Powering LED light displays more efficiently.

[onload]

Dear Engineers,

I am working with LED light displays. There is a series chain of 5 LEDs whose brightness must be controllable. This series chain is across a 24V DC supply. There is a series NPN transistor in series with the LEDs which "burns off" excess volts if the LEDs need to be dimmed. This series NPN has its base fed by an op-amp which has a 0-10V input supply. -The brightness of the LEDs is thus controlled by adjusting the 0-10V supply. -However, the series transistor is "burning off" excess volts inefficiently. -Does anyone know if it is feasible to apply volts to the LEDs using a switch mode convertor?

Their are hundreds of these series LED chains and i think it would be saving electricity to use a controllable switch mode convertor to apply volts to the series bank of LEDs.

Any thoughts greatly sppreciated.

 

[onload]

Hi,

I'm afraid i can't say who i work for..(for one thing, it would identify me and my boss might read my embarassing slip above about V versus I curves for diodes!)

Suffice to say we have high sales volumes.

Anyway, i was interested in the above post which said that varying LED brightness by varying the voltage across them is "bad design"...we do this in all of our products. -To be honest, it looks OK....you just vary the voltage till you get the appearance/brightness that you want....i know brightness does not vary linearly with linear voltage variation...but to my eye it seems smooth and linear enough. We don't think about keeping the current constant through the LEDs either.

Our linearly regulated circuits are in current use and work fine....i was just wondering if they may fail (even) less often if they ran cooler by employing switch mode regulation. (Not the 24V power supply...that IS a switch mode power source for us...but the individual regulators in the hundreds of series LED chains are linear type, as explained above).

In the products, there are series chains of red, blue and green LEDs....-one then varies the overall colour by dimming red, green and blue to the required extent.

I am wondering if the ON/OFF method of dimming may not give the same range of brightness levels? -maybe the linear regulation method that we currently use gives a wider variation in brightness range?

I am also wondering whether having so many ON/OFF dimmers could cause havoc with RFI -specially since the displays are often/usually in public places.

The ON/OFF dimming solution above (with 555 timer) looks cheap and intrigues me......its good to keep cost down especially since new customers are often not really sure if they really want a light display -and higher cost can swing them away from a purchase.....but i think the "555 into a FET" dimming method would be cheap -any guess's as to the relative expense would be greatly appreciated.

 

[IRstuff]

The brightness range is at least as good as the analog approach, but with less overall static power dissipation. In some cases, you can even overdrive the LEDs beyond what you would be allowed to do statically, since there is time for the LED to recover.

As for RF, that's really an EMI control question. The repetition rates aren't necessarily that high, and some amount of sloping of edges would certainly mitigate that further. After all, your SMPS is switching larger currents in the hundred kHz range and there's no complaint, right?

TTFN

 

[LionelHutz]

I'd first like to point out that your present circuit is a simple linear current controller (or variable current regulator) and not a voltage controller.

The simplest way to do this would be to feed a 10V sawtooth waveform of the appropriate frequency into the negative input of your op-amp and keep the 0-10V control signal going into the positive input. Move the emmitter resistor to the collector side. I'd change the transistor to a small FET and add a gate resistor.

In practice, I would use a comparator instead of the opamp which is better for this appliation. Use one common sawtooth waveform for all circuits. Cost = one sawtooth waveform generator circuit.

 

[itsmoked]

If the device is in motion even 60Hz is too slow.

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[richs]

Hiya-

I concur with itsmoked (nothing new here). I can testify
that if the display is moving, 60Hz would cause a noticeable
flicker. A couple of KHz should be fine.

Your comment:

"In the products, there are series chains of red, blue and
green LEDs....-one then varies the overall colour by
dimming red, green and blue to the required extent."

If you have these aforementioned "hundreds of chains"
does each individual chain require seperate control? Are
these chains close together physically? You mentioned:

"In a typical display unit, there would be 160 of these series LED chains with the NPN."

Keying off the "display" unit, I'm thinking along different
lines now.

Definitely would go with a micro here. Wouldn't be passing
no analog signals to 160 different op amps. No siree bob!

My favorite microcontroller is the Microchip PIC, but others
might go with an ATMEL. Both are fine processors.

Now, just off the cuff, and some real loose guessing, I'd
be able to most likely do 16 PWM sticks per 16F877 type
chip, and be able to get away with it pretty well running
the micros at 20MHz. That's about 10 PICs. Most likely
with a little bit of figuring, I could probably run them
off some cheap 20MHz crystals. They usually run about
$0.69 cents US. The PIC microcontrollers are about $5.00
in quantity 100's from mouser (again US). The NPN
transistors about $0.20 or so. Need a series resistor
from each of the PIC outputs to the NPNs, another $0.03
each. A dropping resistor in each one of the sticks,
another $0.03 each.

So (mumble mumble) for 16 "sticks" that's about $9.85
not including the cost of the LEDs.

Yipes, you might say. Parts cost might be higher, but
the overall manufacturing costs would most likely be lower,
as, I guessing, there is some "tuning" that has to be
done with the present design?

Batch to batch brightness levels on the LEDs could, I
believe, cause noticable color shift (there I go again with
the US spelling). These could be mitigated with PWM values
held in the micro's non volitle storage, and adjusted
via computer control at the factory during testing and
once set, would not drift, unlike pots maybe getting
bounced in shipment, causing a change in resistance.

Funny how the two little words, "display" and "colour"
can modify the overall system. In fact, I'd probably
have the micros controlled with digital signalling rather
than 160 analog signals going out on the display.

RFI might indeed be a problem. Something that has to be
looked at. But, I sudder to think of 160 555 timers
in a system and the poor technician who would have to
troubleshoot them.......

Hope this helps.

Cheers,

Rich S.

 

[itsmoked]

And... make each PIC and... each LED addressable and.... hang them all on an RS485 buss.

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[MacGyverS2000]

Sorry kids, Comcast has had my net connection (and incidentally my IP phone connection) FUBAR'ed for the last 4-5 days, so I haven't been able to reply...

The circuit shown is current control, not voltage, as mentioned earlier. This is the main reason why your voltage adjustments do not make large leaps in brightness. Still, it's not linear in nature, though more so than varying the supply voltage. Also, don't confuse the terms "linear" and "smooth"... until you get into digital control (or hit a knee voltage), it should look as smooth as the precision of your controlling pot/voltage.

For PWM, a refresh rate of 100+ Hz should be good enough for motion, but some more leeway is ideal for those with really sharp vision. Since you more than likely don't need precision timing, you could go the really cheap route and pick up some 6-8 pin PICs and use their internal oscillators for individual line control, or go with a slightly larger one to handle a handful of lines. Production quantity would bring the cost down to around $0.50/line, probably quite a bit more given enough volume.

Provide caps on the lines and alternate when a bank of chips is switching to reduce EMI/RFI and you'll get more noise out of the comm line than the LED lines. Brightness levels isn't an issue either... the average person would be hard pressed to notice more than 128 levels, so go with 256 levels of PWM and be done with it. This is easily done with any 8-bit micro like the PIC and makes the math simple.

To waste that much power and/or take the viewpoint that going the switching/PWM route is too expensive is a project with the Ostrich Syndrome.

Dan - Owner

http://www.Hi-TecDesigns.com

 

[IRstuff]

Some good reading for you:

http://www.maxim-ic.com/appnotes.cfm/appnote_number/1883/

TTFN

 

[richs]

Hiya-

Yep. I made it out roughly $.060 per line with the 877 style
PICs. macgyvers2000 is right. Most would be hard pressed
to differentiate 128 levels of brightness.

Even most professional "squinters" have trouble when pressed.
I've done it as a video engineer at a TV station as well as
color photography so I do "squint" at the color balance.
128 should be enough.

I make it out a little bit more than $0.60 cents a line using
the smaller pics however. The bottom end pics cost about
$0.45 each in thousands, but they only have 4 i/o lines.
I'm guessing that it's just about a wash either way.
STILL much cheaper than a 555 with the associated passive
components after assembly. So I guess that's the main
issue.

You could do a multiplex output, but then you are trading
i/o pins for X number of bit latches. Almost a wash. Besides, really trying to do PWM on more than 16 lines
at a time, and I'd get a little bit nervous. Even
running the PICs at the 20MHz top end of the 16FXXX
line. Besides, with those bigger PICs, you get the
larger program flash and all that nice EEPROM to store
the "adjustment" values in.......

Six of one, half dozen of the other.

Hope we haven't lost him by talking about micros. It can
be an "interesting" learning curve.

Cheers,

Rich S.

 

[onload]

Hello,

RichS….reference your input from a few posts back...

Please may I give an example of a single “Lamp” (there would be many lamps in a product).

All LEDs and components would be very phsically close in a Lamp.

Please may I initially hereby define a “Stick” to be a series chain of X LEDs in series with an NPN transistor, with also a series resistor in this “stick”. This “stick” then sits across the 24V supply.

In the current linearly regulated lamp, we have….

7 sticks of five_in _series green LEDs…..the NPNs of each of these sticks is controlled by one op-amp (the input to this op-amp is, as mentioned before, a 0-10V variable supply)

3 sticks of 9_in_series red LEDs…. the NPNs of each of these sticks is controlled by one op-amp (the input to this op-amp is, as mentioned before, a 0-10V variable supply)

3 sticks of 6_in_series blue LEDs…. the NPNs of each of these sticks is controlled by one op-amp (the input to this op-amp is, as mentioned before, a 0-10V variable supply)

…..So, you can see we have three separate 0-10V supplies feeding the three op-amps, and thus the green, red and blue LEDs are dimmed independently.

The three op-amps are on the same chip (there is another op-amp on this chip which is used to switch on the fans.)

Given this arrangement, I wondered if you would suggest the PIC method (Richs) or the method involving a TL494 into a FET (RajeevCell). –Or would you suggest sticking with the currently used linear regulation method for dimming the LEDs?

One disadvantage with our current method is that there is some pot-twiddling required during production test (well spotted Richs!). That is, there is a pot on the input of each of the op-amps (i.e. three pots all told) which has to be twiddled until the total current drawn by the red, green and blue LEDs is respectively X, Y and Z milliamps respectively. (The 0-10V supply is set at 10V for this pot-twiddling).
-That is, this pot twiddling ensures that at maximum input volts (10V) the current through the LEDs is not above the rated level for the LEDs.

-Not only this, but the inverting input of each op-amp is tied very close to 0V by a potential divider across the 24V supply. Sometimes a resistor in this potential divider needs changing if
1. Any colour group of LEDs is ON when a certain minmum voltage is set with the respective 0-10V supply.
2. Any colour group of LEDs is OFF when a voltage just above this minimum voltage is applied.

These last two steps are all about ensuring that we have a known minimum voltage level for each op-amp input, where we can be sure that the respective three colours are just about ON.
Also, we need (for all three op-amps) to be able to have a certain (defined) minimum voltage where we know for sure that red, green and blue LEDs respectively are definetely OFF.

That is, this is all about controlling the LED brightness’s and ON/OFF levels for red, green and blue LEDs respectively….e.g, so we can raise the input voltages from LOW and know for sure that we bring all colours ON at the same/required time for example.

As you can tell, this takes up labour time in production test. Given these requirements, do you still favour the above PWM methods and can you think of ways of cutting down this pot-twiddling and resistor changing time?

Any thoughts greatly appreciated.

 

[itsmoked]

There are color measuring sensors that you would hook to a test set up. You then have the test "jig", (as they are called in the trade), talk to the embedded controller that is running the LED sticks, arrays, etc. The test jig would talk to the embedded controller to tell it what calibration settings to remember in its EEPROM so it's always available during start up.

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[MacGyverS2000]

With PWM, there would be no need at all for twiddling for just-on or just-off thresholds. With PWM, the LEDs are either full on or full off for varying amounts of time. The more I read about your current setup, the more I cringe... wasted power, wasted components, wasted setup time.

Dan - Owner

http://www.Hi-TecDesigns.com

 

[richs]

Hiya-

O.K. I printed out onload's last post and have been reviewing it for a little while.

I think that I have a couple more questions, if you don't
mind. It will change the selection of the PIC that I
suggest.

I am thinking that your phrase:
"(The 0-10V supply is set at 10V for this pot-twiddling).
-That is, this pot twiddling ensures that at maximum input volts (10V) the current through the LEDs is not above the rated level for the LEDs."

I would interpret as:
"the INPUTS to the three channels to each of the op amps
is set to 10Volts, and the current verified to not exceed
the maximum input current to the LEDs".

In other words, in normal operation, there are 3 analog
inputs from the "outside world" (or other parts of the
product), that have a certain transfer function where a
set of input voltages would correspond to a "color" and
an "brightness" of the displays".

Can we characterize the number of different levels of
brightness and color during normal operation?

As mentioned in earlier posts, the "off" state of the
LEDs will not be a problem. Indeed, we can guarantee
an off condition with a PWM driven circuit. Just don't
supply any current to the NPN transistors, and they
stay in cutoff, hense no current through LEDs at all.
Conversely, in the "full on", or maximum brightness of
the LEDs, one would typically find that it is some duty
cycle/period of drive that would, in general, allow the
LEDs to be ON and draw rated current without adjusting
any pots at all. In other words, after design, there
would be no need to adjust pots at all. Typically, with
this large a number of LEDs, they would tend to "average
out" themselves. Even if they don't the current design
will not really allow for individual adjustment of
single LEDs.

That being said here are some caveats that I hope that
you will take to heart and at least think about.

1. I would "drag out the soldering iron" at this point.
We have chatted quite a bit with all this talk and I think
that further talk will not really add much to the
information base.

a. If you have a square wave function generator with a
variable output duty cycle around the company, or have a
friend who has one, or one you could borrow from a local
school. Hook the output of that in series with a
resistor to the base of one of the NPN transistors and a
stick of LEDs (any color) with a series resistor to limit
the current in the collector circuit of the NPN to the
design spec of the LEDs. Vary the frequency and period
of the output of the function generator and watch the
effects. If you feel that you can pick a frequency and
brightness suitable for your needs then you have some
very valuable data for your reference.

b. If you don't have a square wave function generator
then you can check the internet for a 555 or 556 circuit
(check some of the ham radio pages) for a small circuit
in the "audio frequency" range that will replace the
function generator. You are building a "poor man's"
square wave geneator. You can get a rough idea of the
frequency of interest by hooking up an earphone to the
output of this circuit and listen to the tone. I would
strive for a frequency of about 1KHz. That will get
you in the ball park. Continue with (a) above. If you
have an oscilloscope, then you, of course, don't need
the earphone setup.

2. Find an "output transfer curve" or sets of data for your
brightness levels in normal operation. This will give you
a rough idea on the various levels you will need to shoot
for in the final digital design, either the 555, TL494,
PIC, ATMEL, or whatever design. This again, is very
valueable data.

3. Deduce, by whatever means necessary (photo exposure
meter), reference to exisiting documentation, eyeball
guess, the voltage levels for the "input transfer curve".
In other words, during normal operation, what voltage
level curves do you need for each of the 3 "channels" of
operation.

4. Take the "input transfer curve" and "output transfer
curve" and formulate a "transfer curve" by whatever means
necessary.

5. NOW we are ready to start looking at alteratives. For
the purposes of this reply, I will limit my response to the
PIC alternative. Now, I do not work for Microchip or any
of it's affiliates, nor do my thoughts represent anything
other than my own opinions. If you decide to go down this
path, please be prepared for quite a learning curve if you
have not done computer programming. That being said:

a. I would start off with the Microchip Debug Express 2
development programmer/debugger. Part Number: DV164121.
This comes with a little programmer and an evaluation board.
Please note that the processor on the demo board is NOT
the final chip. It is overkill (most likely) for your
needs, but it provides a good starting point for the
final design. You can go cheaper, but the learning curve
will be starting at an earlier point with less "knowns" in
the mix.

b. Go through the lesson plans for this product. There
are a number of "canned" programs that guides one through
the process of embedded computer design with this package.

c. This will give you the "toolchain" to develop your
product.

d. Build your program up in steps. I would first, after
the lesson plans, do an "output transfer program" that would
cycle through the desired range of brightness for the
display.

e. Evaluate your input requirements. For example, if the
input range of analog signals can be broken down into
"chunks" or ranges, then you might have some alternatives.
I would (your mileage may vary) go with a 16F648A chip
and use the voltage comparator as a poor man's a/d
converter. Another candidate might be the 16F506 with an
8 bit a/d converter and 3 input channels. If you need
more than 8 bits (256 levels of brightness) then you can
go with a 10 bit a/d. A simple voltage divider on each
of the 3 input signal lines to divide the input voltage
down from 10V max to Vdd (possibly 5V, a very typical
supply voltage for digital circuitry) to convert the input
signal to a digital value.

d. Build yourself a program to take the input signals and
output them to a display device representing a digital
output condition. This could be to sample one input
channel, and output clock and data to a pair of digital
outputs. Don't laugh, I just got through with this
exercise myself this week. I'm working with a new PIC
chip for me (10F22X), and I output the value of the a/d
out and look at the value from the digital a/d. I clock
out the data and put the clock on a seperate output. I
trigger on the clock being asserted and read the values
off on the scope. This is all part of the "learning curve"
of learning a new chip. Some of the other variables that
you will have to determine is how often you really need
to sample the inputs. Here again, you don't need to
sample all that often (relatively speaking). You can
most likely sample each channel at 20Hz, leaving the input
sample rate at about 60Hz. I don't think that the
customer would notice (or might notice, but would not
object) to a change in brightness/color at a faster rate.
There is no reason that it couldn't be faster, but why
waste resources and it might make the programming of the
chip harder. Here's just an experienced guess on my part.

e. Add additional functionality as required. You mentioned
fans in the product. Here too, the PIC can control the
output of the fans based upon some sort of temperature
transducer input. I would suggest that you get the above
steps working first, then you can add that sort of
functionality.


Whew! Quite a response to your post here. THIS IS QUITE
A PARADIGM SHIFT in the design process. HOWEVER, it might
prove to be quite useful to both you in your career and to
the company that you work for. By shifting into embedded
processor control, you (both you and the company) have
greatly increased your design options, both now and in the
future. With this toolchain under control, you can apply
completely new design processes to your engineering efforts.

Applications of this can be applied to not only the
product itself, but in the manufacturing of products.

For my own example, bearing in mind that I have my
toolchain pretty much in order already. I am working on
a design that requires the monitoring of an Li-ion battery.
I have to use the SMALLEST possible package and only have
to worry about one input and one output. We will have
prototypes shortly after Christmas. These prototypes will
not have the small package (SOT-23s), but will be using
8 pin dips "dead bugged" onto the board. When the SOT-23
chips come in, we'll be ready to run. In production,
the circuit will be less than one dollar each. The PIC in
question is a new one for me, so I had to go through the
drill from "hello world" (blink an led) on up. The code
is written in C. The pickit 2 debug express was bought
to take care of a programming problem with my existing
development hardware (drat it!).

I was made aware of this "fast track" project on Dec. 4th.

Hope this helps.

Cheers,

Rich S.

 

[jimkirk]

People are talking about two very different things here, Switch Mode Power Supplies and Pulse Width Modulation. If you're keeping the series resistor and simply turning the transistor full on and off, you won't be saving much power. LEDs are pretty close to constant voltage devices, so lets consider one of your strings of 5 green LEDs.

The total forward voltage is about 5 * 2.2 volts, or 11 volts. With a 24 volt supply, at 25 mA, the resistor dissipates about .325 watts. At half current it dissipates about .1625 watts.

Now try it with PWM. At 100% duy cycle the resistor dissipates the same .325 watts. At 50% duty cycle, it will dissipate .325 watts half the time, back to .1625 watts. No power savings. Again, this is assuming a constant voltage load, which you pretty much have. There will be some difference in the real world, but you're not going to save much.

And consider your string of 6 blue LEDs. With a forward drop of about 3.6 volts, you have a total of 21.6 volts drop. With a 24 volt supply that's already on the order of 90% efficient, even a SMPS isn't going to be a lot better than that.

And I'd guess that whatever twiddling you need to do adjust the maximum current in your linear circuit (more about that later), you'd still be doing with a PWM circuit. All PWM gets you is turning your driver stage full on (which you seem to need to tweak) or off. The tweak doesn't inherently go away with PWM.

On the other hand, if you design a true SWPM, with an energy storing inductor, then you can indeed get in the neighborhood of 90% efficiency. You also have the benefit of being more immune to supply sags. If your 24 volts drops 20% your blue string will not get to maximum brightness with your current circuit.

You need a SMPS boost mode current source. Start with a standard SMPS voltage supply. Let's say the regulator has a reference voltage of 1 volt. To build a voltage supply you'd drive that with a resistor divider across the output, so for example 40K resistor from the output with a 10K ressitor to ground will force a 5.0V output to keep the feedback input equal to the reference 1 volt. Now replace the 40K resistor with the LED string and the 10K resistor with 40 ohms. Now the voltage will be whatever is necessary to force 25 mA through the 40 ohm resistor, matching the 1.0 volt reference, and that current has to go through the diodes. That 1 volt drop across the resistor will eat into your efficiency budget. Add to that the switching loses, and supply current for the rest of the circuitry. The lower the reference voltage, the better the efficiency, but errors and offsets will also have a greater effect.

Now that you've done all that, you want to vary the output. You have at least two choices.

One, scale the reference voltage down from your 0 to 10 volt control. This is pretty easy, there will be no flicker as the LEDs will always be on, but at variable current, however sometimes SMPS loop stability can vary over such a wide range of operating conditions.

Two, PWM the reference voltage based on the 0 to 10 volt control. Choose a high enough frequency so that the eye can't discern it, but a low enough one so that the SMPS can stabilize. (Many designs just low pass filter the PWM signal, giving a variable DC reference anyway.)

Will there be a difference in efficiency? With the linear control voltage, our 40 ohm resistor will dissiapate 6.25 mW at half brightness. With PWM it will dissipate 25 mW half the time. The linear control will be slightly more efficient in that regard, but it may well be swamped out by the rest of the SMPS circuitry.

Now, why do you need to twiddle your maximum current? I wonder, in your circuit you posted, what drives the inverting input of your opamp? If it's derived from the voltage across the emitter resistor, and you have fairly decent components (1% tolerance and such), why can't you design a known maximum current with 10.00 volts reference to within a few percent? If it isn't derived from that voltage, why not? That's how you'd get a well controlled current trough the LED string.

So, I'd recommend using a SMPS to increase your efficiency, but much of this talk of PWM is slightly off course, as it doesn't seem to be talking about the means by which a SMPS control loop may be varied, but is simply turning your output transistor on and off, while still dissipating the exact same level of power in your series resistor.

If you want a switch mode supply, controlled by a microprocessor, consider Microchip's MCP1630. This is the heart of a SMPS that interfaces easily with a processor. It needs a clock, whose duty cycle can control the maximum on time of the supply, which allows an easy soft start, it might even be all you need to control the brightness, though I don't think it will get you all the way down to zero, I think the minimum on time is 10 nS. Even if it can't do that, it a versatile chip. Just add appropriate inductors, FETs, Schottky diode and capacitors. (No, I don't work for Microchip.)

 

[jimkirk]

Typo, SWPM should be SMPS

 

[richs]

Hiya-

A couple of comments on Jimkirk's response.

Although a boost circuit power supply would be an option,
please consider:

1. That would also consitute a major redesign of the system.
2. It is possible that the +24V power supply might be
sourceing other subsytem elements in the system.
3. Building a switching power supply is fraught with peril.
An older power supply designer mentioned (and I'm
paraphrasing):
"A linear power supply is one step away from an
oscillator, a switcher removes that option".


With the micro controling the system, there is much greater
flexability on the input side of things. Not knowing what
the rest of the subsystem is, removal of the op amps would
remove the tweaking requirements. It might be leveraged
to utilize the fan controller that the original poster
mentioned. Maybe not.

Although I agree with your statements of the power
resistor in the collector of the NPN, I can see where
just lowering the voltage would save some power. By
phasing the duty cycle of the 3 sets of sticks, the power
requirements in Watts don't change, however, the
instantainious current requirement for the supply is
lowered. THAT might be a significant exchange, but there
again, we are changing major portions of the system.

Per usual, with this type of question, we always get down
to the question of not having all the information.

IF we can change the 24V supply and not have it interfere
with other portions of the system, then by all means change
it. While we are at it, why don't we redesign the LED
sticks and go with a series parallel? How about 3 switcher
supplies tuned to the color sticks themselves?

Shoot, we don't even have data sheets on the LEDs that he
uses. We are making wild assumptions on the requirements.

BUT, in summary, the original poster has been left with
several options he can try:

1. 555 timer type circuits. Pulse modulation. Easy to
hook up and play with.
2. A micro. Well, everybody throws a micro into their
designs nowadays. BUT there is a huge learning curve.
A combination of frequency modulation and/or pulse width
modulation.
3. Switched mode supplies. Obviously the most efficient
but the RFI concerns would be a little more interesting.
You do realize that the data sheet for the MCP1630 reads
"High-Speed, Microcontroller-Adaptable Pulse Width
Modulator"? I haven't read the data sheet in detail,
however, I get the idea that it is suggested by
microchip to be used with a micro thrown in there.....
Indeed the evaluation board uses a 10F200....

From the original post:
"However, the series transistor is "burning off" excess volts inefficiently. -Does anyone know if it is feasible to apply volts to the LEDs using a switch mode convertor?"

I think that we have pretty much covered all the bases.
I suggest the original poster try some of these suggestions
and then if he has additional questions, he can direct them
with more detail.

I don't mean to belittle jimkirks efforts and suggestions,
nor any of the other posters. I detest flame wars, and
please excuse me if I even sounded like any baiting was done
on my part. It was purely unintentional and I aplogize
in advance.

And I think that it is well know now that neither jimkirk
nor I work for Microchip....... ;-)

Cheers,

Rich S.

 

[jimkirk]

Hey Rich,

Thanks for pointing out that for onload's needs, he does in fact need a buck converter powered by his 24 volt supply, *as long as the 24 volts doesn't sag too much*. I've been designing LED drivers operating from a much lower voltage, hence my obsession with boost converters. My mistake. No flames detected. (And I won't even mention a SEPIC topology. )

Still, the thrust of my posting was that if you have a true switch mode power supply, with energy storing inductor, you can get an efficiency boost. But if you just do a pulse width modulation, dropping excess voltage across the resistor, you're pretty much no better off in efficiency than a straight linear. In that I think we're in agreement. It appeared to me there was some implication in some of the replies that simply pulse width modulating the 24 volt supply would save power. I also agree with you that SMPS design is a non-trivial undertaking. It can be a specialty in itself, and you really need to evaluate the costs and benefits to go that route.

Cheers.

 

[richs]

Cool!

I too play in the boost range a bit and the efficiency is
real important with low voltage designs. Don't want to
be wasting energy as heat through a resistor.

Trying to get every last erg out of a battery to let that
light shine as long as possible can be a real undertaking.

Well, the original poster has plenty of things to play with
and I hope that he keeps us informed at to his progress.

I've got an insturmentation amp design that is calling to
me and a bit bang serial routine that will not get written
by itself.

Cheers,

Rich S.