Author Options:

A More efficient joule thief Answered

I need to make a joule thief type of thing that can up the power from a 0.55V source to a simple LED. I know that the joule thief is not the most efficient circuit which is why I am asking if you know of an efficient circuit I can use.


Reduce losses, better core, bifilar windings, biggest wire you can get away with - all tricks that will increase efficiency, but low voltages do not make for highly efficient switcher topologies, so the Joule Thief is a darned good little circuit.

Well then what can be the most efficient ferrite coil specs (i.e. windings, wire gauge) ?

Make the core bigger, so you can get the same number of turns of heavier wire. You ARE ultimately constrained by the fact you can't get good turn on from the transistor with such low biasing voltages. There are very low start voltage switch mode regulators from the likes of Maxim that bootstrap themselves to increase efficiency at low input voltages.

This is an interesting paper on SMPSU efficiency.
and this one is the datasheet for a chip that allows you to build a VERY efficient
0.7V starting PSU

I've been playing with the circuit and believe I've found some indications as to how to reduce the current used. However, I have not found how to use a battery below ~ 0.6V since the circuit needs somewhere around 0.6V to start. Once started it can run to about 0.4V. You could use two batteries at 0.55V in series. See my site http://cs.yrex.com/ke3fl and search for "Joule Thief" to find the link to my JT page to see what I've found so far. I'm still working on it. -March 2014

Check out my web page on this, it was a question I was interested in and have tried a few things to see what makes it better, uses less current, to run the white LED. I'm not done yet and will be trying a combination of changing the total number of windings and then also decreasing the number of windings going to the base of the transistor. My web page is: http://cs.yrex.com/ke3fl then search on "joule thief"

If you're really interested in a high efficiency JT, you can see the difference between the ones I built.  I call it the Supercharged Joule Thief.  Click here for the original circuit and click on the link in my blog for further info.

This Supercharged Joule Thief circuit works incredibly well and is worth the extra components. I have been building these circuits since it was published on WatsonsEblog back in 2009. (That was a great blog, but it disappeared for some reason.) I build this circuit using all SMD components for ease of assembly and a compact footprint.

It's me, Watson.  You can read about it in my web page rustybolt.info.  My blog with much of what I had in my watsonseblog is now rustybolt.info/wordpress/ and the comparison of a conventional JT and my SJT are here.  I think you're the first person (other than Quantsuff) who has said that they use my circuit.  I'm glad someone else realizes how much better, efficiency wise, it is compared to the conventional JT.

If you want a conventional JT to work at a voltage as low as 0.55 volt, put five or more times the number of turns on the feedback winding and decrease the resistor to zero ohms.  Also disconnect the cathode (flat spot) of the LED from negative and connect it to the base of the transistor.  There is a video on Youtube of a JT like this that can operate at less than a quarter of a volt.

Watson, I found your rustybolt.info site a while ago and visit every few months to catch up on your latest experiments.  I have been busy experimenting myself, and have made a lot of interesting Joule Thief devices.  Here is an example of the Supercharged Joule Thief that I built with SMT components and a hand wound toroid. 

I have been considering an Instructable on building one of these, but haven't had the time to document the process properly.  I have the PCB design perfected, and the component selection has been refined enough to drain a cell to 0.55V-0.57V.  It is incredibly easy to use, and the connections are nearly foolproof.

A joule thief is the most efficient homebrew solution you'll find -- otherwise you need a commercial power supply option "boost converter".

Where did you read that they are not very efficient?  buck converters are 95%+ efficient and boost are usually 70-95% efficient.

A conventional Joule thief with the most commonly used components has an efficiency between 40 and 70 percent, typically a little over 50%.  My experiments with many conventional and my supercharged JTs confirm this.  If you want to see some data, go here.

FYI, efficiency depends on load. Efficiency worsens on low loads with some topologies.