Driving LEDs with a battery pack

[fungus]

On Jul 10, 1:13 pm, greg <g...@cosc.canterbury.ac.nz> wrote:

fungus wrote:
With three batteries and four turns my little transistor heats up
rapidly

That suggests the circuit is operating inefficiently
due to the transistor switching time being a significant
fraction of the pulse time.

I imagine in my ignorance that the transistor heats up
most when it's half-on (still has some internal resistance),
is that right? Slow switching means it spends more time
in that zone so it heats up more...?

Also, do you know what kind of ferrite your ring is
made of?

No idea, sorry. I pulled it out of a junker PC.

 

[fungus]

On Jul 10, 1:45 pm, default <defa...@defaulter.net> wrote:

They are damn bright.  

They also have some impressive mcd numbers ALONG with fairly large
beam spreads.

Beam spread is ok (might even be better for what I'm doing...)

 

[default]

On Fri, 10 Jul 2009 05:46:34 -0700 (PDT), fungus
<openglMYSOCKS@artlum.com> wrote:

On Jul 10, 1:45 pm, default <defa...@defaulter.net> wrote:

They are damn bright.  

They also have some impressive mcd numbers ALONG with fairly large
beam spreads.


Beam spread is ok (might even be better for what I'm doing...)

That's the nature of the millicandella rating though. It is easy to
make a poor performer sound good simply by narrowing the focus of the
spot of light.

The same total flux of light from a led, is brighter when the light
energy is concentrated in a smaller beam. The led chip and total flux
output can be the same, and the manufacturer just offers different
beam angles by changing the shape of the lens.

You, no doubt, have an idea of the size of the area you want to
illuminate. You select based on the distance from the source and
illumination needed and choose the angle and candela output to match.

I have some 25 leds up on a pole illuminating a ramp I use to load my
kayak in the dark - narrow angle, high output leds work better for
that application.
--

 

[Jon Kirwan]

On Fri, 10 Jul 2009 23:31:50 +1200, greg <greg@cosc.canterbury.ac.nz>
wrote:

Jon Kirwan wrote:

But that may actually be because of saturation in the core,
itself. It turns out that it is the area underneath your volt curve
(volt-seconds, Webers, etc.) that saturate the core. With low
frequencies, the applied voltage hangs around longer and therefore the
volt-seconds goes up. That's bad, mostly because at some point the
whole thing just stops accepting any more volt-seconds and to fix that
the voltage goes to zero (and this means your transistor gets the
entire voltage applied to it and that makes it very hot if it stays ON
for long.)

Hang on a moment -- unless I'm mistaken, this type of
circuit *relies* on saturation of the core for its
operation. As soon as saturation occurs, the transistor
turns off.

So saturation isn't bad, it's necessary! The only
question is how long you want to let the inductance
charge up before it occurs.

I don't think so. I can do the calculations, and achieve accurate
results from them, without taking saturation into account. I'll try
and give a more detailed account than I have, if interested.

Jon

 

[Jon Kirwan]

On Fri, 10 Jul 2009 05:16:40 -0700 (PDT), fungus
<openglMYSOCKS@artlum.com> wrote:

On Jul 10, 4:33 am, Jon Kirwan <j...@infinitefactors.org> wrote:

snip
That means your poor BJT is zenering, I think.

Don't know what that implies ... but it sounds scary

"Breaking down" and going bzzzzt internally because the voltage is too
much. It doesn't make the transistor better.

Jon

 

[Jon Kirwan]

Okay. Here's another shot at the general idea. I'll write this in
ASCII characters. You'll need to use a fixed-spaced font to read it,
such as Courier or Courier New or Prestige Elite. Notepad will do
that, so will many newsgroup readers. Even google can do it, if you
look around for the option on the page.

I'll assume 6 LEDs and 3 batteries:

: ,--------,
: | |
: | ---
: | \ / D6
: | ---
: \ |
: ,---------+----------, / R2 |
: | | | \ 1 ---
: | | | / \ / D5
: | | | o | ---
: | )| )| | |
: --- )| L1 )| L2 | |
: - V1 )| )| | ---
: --- 1.5V )| )| | \ / D4
: - | o | | ---
: | | | D8 | |
: | | +--|>|---+ |
: | | | | ---
: --- \ | | \ / D3
: - V2 / R1 | | ---
: --- 1.5V \ | | |
: - / | | |
: | | | --- C1 ---
: | | |/c Q1 --- 10uF \ / D2
: | +--------| | 50V ---
: --- | |>e | |
: - V3 | | | |
: --- 1.5V _|_ D7 | | ---
: - /_\ | | \ / D1
: | | | | ---
: | | | | |
: | | | | |
: gnd gnd gnd gnd gnd

Tweek your transformer to get in the vacinity of 20-50kHz. The LED
current will vary somewhat less, closer to 50kHz, but down at 20kHz it
should look fine, as well. C1 should be at least 35V, but 50V is
safer.

The value for R1 can be as you see fit. I'd stay at or above 1k and
you can try it upwards of 3.3k. I tweeked things so that 2.7k appears
to work fine.

LED current will definitely droop as the battery voltages also droop.
Nothing in there to help that problem. As battery voltage goes down,
so does base current. Lower base current, lower peak collector
current. Lower peak collector current, lower energy stored and
transferred. Etc. But I think it will move down by the ratio of
voltages, so no worse than 1.1/1.5 or about 3/4ths.

What you need to add is C1, R2 (not strictly necessary, I just added
it so you could play with some values there), and D8. Some folks may
recommend a schottky diode for that, because they present a somewhat
lower voltage and they switch fast. But probably you can use what you
have available.

Oh, and D7!! I added that to protect your Q1 base-emitter from
reverse voltages. That will help protect your poor BJT.

If you try this, let me know what happens!

Jon

 

[Jon Kirwan]

On Fri, 10 Jul 2009 17:05:05 GMT, Jon Kirwan
<jonk@infinitefactors.org> wrote:

Tweek

Tweak!

Jon

 

[Jon Kirwan]

On Fri, 10 Jul 2009 09:23:50 -0700 (PDT), fungus
<openglMYSOCKS@artlum.com> wrote:

On Jul 10, 5:00 pm, default <defa...@defaulter.net> wrote:

That's the nature of the millicandella rating though.  It is easy to
make a poor performer sound good simply by narrowing the focus of the
spot of light.

The same total flux of light from a led, is brighter when the light
energy is concentrated in a smaller beam.

Obviously...

So you think the LEDs this guy sells are just narrower
focus than the ones I have, ie. total number of photons
is approx. the same for all the 2.0V @ 20mA LEDs out
there?

If you read 'default' more closely, you'll see him write, "They also
have some impressive mcd numbers ALONG with fairly large beam
spreads." Note that he includes "___ALONG with___" here and describes
the idea of large beams. So, actually, I think 'default' is saying
that these seem to be pretty impressive not just for having high
numbers but also for having those numbers while also having wide
spreads.

So to the opposite, I think. I don't think he thinks they get their
numbers from a narrower focus. He feels pretty positive, I gathered.

Jon

 

[fungus]

On Jul 10, 5:00 pm, default <defa...@defaulter.net> wrote:

That's the nature of the millicandella rating though.  It is easy to
make a poor performer sound good simply by narrowing the focus of the
spot of light.

The same total flux of light from a led, is brighter when the light
energy is concentrated in a smaller beam.

Obviously...

So you think the LEDs this guy sells are just narrower
focus than the ones I have, ie. total number of photons
is approx. the same for all the 2.0V @ 20mA LEDs out
there?

 

[Jon Kirwan]

On Fri, 10 Jul 2009 23:31:50 +1200, greg <greg@cosc.canterbury.ac.nz>
wrote:

Jon Kirwan wrote:

But that may actually be because of saturation in the core,
itself. It turns out that it is the area underneath your volt curve
(volt-seconds, Webers, etc.) that saturate the core. With low
frequencies, the applied voltage hangs around longer and therefore the
volt-seconds goes up. That's bad, mostly because at some point the
whole thing just stops accepting any more volt-seconds and to fix that
the voltage goes to zero (and this means your transistor gets the
entire voltage applied to it and that makes it very hot if it stays ON
for long.)

Hang on a moment -- unless I'm mistaken, this type of
circuit *relies* on saturation of the core for its
operation.

I think you _are_ mistaken.

As soon as saturation occurs, the transistor turns off.

No. Saturation of the core isn't required. This would work on an air
core transformer, I believe.

.....

The base current starts out unaided by any induced voltage on the base
winding, but still turns the BJT on. Once that happens, though, the
collector drops immediately to near zero (there is no collector
winding current to speak of, at that point.) This places a near-fixed
voltage across the collector winding, which allows the collector
current to rise according to V/L. Almost from the first moment this
takes place, there is an induced voltage in the base winding due to
the rate of flux change on the collector winding which adds to the
battery voltage (if wired with the correct orientation, of course) and
this increases the base current to it's initially 'highish' value of
(2*V_bat-Vbe)/R.

At first, the Vce of the BJT remains very close to zero because the
Ic/Ib is well below 10. (The BJT is severely saturated.) But as Ic
rises along the ramp of V/L, while at the same time Ib remains close
to fixed, it eventually reaches the point where Ic/Ib goes over 1,
then goes over 10, then goes over 20, etc. During this time, the Vce
rises, too. As that happens, the induced voltage on the base winding
declines due to a falling rate of flux change. That reduces the base
current, but does so exactly at the point when higher Ic requires
more, not less, Ib. In other words, the aiding voltage by the base
winding is falling and _reducing_ base current right at the point
where Vce is rising and reducing the voltage across the collector
winding.

Suddenly, the beta just isn't enough and the BJT attempts to reduce
Ic. As soon as it 'tries' to do that, though, the collector winding
immediately responds by reversing its polarity as the only possible
response to allow a reduction in Ic (V/L must reverse its sign.) But
that immediately causes the base winding to also reverse its polarity
and __oppose__ the battery voltage that is struggling to drive current
into the base. The whole thing collapses with the battery voltage
opposed by an overwhelming reverse polarity, base current goes to
zero, base voltage goes below ground, etc. The BJT is off at this
point.

After the collector winding reverses voltage, current is driven now
through the LED as the reversed-sign V/L allows the current through it
to decline along a ramp (the LED maintains a fairly fixed, but
gradually declining V/L.) At some point, the LED is no longer able to
accept much current (non-linear decline, as well) and the collector
winding's field entirely collapses and has no energy remaining. It's
voltage goes to zero, so does the induced voltage on the base winding,
the battery is now able to generate some current into the base of the
BJT, the BJT turns back on, a voltage is applied to the collector
winding, the collector winding induces a renewed aiding voltage on the
base winding, the base current rises a bit, and the whole cycle
repeats.

No saturation of a core invoked here.

So saturation isn't bad, it's necessary! The only
question is how long you want to let the inductance
charge up before it occurs.

But I disagree. And, it appears, so does LTSpice where I don't have
to add any saturation effects to get it to oscillate just fine.

Hope that helps. Or, if you find good fault with my reasoning, it
will help me. Either way, it's all good.

Jon

 

[fungus]

On Jul 10, 7:05 pm, Jon Kirwan <j...@infinitefactors.org> wrote:

Okay. Here's another shot at the general idea.

What you need to add is C1, R2 (not strictly necessary, I just added
it so you could play with some values there), and D8. Some folks may
recommend a schottky diode for that, because they present a somewhat
lower voltage and they switch fast. But probably you can use what you
have available.

I added D8 and a big scary electrolytic capacitor* I found
in my box at C1 (I don't exactly have a big selection of
components to play with and that's the only capacitor
which looked likely to do anything...)

The capacitor adds a cool effect - the LEDs fade up/down
when I switch the thing on/off.

But ... it works! I think I'm pretty much at full brightness
with the capacitor in there.

Looking on the scope, the voltage at the first diode is
perfectly flat (to be expected).

Best of all ... I measured the output current and with 4.2V
input it was 106mA for six LEDs, that's an average of 17.6mA
each - right on target.

With some half-depleted batteries I got 104mA across the
LEDs from 3.8V input.

With two half-depleted batteries I got 70mA from 2.5V
- still quite respectable.

Oh, and D7!! I added that to protect your Q1 base-emitter from
reverse voltages. That will help protect your poor BJT.

I'll be thinking of him...

If you try this, let me know what happens!

So far so good... now I want to work on the size of the
ferrite bead (I want this to fit in a more discrete package).


[*] Electrolytic capacitors always make me nervous...

 

[fungus]

On Jul 10, 7:09 pm, Jon Kirwan <j...@infinitefactors.org> wrote:

So to the opposite, I think.  I don't think he thinks they get their
numbers from a narrower focus.  He feels pretty positive, I gathered.

I'm definitely interested...

 

[fungus]

On Jul 10, 10:03 pm, fungus <openglMYSO...@artlum.com> wrote:

I added D8 and a big scary electrolytic capacitor

Which leads to another question before I go parts-shopping
tomorrow ... what sort of capacitor is best for this? There's
a zillion different types.

 

[Jon Kirwan]

On Fri, 10 Jul 2009 15:44:54 -0700 (PDT), fungus
<openglMYSOCKS@artlum.com> wrote:

On Jul 10, 10:03 pm, fungus <openglMYSO...@artlum.com> wrote:

I added D8 and a big scary electrolytic capacitor

Which leads to another question before I go parts-shopping
tomorrow ... what sort of capacitor is best for this? There's
a zillion different types.

Electrolytic for cheapness, 10uF or bigger, 35V or higher. I don't
think it matters that much, otherwise. You can strip something out of
some electronic device (the same one you got the transformer core out
of?) Don't shop if you don't have to....

Jon

 

[greg]

fungus wrote:

I imagine in my ignorance that the transistor heats up
most when it's half-on (still has some internal resistance),
is that right? Slow switching means it spends more time
in that zone so it heats up more...?

Yes, that's right. Which means you need to keep the
frequency low enough so that the switching time is
small compared to the on time.

If you have a scope, you could look at the voltage
between the emitter and collector. It should be
close to a square wave, with steep slopes on the
rising and falling edges, for best efficiency.

Also, do you know what kind of ferrite your ring is
made of?

No idea, sorry. I pulled it out of a junker PC.

What part of the PC? If it was somewhere in the
power supply, it's probably okay.

--
Greg

 

[greg]

fungus wrote:

That means your poor BJT is zenering, I think.

Don't know what that implies ... but it sounds scary

It means that there is a high enough reverse voltage
being applied to the base-emitter junction for it
to break down like a zener diode, which could damage
the transistor.

--
Greg

 

[greg]

Jon Kirwan wrote:

During this time, the Vce
rises, too. As that happens, the induced voltage on the base winding
declines due to a falling rate of flux change.

Interesting -- there appear to be more subtle things
going on in this circuit than I thought!

But it seems like saturation would have much the same
effect -- the flux suddenly stops rising, causing the
base voltage to fall to the point where transistor
can't sustain the current and a flip-over occurs.

So the circuit will still work if saturation occurs,
just in a slightly different mode.

It might even work slightly better in that mode, if
it causes the transistor to cut off more sharply.

In any case, it seems like it would operate more
predictably, since the period and maximum current
would depend on the saturation flux of the core
rather than some rather uncertain transistor parameters.

If you just pick some random transistor, then it's only
by luck that you avoid overdriving the LED.
Seems like a rather hairy way to design a circuit to me!

Another thought: For driving several LEDs in series,
maybe it would help to use a separate secondary winding
with more turns for the output. That would give a
current step-down relative to the transistor current
and allow the output duty cycle to be increased for
the same output current.

E.g. suppose you have 4 LEDS in series and want to
drive them at 20mA max. If you use a 4:1 turns ratio
and arrange things so that the primary charges up
to 80mA over an on-time of t, it will then deliver
20mA initially to the secondary, ramping down to 0
over a time of 4t before the stored energy runs out.
So the LEDs are driven at a duty cycle of 80%.

In contrast, without the current step-down, the peak
transistor current has to be limited to the maximum LED
current, and the LED duty cycle falls in proportion to
the number of LEDs -- which means you can never get more
light out of series LEDs than you could from a single LED.

Does any of that make sense?

--
Greg

 

[default]

On Fri, 10 Jul 2009 09:23:50 -0700 (PDT), fungus
<openglMYSOCKS@artlum.com> wrote:

On Jul 10, 5:00 pm, default <defa...@defaulter.net> wrote:

That's the nature of the millicandella rating though.  It is easy to
make a poor performer sound good simply by narrowing the focus of the
spot of light.

The same total flux of light from a led, is brighter when the light
energy is concentrated in a smaller beam.


Obviously...

So you think the LEDs this guy sells are just narrower
focus than the ones I have, ie. total number of photons
is approx. the same for all the 2.0V @ 20mA LEDs out
there?

No. The leds this guy sells specify beam angle, and has a range of
beam widths.

I'm very happy with the tricolor leds I got for a project. I'm using
them for power indicators and the wide beam lets me see them from
anywhere in the room with any ambient light level.

His price was 60 cents a led versus $2.80 from a US importer of the
same part.

Only negative comment I wish he'd link to the actual data sheets for
the Leds instead of the abbreviated ones he provides.
--

 

[fungus]

On Jul 11, 4:19 am, Jon Kirwan <j...@infinitefactors.org> wrote:

Electrolytic for cheapness, don't shop if you don't have to....

There's dozens of capacitors within easy reach but they're
all soldered to PCBs and only have 1mm legs... :-(

 

[fungus]

On Jul 11, 11:40 am, greg <g...@cosc.canterbury.ac.nz> wrote:

Also, do you know what kind of ferrite your ring is
made of?
No idea, sorry. I pulled it out of a junker PC.

What part of the PC? If it was somewhere in the
power supply, it's probably okay.

Wasn't from the PSU. I just opened a PSU and
pulled a couple of smaller ones out. They're made
of blue ceramic or something and they stick a lot
harder to a magnet than my big one does. I'll give
them a try when I can get some better wire.


PS: Hard disk magnets ... crazy strong. I nearly
lost a finger.