I have a new client with an interesting product. They are driving large quantities of LEDs (in the hundreds, with future products to be in the thousands) in banks of red, blue, and green (with, of course, different voltage/current requirements for different colors). They want be able to dim each color using a microcontroller
-- PWM is OK as long as the human eye doesn't see any flicker. Maximum efficiency is important: they want it to run cool and to save energy. They are looking at a first run of 100 units to test the market.
How would you go about meeting these requirements? An off-the-shelf constant-current power supply and some uC-controlled switching transistors is the first thing to come to mind for such small unit quantities, assuming that I can find a power supply that is happy having the load go from zero to max and back every few milliseconds. Anyone have any better ideas?
How many watts total? It might make more sense to dump a bit of voltage in resistors and use better quality LEDs. Can you change the number of LEDs per color to match the brightness to a first approximation?
So your controller really only has three outputs? Sounds trivial. Best regards, Spehro Pefhany
--
"it\'s the network..." "The Journey is the reward"
speff@interlog.com Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com
How about three cheap (MeanWell are good) constant-voltage offline switchers? LEDs aren't zero-impedance loads (some people run them constant-voltage!) so aren't all that pickey. You could add a current-sense resistor somewhere in each bank and have a uP measure that and fake out the remote-sense inputs of the supply to servo the current.
We buy a nice 150 watt MeanWell, all approved, emi filtered, PFC switcher for under $40.
Oh, I have a question for you. Could you email me?
Each color should be separate. Each color should be made up of series, parallel strings operating on as high of voltage as reasonable, say 140 Volts. Each series string would connect to this 140 V. rail through a small ballast resistor. The number of LED's in series depends on their operating voltage at the current of interest.
For example, Say the voltage of each LED was 2.2 volts at 20 mA., and, you want 10 Volts of ballast voltage. Then the string voltage would be 130 Volts and there would be 130/2.2 = 59 (60 is ok) diodes in series. The ballast resistor would be 10/.02 = 500 Ohms.
As many parallel strings of 60 diodes each are used to make the total required. Each series string has its own ballast resistor. Each color is made up similarly but has its own number of LED's in series and ballast resistor values for their specific voltage and current
An appropriate high voltage FET or power transistor of sufficient current rating connects each color group to ground when on. These FETS or transistors are driven from microprocessor ports. They are switched on and off in a PWM fashion to control the brightness. Code in the processor controls this at a frequency high enough to avoid any visible flicker, say above 1kHz.
The only regulator necessary is for microprocessor power. The ballast resistors perform the current equalization function, no regulation is required for the LED's. Current feedback could be applied to the micro to control the PWM for current regulation if necessary.
I have no idea, but it might be illuminating to see what's already being done and what technology is being used. For example, most of the signs in Tokyo and Times Square have gone to LED's. Also statium scoreboards:
(note the power supplies)
Here's an article on the New Years Times Square ball, run by LED's:
--
Jeff Liebermann jeffl@cruzio.com
150 Felker St #D http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558
Check the forward voltage vs. junction curve of an LED ( there is a curve at [
formatting link
]), then calculate the voltage drop of the string at startup on the coldest day that the LEDS will experience and the voltage drop at full power on the hottest day that the LEDS will experience. I believe that you will not be happy with the difference in current with a 130V string and a 10V ballast resistor. And even at the temperature where it is at 130V/10V, I would be wasting 7.7% of the power used heating up resistors. This is why I am looking into a switching poer supply with constant current output rather than constant voltage output.
There are some guidelines for driving LEDs here: [
So the current rises until the level comparator triggers, assuring always the same current. I do not know what this looks like visually, as I have not tried. But is has near zero losses, so even the RED LEDs are green :-)
100 watts for the first model, 1000 watts for a future model if the 100W one sells well enough. Some users will be running over a hundred 1KW units, so I do want to make it as energy efficient as possible given unit cost constraints.
It might make more sense to dump a bit of
Certainly. I just started working on this yesterday, and my first step is LED selection, making tradeoffs between cost, efficiency, number of LEDs, etc. I am not limited to using the same number of LEDs per color (it all gets combined into a single diffused light source), so I will make it so that all lamps at max output equal white, and use dimming to get the other colors. I need to look into the customer needs to see whether having pure red or pure blue at a lower output level is acceptable, otherwise I will have to dim down the white so as to match the intensity of the blue and red.
/me thinks that the first time you shut off the transistor, the inductor would spike and blow the transistor up.
Based on a standard SMPS design, how about this?
+140 -------transistor---+ | | +--------+ | | | LEDS | | | IND DIODE - | ^ +--- to comparator | | | RES - 1 ohm to sense current with | | GND ------------+--------+---
When there isn't enough current, the voltage across the sense resistor is too low, and the transistor turns on. This causes the current in the inductor to slowly increase ("charge") until the comparitor senses a high enough voltage and shuts off the transistor. Now current continues to flow through the inductor and LEDS, but this time through the diode. This continues until the inductor "discharges" enough to lower the current, and the cycle starts again.
That looks quite promising. With uC control of the current instead of just PWM, I can even add temerature compensation.
Now nobody flame me if this is a stupid idea, but it occured to me that I could maks a simple switcher where I build up the field in an inductor and then let it discharge into a string of LEDs. Just an idea; I haven't worked out the details.
You are correct. But switching rate will be dependent on comparator hysteresis, inductor value and total load. Look into PWM'ing.
...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: "skypeanalog" | |
| Voice:(480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |
America: Land of the Free, Because of the Brave
To some degree, but keep in mind that typically you won't be able to maintain isothermal conditions within your array, so the inner parts will get hotter, and thus hog more current, and get a bit hotter again.
Yes, it's certainly possible. You'd probably want to heavily LC filter the output current so that it doesn't radiate EMI like it does light. It's not as amenable to inexpensive COTS solutions though.
BTW, the early HP LED-display calculators used a similar method to conserve battery power (an inductor per segment, IIRC, so 8 inductors). Best regards, Spehro Pefhany
--
"it\'s the network..." "The Journey is the reward"
speff@interlog.com Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com
There are a lot of LED drivers available, and new ones being introduced. Most of them assume input voltage higher than needed for the entire string, and then use an inductor in a switching circuit as depicted above to get a specific current at high efficiency, since the only losses are the resistance of the inductor, MOSFET losses, diode losses, and the sense resistor. All of those can be optimized with tradeoffs in cost and size, to get efficiencies approaching 98%. The actual switching rate is several hundred kHz to 1 MHz, but then an additional PWM is introduced at about 100 Hz, which can adjust brightness by a factor of 1000:1, by modulating bursts of the full current output. Zetex has a 60 Volt 1 A driver ZXLD1362 that I was looking into last September, but at that time it was made of Unobtainium.
For my project, I needed to start with a nominal 12 VDC SLA battery source, and drive a string of 7 or 13 high power white LEDs at up to 800 mA, and the entire circuit had to fit inside a tube with 1" ID, and about 2.5" long. The 13 LED cluster is about 35 watts, and this has proven to be a bit of a challenge, especially trying to get high efficiency. I started with a design using a Microchip PIC16F684 with PWM and a standard one-inductor boost circuit, and it worked reasonably well. Then I was urged to use a current mode regulator, so I made a circuit with a TI UC1843a, but I had problems with stability at higher output power (mostly from board layout and insufficient trace width and thickness on my cheap Chinese boards). But also the UC1843a could not drive the gate of larger MOSFETs, and the current sense voltage is too high to allow a small resistor to be used except by playing tricks with the circuit.
Now I am redesigning it with a PIC16HV616, and I'm about ready to send out for more boards. The power of the circuit will be limited to about 40 watts in the volume I have to work with, about 2 cubic inches, but there is provision for larger external inductors and a TO220 MOSFET which could be used for higher power (although track size and spacing would probably limit this to 100 watts or so). The PIC can be programmed to provide variable brightness levels, although probably about a 20:1 factor would be maximum. I have a large output capacitor which produces a smooth output waveform, which is really not necessary for LEDs, and slows down the response time. There are still some details to be determined when I get these boards done, as I have added a source resistor connected to the PIC's comparator to limit the inductor current on a pulse-by-pulse basis.
If you are interested in more details, please contact me directly, and we can work something out. I am designing this for a client, so I would need to clear any technology sharing with him, but I plan to order about 100 boards as a first shot, and they will be less than $2 each. If this might be helpful to you, let me know.
Without reading any further i suggest one current source for each color, current mode DACs to adjust current level for the number of LEDs lit on a per pulse basis, and 1 ms PWM slots, and 10 way temporal multiplexing.
You need a power supply to start with anyhow. The power supply already has a servo loop inside. It's free, so it makes sense to try to use it, rather than adding another full set of power switching and control stuff downstream.
It's hard to beat having all the power stuff done for you for 25 cents a watt.
That's exactly what I had in mind. No sense building one out of seperate parts if an IC exists that does what I want. Thanks! I especially like the "Internal 8.0 to 450V linear regulator"...
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.