Joule Thief - still not working....

[ehsjr]

David Eather wrote:

Jon Kirwan wrote:

On Sun, 26 Jul 2009 17:06:24 GMT, ehsjr <ehsjr@NOSPAMverizon.net
wrote:

fungus wrote:

On Jul 26, 8:57 am, ehsjr <eh...@NOSPAMverizon.net> wrote:

I don't _know_ if it qualifies as "a whole lot" better, but
available one chip solutions can meet the op's stated requirement
of keeping the current at 15-20 mA, and the joule thief cannot.

Can you maybe recommend one...?

Manufacturer chips posted below are just the first few found by a
Google search with "led boost drivers" in the search box.

National recommends their LM3410X for this.
http://www.national.com/ds/LM/LM3410.pdf


$2.50-$3 each. Lots around.

TI shows the TPS61160 meeting the requirements.
http://focus.ti.com/lit/ds/symlink/tps61161a.pdf


Hmm. Cheaper. $2 each. Lots around.

Onsemi has the CAT3606-D
http://www.onsemi.com/pub_link/Collateral/CAT3606-D.PDF


Couldn't find the -D around anywhere. But did find CAT3606HV4-T2 at
Digikey for $1 (and at only two other places.) This device cannot
handle more than 4.2V input and must have at least 3V. It's designed
for Li-ion sources and can run in either 1X or 1.5X mode. I'm not
hyped on this as a 'solution.' It's a charge pump with regulation on
the current, I think.

Linear's LT3598 will do it:
http://cds.linear.com/docs/Datasheet/3598fa.pdf


Mucho expensive. I found them for over $7 each! (Some at under $5,
too.) Only a few places carry them.

....

TI seems to be the one out of the above I'd focus more on. Looks nice
and seems to do the right job for a reasonable price and is at various
stores, as well.

Jon

I'm not recommending any one of those over any other,
and there are other chips from those manufactures and
others that may suit your needs.

Ed


Um, all those chips are in surface mount packages - the OP has bugger
all chance of being able to solder them.

He can use a TL499 in PDIP, which is a boost switcher and
voltage regulator, with current limiting. The point was to show
that there are one chip solutions, better than the joule thief
because they meet the OP's requirements, not to provide every
possible chip he could use. There's a lot of chips available.

What's nice is you came up with the 3909 & he came up with the
3914, and the 499 is DIP so he go can SMT or through hole,
whichever way he wants. If this becomes a hobby for him, he'll
likely end up using SMT some day.

There are also preassembled solutions. For example:
http://www.pololu.com/catalog/product/799

I suspect that many/most/all of the chips will beat the joule thief
in terms of efficiency, as well as being better because they meet the
op's requirements and the thief doesn't. I'd like to hear what others
have measured for joule thief efficiency. Jon mentioned spice showing
80 - 85%. I find that hard to believe. I want to hear what people
actually measured, not a simulated figure. I suspect that 85% is peak
efficiency, which will drop as Vin drops - but I find that peak
suspicious even with a brand new battery.

Ed

 

[David Eather]

ehsjr wrote:

David Eather wrote:
Jon Kirwan wrote:

On Sun, 26 Jul 2009 17:06:24 GMT, ehsjr <ehsjr@NOSPAMverizon.net
wrote:

fungus wrote:

On Jul 26, 8:57 am, ehsjr <eh...@NOSPAMverizon.net> wrote:

I don't _know_ if it qualifies as "a whole lot" better, but
available one chip solutions can meet the op's stated requirement
of keeping the current at 15-20 mA, and the joule thief cannot.

Can you maybe recommend one...?

Manufacturer chips posted below are just the first few found by a
Google search with "led boost drivers" in the search box.

National recommends their LM3410X for this.
http://www.national.com/ds/LM/LM3410.pdf


$2.50-$3 each. Lots around.

TI shows the TPS61160 meeting the requirements.
http://focus.ti.com/lit/ds/symlink/tps61161a.pdf


Hmm. Cheaper. $2 each. Lots around.

Onsemi has the CAT3606-D
http://www.onsemi.com/pub_link/Collateral/CAT3606-D.PDF


Couldn't find the -D around anywhere. But did find CAT3606HV4-T2 at
Digikey for $1 (and at only two other places.) This device cannot
handle more than 4.2V input and must have at least 3V. It's designed
for Li-ion sources and can run in either 1X or 1.5X mode. I'm not
hyped on this as a 'solution.' It's a charge pump with regulation on
the current, I think.

Linear's LT3598 will do it:
http://cds.linear.com/docs/Datasheet/3598fa.pdf


Mucho expensive. I found them for over $7 each! (Some at under $5,
too.) Only a few places carry them.

....

TI seems to be the one out of the above I'd focus more on. Looks nice
and seems to do the right job for a reasonable price and is at various
stores, as well.

Jon

I'm not recommending any one of those over any other,
and there are other chips from those manufactures and
others that may suit your needs.

Ed


Um, all those chips are in surface mount packages - the OP has bugger
all chance of being able to solder them.

He can use a TL499 in PDIP, which is a boost switcher and
voltage regulator, with current limiting. The point was to show
that there are one chip solutions, better than the joule thief
because they meet the OP's requirements, not to provide every
possible chip he could use. There's a lot of chips available.

What's nice is you came up with the 3909 & he came up with the
3914, and the 499 is DIP so he go can SMT or through hole,
whichever way he wants. If this becomes a hobby for him, he'll
likely end up using SMT some day.

There are also preassembled solutions. For example:
http://www.pololu.com/catalog/product/799

I suspect that many/most/all of the chips will beat the joule thief
in terms of efficiency, as well as being better because they meet the
op's requirements and the thief doesn't. I'd like to hear what others
have measured for joule thief efficiency. Jon mentioned spice showing
80 - 85%. I find that hard to believe. I want to hear what people
actually measured, not a simulated figure. I suspect that 85% is peak
efficiency, which will drop as Vin drops - but I find that peak
suspicious even with a brand new battery.

Ed

Fungus already reported 55% efficency

 

[John Larkin]

On Wed, 29 Jul 2009 19:34:16 GMT, Jon Kirwan
<jonk@infinitefactors.org> wrote:

On Wed, 29 Jul 2009 07:56:36 -0700 (PDT), fungus
openglMYSOCKS@artlum.com> wrote:

On Jul 29, 7:15 am, Jon Kirwan <j...@infinitefactors.org> wrote:

There's a noticeable roundedness towards the end of
the on period, which might be contributing to your
heating problem.

I agree with you.  Most of the heating takes place at that rounded
'corner' as Vce rises.

??

A few posts higher up we were discussing ways to make the
transistor turn off faster by sucking the electrons out of the
base...are we sucking too hard? :-S

I don't think that's an issue we can much control. I think John's
point about using a thevenin arrangement on the base is a good one --
he was the one bringing up that point. But it seems one of the
primary advantages to it is that you don't need the protection diode
at the base. There is a lower impedance (but not miraculously lower)
to allow somewhat faster turn off, and it will shorten up the time
during which the collector winding drives the collector negative (that
odd bump below zero in your GIF.) These are all 'good things.' But I
don't think that failing to do ao causes much of a change in total BJT
dissipation.

To use BJTs and still get it to turn off faster, I mean really faster,
would seem to require another BJT (or two) to my hobby level of
understanding.

Bypass R2.

John

 

[fungus]

On Jul 29, 10:58 pm, fungus <openglMYSO...@artlum.com> wrote:

But ... there may be a possibility that my "mega-bead"
might provide high current at low frequency.

Ok, in the name of science I gritted my teeth and made
one with the mega-bead. Winding lots of turns onto a
tiny ring isn't fun... :-(

But ... it works! I got until I couldn't physically add any
more turns but the output current was still going up
with every turn - I obviously didn't reach the limit yet.

At the end I could still use R1 to dial current up/down,
I got 20mA at around 600 ohms.

Operation frequency was about 16kHz, ffficiency is
around 60%.

Transistor waveforms look cleaner (especially at the
base) and it's been running for a few minutes now
without getting hot. I'm going to leave it running to
see if anything bad happens (and also to see the
current/drightness dropoff as the battery runs down).

 

[fungus]

On Jul 30, 12:40 am, Jon Kirwan <j...@infinitefactors.org> wrote:

I've never been able to get enough output current with a
lower frequency circuit.

I think this is because you are using one of those RF suppression
things with permeability to the max and really low saturation levels
designed at the outset to "kill signals."

That big one is, but I've tried two others which I pulled out of
a PSU. One was a much smaller ferrite ring and one was an
"iron-powder" type. All of those behaved badly.

If it works, the next step would be to figure out what's
different about that bead.

Saturation level, mu, and l_m?

What it means is that I can't just go around pulling beads
out of things and expecting them to work.

 

[fungus]

On Jul 30, 1:34 am, ehsjr <eh...@NOSPAMverizon.net> wrote:

He can use a TL499 in PDIP, which is a boost switcher and
voltage regulator, with current limiting

What's nice is you came up with the 3909 & he came up with the
3914, and the 499 is DIP so he go can SMT or through hole,
whichever way he wants.

I like the 3914 because the number of external components
is minimal (a couple of resistors in the simplest case). The
other ones need quite a few external components, including
external inductors.

I suspect that many/most/all of the chips will beat the joule thief
in terms of efficiency

The 3914 datasheet says it's around 90% efficient.

as well as being better because they meet the
op's requirements and the thief doesn't.

I'm still holding out for sound activated lights using the 3914. I've
been googling and it looks like a small electret microphone
could do the job fairly easily (...but I'm the guy who thought
lighting a LED with a battery would be easy so what do I know?)

 I'd like to hear what others have measured
for joule thief efficiency.  Jon mentioned spice showing
80 - 85%. I find that hard to believe.

My latest one measured around 60%. I could probably have
improved improved that with more turns, but that's all that
fit on the bead. Maybe there's better beads out there as well...

 

[Jon Kirwan]

On Wed, 29 Jul 2009 17:43:47 -0700, John Larkin
<jjlarkin@highNOTlandTHIStechnologyPART.com> wrote:

On Wed, 29 Jul 2009 19:34:16 GMT, Jon Kirwan
jonk@infinitefactors.org> wrote:

On Wed, 29 Jul 2009 07:56:36 -0700 (PDT), fungus
openglMYSOCKS@artlum.com> wrote:

On Jul 29, 7:15 am, Jon Kirwan <j...@infinitefactors.org> wrote:

There's a noticeable roundedness towards the end of
the on period, which might be contributing to your
heating problem.

I agree with you.  Most of the heating takes place at that rounded
'corner' as Vce rises.

??

A few posts higher up we were discussing ways to make the
transistor turn off faster by sucking the electrons out of the
base...are we sucking too hard? :-S

I don't think that's an issue we can much control. I think John's
point about using a thevenin arrangement on the base is a good one --
he was the one bringing up that point. But it seems one of the
primary advantages to it is that you don't need the protection diode
at the base. There is a lower impedance (but not miraculously lower)
to allow somewhat faster turn off, and it will shorten up the time
during which the collector winding drives the collector negative (that
odd bump below zero in your GIF.) These are all 'good things.' But I
don't think that failing to do ao causes much of a change in total BJT
dissipation.

To use BJTs and still get it to turn off faster, I mean really faster,
would seem to require another BJT (or two) to my hobby level of
understanding.

Bypass R2.

I did that in a few examples I posted. In fact, before you'd even
suggested either the thevenin approach or this bypass cap I had
already posted a schematic that included both as a possible
suggestion. That doesn't mean I'm smart about these things, but it
does mean I have at least a little intuition -- even if it's not
justified by a clear theoretical understanding.

I'm very glad you talked more about it, though. It helps focus my
attention and I'm learning.

I took some notes about using the exact same core (L2=120uH and
l1=7.5uH with a coupling constant of 0.9) and same divider (R1=1k and
R2=1.2k.) I assumed a standard silicon diode for the flyback current
instead of a schottky, which would be better. The diode losses aren't
list here but the output power includes that loss.

R2 ----- BJT ----- Input Oper -------- Output -----
Bypass C-E B-E Power Freq Vout Iout Power %eff
----------------------------------------------------------------------
10pF 52.26mW 2.79mW 721.9mW 72kHz 20.32V@30.73mA=624.6mW 86.5%
100pF 49.07mW 3.14mW 717.7mW 72kHz 20.32V@30.67mA=623.6mW 86.8%
1nF 49.02mW 4.28mW 711.4mW 73kHz 20.30V@30.40mA=617.1mW 86.7%
10nF 44.91mW 5.25mW 682.1mW 76kHz 20.22V@29.31mA=592.5mW 86.9%
100nF 25.20mW 5.88mW 410.8mW 126kHz 19.38V@18.19mA=352.5mW 85.8%
1uF 15.65mW 6.69mW 258.1mW 200kHz 18.86V@11.46mA=216.3mW 83.8%

As the operating frequency pulls away from what you'd expect from the
collector winding itself, you also get these nasty looking Vce ringing
things that just ring and ring and ring throughout the flyback period
-- in the tens of MHz area. And they don't damp out fast. I felt
pretty uncomfortable with that behavior and the fact remains that I
haven't completely analyzed this effect and how to compute values on
the basis of a desired output. If you can explain it and puts some
sense to it, that would really help me.

Anyway, that's part of why I didn't focus more on the bypass cap.

Jon

 

[Jon Kirwan]

On Wed, 29 Jul 2009 23:34:58 GMT, ehsjr <ehsjr@NOSPAMverizon.net>
wrote:

I suspect that many/most/all of the chips will beat the joule thief
in terms of efficiency, as well as being better because they meet the
op's requirements and the thief doesn't.

Yes. They include a LOT more parts inside, so they can compensate
'this and that' until the cows come home.

I'd like to hear what others
have measured for joule thief efficiency. Jon mentioned spice showing
80 - 85%. I find that hard to believe.

It's a best case, no doubt, assuming perfection everywhere. I would
consider it to be an absolute best case that could never really be
achieved in life.

I want to hear what people
actually measured, not a simulated figure. I suspect that 85% is peak
efficiency, which will drop as Vin drops - but I find that peak
suspicious even with a brand new battery.

I was using 4V as the input voltage. Not 1.5. No core losses or wire
losses in the transformer, either. So it's mostly for comparison with
other simulations.

Jon

 

[greg]

Jon Kirwan wrote:

On Mon, 27 Jul 2009 00:27:26 -0700 (PDT), fungus
openglMYSOCKS@artlum.com> wrote:

On Jul 27, 6:39 am, greg <g...@cosc.canterbury.ac.nz> wrote:

With the output capacitor, it actually carries *more*
current than the LEDs.

I'm not sure I followed greg's comment above, myself.

All I meant is that the average transistor current is
greater than the average LED current. It must be, due
to energy conservation and the fact that Vout > Vin.

--
Greg

 

[greg]

fungus wrote:

A few posts higher up we were discussing ways to make the
transistor turn off faster by sucking the electrons out of the
base...are we sucking too hard? :-S

Possibly you're still not sucking hard enough.

You could try experimenting with various values of
capacitor across the base resistor to see if you
can get the rounded part of the waveform to square
up more.

--
Greg

 

[default]

On Wed, 29 Jul 2009 07:56:36 -0700 (PDT), fungus
<openglMYSOCKS@artlum.com> wrote:

On Jul 29, 7:15 am, Jon Kirwan <j...@infinitefactors.org> wrote:

There's a noticeable roundedness towards the end of
the on period, which might be contributing to your
heating problem.

I agree with you.  Most of the heating takes place at that rounded
'corner' as Vce rises.


??

A few posts higher up we were discussing ways to make the
transistor turn off faster by sucking the electrons out of the
base...are we sucking too hard? :-S

I doubt that's possible. There's something called the Baker Switch -
uses an active shut off that returns the base to a voltage lower than
the emitter (drives it below ground to a negative supply) - just to
buy a little faster turn off and a little better efficiency.

This is just a blocking oscillator - don't expect too much from it.
You should expect the transistor to run cool - but gee whiz efficiency
or clean wave forms are not too likely.

Its main claim to fame is its simplicity - self oscillation with a
single transistor, and great big voltage spikes that can be used to
boost output voltage.

Switched capacitor voltage converters are a better bet for efficiency
and clean wave forms - at the cost of larger size and complexity
(unless you buy a chip with the works already on it).
--

 

[default]

On Wed, 29 Jul 2009 07:56:36 -0700 (PDT), fungus
<openglMYSOCKS@artlum.com> wrote:

On Jul 29, 7:15 am, Jon Kirwan <j...@infinitefactors.org> wrote:

There's a noticeable roundedness towards the end of
the on period, which might be contributing to your
heating problem.

I agree with you.  Most of the heating takes place at that rounded
'corner' as Vce rises.


??

A few posts higher up we were discussing ways to make the
transistor turn off faster by sucking the electrons out of the
base...are we sucking too hard? :-S

Some basics on blocking oscillators

Easy to grok, and well written:
http://mysite.du.edu/~etuttle/electron/elect37.htm

Basic, almost useless:
http://en.wikipedia.org/wiki/Blocking_oscillator

Good explanation and historical:
http://www.vias.org/eltransformers/lee_electronic_transformers_11_01.html
--

 

[fungus]

On Jul 30, 12:40 am, Jon Kirwan <j...@infinitefactors.org> wrote:

If it works, the next step would be to figure out what's
different about that bead.

Saturation level, mu, and l_m?

The real question is, if I were to order some from RS, what
would I order? RS stocks these: http://tinyurl.com/mpcjx7

Do I want low or high permeability? Maybe there's some
out there that work even better than the megabead...

{My criteria at the moment is "The harder it sticks to a
magnet, the better it'll work" ... but I'm not sure RS has
a search option for that... }


PS: Is RS "Radio Shack"?

 

[fungus]

I discharged a set of batteries through the latest JT and added the
results to my input/output graph:

http://www.artlum.com/jt/jt_vs_res.gif

The low end of the voltage range is done with one/two batteries.

 

[Jon Kirwan]

On Thu, 30 Jul 2009 13:01:09 -0700 (PDT), fungus
<openglMYSOCKS@artlum.com> wrote:

On Jul 30, 12:40 am, Jon Kirwan <j...@infinitefactors.org> wrote:

If it works, the next step would be to figure out what's
different about that bead.

Saturation level, mu, and l_m?

The real question is, if I were to order some from RS, what
would I order? RS stocks these: http://tinyurl.com/mpcjx7

Do I want low or high permeability? Maybe there's some
out there that work even better than the megabead...

{My criteria at the moment is "The harder it sticks to a
magnet, the better it'll work" ... but I'm not sure RS has
a search option for that... }

Well, I'm just learning this stuff, myself.

Let's go over what we think we know (I'm a hobbyist, not an expert on
this stuff, so it's good to sit back for a moment and summarize.)

(1) Recoverable energy is stored as I^2*L/2.
(2) Saturation causes L to rapidly change from a high value towards a
very low (air core) value. The result of this fact is that the
current goes upwards __very fast__.
(3) After saturation the L is quite different, so blind calculation
of energy using a fixed L will yield the wrong answer once saturation
occurs.
(4) Power to the load appears to depend upon Ic_peak, not L.
(5) But frequency of operation does depend on L and we can't run the
BJT too fast (core losses, other than air core, become significant at
about 200kHz _and_ the BJT itself has 'turn around' difficulties,
too.)
(6) Most of the power losses (heating) in the BJT are due to the last
moments during its ON time because of Vce rise right at the same
moment the collector current is highest. Lost energy here is the
product of Vce and Ic and time.

Okay.

The upside of using air core inductors are (a) no core losses to worry
about; and, (b) no need to go buying something, as air is free; and,
(c) no saturation effect to complicate analysis. The downside of
using air core inductors are (d) lots and lots of turns to get the
frequency down to mitigate BJT losses at higher frequencies; and, (e)
poor flux linking resulting in lots of leakage inductance without
using special wiring techniques (and even then, expect more than what
you'd have with a manufactured core.)

Air core is acceptable and you've already got some thoughts on that
here. If you have ready access to that CAT5 or CAT6 cabling, that may
work. The downside there is that the wire itself is quite thick and
your transformer will be sizeable. If you have fine-gauge magnet wire
(did I read that you said you had some?), you might haul out a long
length of it, fold it over, and try and impart a uniform number of
twists per foot/meter in it and use it as a bifilar pair. Or use
google and see what other good ideas you can find out there for
winding good air core transformers with excellent flux linking. I am
not experienced at this (more experienced at winding on manufactured
cores, but still not very experienced there, either.)

Your question is about selecting a core, so let's go there and bypass
any further air core discussion.

One thing about anything other than air is saturation effects. Are
these 'good' or 'bad?' Well, to answer that I'm going to have to draw
a picture:

http://www.infinitefactors.org/misc/images/saturation.gif

I've drawn an energy curve, which is current verses volt-seconds. It
does NOT include core losses, but that's not helpful right now. It
also does not include the whole picture that lots of web sites and
books will include, as I've just focused on one aspect here.

The blue painted region is the recoverable energy -- what you get back
when the BJT turns off.

There are two noticeable parts to that energy. One that lies to the
left of a steeper sloped part of the red line and a tiny remainder
that lies also to the left of that almost imperceptably sloped red
line. The total area of the blue region is your energy.

Okay. So what happens under saturation? It's not entirely bad. There
is still energy to be recovered. The problem is that the current
rises upwards like a veritable rocket ship (causing the BJT to shut
off quicker) and for all that current there isn't all that much energy
added for the trouble.

Now, if you had access to super-conducting wires with zero ohms and
other idealist electronic parts like BJTs with an infinite capacity to
handle current, you wouldn't care. Air core would be fine. But
reality impinges. Wires have resistance, electronic parts that handle
huge currents are big and cost a lot, etc.

So current matters because real parts dissipate energy when passing
large currents. And so does frequency because real parts tend to
dissipate energy when dealing with high frequencies, as well as
behaving differently, too.

One thing we've already worked out for you is this:

Ic_peak = 2*Iout*(1+Vout/Vin)

So we know what that current should peak out at and we really don't
have a choice about this. Although it might be nice to figure out how
to use a lower Ic_peak, we don't have that option. It's cast in
stone. Frequency is something we can control, though. Via the
inductance of the collector winding in this circuit.

So what does saturation do? One thing is that it reduces the time
required to reach Ic_peak. In doing that, it increases the frequency
of operation. We generally don't want too much of that.

Are there any positive benefits to saturation in this case? Perhaps.
If it is a very controlled amount. Most of the BJT dissipation, in
watts, takes place towards the very end of the ON time. If we could
shorten up the ON time, then the total energy (watts times time)
wasted there would be less. If the transformer could be designed to
have a controlled point of saturation such that the knee (at point A
in the curve mentioned above) takes place close to Ic_peak, then the
current would rapidly rise and, in doing so, shorten the duration that
the BJT's Vce soaks up power and that would reduce wasted energy in
the circuit. The cost of being wrong, and having that point A be far,
far too soon, is that the frequency of operation would go way up and
we'd lose out on that score.

So perhaps it would be a good idea to use a planned design where the
Bmax value is used with the idea that if we are a little bit wrong
about it, we can adjust winding turns and tweak towards an even more
efficient design.

Where does point A take place? I'll use terms found at that Fair-Rite
web site thing you pointed at. It's about here:

Imax = Bmax * le / (u0 * ui * N)

where N is the number of turns you use. (ui is what they use for the
relative permeability of the material, u0 is the permeability of air
[4*pi*1e-7], and le is what they use for the effective magnetic loop
length.) In the above equation, using the u0 figure I gave, le should
be given in meters and Bmax in Teslas, not Gauss.

Note that larger N and larger absolute permeability of the core reduce
the current level we can reach before saturation. Longer magnetic
path lengths increase it.

Bmax is pretty much set. Ferrites will be in the 0.1 to 0.3 Teslas.
Iron cores maybe go up to 2T.

All this points up one of the reasons why I think that lower
permeability cores are better here. And it also shows why you
discovered that fewer windings worked better (when you had too many, N
was larger and Imax became less than your needed Ic_peak, the power
delivered went down, and probably your frequency went up, too.)

Let's call Ic_peak (the one you _want_ to have) is about 350mA. We
will set Imax to that value. Let's start with the number of turns and
set it to 50, for now. We will assume ferrite. Your web site uses
Gauss and Oersteds.

Take a look at their 601 sized type 67 core. I clicked on the first
tab at the site you mentioned (type 67 tab) and then looked down to
find their part numbers and the one ending in 601. Click on that.

Look down the page and see Hysteresis Loop chart. The 25C curve
starts to 'softly' bend over somewhere between 1000 and 1500 Gauss.
That's the beginning of saturation. On ferrites, this is usually a
slow process.

Now go back and pick out the same sized type 61 core. (Bigger
selection to choose from, size wise.) Find the 601 again and look at
the Hysteresis Loop again. Yup. About 1500 Gauss, again. But this
curve looks less gradual as it goes into saturation -- more
pronounced.

Now same sized type 43. Same area where it starts to bend over,
though there is 'something weird' going on at about 1 Oersted. Also
take a look at the Initial Permeability vs. Temperature curve. That
one shows a lot of change over termperature. Hold that thought, look
back at the prior two, and go over to type 75 and the 601 part there
and look at its temp curve, too.

I don't like that variation over temperature. So already besides the
fact that we already know that lower permeability allows a higher Imax
(which we may indeed want), temperature variation also adds another
encouragement.

By the way, if we use lower perm materials we will need more turns to
get the frequency we want. But this is fine because inductance goes
down proportionally to permeability but up by the number of turns
squared. So just a few extra windings can compensate, easily.

Let's focus on Type 61. It's temperature curve looks nice and its
saturation looks predictably smooth. Looking at the saturation curve
(and noting what it does at 100C as well as 25C), we need to figure
out H, in Oersteds.

H = 0.4*pi*N*I/le, using le in centimeters.

With le=5.2cm, we get 4.23 Oersteds. Definitely starting to saturate
there. But as I discussed above, that's not necessarily a bad thing.
It actually might help just a little in shortening up the time where
the BJT is dissipating faster. So let's say that looks pretty darned
good, as a first shot at this!

What's the relative permeability? Well, it's dB/dH in the cgs system.
At 4 Oersteds I read out 1900 Gauss and at 0 Oersteds I read out 1250.
This works out to (1900-1250)/(4-0) or about 162.5 (at 10kHz
operation.) So using the design figure of 125 and our computed
estimate aren't too far apart.

What's the inductance for this core with 50 turns for the collector
winding? Well,

L = u0*ui*Ae*N^2/le

But the figures are in meters, etc. So from the web site we see that
Ae=.243 cm^2, which is 2.43e-5 m^2. Also, le=5.2cm or .052m. So, L
is about 185uH, using the ui=125 figure. Or getting close to 240uH
with the estimate we made at 4 Oersteds. Turns out, this is probably
pretty good for a usable frequency, too.

The only question now is ... can you actually wind 50 turns for the
collector and still have room enough for your secondary winding.

PS: Is RS "Radio Shack"?

I don't know. It's what I'd assume. But I can't recall them having
any cores. Or if they do, it'll not be very many types or sizes.

Jon

 

[Jon Kirwan]

On Fri, 31 Jul 2009 10:44:16 GMT, Jon Kirwan
<jonk@infinitefactors.org> wrote:

If we could shorten up the ON time

insert,
[right at this point of operation]

, then the total energy (watts times time)
wasted there would be less.

Just in case the point wasn't clear.

Jon

 

[John Larkin]

On Thu, 30 Jul 2009 23:27:12 -0700 (PDT), fungus
<openglMYSOCKS@artlum.com> wrote:

I discharged a set of batteries through the latest JT and added the
results to my input/output graph:

http://www.artlum.com/jt/jt_vs_res.gif

The low end of the voltage range is done with one/two batteries.

Rotten excuse for regulation.

John

 

[fungus]

On Jul 31, 12:44 pm, Jon Kirwan <j...@infinitefactors.org> wrote:

Air core is acceptable and you've already got some thoughts on that
here.

Does air have high or low permeability.

All this points up one of the reasons why I think that lower
permeability cores are better here.  And it also shows why you
discovered that fewer windings worked better (when you had too many, N
was larger and Imax became less than your needed Ic_peak, the power
delivered went down, and probably your frequency went up, too.)

On my big ferrite ring, adding too many turns had a negative effect
on output current, yes. On the good rings I could keep on adding
turns until the ring was physically full and output current kept
on going up while frequency was going down (I assume I never
reached saturation but that eventually I would if I could add
enough turns).

I'm still a bit unsure of the relationship between permeability and
saturation. Low permeability means magnetic flux has a harder
time penetrating so it takes longer to align the atoms inside it,
right? This seems at odds with my observation that the best ones
for a Joule Thief are the ones which stick hardest to a magnet.
Surely they would have more iron content and thus higher
permeability.

Probably my monkey-brain is missing something obvious.

The only question now is ... can you actually wind 50 turns for the
collector and still have room enough for your secondary winding.

Easily ... in fact I got nearly 100 turns on something that would fit
through the middle of a 601. I'd be tempted to go with the 301 size.
The numbers look identical as the 601 for the same material.

PS: I wind primary/secondary at the same time using a double
strand of wire.

 

[Jon Kirwan]

On Fri, 31 Jul 2009 06:35:38 -0700 (PDT), fungus
<openglMYSOCKS@artlum.com> wrote:

On Jul 31, 12:44 pm, Jon Kirwan <j...@infinitefactors.org> wrote:

Air core is acceptable and you've already got some thoughts on that
here.

Does air have high or low permeability.

Air has the lowest.

All this points up one of the reasons why I think that lower
permeability cores are better here.  And it also shows why you
discovered that fewer windings worked better (when you had too many, N
was larger and Imax became less than your needed Ic_peak, the power
delivered went down, and probably your frequency went up, too.)

On my big ferrite ring, adding too many turns had a negative effect
on output current, yes. On the good rings I could keep on adding
turns until the ring was physically full and output current kept
on going up while frequency was going down (I assume I never
reached saturation but that eventually I would if I could add
enough turns).

Probably so.

I'm still a bit unsure of the relationship between permeability and
saturation. Low permeability means magnetic flux has a harder
time penetrating so it takes longer to align the atoms inside it,
right? This seems at odds with my observation that the best ones
for a Joule Thief are the ones which stick hardest to a magnet.
Surely they would have more iron content and thus higher
permeability.

I'll start with the last comment here and work outwards from there.

Iron has a higher Bmax, by a factor of 10 or so. So you can 'run it
up' more. Magnets do stick to iron. Ferrite Bmax is usually in the
1000-3000 Gauss area, Iron can be more like 20000 Gauss. That's the
factor of 10 or so. I need to learn more about materials, like the
difference between iron and some iron oxides and other additives which
include nickel, to know more about the details why. But it seems to
be a general rule that there are two 'classes' of saturation -- your
basic iron core, which saturates at much higher levels, and ferrites
which hinder eddy currents and saturate at a factor of 10 below iron.
So for now, I just figure this as two categories with a general rule
about each. I'm sure there is a continuum of sorts that a materials
scientist could go on and on about. But the practical division seems
apparent to me, right now.

Getting to your question about what low permeability "means," at our
macroscale level of experience it's just a constant (or a function)
used in some equations. You go look at charts and pick off values.
That's about it.

I gave an explanation (that I like, haven't seen elsewhere, and yet
helps makes sense to me) already about it, though. I think of empty
space (vacuum) as being the ONLY place where magnetic energy can be
stored. Atoms can sit within that space and act as a short circuit so
that it takes no energy to cross through them. I kind of think of
these atoms as 'carriers' that provide a free ride across a small
vacuum gap. And because it takes no energy, no energy gets stored
there. It's only stored in the leaps the field must make when the
free ride ends and across to the next free ride at a nearby atom.
Feynman (if I may dare speak for him) might suggest that a specific
magnetic surface bubble is a fiction and that actually the magnetic
force takes all possible pathways simultaneously but that the sum of
the other longer pathways cancel each other out leaving only the
minimum energy pathway left for us to worry over [for every longer
path, there is always another longer path with opposite phase to
cancel it out; but for the minimum energy path there is only the one
and nothing cancels it.] In that view, low permeability just means
fewer 'free rides' and more vacuum space that has to be hopped. In
the case of an air core, it's 'vacuum all the way down' and there are
no free rides. Because of that, flux density is very low and the
field is forced to bloom out and take up a lot of volume. Spread out
like that, coupling between two nearby coils (transformer) is poorer.
Which is one reason why it's nice to have a low reluctance pathway --
it allows a higher concentration within a known volume and the wires
are linked much better. Even with low permeability materials, because
while there is (according to me) more vacuum that must be hopped, the
nifty arrangement we've made of the solid material provides a lot of
short hops that the magnetic field uses. Imagine you and a bunch of
others are in a bicycle race from point A to B. Between the two
points is a lot of rough ground. But... someone has placed a whole
lot of short strips of smooth pavement "aligned" A to B. You can use
them, if you want, but they end and you still have to cover some rough
ground to get to the next short strip. Now, do you think the racers
will take advantage of these short hops or not? I think they would.
Because the alternative (avoiding the useful, short strips) would be
stupid. Might as well get the easy bits you can, yes? Now if you
spaced them further apart, the energy required to get from A to B
might increase, but the riders would still use the strips all the same
because it is still easier that way. And that also means that besides
improving the energy required (reducing the energy involved), the
pathways also mean that you can predict where all those bicycles are
at better, too. (read: better flux linking.)

That's kind of how I like to see it.

I'll divert for a moment. Maxwell combined magnetics and electrics
into electromagnetism and showed (actually, he guessed at it because
the two constants yielded the speed of light and he felt that it was
too much of a coincidence to not be cause/effect) that the speed of
light itself depends upon two constants, the permitivity and
permeability of free space. The permeability, in SI units, is
4*pi*1e-7. And it is a constant of proportion in the equation B=mu*H,
which relates magnetic flux to the perpendicular magnetic field's
equipotential surfaces.

(Sometimes, it's kind of hard visualizing this. One way is to imagine
the surface of a ball that has hairs on it. Now imagine that the
hairs on the surface must always be exactly perpendicular to the
surface of the ball. Now further imagine that each hair is actually
just a short segment of a loop that penetrates the ball, follows some
pathway through the interior, comes out the other side and loops out
high above the ball [or far below it], gradually curls back around to
the original hair. Now also imagine that the ball isn't just one
surface, but a stack of surfaces below and above. If you can get all
that in your head, you are imagining something not unlike the
difference between magnetic fields and magnetic flux in 3D space.)

Permitivity relates to the electric field and flux and has it's own
(different) mu value, in D=mu*E.

If you aren't confused yet, let me add something else. These two 3D
things are actually related to each other via Maxwell's equations (4
of them, which relate D, E, B, and H.) And even Maxwell is behind the
times. Radioactivity (weak force) was combined into the other two.
And I believe that the strong nuclear force has also been added (pi
mesons and the like.)

It gets worse. There is a difference between a micro-scale view and a
macro-scale view and the differences gets important when trying to
explain things like negative refraction. Macro views are essentially
some kind of averaging (usually spatial) and new physical laws arrive
at those levels. An example of this effect (macro vs micro) are the
concepts of temperature, entropy, and energy. Energy is a micro and
macro scale concept. But temperature and entropy are NOT micro scale
and only arrive as 'entities' at a macro scale view.

Probably my monkey-brain is missing something obvious.

More likely my monkey-brain is making a darned mess of things.

The only question now is ... can you actually wind 50 turns for the
collector and still have room enough for your secondary winding.

Easily ... in fact I got nearly 100 turns on something that would fit
through the middle of a 601. I'd be tempted to go with the 301 size.
The numbers look identical as the 601 for the same material.

PS: I wind primary/secondary at the same time using a double
strand of wire.

Okay. I want to encourage you to try out the other schematic I gave,
which requires a 4:1 winding ratio. Later, though. For now, you are
right. Just get the original method down well.

In any case, you can try out the equations I discussed in selecting a
different core, too. Pick another one (one you can get ahold of?) and
see where the numbers put you. If they are about right (collector
winding in the area of more than 100uH, let's say), then you've got a
good shot at it.

Jon

 

[Jon Kirwan]

On Fri, 31 Jul 2009 14:38:58 -0700 (PDT), fungus
<openglMYSOCKS@artlum.com> wrote:

On Jul 31, 12:44 pm, Jon Kirwan <j...@infinitefactors.org> wrote:

Are there any positive benefits to saturation in this case?  Perhaps.
If it is a very controlled amount.

My ham-fisted experiments seem to show that R1 can't
be used to control the output of a saturated circuit.

R1 can't in that case. Did I seem to give a different impression?

Jon