hi , i have 19 rows of four , 10mm white leds. would it be better to
power them straight from dc or a pwm supply ?
thanks

mark k

mark krawczuk wrote:
hi , i have 19 rows of four , 10mm white leds. would it be better to
power them straight from dc or a pwm supply ?
thanks

Define "better".

--
Regards,

John Popelish

"mark krawczuk" wrote in message
...
hi , i have 19 rows of four , 10mm white leds. would it be better to
power them straight from dc or a pwm supply ?
thanks

mark k

The LEDs each have a more or less constant forward voltage so you have to
arrange for the current to be limited to the value stated in the data sheet.

Traditionally this has been done with a series current limiting resistor,
but this wastes energy as heat in the resistor, the more LEDs you can chain
in series within the supply voltage the smaller the resistor and energy
waste - but this arrangement is not very tolerant of variations in supply
voltage.

Increasing numbers of wall-wart power packs are SMPSU, but that doesn't
necessarily mean good regulation - try a few off load and loaded at rated
power.

A SMPSU will at least minimise dissipation as compared to that in a linear
PSU.

A current regulated SMPSU isn't trivial, I'm currently experimenting with a
buck regulator to drive a single 1W TO220 white led from a 6V SLA battery,
early tests with a couple of ordinary LEDs drawing 20mA gave an efficiency
less than 30%, drawing higher current improved regulator efficiency but its
still only a small improvement on a series resistor.

"ian field" wrote in message
...

"mark krawczuk" wrote in message
...
hi , i have 19 rows of four , 10mm white leds. would it be better to
power them straight from dc or a pwm supply ?
thanks

mark k

The LEDs each have a more or less constant forward voltage so you have to
arrange for the current to be limited to the value stated in the data
sheet.

Traditionally this has been done with a series current limiting resistor,
but this wastes energy as heat in the resistor, the more LEDs you can
chain in series within the supply voltage the smaller the resistor and
energy waste - but this arrangement is not very tolerant of variations in
supply voltage.

Increasing numbers of wall-wart power packs are SMPSU, but that doesn't
necessarily mean good regulation - try a few off load and loaded at rated
power.

A SMPSU will at least minimise dissipation as compared to that in a linear
PSU.

A current regulated SMPSU isn't trivial, I'm currently experimenting with
a buck regulator to drive a single 1W TO220 white led from a 6V SLA
battery, early tests with a couple of ordinary LEDs drawing 20mA gave an
efficiency less than 30%, drawing higher current improved regulator
efficiency but its still only a small improvement on a series resistor.


"ian field" wrote in message
...

"mark krawczuk" wrote in message
...
hi , i have 19 rows of four , 10mm white leds. would it be better to
power them straight from dc or a pwm supply ?
thanks

mark k

The LEDs each have a more or less constant forward voltage so you have to
arrange for the current to be limited to the value stated in the data
sheet.

Traditionally this has been done with a series current limiting resistor,
but this wastes energy as heat in the resistor, the more LEDs you can
chain in series within the supply voltage the smaller the resistor and
energy waste - but this arrangement is not very tolerant of variations in
supply voltage.

Increasing numbers of wall-wart power packs are SMPSU, but that doesn't
necessarily mean good regulation - try a few off load and loaded at rated
power.

A SMPSU will at least minimise dissipation as compared to that in a linear
PSU.

A current regulated SMPSU isn't trivial, I'm currently experimenting with
a buck regulator to drive a single 1W TO220 white led from a 6V SLA
battery, early tests with a couple of ordinary LEDs drawing 20mA gave an
efficiency less than 30%, drawing higher current improved regulator
efficiency but its still only a small improvement on a series resistor.


Little wonder: the current sensor with a 1.25v feedback will eat up 35% of
the efficiency. You need to stick an amp in there to lower the drop. Or try
the arrangement shown on the third ('Hack') circuit he
http://www.quantsuff.com/LED2.htm.

"QS" wrote in message
...

"ian field" wrote in message
...

"mark krawczuk" wrote in message
...
hi , i have 19 rows of four , 10mm white leds. would it be better to
power them straight from dc or a pwm supply ?
thanks

mark k

The LEDs each have a more or less constant forward voltage so you have to
arrange for the current to be limited to the value stated in the data
sheet.

Traditionally this has been done with a series current limiting resistor,
but this wastes energy as heat in the resistor, the more LEDs you can
chain in series within the supply voltage the smaller the resistor and
energy waste - but this arrangement is not very tolerant of variations in
supply voltage.

Increasing numbers of wall-wart power packs are SMPSU, but that doesn't
necessarily mean good regulation - try a few off load and loaded at rated
power.

A SMPSU will at least minimise dissipation as compared to that in a
linear PSU.

A current regulated SMPSU isn't trivial, I'm currently experimenting with
a buck regulator to drive a single 1W TO220 white led from a 6V SLA
battery, early tests with a couple of ordinary LEDs drawing 20mA gave an
efficiency less than 30%, drawing higher current improved regulator
efficiency but its still only a small improvement on a series resistor.


"ian field" wrote in message
...

"mark krawczuk" wrote in message
...
hi , i have 19 rows of four , 10mm white leds. would it be better to
power them straight from dc or a pwm supply ?
thanks

mark k

The LEDs each have a more or less constant forward voltage so you have to
arrange for the current to be limited to the value stated in the data
sheet.

Traditionally this has been done with a series current limiting resistor,
but this wastes energy as heat in the resistor, the more LEDs you can
chain in series within the supply voltage the smaller the resistor and
energy waste - but this arrangement is not very tolerant of variations in
supply voltage.

Increasing numbers of wall-wart power packs are SMPSU, but that doesn't
necessarily mean good regulation - try a few off load and loaded at rated
power.

A SMPSU will at least minimise dissipation as compared to that in a
linear PSU.

A current regulated SMPSU isn't trivial, I'm currently experimenting with
a buck regulator to drive a single 1W TO220 white led from a 6V SLA
battery, early tests with a couple of ordinary LEDs drawing 20mA gave an
efficiency less than 30%, drawing higher current improved regulator
efficiency but its still only a small improvement on a series resistor.


Little wonder: the current sensor with a 1.25v feedback will eat up 35% of
the efficiency. You need to stick an amp in there to lower the drop. Or
try the arrangement shown on the third ('Hack') circuit he
http://www.quantsuff.com/LED2.htm.



That's not a buck regulator, there's no current control and whether the LED
is overdriven or not is more by luck than judgement.

"mark krawczuk" wrote in message
...
hi , i have 19 rows of four , 10mm white leds. would it be better to
power them straight from dc or a pwm supply ?
thanks

mark k


Check out http://home.comcast.net/~theledguy/home.htm for LED drivers.

hi, more efficent / brighter ?
mark k


"John Popelish" wrote in message
...
mark krawczuk wrote:
hi , i have 19 rows of four , 10mm white leds. would it be better to
power them straight from dc or a pwm supply ?
thanks

Define "better".

--
Regards,

John Popelish

mark krawczuk wrote:
"John Popelish" wrote:
mark krawczuk wrote:
hi , i have 19 rows of four , 10mm white leds. would it be better to
power them straight from dc or a pwm supply ?
thanks

Define "better".

hi, more efficient / brighter ?

Since you don't mention cost, size or complexity, you should
look into switching regulator concepts. Current or power
controlling switching regulators/converters will produce
more light per watt input than passive (series resistors) or
linear regulators, in most situations (ability to handle
voltage and temperature variations), since switchers have no
intentional power wasting parts controlling the current.
Only energy transfer and storage elements and switches.

Unfortunately, to control all LED currents, simultaneously,
the simplest approach is to connect them all in series and
regulate the single current through that string. But that
will require the regulator to produce about 260 volts for
your number of LEDs.

Of course, you can divide that voltage by the number of
separate regulators you make.

--
Regards,

John Popelish

On Tue, 26 Aug 2008 23:30:11 +0930, "mark krawczuk"
wrote:

hi , i have 19 rows of four , 10mm white leds. would it be better to
power them straight from dc or a pwm supply ?
thanks

mark k


Use 20 NPN transistors to make a current mirror. Hook their emitters
to ground, their collectors each to a string of 4 LEDs that are
connected to Vcc, and their bases together. Put a pot between Vcc and
the collector of the remaining one.

Then, you'll be able to control the brighness by changing the
resistance of the pot.

Google "current mirror".

Regards,
Bob Monsen

Robert Monsen wrote:

On Tue, 26 Aug 2008 23:30:11 +0930, "mark krawczuk"
wrote:


hi , i have 19 rows of four , 10mm white leds. would it be better to
power them straight from dc or a pwm supply ?
thanks

mark k



Use 20 NPN transistors to make a current mirror. Hook their emitters
to ground, their collectors each to a string of 4 LEDs that are
connected to Vcc, and their bases together. Put a pot between Vcc and
the collector of the remaining one.

Then, you'll be able to control the brighness by changing the
resistance of the pot.

Google "current mirror".

Regards,
Bob Monsen
Err...a (randumb) handful of transistors will have a fair range of
Vbe mismatch which canlead to some LEDs getting twice (or more) the
current than others.
Better put equal value of resistors in series with the emitters..
0.5V drop on the resistors will do wonders.

"Robert Baer" wrote in message
net...
Robert Monsen wrote:

On Tue, 26 Aug 2008 23:30:11 +0930, "mark krawczuk"
wrote:


hi , i have 19 rows of four , 10mm white leds. would it be better to
power them straight from dc or a pwm supply ?
thanks

mark k


Use 20 NPN transistors to make a current mirror. Hook their emitters
to ground, their collectors each to a string of 4 LEDs that are
connected to Vcc, and their bases together. Put a pot between Vcc and
the collector of the remaining one.

Then, you'll be able to control the brighness by changing the
resistance of the pot.

Google "current mirror".

Regards,
Bob Monsen
Err...a (randumb) handful of transistors will have a fair range of Vbe
mismatch which canlead to some LEDs getting twice (or more) the current
than others.

The problem can be reduced somewhat by using monolithic transistor arrays.

On Wed, 27 Aug 2008 22:53:12 -0700, Robert Baer
wrote:

Robert Monsen wrote:

On Tue, 26 Aug 2008 23:30:11 +0930, "mark krawczuk"
wrote:


hi , i have 19 rows of four , 10mm white leds. would it be better to
power them straight from dc or a pwm supply ?
thanks

mark k



Use 20 NPN transistors to make a current mirror. Hook their emitters
to ground, their collectors each to a string of 4 LEDs that are
connected to Vcc, and their bases together. Put a pot between Vcc and
the collector of the remaining one.

Then, you'll be able to control the brighness by changing the
resistance of the pot.

Google "current mirror".

Regards,
Bob Monsen
Err...a (randumb) handful of transistors will have a fair range of
Vbe mismatch which canlead to some LEDs getting twice (or more) the
current than others.
Better put equal value of resistors in series with the emitters..
0.5V drop on the resistors will do wonders.

true. you can also decrease the steering current by using a larger
emitter resistor with the steering transitor.

OTOH, I just built one with a (randumb) handful of ebay 2N3904
transistors, using the original circuit I proposed, and the currents
were all within about 10%. This is with fairly small collector
currents of 3mA. More current may skew the result. It should also be
possible to pick transistors to match. I tested the Hfe of these
transistors, and they varied quite a lot, but tended to cluster, so
just picking clustered transistors would probably make it much more
consistent.

Regards,
Bob Monsen

Robert Monsen wrote:
On Wed, 27 Aug 2008 22:53:12 -0700, Robert Baer
wrote:
(snip)
Better put equal value of resistors in series with the emitters..
0.5V drop on the resistors will do wonders.

true. you can also decrease the steering current by using a larger
emitter resistor with the steering transitor.

OTOH, I just built one with a (randumb) handful of ebay 2N3904
transistors, using the original circuit I proposed, and the currents
were all within about 10%. This is with fairly small collector
currents of 3mA. More current may skew the result. It should also be
possible to pick transistors to match. I tested the Hfe of these
transistors, and they varied quite a lot, but tended to cluster, so
just picking clustered transistors would probably make it much more
consistent.

Dropping as little as 0.1 volt with emitter resistors will
greatly narrow the bell curve.

--
Regards,

John Popelish

Robert Monsen wrote:

On Wed, 27 Aug 2008 22:53:12 -0700, Robert Baer
wrote:


Robert Monsen wrote:


On Tue, 26 Aug 2008 23:30:11 +0930, "mark krawczuk"
wrote:



hi , i have 19 rows of four , 10mm white leds. would it be better to
power them straight from dc or a pwm supply ?
thanks

mark k



Use 20 NPN transistors to make a current mirror. Hook their emitters
to ground, their collectors each to a string of 4 LEDs that are
connected to Vcc, and their bases together. Put a pot between Vcc and
the collector of the remaining one.

Then, you'll be able to control the brighness by changing the
resistance of the pot.

Google "current mirror".

Regards,
Bob Monsen

Err...a (randumb) handful of transistors will have a fair range of
Vbe mismatch which canlead to some LEDs getting twice (or more) the
current than others.
Better put equal value of resistors in series with the emitters..
0.5V drop on the resistors will do wonders.


true. you can also decrease the steering current by using a larger
emitter resistor with the steering transitor.

OTOH, I just built one with a (randumb) handful of ebay 2N3904
transistors, using the original circuit I proposed, and the currents
were all within about 10%. This is with fairly small collector
currents of 3mA. More current may skew the result. It should also be
possible to pick transistors to match. I tested the Hfe of these
transistors, and they varied quite a lot, but tended to cluster, so
just picking clustered transistors would probably make it much more
consistent.

Regards,
Bob Monsen
It is NOT the beta that controls the current matching (oe
mis-matching to be more accurate), but the Vbe matching at a given current.

On Thu, 28 Aug 2008 20:26:27 -0400, John Popelish
wrote:

Robert Monsen wrote:
On Wed, 27 Aug 2008 22:53:12 -0700, Robert Baer
wrote:
(snip)
Better put equal value of resistors in series with the emitters..
0.5V drop on the resistors will do wonders.

true. you can also decrease the steering current by using a larger
emitter resistor with the steering transitor.

OTOH, I just built one with a (randumb) handful of ebay 2N3904
transistors, using the original circuit I proposed, and the currents
were all within about 10%. This is with fairly small collector
currents of 3mA. More current may skew the result. It should also be
possible to pick transistors to match. I tested the Hfe of these
transistors, and they varied quite a lot, but tended to cluster, so
just picking clustered transistors would probably make it much more
consistent.

Dropping as little as 0.1 volt with emitter resistors will
greatly narrow the bell curve.

Two transistors whose VBE's, when at the same current, differ by VOS,
add emitter resistors, R, nominal operating current, I, then...

delta (fractional current mismatch) = VOS/(IR + kT/q)

kT/q for non-semiconductor types ~ 26mV at 27°C ;-)

With VOS = 10mV, IR = 100mV, delta = 8%

...Jim Thompson
--
| James E.Thompson, P.E. | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona 85048 Skype: Contacts Only | |
| Voice480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |

Common Values
Common Purpose
Common Buzzwords
Common Ignorance

Jim Thompson wrote:
John Popelish wrote:

Dropping as little as 0.1 volt with emitter resistors will
greatly narrow the bell curve.

Two transistors whose VBE's, when at the same current, differ by VOS,
add emitter resistors, R, nominal operating current, I, then...

delta (fractional current mismatch) = VOS/(IR + kT/q)

kT/q for non-semiconductor types ~ 26mV at 27°C ;-)

With VOS = 10mV, IR = 100mV, delta = 8%

Thank you. I think an 8% spread is appropriately
significant compared to the thermal differences in a bunch
of hot LEDs. Getting the currents to match to, say, 0.8%,
instead, is pretty pointless, I suspect.

--
Regards,

John Popelish

On Fri, 29 Aug 2008 18:22:42 -0400, John Popelish
wrote:

Jim Thompson wrote:
John Popelish wrote:

Dropping as little as 0.1 volt with emitter resistors will
greatly narrow the bell curve.

Two transistors whose VBE's, when at the same current, differ by VOS,
add emitter resistors, R, nominal operating current, I, then...

delta (fractional current mismatch) = VOS/(IR + kT/q)

kT/q for non-semiconductor types ~ 26mV at 27°C ;-)

With VOS = 10mV, IR = 100mV, delta = 8%

Thank you. I think an 8% spread is appropriately
significant compared to the thermal differences in a bunch
of hot LEDs. Getting the currents to match to, say, 0.8%,
instead, is pretty pointless, I suspect.

Without the resistors you would be close to 40%, pretty nasty!

I generally recommend IR = 0.25V, which takes you down to 3.62% in the
VOS = 10mV case... 1.8% if VOS = 5mV!

You could pre-sort the transistors, but I'd still use ballast
resistors. What we still haven't analyzed is asymmetric heating in
the transistors :-(

...Jim Thompson
--
| James E.Thompson, P.E. | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona 85048 Skype: Contacts Only | |
| Voice480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |

Common Values
Common Purpose
Common Buzzwords
Common Ignorance