Dynohub + LED

08 Sep 2004 12:27:39 -0500, Jim Smith:

LEDs ideally like a constant current supply. A
voltage regulator and a resistor is a common approximation of this.
Something like the zeners and a Luxeon with a current limiting
resistor would probably work and be safe. ISTR that one can buy
little modules for about $10 that are designed to serve as a constant
current source for LED lights. Should be more efficient than the
zener and resistor.

Bicycle generators are current limited - not voltage limited. Just use a
rectifier and LED(s) which can handle the 500-600mA (Luxeon 3W or 5W
type), maybe add a fat capacitor for smoothing and a goldcap for
"standlight" and you are done. A voltage regulator is only a waste. If you
want to use "classic" 20-50mA LEDs the best way is to hook two in series,
add a small series resistor and bundle 10 to 25 of this strings.

Also, white LEDs usually drop about 3.5 volts, so
maybe the generator can drive two in series.

Voltage doesn't matter. A bicycle dynamo will provide 3Volt to a 6 Ohm
load, 6Volt to a 12 Ohm load, 12 Volt to a 24 Ohm load (if speed is high
enough).

Andreas

08 Sep 2004 12:27:39 -0500, Jim Smith:

LEDs ideally like a constant current supply. A
voltage regulator and a resistor is a common approximation of this.
Something like the zeners and a Luxeon with a current limiting
resistor would probably work and be safe. ISTR that one can buy
little modules for about $10 that are designed to serve as a constant
current source for LED lights. Should be more efficient than the
zener and resistor.

Bicycle generators are current limited - not voltage limited. Just use a
rectifier and LED(s) which can handle the 500-600mA (Luxeon 3W or 5W
type), maybe add a fat capacitor for smoothing and a goldcap for
"standlight" and you are done. A voltage regulator is only a waste. If you
want to use "classic" 20-50mA LEDs the best way is to hook two in series,
add a small series resistor and bundle 10 to 25 of this strings.

Also, white LEDs usually drop about 3.5 volts, so
maybe the generator can drive two in series.

Voltage doesn't matter. A bicycle dynamo will provide 3Volt to a 6 Ohm
load, 6Volt to a 12 Ohm load, 12 Volt to a 24 Ohm load (if speed is high
enough).

Andreas

Chalo wrote:

Frank Krygowski wrote:

Hmmm. (Thinking slowly, since I'm an ME, not an EE) OK, that makes
some sense to me. But are the differences likely to be significant?
Another post referred to LED flashlights significantly overdriving the
LEDs. If you were closer to spec by my method, would the (apparent)
safety factor in the ratings be good enough?

A problem, I guess, is that failure might be quickly progressive. One
LED might fail and take out the whole array.


Right; the problem relates to an LED's resistance going _down_ as it
gets hotter, just like any other semiconductor's. A small difference
in forward voltage becomes exaggerated in a self-stoking cycle that
can lead to premature failure of the hottest-running LED in a parallel
array. Then the next hottest one goes, and so forth. It's a common
phenomenon among the cheapest and simplest multi-LED flashlights,
accelerated there because the LEDs are overdriven.

OK, a related question: Are current limits for LEDs based entirely on
management of heat, or is there something else besides?

IOW, what can you get away with in duty cycles, where an LED is
significantly overdriven but only for a small percentage of a short
repeating cycle time, and shut off for the rest?

With tungsten filament bulbs, AFAIK, the thermal lag ensures that it's
just RMS current that matters, at least at reasonably high cycle
frequencies. I imagine "reasonably high frequencies" may need to have a
larger minimum for LEDs (since they operate almost instantaneously) but
is the situation similar, or is there some maximum current even for an
infinitesmal duration?

I guess I'm also curious about the relationship between current and
light output - is it directly proportional, or does output increase more
slowly than current?

I suppose I could go look all this stuff up. But Usenet is more fun.


--
Frank Krygowski [To reply, remove rodent and vegetable dot com.
Substitute cc dot ysu dot
edu]

Chalo wrote:

Frank Krygowski wrote:

Hmmm. (Thinking slowly, since I'm an ME, not an EE) OK, that makes
some sense to me. But are the differences likely to be significant?
Another post referred to LED flashlights significantly overdriving the
LEDs. If you were closer to spec by my method, would the (apparent)
safety factor in the ratings be good enough?

A problem, I guess, is that failure might be quickly progressive. One
LED might fail and take out the whole array.


Right; the problem relates to an LED's resistance going _down_ as it
gets hotter, just like any other semiconductor's. A small difference
in forward voltage becomes exaggerated in a self-stoking cycle that
can lead to premature failure of the hottest-running LED in a parallel
array. Then the next hottest one goes, and so forth. It's a common
phenomenon among the cheapest and simplest multi-LED flashlights,
accelerated there because the LEDs are overdriven.

OK, a related question: Are current limits for LEDs based entirely on
management of heat, or is there something else besides?

IOW, what can you get away with in duty cycles, where an LED is
significantly overdriven but only for a small percentage of a short
repeating cycle time, and shut off for the rest?

With tungsten filament bulbs, AFAIK, the thermal lag ensures that it's
just RMS current that matters, at least at reasonably high cycle
frequencies. I imagine "reasonably high frequencies" may need to have a
larger minimum for LEDs (since they operate almost instantaneously) but
is the situation similar, or is there some maximum current even for an
infinitesmal duration?

I guess I'm also curious about the relationship between current and
light output - is it directly proportional, or does output increase more
slowly than current?

I suppose I could go look all this stuff up. But Usenet is more fun.


--
Frank Krygowski [To reply, remove rodent and vegetable dot com.
Substitute cc dot ysu dot
edu]

Chalo wrote:

Frank Krygowski wrote:

Hmmm. (Thinking slowly, since I'm an ME, not an EE) OK, that makes
some sense to me. But are the differences likely to be significant?
Another post referred to LED flashlights significantly overdriving the
LEDs. If you were closer to spec by my method, would the (apparent)
safety factor in the ratings be good enough?

A problem, I guess, is that failure might be quickly progressive. One
LED might fail and take out the whole array.


Right; the problem relates to an LED's resistance going _down_ as it
gets hotter, just like any other semiconductor's. A small difference
in forward voltage becomes exaggerated in a self-stoking cycle that
can lead to premature failure of the hottest-running LED in a parallel
array. Then the next hottest one goes, and so forth. It's a common
phenomenon among the cheapest and simplest multi-LED flashlights,
accelerated there because the LEDs are overdriven.

OK, a related question: Are current limits for LEDs based entirely on
management of heat, or is there something else besides?

IOW, what can you get away with in duty cycles, where an LED is
significantly overdriven but only for a small percentage of a short
repeating cycle time, and shut off for the rest?

With tungsten filament bulbs, AFAIK, the thermal lag ensures that it's
just RMS current that matters, at least at reasonably high cycle
frequencies. I imagine "reasonably high frequencies" may need to have a
larger minimum for LEDs (since they operate almost instantaneously) but
is the situation similar, or is there some maximum current even for an
infinitesmal duration?

I guess I'm also curious about the relationship between current and
light output - is it directly proportional, or does output increase more
slowly than current?

I suppose I could go look all this stuff up. But Usenet is more fun.


--
Frank Krygowski [To reply, remove rodent and vegetable dot com.
Substitute cc dot ysu dot
edu]

Frank Krygowski wrote:

Chalo wrote:

A small difference
in forward voltage becomes exaggerated in a self-stoking cycle that
can lead to premature failure of the hottest-running LED in a parallel
array.

OK, a related question: Are current limits for LEDs based entirely on
management of heat, or is there something else besides?

My understanding is that it is solely a heat transfer issue. However,
the critical junction is between the LED die (chip) and its substrate,
which imposes an upper power limit even with infinite heatsinking.

IOW, what can you get away with in duty cycles, where an LED is
significantly overdriven but only for a small percentage of a short
repeating cycle time, and shut off for the rest?

That method is used to make blinky lights flash more brightly than
they could burn steadily. As long as the LED die remains within its
tolerable temperature limits, no problem.

With tungsten filament bulbs, AFAIK, the thermal lag ensures that it's
just RMS current that matters, at least at reasonably high cycle
frequencies. I imagine "reasonably high frequencies" may need to have a
larger minimum for LEDs (since they operate almost instantaneously) but
is the situation similar, or is there some maximum current even for an
infinitesmal duration?

At some current I imagine the resistance in the tiny leads would
impose a limit. But your question exceeds my understanding of the
relevant issues.

I guess I'm also curious about the relationship between current and
light output - is it directly proportional, or does output increase more
slowly than current?

Efficiency drops off with increasing power, within normal power
ranges. You get more lumens per watt by underdriving LEDs than you do
by running them to spec, and overdriving them yields much less change
in luminosity than the change in power input.

Overdriving LEDs is a cost-cutting measure (fewer LEDs to get a
desired light output) for manufacturers, who are not feeding
replacement batteries to the devices after all.

Chalo Colina

Frank Krygowski wrote:

Chalo wrote:

A small difference
in forward voltage becomes exaggerated in a self-stoking cycle that
can lead to premature failure of the hottest-running LED in a parallel
array.

OK, a related question: Are current limits for LEDs based entirely on
management of heat, or is there something else besides?

My understanding is that it is solely a heat transfer issue. However,
the critical junction is between the LED die (chip) and its substrate,
which imposes an upper power limit even with infinite heatsinking.

IOW, what can you get away with in duty cycles, where an LED is
significantly overdriven but only for a small percentage of a short
repeating cycle time, and shut off for the rest?

That method is used to make blinky lights flash more brightly than
they could burn steadily. As long as the LED die remains within its
tolerable temperature limits, no problem.

With tungsten filament bulbs, AFAIK, the thermal lag ensures that it's
just RMS current that matters, at least at reasonably high cycle
frequencies. I imagine "reasonably high frequencies" may need to have a
larger minimum for LEDs (since they operate almost instantaneously) but
is the situation similar, or is there some maximum current even for an
infinitesmal duration?

At some current I imagine the resistance in the tiny leads would
impose a limit. But your question exceeds my understanding of the
relevant issues.

I guess I'm also curious about the relationship between current and
light output - is it directly proportional, or does output increase more
slowly than current?

Efficiency drops off with increasing power, within normal power
ranges. You get more lumens per watt by underdriving LEDs than you do
by running them to spec, and overdriving them yields much less change
in luminosity than the change in power input.

Overdriving LEDs is a cost-cutting measure (fewer LEDs to get a
desired light output) for manufacturers, who are not feeding
replacement batteries to the devices after all.

Chalo Colina

Frank Krygowski wrote:

Chalo wrote:

A small difference
in forward voltage becomes exaggerated in a self-stoking cycle that
can lead to premature failure of the hottest-running LED in a parallel
array.

OK, a related question: Are current limits for LEDs based entirely on
management of heat, or is there something else besides?

My understanding is that it is solely a heat transfer issue. However,
the critical junction is between the LED die (chip) and its substrate,
which imposes an upper power limit even with infinite heatsinking.

IOW, what can you get away with in duty cycles, where an LED is
significantly overdriven but only for a small percentage of a short
repeating cycle time, and shut off for the rest?

That method is used to make blinky lights flash more brightly than
they could burn steadily. As long as the LED die remains within its
tolerable temperature limits, no problem.

With tungsten filament bulbs, AFAIK, the thermal lag ensures that it's
just RMS current that matters, at least at reasonably high cycle
frequencies. I imagine "reasonably high frequencies" may need to have a
larger minimum for LEDs (since they operate almost instantaneously) but
is the situation similar, or is there some maximum current even for an
infinitesmal duration?

At some current I imagine the resistance in the tiny leads would
impose a limit. But your question exceeds my understanding of the
relevant issues.

I guess I'm also curious about the relationship between current and
light output - is it directly proportional, or does output increase more
slowly than current?

Efficiency drops off with increasing power, within normal power
ranges. You get more lumens per watt by underdriving LEDs than you do
by running them to spec, and overdriving them yields much less change
in luminosity than the change in power input.

Overdriving LEDs is a cost-cutting measure (fewer LEDs to get a
desired light output) for manufacturers, who are not feeding
replacement batteries to the devices after all.

Chalo Colina

lance house wrote:


hi

has anyone tried building an LED lamp to be powered by the old
Sturmey-Archer Dynohub? I found one in fantastic condition and ,ahem,
fried one of my LED front lights on an, ah, test ride. I'm figuring now
that I definitely need a bridge rectifier and maybe some batteries as a
kind of voltage sink. Can I just rig it up with about the right number
of LEDs in series to accommodate the voltages which I expect to get out
of it?

Lamps designed for dynamos (like B&M Lumotec) have built-in voltage
limiting circuitry (zener diode?), which really does work. You may be
able to cannibalise a B&M lamp - the basic Lumotecs are quite cheap.