the pot the other way it will get brighter to a point as the amps drop, then the brightness will drop with the amps after that point. You can find the same point with a collector to base led. as you add resistance it will dim. the point it goes out at is not quit max brightness on the collector to emitter led, but it will be about the best amp to lumen ratio you will get..


Also worth noting, is that you get some fairly high voltage gains if you don't have a load. I'm seeing up around 18.9v on my scope from a 1.1v AA, with neg spikes on the base to emitter at equal size and the same time.



I have a new coil and a few other interesting findings to share. I'll add a link in this thread when I get the next video up.
 

Mongrel Shark

Jun 6, 2012
260
Joined
Jun 6, 2012
Messages
260
Lol. I thought I posted this earlier. Been sitting in a tab waiting for me to post it :D


I have a lux meter. Was thinking about getting a 4 pack off AA's and putting 2 in series to a LED and 2 in parallel to the JT... Might need to add a pot to the standard light to get the same LUX, or just set by the amps draw. Although most DMM's are not good for measuring amps to a Joule Thief. As the sample rate is way to low.

Also at what point do you declare a winner. The joule thief is going to be on longer. but how bright? I'm goint to have to make a cut off lux level. First LED below that is the looser..

Then People will say I should have used 4.5v and a resistor to run the control. Which I disagree with as the resistor is added loss.

Need to get both circuits as evenly matched as possible, and that's trickyer than it first seems...


I like the 3v series vs 1.5v parallel test because the LED in the 3v will be dimmed a little. as it needs 3.4v for full power. The JT LED will also be not at full brightness. The JT will be using twice the amps. while the control will be using twice the volts... Normally I'd have my money on the control and higher volts. But I like the look of the JT's pulses. I can get them pretty bright and only actually on 1 8th of the time. The leakage seems to be small in between pulses.

If you like I can have such a test up and running in a few days. I'd video the start and upload it then do update videos every day or so...
 

Raven Luni

Oct 15, 2011
798
Joined
Oct 15, 2011
Messages
798
Ok - I've been experimenting with my crude adjustable transformer (ferrite rod with 6 windings of 10 turns each) and have concluded that the ratio has very little to do with efficiency. The higher the ratio the more current drawn but after about 2:1 the light starts dimming again so theres wasted energy there.

Still - nothing beats my original bifilar wound toroid. I'd say the performance is more to do with induction, core saturation etc ie. the number of turns and nothing to do with the ratio. Im not sure if the bifilar winding has anything to do with it - any experts on the subject?

Also, 10nF is the only fixed value capacitor that seems to give an improvement. The transformer and resistor have no effect on this particular variable. This only leaves the putput voltage which has always been about 12V (3x 4Vf LED). The ratio of input to output voltage could possibly be a factor in the overall resonance. I've tested with 1.5V and 3V and the optimum capacitor value is the same - but one is twice the other and if the frequency is proportional then one would be a harmonic of the other (and since its a DC circuit there will be even harmonics).

Also worth noting that in the 'common' circuit (with the resistor connected to the transistor base and not the transformer) the capacitor goes across the resistor.
 
Last edited:

Mongrel Shark

Jun 6, 2012
260
Joined
Jun 6, 2012
Messages
260
Also worth noting that in the 'common' circuit (with the resistor connected to the transistor base and not the transformer) the capacitor goes across the resistor.

I've never made one with the resistor at the coil junction, been on my to do list for a while though. I find it interesting that it would change the placement of the cap so much.

All my experiments are with batterys of around 1.2v or less. As that's when my kids toys and tv remotes stop working, and I want to keep using all the power from a battery. Although 3v efficiency are still very interesting. I really like playing with circuits that run from >.7v and in the micro amps range. Every one I have seem is dim, but they go for ages, and you can still read from the light. I think that needs to be considered in any efficiency tests done. Having a book light that works for weeks on a single AAA seems like a pretty handy thing to me...


Without having done a proper test yet. My current hypothesis is that joule circuits are less eficient than non-pulse or even cap based pulse systems at high brightness. Although low brightness is another matter. I think end battery voltage should be considered in any efficiency testing done. ie I have been able to drain a AA to zero volts (well it went to .0025v with reversed polarity), and have it stay there. If you take say a 1.2v AA and drain it to around .5v then sit it on the shelf for a few days. It will be back up around 1.2v again. It's surprising how many times you can do this. The battery will die in the end. I'm just a little surprised how many times you can get a few hours reading light from a single flat battery...

I've been looking at a few things this week. Found I can charge a cap to around 270v with some coil/transistor combos, lower figures on other coils/trany combos. This works best with a really long coil to the collector. Although I suspect better results could be had with the longer coil on the base, as well as using the base as the negative. I've only just got this to work and am not commenting on it yet., The peak to peak volts suggests this is the best place for high voltage though.. The coil I am playing with today gives a 300v+ peak on the collector to emitter at the same time as a -20v peak on the base to emitter... I'm putting the = of the cap on the collector, and the neg of the cap to the anode of a diode which goes to either the emitter or base. I should note this experiment has very little to do with efficiency as it uses a lot of milli amps. Also, if anyone tries it, hearing and eye protection is highly recommended. I wear a welding mask to protect my eyes from the flash if a cap discharges (I may or may not be doing this for amusement). I find it interesting because it means you can charge a battery from another battery with less charge! Very handy trick in the bush!

Also found you can run the LED on many different positions on the transistor. Even anode on emitter cathode on collector!.... About the only position the led wont run is with the anode on the base.... This did my head in for a bit, but I think I get it now... some positions can be very efficient, Check out this guys video. I notice the LED often starts to light with only one side connected, I think this may be basic wireless transmission. Although I'm not 100% yet as my scope being connected is making it happen easier...

Tried to do a video a few times. I have done to much since my last one, and I kept going over 20 min with not all the info I want to include. Going to try and film a multi part series today to keep videos under 10-15min. Fingers crossed I will have something useful up in the next few hours...
 

Raven Luni

Oct 15, 2011
798
Joined
Oct 15, 2011
Messages
798
I was having a play around with 'shorted' configurations. Some people claim improvements by adding a single loop of fire to 'short' the transformer - mostly you tube videos by space hippies offering no explanation or instruction.

However - I did actually notice something interesting. A small loop of wire will reduce brightness and is pretty uselsss, but a longer piece will yield a slight increase. A very long piece has no effect. The shape and position has no effect. The medium piece coiled and stuffed inside the toroid was no different than hanging out straight.

Unfortunately I couldnt get satisfactory current readings. Although the shorted version showed avout a 5mA drop, the light was dimmer probably due to the effects of the meter's resistance on the circuit.
 

Mongrel Shark

Jun 6, 2012
260
Joined
Jun 6, 2012
Messages
260
There is one good video where someone does this with very improved results. about 3x as bright at a 10th the power, but the circuit is all backwards, and has a variable cap, and some rare earth magnets which he claims to be tuned.. although he ignored my request for a schematic, you can see most of the circuit, but some parts could be clearer. only some coils seem to be affected. I have not been able to replicate, although I haven't tried much. it's on the to do list with a ton of other tests...

However - I did actually notice something interesting. A small loop of wire will reduce brightness and is pretty useless, but a longer piece will yield a slight increase. A very long piece has no effect. The shape and position has no effect. The medium piece coiled and stuffed inside the toroid was no different than hanging out straight.

So it's a secondary winding that is not connected to anything.... Like a Tesla coil.. It's probably sending or receiving low amounts of wireless electricity or flux or whatever you call it when it has no amps..... Try measuring an antenna for 50hz (Or 60hz if that's your mains power) 50hz is 2cm wavelength, I'm not good with antennas, but I'm guessing 4,6,8,10cm etc at as precise a measurement as possible may act as a "receiver coil" for mains power.... I get some pretty big 50hz waveforms (Modified by coil but still 50hz) from my coils with no power connected. one goes to 6v + & - some coils don't go to equal levels on pos and neg sides of the waveform. I haven't played around much, but I intend to look at it all more when I find enough time. A flyback picks up as much as 30v from 50hz btw, in the few tests I have done, the voltage and waveform vanish if I try to measure amps...


I got two videos up on my channel the other day too (still planning more, including a series on everything I know about the JT), if you haven't seen them. I have some scope shots you may find helpfull I even get light with the LED anode on the emitter and cathode on the collector. Did my head in for a bit, then I noticed the -3.4v spike on the collector... Then I had one of those Ahhh moments. It runs on negative power.... Battery voltage can increase too, I don't think I have overunity (doubt that's even possible, although some interesting things have been discovered by those that try, so it on my list), battery drops back to less than start voltage once disconnected. It may improve efficiency though.

In the other video I did I charge a microwave cap to 250v from 2 AA's. I was going to do one with more farads but it takes too long to charge... Scope shot of 300v too, although my scope is not calibrated and a little optimistic. I think it exaggerates by about 12% although it seems to vary a bit. I got 280v in a 270uf cap the other day from 200ma from 1.5v. So i'd say that's about all I'm getting. Just short of the 300-310v seen on the scope...


If anyone thinks my videos are too space hippie please let me know, as well as giving a few suggestions on how to improve them. I try to explain everything, and can provide schematics if I feel it's greatly different from other experiments or upon request. I know my videos are not first class, due to all the clip leads etc, but I do try to make everything replicate-able...




I have been looking at amp draw more this week too. I'm certain now it goes up and down with the voltage spikes. It does with the ones that flash. I first noticed this with a "Penny oscillator" which claims to run on 1uA, it does only draw 1uA when transistor is on, but goes up to 40uA when the emf breaks down. Only my fastest meter can see this as this happens in a few micros seconds, other meters just say 1uA all the time.. I doubt my meter saw the full spike.... Can anyone suggest a piece of kit than can track amps at around 35msps? My best trms meter only does 300sps.... I'll be getting some needle type amp meters next visit to the electronics store. They will be better than a digital display, but still only going to detect pulses up to a few hz. after that they will just give an average.... Suggestions anyone?
 

Mongrel Shark

Jun 6, 2012
260
Joined
Jun 6, 2012
Messages
260
Uh, try 6000000 meters in wavelength, only off by about 9 orders of magnitude...
bob


Well I did say I wasn't good with antennas. LOL. See what happens when I use google instead of starting a new thread...

A lot of people are saying it's 22 feet too. But they are using the speed of sound as part of the formula...

Rather than start a new thread could someone suggest a few size antennas (when I say antenna I mean a piece of wire) that would be tuned to 50hz?
 
Top