Joule Thief - still not working....

[default]

On Sat, 8 Aug 2009 23:37:53 -0700 (PDT), fungus
<openglMYSOCKS@artlum.com> wrote:

The PIC chip is close to a good idea. Maybe you could look at the PICAXE
system / chip? The software and documentation is free and includes a
software simulator. A PICAXE chip programs in a simplified BASIC
language and the programmer is a serial cable and two resistors which
you can buy or make at home.

It seems like overkill, expensive too.

$3 is too much?

If you're in the US there's lots of places to get them with low
shipping charges. Anderson even supplies the programming resistors
with the chips for $2.95.

I lack programming experience and found them easy to use. I had my
first M08 out of the box and sequentially flashing four leds in about
ten minutes. The on line forum is great for help applying them.

It was harder to get windows to accept the USB to serial cable I
needed, than figuring out the simple program the picaxe needed.
--

 

[default]

On Sun, 09 Aug 2009 15:20:19 +1000, David Eather <eather@tpg.com.au>
wrote:

The PIC chip is close to a good idea. Maybe you could look at the PICAXE
system / chip? The software and documentation is free and includes a
software simulator. A PICAXE chip programs in a simplified BASIC
language and the programmer is a serial cable and two resistors which
you can buy or make at home.

David, I think it was you who first mentioned the picaxe here. I got
a few and have had nothing but fun with these things.

Thank you.
--

 

[fungus]

On Aug 9, 1:20 pm, default <defa...@defaulter.net> wrote:

$3 is too much?  

Oh, I didn't see the prices for individual chips, only
the ones with PCBs.

 

[fungus]

On Aug 9, 1:32 pm, default <defa...@defaulter.net> wrote:

David,  I think it was you who first mentioned the picaxe here.  I got
a few and have had nothing but fun with these things.  

Yes, they look fun.

 

[fungus]

On Aug 9, 4:01 pm, fungus <openglMYSO...@artlum.com> wrote:

On Aug 9, 1:32 pm, default <defa...@defaulter.net> wrote:

David,  I think it was you who first mentioned the picaxe here.  I got
a few and have had nothing but fun with these things.  

Yes, they look fun.

I've just been reading a bit more about them ... I
bet they're a lot of fun to play with.

But ... I don't think they do any kind of current
regulation on the outputs so they don't solve my
basic problem - how to keep LEDs at (or near) 20mA
independently of battery voltage.

(I know the joule thief doesn't either but the
current dropoff from a JT *is* less than with a
simple LED/resistor).

 

[default]

On Sun, 9 Aug 2009 07:01:28 -0700 (PDT), fungus
<openglMYSOCKS@artlum.com> wrote:

On Aug 9, 1:20 pm, default <defa...@defaulter.net> wrote:

$3 is too much?  


Oh, I didn't see the prices for individual chips, only
the ones with PCBs.

That's the beauty of them. Only need a solderless breadboard and

maybe a 9 pin RS232 connector plus 2-3 resistors.

I bought a board first time around - and still haven't used it, and
now the parts for the kit have already found their way into other
projects. I've got them working two time lapse cameras, two of the
range controls on my stove, and a computer fan and temperature alarm.
But I can have fun for hours just fooling around with the breadboard
and trying out different ideas.

To get into pics? ordinarily you might buy a programmer - for
$100-300 a software suite for another $300 and then learn some pretty
esoteric machine code to apply them. Picaxe does it all with minimal
overhead. There are a few other hobbyist pic solutions out there that
don't cost an arm and leg, but the 'axe is the cheapest and has the
best support. There are grade school kids learning to use 'axes.
--

 

[default]

On Sun, 9 Aug 2009 07:58:25 -0700 (PDT), fungus
<openglMYSOCKS@artlum.com> wrote:

On Aug 9, 4:01 pm, fungus <openglMYSO...@artlum.com> wrote:
On Aug 9, 1:32 pm, default <defa...@defaulter.net> wrote:

David,  I think it was you who first mentioned the picaxe here.  I got
a few and have had nothing but fun with these things.  

Yes, they look fun.

I've just been reading a bit more about them ... I
bet they're a lot of fun to play with.

But ... I don't think they do any kind of current
regulation on the outputs so they don't solve my
basic problem - how to keep LEDs at (or near) 20mA
independently of battery voltage.

(I know the joule thief doesn't either but the
current dropoff from a JT *is* less than with a
simple LED/resistor).

Actually they could do that. A simple two transistor current source
is lots easier and cheaper though.

All the axes have one or more A/D converters built in that could sense
current and pulse width modulation outputs to adjust brightness or
current. The A/D are 8 bit or 10 bit resolution (set in the program).
They have no analog voltage out, but the pwm can be fed into a simple
RC integrator to get a voltage level out.

They work over 2 volts to 5 volts - and don't tolerate much over 5 on
an input output or power source. My BB has two or three AA batteries
and a switch to select off, 3 or 4.5 volts. Current use is in
microamps, unless you drive loads with the outputs (sink or source 20
ma directly).
--

 

[Jon Kirwan]

On Sun, 09 Aug 2009 07:40:46 GMT, "Paul E. Schoen" <paul@peschoen.com>
wrote:

"Jon Kirwan" <jonk@infinitefactors.org> wrote in message
news:snds75da597nbbfk66rr3o4s8mha96nor1@4ax.com...

Couldn't you get by with fewer? Just parallel up the BJTs bases and
emitters and use their separate collectors for the LEDs. Collector
current is determined by Vbe (ignoring Early effect among other
things.) If you wanted to be a little extra careful you could add
some low-valued emitter resistors, I suppose. BJT count is then 1
plus number of LEDs, resistor count is 2 plus number of LEDs, etc.

Something like:

: Vcc Vcc Vcc
: | | |
: | | |
: \ | |
: / Rx --- ---
: \ 1k \ / D1 \ / D2
: / --- Red 2V --- White 3V
: | | |
: | | |
: +------------+--------------+---------------> etc
: | | | | |
: | | | | |
: | | |/c Q1 | |/c Q2
: | '--| '--|
: | |>e |>e
: | | |
: | | |
: | \ \
: | / R1 / R2
: | \ 10 \ 10
: | / /
: Qx c\| | |
: 2N3904 |----------+----+--------------+----------> etc
: e<| |
: | |
: | \
: | / Rz
: gnd \ (0.7V/20mA/N)
: /
: |
: |
: gnd

Rx is plenty low enough in your original schematic to drive a host of
LEDs, so I'd leave that value alone. Qx will suck away whatever isn't
needed, anyway. Rz is supposed to drop enough to drive Qx, which is
about 0.7V. At 20mA per LED, the value to be used is easy to compute.
R1, R2, ... Rn are just there to normalize out BJT variations. At 10
ohms, even a 2N2222 and 2N3904 can be paired up with good results.

Of course, Vcc needs to be enough to drive the highest-V LED, or else
the other LEDs will get a bit of a boost due to the fact that Rz's
current needs to be where it is designed to be.

I made some changes above that are even simpler. It might not be very
stable with temperature and it is not tightly regulated, but not bad for
such a simple circuit. Germanium transistors would work at even lower
voltages, but are hard to find. A 1VDC LDO would be better than the two
diodes. There is a TI TPS71701 that can be set to 1.00 VDC or even 0.9 VDC.

The LTspice [replaced with ASCII] for this circuit follows.

: Vcc Vcc Vcc
: | | |
: | | |
: \ | |
: / Rx --- ---
: \ 100 \ / D1 \ / D2
: / --- Red 2V --- White 3V
: | | |
: | | |
: +------------+--------------+---------------> etc
: | | | | |
: | | | | |
: | | |/c Q1 | |/c Q2
: _ '--| '--|
: \ / |>e |>e
: --- 1N4148 | |
: | | |
: - \ \
: \ / / R1 / R2
: --- 1N5818 \ 15 \ 15
: | / /
: | | |
: |-----------------+--------------+----------> etc
: |
: gnd

That's fine enough. I had originally used separate resistors but then
decided kept a single one to allow a single point of setting overall
current rather than different ones. This is fine, using each driver
BJT as emitter followers with a current-set resistor in the emitter
leg and a voltage drop across each determined by some 'softish' diode
drop which you've stiffened once the voltage gets high enough by
passing rather huge currents through Rx. I kind of liked the other
one a little better in that regard. There's another reason I liked
the other one better and that's that this one uses only a small
schottky voltage across each emitter leg's resistor and this voltage
rises with increasing current through it. Plus, the re in each BJT is
about 26mV/Ic or over an ohm here; not too close to the 15 ohms, but
close enough I feel a little uncomfortable. I kind of don't like it.

The earlier one you posted up hits its plateau at about 0.7V above the
required LED voltage. For a red LED, call that 2.7V. This one you
suggested above works with slightly lower requirements, but rises with
higher supply voltage as the schottky voltage increases from the
increased current. The change in diode current will be the change in
supply voltage divided by Rx, so a change from 3V to 6V will yield
3V/100 or 30mA change through the diodes. Since at 3V, the current
would be around 20mA, this new current of 50mA is about 2.5X higher.
The 1N4148 has its emission coefficient at about 1.75 so this suggests
about a 60mV increase across the 15 ohm emitter resistor or about 4mA
change in the LED current for a 3V change. Maybe a little more
perhaps added to that, because of the schottky. Perhaps 5mA in total?
But that gives an idea. And that is on top of the 15mA at 3V, which
is an increase of 30+%.

Jon

 

[David Eather]

default wrote:

On Sun, 09 Aug 2009 15:20:19 +1000, David Eather <eather@tpg.com.au
wrote:

The PIC chip is close to a good idea. Maybe you could look at the PICAXE
system / chip? The software and documentation is free and includes a
software simulator. A PICAXE chip programs in a simplified BASIC
language and the programmer is a serial cable and two resistors which
you can buy or make at home.

David, I think it was you who first mentioned the picaxe here. I got
a few and have had nothing but fun with these things.

Thank you.

Welcome!

 

[Jon Kirwan]

On Sun, 9 Aug 2009 19:41:34 -0700 (PDT), fungus
<openglMYSOCKS@artlum.com> wrote:

On Aug 9, 5:11 am, Jon Kirwan <j...@infinitefactors.org> wrote:

:           Vcc               Vcc            Vcc
:            |                 |              |
:            |                 |              |
:            \                 |              |
:            / Rx             ---            ---
:            \ 1k             \ / D1         \ / D2
:            /                --- Red 2V     --- White 3V
:            |                 |              |
:            |                 |              |
:            +------------+--------------+---------------> etc
:            |            |    |         |    |
:            |            |    |         |    |
:            |            |  |/c Q1      |  |/c Q2
:            |            '--|           '--|
:            |               |>e            |>e
:            |                 |              |
:            |                 |              |
:            |                 \              \
:            |                 / R1           / R2
:            |                 \ 10           \ 10
:            |                 /              /
:         Qx c\|               |              |
:     2N3904   |----------+----+--------------+----------> etc
:            e<|          |
:            |            |
:            |            \
:            |            / Rz
:           gnd           \ (0.7V/20mA/N)
:                         /
:                         |
:                         |
:                        gnd

Rx is plenty low enough in your original schematic to drive a host of
LEDs, so I'd leave that value alone.  Qx will suck away whatever isn't
needed, anyway.  Rz is supposed to drop enough to drive Qx, which is
about 0.7V.  At 20mA per LED, the value to be used is easy to compute.
R1, R2, ... Rn are just there to normalize out BJT variations.  At 10
ohms, even a 2N2222 and 2N3904 can be paired up with good results.

I just tried doing that one in LTSpice for a bit of practice
and it seemed to work. LTSpice only has about three LEDs in it
though so I'm not sure how well it represents real LEDs.

Yeah. The LED count is very small, there. You can add your own by
pasting in a 'S' directive (just type the letter S and a dialog box
pops up.) Add:

.model MYLED D(Ron=12.6 Vfwd=3)

for example. When you place an LED on the schematic, there is a 'D'
that displays with it. If you right click on the LED's displayed 'D'
value, another little dialog box comes up with the letter D set there
and you can change it to 'MYLED', if you like. That should then show
up on the schematic, too. When you run the simulation, LTSpice will
go find the .model directive and use it. You can add a bunch of the
..model directives with different names to get different responses for
each.

Of course, Vcc needs to be enough to drive the highest-V LED, or else
the other LEDs will get a bit of a boost due to the fact that Rz's
current needs to be where it is designed to be.

I hacked the database to add a 2.2V LED (eg. red), I don't know
what most of the LED parameters in the file were but I could
set voltage and current.

As the voltage drops below 4V the difference in current
between the built-in LED and my hacked LED becomes much
larger, you end up with some LEDs at 15mA and some at 5mA.

No, that's the right behavior. Some of the LEDs will require more
headroom than others. When the voltage is below what it takes, those
sections 'shut down' leaving more current for the other LEDs. The
difference in current becomes almost night and day. The main point is
that if the voltage is enough to drive the most voltage-hungry of the
LEDs, then it will run all of them close to the same current.

(Of course I might be doing it all wrong...)

There's also quite a big voltage drop across the BJT,
0.6V or so. I assume this is wasted power so I'm not
sure the transistors wouldn't heat up.

Yes, the transistors will heat up. For any given voltage, their Vce
will have to be enough to 'suck up' whatever voltage remains after the
Vbe of Qx is handled and the LED itself is satisfied. So if you are
running 2V red LEDs off of a 4.5V supply, expect to see Vce=1.8V
(which is 4.5V minus 2V minus .7V.) That BJT will dissipate 1.8V*20mA
(assuming 20mA is the set point) or 36mW. That's not that bad,
really. And since there is a separate BJT for each LED, it's not much
of a problem.

Jon

 

[fungus]

On Aug 9, 5:11 am, Jon Kirwan <j...@infinitefactors.org> wrote:

:           Vcc               Vcc            Vcc
:            |                 |              |
:            |                 |              |
:            \                 |              |
:            / Rx             ---            ---
:            \ 1k             \ / D1         \ / D2
:            /                --- Red 2V     --- White 3V
:            |                 |              |
:            |                 |              |
:            +------------+--------------+---------------> etc
:            |            |    |         |    |
:            |            |    |         |    |
:            |            |  |/c Q1      |  |/c Q2
:            |            '--|           '--|
:            |               |>e            |>e
:            |                 |              |
:            |                 |              |
:            |                 \              \
:            |                 / R1           / R2
:            |                 \ 10           \ 10
:            |                 /              /
:         Qx c\|               |              |
:     2N3904   |----------+----+--------------+----------> etc
:            e<|          |
:            |            |
:            |            \
:            |            / Rz
:           gnd           \ (0.7V/20mA/N)
:                         /
:                         |
:                         |
:                        gnd

Rx is plenty low enough in your original schematic to drive a host of
LEDs, so I'd leave that value alone.  Qx will suck away whatever isn't
needed, anyway.  Rz is supposed to drop enough to drive Qx, which is
about 0.7V.  At 20mA per LED, the value to be used is easy to compute.
R1, R2, ... Rn are just there to normalize out BJT variations.  At 10
ohms, even a 2N2222 and 2N3904 can be paired up with good results.

I just tried doing that one in LTSpice for a bit of practice
and it seemed to work. LTSpice only has about three LEDs in it
though so I'm not sure how well it represents real LEDs.

Of course, Vcc needs to be enough to drive the highest-V LED, or else
the other LEDs will get a bit of a boost due to the fact that Rz's
current needs to be where it is designed to be.

I hacked the database to add a 2.2V LED (eg. red), I don't know
what most of the LED parameters in the file were but I could
set voltage and current.

As the voltage drops below 4V the difference in current
between the built-in LED and my hacked LED becomes much
larger, you end up with some LEDs at 15mA and some at 5mA.

(Of course I might be doing it all wrong...)

There's also quite a big voltage drop across the BJT,
0.6V or so. I assume this is wasted power so I'm not
sure the transistors wouldn't heat up.

 

[fungus]

On Aug 9, 9:40 am, "Paul E. Schoen" <p...@peschoen.com> wrote:

I made some changes above that are even simpler. It might not be very
stable with temperature and it is not tightly regulated, but not bad for
such a simple circuit. Germanium transistors would work at even lower
voltages, but are hard to find. A 1VDC LDO would be better than the two
diodes. There is a TI TPS71701 that can be set to 1.00 VDC or even 0.9 VDC.

This one totally lost me - there doesn't seem to be
any feedback mechanism to regulate anything (but then
again I've got no idea what the diodes are doing -
seems like they're just acting as resistors but I'm
guessing it's something to do with electrons escaping
through the base of the transistor).

 

[Jon Kirwan]

On Sun, 9 Aug 2009 19:48:04 -0700 (PDT), fungus
<openglMYSOCKS@artlum.com> wrote:

On Aug 9, 9:40 am, "Paul E. Schoen" <p...@peschoen.com> wrote:

I made some changes above that are even simpler. It might not be very
stable with temperature and it is not tightly regulated, but not bad for
such a simple circuit. Germanium transistors would work at even lower
voltages, but are hard to find. A 1VDC LDO would be better than the two
diodes. There is a TI TPS71701 that can be set to 1.00 VDC or even 0.9 VDC.

This one totally lost me - there doesn't seem to be
any feedback mechanism to regulate anything (but then
again I've got no idea what the diodes are doing -
seems like they're just acting as resistors but I'm
guessing it's something to do with electrons escaping
through the base of the transistor).

Hehe. You aren't too far off, really. Imagine that the two diodes
are forward biased (conducting normally.) They exhibit a voltage drop
that doesn't change a lot with changes in current.

I'm going to go off on a slightly helpful side track, now. The bog
standard figure is kT/q (Boltzman's constant times Kelvin temperature
divided by the charge on an electron) or about 26mV times the natural
log of the current ratio. So if you compute, 26mV*ln(10/1) for a
10-fold increase in current, you will see about 60mV change. That's
with the emission coefficient at 1 (typical for small signal BJTs, but
NOT typical for diodes.) For many diodes like the 1N4148, the
emission coefficient is more than 1 -- the 1N4148 is about 1.75. That
just means you multiply by that factor, as well. So in the case of
the 1N4148, you would see about 100mV change with a 10-fold change in
current. The basic point here is that the voltage doesn't change a
lot and the circuit that Paul posted depends on that fact.

Now, Paul pasted in a 100 ohm resistor as Rx. Temporarily assume that
the 1N4148 has 0.7V (roughly) across it and the schottky diode has
another 0.3V across it (not uncommon.) That's 1V. All the rest of
the battery voltage then appears across the 100 resistor. So the
current arriving from the 100 ohm resistor will be about (V-1)/100 in
amps. For V=5, this is 40mA. A lot! The BJT (or two or three or
four...) will not require much of that. For Ic=20mA and where the
battery voltage is high enough that the poor BJT isn't in saturation,
perhaps only 100uA each. So most of the 40mA winds up following the
one easy path through the diodes to ground. That current yields a
voltage which hauls the BJT's base upwards to about 1V, give or take a
little. The BJT's emitter then "follows" that voltage, being less by
one Vbe's worth -- say 0.7V. so the emitter should be about 0.3V or
so (which is the voltage across the schottky diode itself.) In effect
the 1N4148 jacks the BJT base up high enough so that the BJT can at
least 'turn on' and the schottky jacks it up just a little further so
that there is a 'known' voltage across the emitter resistor. That
voltage causes a current and pretty much all of the emitter current
must go through the collector and the LED. About 0.3V/15, or 20mA.

Why not use two 1N4148's, instead? You could. The base would be a
little higher up, you'd need a little more battery voltage for
everything to work according to easier design rules, that's all. There
are other differences which relate to how the 1N4148, BJT, and
schottky respond to temperature variations, though. And in some
circuits, the use of very specific currents and components are
combined to make circuits that exhibit supurbly consistent behavior
even when the temperature varies over a very wide range. I've seen
schottky's used in those cases, for example. Paul can probably
explain his exact reasoning for using one here, but don't get too
caught up in the fact that there is one there. There are many
alternatives that rely upon the same basic idea, but just get there in
different ways.

The obvious problem in using just one 1N4148 is that it's own diode
voltage (because it has a higher Is figure, by orders of magnitude)
will be lower than what the BJT requires to get started. When it is
conducting by itself, there isn't enough voltage there to turn the BJT
on -- not much, anyway. Even with 40mA speeding through it and only
100uA going through the BJT, you might only see a difference in
voltage of maybe 0.1V or less. Might be close to zero, even. Which
would leave almost no voltage (and certainly not a reliable one) left
over at the BJT emitter to create a current through that 15 ohm
resistor there. Another diode helps. Some. It's still not my
favorite way to go.

Your point about the lack of feedback is loosely correct. There is a
voltage developed across the diodes. That voltage isn't adjusted much
on the basis of anything happening later on in the circuit, so it is
effectively 'open loop' and set up or down a little by tweaking Rx.
Once you have that voltage, hanging the BJT on as an emitter follower
is also basically open loop without feedback. It all depends on the
voltage developed across the two diodes to set the current.

There is a very tiny amount of negative feedback. If the voltage at
the diodes goes upwards for some reason, the emitter will follow. That
will cause a slightly higher Ic in the BJT and require a slightly
higher base current to maintain it. The slightly higher base current
will require a very slightly higher Vbe, thus just very slighly
dropping the emitter downwards to oppose the change. It's not much
and it won't come anywhere close to compensating for the larger
change. But it is a tiny effect that one could hang there hat on as a
'negative feedback.' But in the scheme of things, it's discountable.

There are other ways to replace those diodes -- often used when more
than one or two diode drops are desired -- using a single BJT and a
pair of resistors. It basically leverages the Vbe drop in a fixed way
to set any effective, fixed diode drop you want. It's more often
found in various IC designs than in discrete ones, in my experience
though. And you need voltage headroom for them for the obvious reason
that they can only increase the drop, not decrease it.

I liked his earlier circuit better. A very slightly lower starting
voltage of operation might be had by using a Sziklai pair for his
feedback BJT. Maybe 0.1V-0.15V at most, though. Probably not enough
for the extra trouble.

Jon

 

[fungus]

On Aug 9, 10:24 pm, default <defa...@defaulter.net> wrote:

That's the beauty of them.  Only need a solderless breadboard and
maybe a 9 pin RS232 connector plus 2-3 resistors.

It might be worth getting a proper PCB and a ZIF socket
if you're doing lots of them...

 

[fungus]

On Aug 10, 5:02 am, Jon Kirwan <j...@infinitefactors.org> wrote:

Yes, the transistors will heat up.  For any given voltage, their Vce
will have to be enough to 'suck up' whatever voltage remains after the
Vbe of Qx is handled and the LED itself is satisfied.  So if you are
running 2V red LEDs off of a 4.5V supply, expect to see Vce=1.8V
(which is 4.5V minus 2V minus .7V.)  That BJT will dissipate 1.8V*20mA
(assuming 20mA is the set point) or 36mW.  That's not that bad,
really.  And since there is a separate BJT for each LED, it's not much
of a problem.

 

[fungus]

On Aug 10, 5:02 am, Jon Kirwan <j...@infinitefactors.org> wrote:

Yeah.  The LED count is very small, there.  You can add your own by
pasting in a 'S' directive (just type the letter S and a dialog box
pops up.)  Add:

   .model MYLED D(Ron=12.6 Vfwd=3)

Aha! That makes life a lot easier...!

 

[ehsjr]

fungus wrote:

On Aug 2, 7:20 am, greg <g...@cosc.canterbury.ac.nz> wrote:

No, anything non-metallic and non-conductive --
wood, plastic, cardboard, etc. -- should make
very little difference.



Ok I got a free half-hour so I made an air-core
with a piece of plastic pipe, here's a pic:
http://www.artlum.com/jt/aircore.jpg

It's got about a 4:1 ratio on the windings ... and
it works! I even got 20mA LED current out of it.

I didn't measure efficiency - the frequency is way
too high so the transistor is heating up. I guess
I need a lot more turns on the inductor or a different
size/shape or something. I'll leave that part up to the
theorists.

I tried two batteries and I found that brightness keeps
on increasing as R1 goes down to zero. The LEDs
were still lit when R1 was just a couple of ohms.

To give you an idea of efficiency with the small air core
I mentioned earlier in the thread, here are the ranges I
measured:

Vin 1.50 to .60
Vout 2.24 to 1.88
Iin mA 168.10 to 29.50
Iout mA 30.0 to 2.0
Efficiency % 26.65 to 21.24
Pin mW 252.15 to 17.70
Pout mW 67.20 to 3.76
Freq kHz 190.4 to 469.4

This was the original joule thief circuit with a 1N4148 on the
collector feeding a 100uf, with a single red led. I also had
a 1N4148 on the base to gnd. There's more data points on the
spreadsheet, I took measurements every 100 mv.

The point here is not to indicate that your air core will perform
at the same efficiency. But the circuit will perform the same way.
The voltage delivered to the LEDs will be relatively constant, but
the LED current will vary by a huge percentage as Vin decreases.

When I used the identical circuit but with 3 white LEDs in series,
efficiency improved to 34.44% at 1.5 Vin and was higher than it
was with the single red LED at every data point except .6V
But Iout was much lower with the 3 white LEDs.

I made no attempts to optomize anything, so undoubtedly you can do
better by changing component values, windings etc. But you can't
get around the large variation in LED current as Vin changes with
the joule thief circuit. If Vin decreases, LED current decreases.
A ferrite core won't solve the problem, nor will different value
air core, different values of resistor, capacitor, etc. If you want
to maintain LED current at some level, you will need to add to the
joule thief circuit, or use a different circuit.

Ed

 

[ehsjr]

fungus wrote:

<snip>

As an aside, would a transformer not work for a joule
thief?

Yes. A transformer is what makes the joule thief work. The
output is magnetically coupled to the input by the transformer.
Without that coupling, you would need some other means for the
output to be fed back to the input to create oscillation.
While it may have been called an "inductor" in the thread,
that's just imprecise wording. Throughout the thread, it
has always been a transformer in spite of however it has been
worded, as long as it referred to the component in the joule
thief circuit.

Ed

 

[fungus]

On Aug 10, 7:00 am, Jon Kirwan <j...@infinitefactors.org> wrote:

Hehe.  You aren't too far off, really.  Imagine that the two diodes
are forward biased (conducting normally.)  They exhibit a voltage drop
that doesn't change a lot with changes in current.

Let's see if I can reduce it to terms my monkey brain can understand.

Why not use two 1N4148's, instead?  You could.  The base would be a
little higher up, you'd need a little more battery voltage for
everything to work according to easier design rules, that's all. There
are other differences which relate to how the 1N4148, BJT, and
schottky respond to temperature variations, though.

The two diodes and R1 are chosen so that the voltage across
each diode is just above its turn-on voltage and the voltage
at the bases of the BJTs is enough to allow 20mA through the
LEDs.

As the battery voltage drops... the resistance of the diodes
starts to rises exponentially, thus raising the voltage at
the base of the BJT and allowing more current through the LED.

 

[fungus]

On Aug 10, 9:24 am, ehsjr <eh...@NOSPAMverizon.net> wrote:

fungus wrote:

snip

As an aside, would a transformer not work for a joule
thief?

While it may have been called an "inductor" in the thread,
that's just imprecise wording. Throughout the thread, it
has always been a transformer in spite of however it has been
worded, as long as it referred to the component in the joule
thief circuit.

I guess what I meant was, would a 'transformer'
which is designed for efficient conversion of
voltages be suitable for making efficient joule
thieves. Maybe the characteristics are completely
different.