Lasersaber strikes again. A joule thief king ?

[TinselKoala]

http://www.ebay.com/itm/3-x-MPSA18-NPN-General-Purpose-Transistor-Free-Shipping-/261092142922

Three for a US dollar, delivered to your mailbox anywhere in the world.

In my testing I have compared MPSA18A and BC337-25 (as well as many other NPNs). (I had to buy a lot of 50 of the BC337s to get a reasonable price like a dime apiece.)
The two are very close competitors; in some JTs one works "better" and in other JTs the other one works better. My criteria for "better" include both low operation voltage and LED brightness. Also... be sure to scope your circuit _after the LED goes out_ to see if the thing is still oscillating.  I recommend using a socket for the transistor so you can easily compare types and individuals within a particular type (they do vary). Also try the transistors "backwards" with C and E reversed in the circuit... you may be surprised.

[Vortex1]

Hi Conrad:
Here is a simple circuit I've been playing with on the bench. I went through about 6 different types of simple LED pulser circuits using FET's, SCR's PUT's etc and liked this one for it's simplicity, a single inductor, regenerative latch, timing cap and resistor. Attached is the basic design with no component values, as each version can be tailored for the voltage, duty cycle, and light output amplitude required.

I have souped up lower current versions which I will post later, this is just the basic circuit to play with. This one does not have the advantage that a blocking oscillator has i.e. high current base drive at little expense to the power drain, also it is not in the class of uWatt long runners, but will work down to 1 volt so can be operated off a single 1.5 V battery cell.

P.S. Yes, put a current limiting resistor in series with R1. This is a rudimentary circuit to give the concept and not a production schematic.

[Vortex1]

Here are some numbers you can use to qualify your designs.
They are continuous mode operation rather than pulse mode.

Also these tests are far below the actual Lumens the LED's are capable of, I'm just measuring at the threshold or lower subjective level of light since no one seems to be quantifying and integrating actual light power delivered vs. power usage.

A single bright white LED (old one I had laying around):

3 uA       3V        0.009 mW           Just Visible
10uA      3.3V      0.033 mW           Definitely Visible
100 uA   3.8V      0.38 mW             Bright
1000uA  4.12V     4.12 mW             Very Bright

A cluster LED array like Tinmans latest video:

2uA        3.27V     0.0065mW          Just Visible
10uA      3.42V     0.0342mW          Definitely Visible
100uA    3.53V     0.353mW            Bright but not evenly lit
1000uA  3.69V     3.69mW            Very Bright and fairly even distribution

You can calculate the energy in a 1000 uF  or 3300 uF capacitor charged to a given voltage and divide it by the required power
to see if you are exceeding expected run time and tapping any unknown source.

The problem is that in pulse mode e.g with a duty cycle of 10% on time a LED operated at 100 uA peak will look bright yet only be drawing 10uA average so this must be factored in when computing run times.

I have already done the calculations and run times seem to be right in the same order of magnitude that people are getting e.g. Lasersaber.

E=1/2CV^2

Let's apply a little simple math and see if anything unusual is happening.

I'm going to rerun these numbers using pulse mode, 10% duty cycle.

Maybe someone else can run the numbers also.

[conradelektro]

Here is a simple circuit I've been playing with on the bench. I went through about 6 different types of simple LED pulser circuits using FET's, SCR's PUT's etc and liked this one for it's simplicity, a single inductor, regenerative latch, timing cap and resistor. Attached is the basic design with no component values, as each version can be tailored for the voltage, duty cycle, and light output amplitude required.

@Vortex1: Interesting circuit, the art of multivibrators from the 1970ies. With modern transistors a very low power draw might be possible.

Can the duty cycle be adjusted with a diode and a second lower resistor in parallel to R1? The losses through R1 / C1 (frequency) seem to be the same as in the MAX931 astable multivibrator circuit? The wave form will not be a steep square wave, which will reduce the back EMF in the coil?

If your goal isn't chasing the lowest power still visible LED but to have a useful long lasting light powered by one or a couple of AAA cells then there are some very good options out there.   I think that 10 lumens is kind of the low end of useful light.  With the right LED you can get that on 80mW to the LED and a little more than 100mW drawn from the batteries.  A pair of AAA batteries gets you more or less 2Wh.  So you could go for 200 hours on such a circuit.

I looked at the "TPS61097A Low Input Voltage Synchronous Boost Converter With Low Quiescent Current". It allows to boost 0.9 Volt to 3.3 Volt (just right for a LED). But the losses at 0.9 Volt input are high (65% efficiency, and limitation to 50 mA output). But at 1.5 V to 3 V input (2 batteries in series which can run down quite low) the efficiency is 80% to 90%. With a 3 lumen Super Flux white LED (30 mA, two could be used to get 6 lumen light from 60 mA, three to get about 9 lumen from 90 mA) one could build a conventional LED driver to drive the LEDs continuously at nominal brightness from two batteries in series. Rechargeable batteries (which work at 1.2 V each) could be used as well.

http://at.mouser.com/Search/Refine.aspx?Keyword=TPS61097A (with 1.5 V to 3 V input only 90 mA at 3.3 V output are possible)

http://at.mouser.com/Search/Refine.aspx?Keyword=T4C4PRB (there might be better LEDs, but these have a low price)

Greetings, Conrad

[Vortex1]

Hi Conrad

The circuit I posted was later tried down to 1 Volt and worked well. The duty cycle increases at lower voltage to about 30%.

Besides actual testing on the bench I have also simulated the circuit in LTSpice if anyone is interested.

If one designing for lowest cost for a consumer torch to operate on a single 1.5 volt cell, the parts cost on this one will be hard to beat. A blocking oscillator using a tapped inductor might beat it, but transistors are generally a lot less expensive than coupled inductors.

As I said there are many improvements that can be made, as this circuit is elementary. The art is in making it tick better and at lower power drain. That's for later posts. Once the operation is fully grasped improvements come easy e.g. super beta transistors, speedup network etc.