Joule Thief

[Pirate88179]

Bill:

I can envision what you are talking about.  There is a technical explanation for that that could apply here.  You can excite various components at high frequencies, not necessarily high voltages, and see very fast pulse trains across them.  The AC impedance of the physical part itself comes into play.  So it's almost a "fake" voltage.  You see voltage spikes but that's due to the inductance of the wire in the part.  The actual diode itself will have an AC impedance.  So the LEDs light up, and voltage spikes are kind of "sprinkled on top" of the current flow through the LEDs.  So the assumption is that this process can take place at high voltages also.

I am not challenging your observations about the apparent brightness, just saying that there is a voltage "fuzz" on top of the signal that is actually powering the LEDs.

You would probably need a proper ultra low inductance current sensing resistor to see the actual current flowing through the circuit.

With respect to the question of what waveform makes the LED appear to be bright, there are other issues at play like the longevity of the LEDs.  It may be that a "high spikey" current going through the LED makes it appear to be brighter, but at the expense of the premature shortening of the life of the LED.  I am just speculating of course.  You have to assume that there are diminishing returns as you increase the current flow, and of course eventually heat dissipation becomes a problem.  It makes me curious about what the current waveform output looks like for a "real" LED power supply used for architectural lighting.  There are fixed and dimmable versions.  Will there be tell-tale inductive discharge spikes or will it look like DC?  What about a dimmable LED power supply?  I cheated and peeked at a chip.  It fancier than and better than my suggestion for the previous chip that Farmhand pointed out.  They do variable PWM, pulse width modulation, of a continuous fixed-frequency control signal that goes to the transistor base input.  So there is always a current pulse stream going to the LED, and the size of the pulses is controlled by the PWM signal.  That makes sense, whereas my suggestion for Farmhannd's chip was a bad kludge.

What's interesting is that there is a good chance that the inductor has continuous current flowing through it.  It either goes into the LED or goes into the transistor.  When it goes into the transistor the coil is being energized.  So the LED sees a full-blown pulse train that looks like a square wave going from no current (like "ground") to a certain level.  The LED never sees the current/voltage decay to zero.  That's because below a certain current level, the LED is useless and not doing its job.  So to be more efficient they avoid the decay to zero alltogether.

Even though I am talking about current all the time here, in this case you can look at the voltage across your LED on your scope.  You should (I hope) see a nice pulse train with sharp rise and fall times and a slightly slanted "hat."  That's telling you that there is a sharp on-off current pulse going through the LED at sufficient current level to light it up and no current is being "wasted" with the useless decay to zero.

The key to all of this is to use a relatively small inductor, not sure if there will be a core, and switch that sucker at a relatively high frequency, perhaps between 30 KHz and 60 KHz.  It's small and cheap and does the job.

MileHigh

P.S.:  You would have to be worried about your scope here and make sure that it's not grounded if you were going to peek at the waveform across the LED if in any way the LED was deriving its power from the mains.  Disconnect the third prong.  I assume that all scopes have their own isolation transformers for their main AC power input.  Honestly, I would still buy a real isolation transformer to keep handy all the time.

MH:

Thanks for the reply.  I am glad to see there might be a classical explanation for my observations.  As you know I have not put any of these flash circuits on the scope but, I do check amp draw and the Cree 60 watt equiv. draws 210 mA's, which is really not bad for the amount of light.  It is rated to require much more than that from the mains, of course, and, I am not getting mains brightness for sure.  BUT, given the power used and the voltage supplied by a "dead" AA battery, there is a lot of light here for many, many hours, so, while not OU or anything,a worthwhile effort for me.

I like your explanation here because it seems to fit with my crude description of "fooling" the leds that they have what they need to light up.  Maybe like a surface charge on a battery to show more voltage than is really available.  We all know about duty cycle and how our eyes are fooled into seeing constant light when it is pulsed, so, that is routine with any JT circuit.

What I guess I am trying to say here is that I am lighting led modules with so little power from their rating that, they should not even turn on, but they do and are pretty bright.

TK has some of my circuits and a little led module I sent to him so maybe he can add his thoughts to this if he chooses to do so.  I would be happy to send one to you as well if you would like to see what it can, and cannot do.  I was wanting to market these on my website but, after seeing TK getting shocked by one in a short video he made, I might have second thoughts.  I get nailed all the time by these even though I "know" better.  I can just see the lawsuits, ha ha.

I just think it is funny when I go to Hackaday and see some guy lighting 10 leds using 12 volts, and they think that is great.  They also seem to think the JT is just for lighting 1 led.  I can light hundreds on an AA but they don't seem to be interested.

Thanks for the reply.

Bill

PS  PM me if you would like one of these circuits to play with.

[MileHigh]

Bill:

Thanks for the offer but I am fine.  Let me describe the "Cadillac" Joule Thief "prototype station" or "test bed" for fun.

When you strip a JT to the bare bones what you have is a transistor switch that energizes an inductor, then the switch switches off and the inductor discharges through the LED.

There are several variables that you can change to experiment with the magnitude and shape of the current/voltage pulse generated when the inductor discharges through the LED or LED array.

The coil energizing voltage
The inductance of the coil
The initial current when the coil discharge starts
The amount of time the transistor is on *
The amount of time the transistor is off *

(* aka the frequency and duty cycle of the pulse train that controls the transistor)

In addition to this, naturally you would want to measure the average current consumption so you would know your input power.

You could use a signal generator or a dual 555 setup to have full control over the signal that controls the switching on and off of the transistor.  For sure there are dozens of Arduino programs that will do that for you.

Then you can just play with the variables to your heart's delight, looking for the most bang for your buck.  The whole thing could be done on a single breadboard.  You check the spec sheet for your LEDs to check the recommended operating current and maximum current vs. time for the LED to make sure you don't blow it.

Supposing you find a sweet spot.  Then comes the real interesting challenge.  The challenge is to create a Joule Thief circuit that emulates the coil and timing configuration that you got from your test bed.  So you have to engineer a Joule Thief.  At least you have the coil and the power source already done.  So perhaps by playing with the number of turns in the secondary coil that goes to the transistor base input resistor you will be able to get the timing to match.  Then you would have engineered your very own optimized Joule Thief.

MileHigh

[Pirate88179]

MH:

I like your idea.  I actually have a test rig I made for testing the HV/JT circuits in parallel, which shows a gain of light output for adding the second circuit, the third circuit added actually hurts the light output.  Now I am going to see if I can run two of these circuits in series...I do not know if that is possible but, we will see.

With your testing system, you are right that you could alter each of the variables and hopefully obtain the optimum configuration.  The only thing I would add, as I am now doing, would be to have a trimpot in the circuit so once you run on battery power, you can adjust the circuit to stay in the best conditions as the battery depletes.  I purchased a bunch of these trimpots (20 turn) for this purpose.

Once again, I appreciate your input.  I still believe that we have not reached the end for designing and building the "perfect" JT circuit.  I still think that more research needs to be done using JT's that output at least 350 volts.  I would like to play with the frequency and the duty cycle as you have suggested and your test rig would fit the bill.

Thanks,

Bill

[crowclaw]

Hi Pirate, MH has answered the question exceptionally well. finding the ideal sweet spot by experiment is certainly not an easy task to accomplish, bearing in mind circuit characteristics vary so much and change considerably with applied voltage. In essence providing the spikes are of sufficient amplitude and the current is not allowed to cause thermal runaway led's will provide some degree of brightness. The art is as mentioned finding the ideal circuit balance and efficiency becomes the ultimate challenge. The JT circuit however is not necessary the best circuit for the task from many angles. MH mentioned the use of the popular 555 timer or alternative an arduino. For some time I had outdoor led lighting using switched inductors producing narrow high voltage spike pulses driven from a 555 astable incorporating inbuilt thermal  compensation for outdoor temperature changes.
Crow

[Pirate88179]

Hi Pirate, MH has answered the question exceptionally well. finding the ideal sweet spot by experiment is certainly not an easy task to accomplish, bearing in mind circuit characteristics vary so much and change considerably with applied voltage. In essence providing the spikes are of sufficient amplitude and the current is not allowed to cause thermal runaway led's will provide some degree of brightness. The art is as mentioned finding the ideal circuit balance and efficiency becomes the ultimate challenge. The JT circuit however is not necessary the best circuit for the task from many angles. MH mentioned the use of the popular 555 timer or alternative an arduino. For some time I had outdoor led lighting using switched inductors producing narrow high voltage spike pulses driven from a 555 astable incorporating inbuilt thermal  compensation for outdoor temperature changes.
Crow

I agree with your description of MH's input.  Excellent advice for experimentation.  my problem with 555's and the like is that they need too many volts to run.  (Not for testing mind you)  What I want to learn about is what is under those gray blobs in my solar garden lights.  We have decided (as a group here) that under that blob is the chip that makes up the JT circuit.  The green resistor looking component is an inductor I have been told.  So, I want to know what chip is under there that will operate down to .6 volts and perform like a JT circuit to light an led requiring 3 volts from a 1.2 volt rechargeable battery.  If I could get some of these, I would make circuits using them.  I would also try the things that MH has suggested to perfect it.  I can not find any chips out there that might do this.  Is this why they are hidden under that glob of gray goo?  Possibly I can do some surgery with a dremel to see what it might be.

Anyway,  I have nothing against the chip approach except that the 555's all want like 5 volts or more to run. 

Anyone know what that garden light chip might be?

Bill