DC Boost Converter ("Joule Thief") Efficiency Measurement

A quick look at the input and output power for a simple one-inductor DC boost converter. Comparing actual measurements of an optimised circuit with the theoretical efficiency given by a Spice simulation.
This is a follow-up to a previous video discussing how the circuit works:
• DC Boost Converter - A...

[@benthere8051]

Someone finally did it right with Spice. Thank you. Well done. I like the NPN / PNP approach as opposed to the special transformer. Although the component count is hard to beat with the toroid.

[@poptartmcjelly7054]

I was mucking about with LTspice just now and came up with a dual NPN boost converter with a theoretical efficiency of about 85%. After a little improvement the circuit was able to supply constant regulated 5V at 1A from a 3.3V supply. I guess now it's time for me to test it in real life. :) Especially since i made a MOSFET version of my circuit that supplies much higher current at a slightly lower efficiency. Also raising the inductance of the coil does not necessarily increase efficiency, I found 4,7µH to be just right.

[@poptartmcjelly7054]

I just made my circuit work under the same conditions as yours did in the simulation and with a bunch of fine adjustments got 61% efficiency. :)

[@ThingsWhichArentWork]

You've certainly improved the efficiency of that circuit compared to what you had in the first video where it was pulling about 5x as much current to light the same LED :-)

[@acmefixer1]

I've built dozens of these, each time I've tried to optimize the performance. I've found that the 2N3904 and BC547 are the least capable of good performance. It's better to use PN2222A, PN4401, or slightly better, BC337-25. If you can get some of these, they're even better: C8050, 2SC2500, NTE11, KSD5041. To minimize losses in the coil, use around 100 uH, with a series resistance of 1/4 ohm or less. A high mu ferrite toroid core with a dozen turns works great. You can check out some of the many Joule Thief circuits I've built at rustybolt.info/wordpress/ For some reason, the provider has been getting DoSed, so it may be difficult to get on my blog.

[@bjl1000]

You can connect two together either series or parallel

[@LucCypHer]

Hallo ! Great demonstration and explanations. I did tried about the same, but with a Capacitor instead of R1. I played with values from 1uF, with LED blinking, to 22nF, and currently with 1.5 nF. The R2 I tuned to get a consumption of 1mA ( currently ) from the Battery. With a different Coil : Torroidal / Bifilar / Air. And "something" more... Measurements indicated about 35% to 38% efficiency. You may want to see : Joule Thief With FeedBack - After 35 Hours / 2014-11-08 and Joule Thief With FeedBack - After 46 Hours / 2014-11-09. Would you please let me know what Application are you using as SPICE Simulator ? Thanks a lot, and Good Luck !

[@stmounts]

It's 2020 now, dunno if this channel is still watched... Hve you checked the efficiency of the simple classic Joule Thief circuit??? The simple circuit uses one transistor and a feedback winding from the inductor. I suspect that the simple circuit is more efficient than the one described here.

[@andyharsley]

Yes, efficiency is woeful. My circuit was originally designed as a way of demonstrating the five basic off-the-shelf components, including both NPN and PNP transistors. By using a simple inductor you also don't need to wind one as per the original joule thief design.

[@keithking1985]

What software is that your using?? It's cool..

[@andyharsley]

LTspice from Linear Tech. Very cool, and also free!

[@keithking1985]

LTspice from Linear Tech!!!. Thanks a lot Andy I'll be loading it up so seeing that its free!! ; ) nice one!!

[@jonbigman9723]

I know this was 4 years ago but using a breadboard is not a good setup for efficiency.... Those breadboards are notorious for adding all kinds of noise via stray resistance and capacitance between the boards buss bars...

[@andyharsley]

Good point! I hadn't factored that in.

[@benthere8051]

I build my prototypes on FR-4. You get a great ground plane and few sneaky distributed Rs and Cs.