Very cheap 9W RGB power led driver

Discussion in 'The Projects Forum' started by Patrick Meijers, Feb 20, 2018.

  1. Patrick Meijers

    Thread Starter New Member

    Apr 16, 2017
    13
    1
    Hi all,

    I need a design for a led driver based on following requirements:
    - very cheap (because I need to build many of them; say 0.10-0.15 eurocent per led)
    - 5V power supply
    - pwm control option
    - led specs (for green): 700mA, 3.9-4.2V forward voltage

    I was inspired by a simple solution as a.o. described here:
    http://www.tbideas.com/blog/build-an-arduino-shield-to-drive-high-power-rgb-led/
    [​IMG]

    The voltage drop over R2 is ~0.6V, but I can't get the voltage drop over Q1 small enough to have enough for the green led (4.2V).
    The reason I think is that T1 will reduce the gate-to-source voltage (Vgs) of Q1 - that is actually the main reason to have T1 as a feedback mechanism to stabilize the current. But by dropping Vgs, the mosfet leaves its saturation area where a small resistance (Rds(on)) applies, and thus the drain-to-source voltage (Vds) increases.

    I used some other transistor and mosfet types that I could get very cheaply - I did not test with the types as mentioned in the picture - but I feel that the issue is fundamental. Correct me if I am wrong.

    Anyone a clue/idea how to get a current control function with a max voltage drop of ~0.8V.
    Either based on the concept as shown above or some alternative cheap method.

    Thanks in advance, Patrick
     
  2. Patrick Meijers

    Thread Starter New Member

    Apr 16, 2017
    13
    1
    For completeness, I used 2N2222 for T1 and A03406 for Q1, which are extremely cheap.
    A03406 specs: max Vds (V) = 30V, max Id = 3.6A, Rds(on) < 65mΩ, max dissipation: 1.4W.
    Hence with 700mA and Vds target < 0.2V, this should fit I thought.
    But I cannot get Vds small enough. And this affects both the brightness of the led, as well as the power dissipation of mosfet.
    The latter I could fix by selecting a more powerful TO-220 mosfet. But not sure if this will lead to small Vds and thereby being able to utilize the full intensity of the led.
     
  3. ElectricSpidey

    Member

    Dec 2, 2017
    443
    95
    I don't really see the need for T1, if you already have a dropping resistor in series with the LED.

    Be sure to place both the LED and the resistor in the drain side of the FET.

    And be sure to use a logic level MOSFET when using a 5 volt PWM driver.

    Use this calculator to determine the correct resistor after you have measured the vorward voltage drop of the LED.

    http://www.quickar.com/starresistorcalc.htm
     
    Last edited: Feb 20, 2018
  4. RichardO

    Late Member

    May 4, 2013
    2,274
    890
    I found the data sheet for the BUZ71. It is not a logic FET. It can require as much as 4 volts before it even starts to turn on. Your FET needs a gate to source threshold of no more than 2 volts maximum -- less is better.

    I could not find data on the A03406. Give us a data sheet or a link to one so we can look at it.

    The green LED dropping 4.2 volts may be a problem. The voltage across R2 is set by T1, the current sense transistor. This can take close to 0.7 volts when it is limiting the current. If the 5-volt power supply is just a bit low in voltage or the FET drops more than about 0.1 volts you will not get full current.

    Note that T1 must have a high beta at low current for minimum base to emitter voltage when current limiting. The 2N5088 probably works better than the 2N2222 in this respect.

    What is your PWM frequency? If the frequency is too high the 10K ohm resistor will not turn on the FET fast enough because of the large capacitance between gate and source plus the capacitance between Gate and drain. You probably need a gate driver for fast enough switching otherwise the FET will spend too much time in the linear region while turning on and off.
     
  5. ElectricSpidey

    Member

    Dec 2, 2017
    443
    95
    I think that circuit would make more sense if the voltage was higher, but with a sense resistor that has to dissipate almost as much as a simple dropping resistor, and the difficulty getting the balance correct, I don’t really see the advantage.
     
  6. Sensacell

    Moderator

    Jun 19, 2012
    1,977
    612
    This is not a simple problem, when you have so little voltage "left over" the resistor alone solution produces terrible current regulation.
    Variation of Vf causes too much change in the output current.

    A 700 mA active sink with 0.8 volts burden is also not easy- especially within the budget mentioned by the TS.
     
  7. Patrick Meijers

    Thread Starter New Member

    Apr 16, 2017
    13
    1
    https://doc.xdevs.com/doc/_Datasheets/ALPHA_OMEGA/ao3406.pdf
    Vgs(th): 1.9V typical

    I will try a transistor with better beta like the 2N5088.

    PWM frequency to be determined. I just need PWM for brightness control and therefore expect this to be in sub kHz range.Just as high to not see any flickering.

    Thanks!
     
  8. Patrick Meijers

    Thread Starter New Member

    Apr 16, 2017
    13
    1
    That's why I am seeking for help ... ;-)
     
  9. Patrick Meijers

    Thread Starter New Member

    Apr 16, 2017
    13
    1
  10. ebeowulf17

    Well-Known Member

    Aug 12, 2014
    2,441
    437
    The website where you got that circuit from references a 12V supply for the LEDs, which is why they could potentially get away with standard (not logic-level) MOSFETs.

    It seems to me like you're going to be cutting it close in this circuit even with logic level parts. Do you have to stick with a 5V supply? If so, are there other suitable LEDs with a lower forward voltage? I could be wrong, but I think you need more wiggle room in your design.
     
  11. Patrick Meijers

    Thread Starter New Member

    Apr 16, 2017
    13
    1
    The 5V supplies are considerably cheaper than the 12V variants.
    I could give add little bit of extra headroom by reducing the max current for the green led.
    The red and blue leds are less critical. Their forward voltage is lower.
    I can also consider using 12V for the green led only. I have to check the granularity of the supplies. Planned to have 4x 40A 5V supplies ...
     
  12. ebeowulf17

    Well-Known Member

    Aug 12, 2014
    2,441
    437
    If I were you, I'd take a close look at all the variables that could throw this out of whack. No part comes in exactly on its ideal spec every time, and most parts have variations in performance with different temperatures, some with ageing, etc.

    Your green LEDs have a nominal Vf spec of 4.2V? What's the range they can fall in (there's always a range.) If the range of the Vf spec extends as high as 4.4, that leaves no margin at all between your desired drop across the sense resistor and your LED Vf.

    How about the MOSFET Rds specs - do they vary significantly with temperature? If they warm up with 700mA running through them, will that alter the Rds enough to throw you out of regulation?

    I don't know the answers to these questions, because I'm not digging into the datasheets in detail. It may be that you're totally fine as is, once you get your logic level MOSFETs, but I kind of doubt it.

    And just to be clear, if you plan to make large numbers of these things, don't take any small number of physical tests as a sign of success. You could do proof of concept with a batch of LEDs that have 4.1Vf, pass all your tests with flying colors, then order your production batch of LEDs, receive parts with 4.3Vf, and suddenly find that nothing works at all.

    Make sure you've figured out what tolerances will determine the success or failure of this circuit, and design accordingly. If price is your primary concern, you'll probably be getting parts with relatively wide tolerances, so don't design anything that requires perfection.

    I don't mean to sound negative. Like I said, maybe everything is fine as-is. I would just urge you to make sure.
     
  13. Patrick Meijers

    Thread Starter New Member

    Apr 16, 2017
    13
    1
    Thanks for sharing the concerns.
    I already recognized quite a spread from a the leds that I received. The 4.2V looked to be the worse case of a few samples I took, actually being significantly higher than specs of 3.2-3.4V (chinese site ...).
    My fallback is to simply reduce the max current (and thus brightness somewhat), or take the hit of using 12V supply.

    But feel free to suggest a different design that is more robust (but not significantly more expensive ;-)

    (FYI, I am making mini RGB moving heads, 50 pieces in total. So I need 150 driver circuits and that's why I want to keep the price low.)
     
    ebeowulf17 likes this.
  14. RichardO

    Late Member

    May 4, 2013
    2,274
    890
    You can't use the typical spec for the threshold voltage. You could easily get a max spec part which would barely work even under favorable conditions.

    Even at a few hundred Hertz you may need a driver. This will depend some on how much current the PWM source can deliver.
     
    ebeowulf17 likes this.
  15. RichardO

    Late Member

    May 4, 2013
    2,274
    890
    Again, you are using a typical curve. For instance, the curve is for operation with the _junction_ at 25 degC. I can say with 99% confidence that you will _not_ be operating at 25 degrees.

    I don't see anything on the data sheet saying what happens at other temperatures. :(:eek:
     
    ebeowulf17 likes this.
  16. ebeowulf17

    Well-Known Member

    Aug 12, 2014
    2,441
    437
    Sweet! I'm glad to know you're thinking in these terms, checking for worst case performance, and that the 4.2V number is the worst you expect to face. Nice job on that.

    Now you just have to apply that same mindset to the transistor and MOSFET specs, the temperature performance variations, the resistor tolerances, etc.

    I'm sure there are more specs to account for, beyond what I've thought of, but starting with the ones @RichardO brought up is a really good starting point.

    As for different voltages, I wouldn't jump straight to 12V if you don't have to. If I'm understanding your application correctly, you need one set of drivers for each RGB LED - in other words, you wouldn't be able to put LEDs in series with each other, right? If the LED is common anode, you won't have individual access to separate anodes for each LED, making series connection impossible.

    Assuming that's true, the excess voltage will all be dropped across the MOSFET, meaning it'll run a lot hotter. If you can get some voltage in between 5 and 12, high enough to give you good control with a solid margin for error, but low enough to not waste so much heat, that would be ideal.

    Without doing all the research and math, this is just a guess, but I'd imagine 7-9V would get the job done.
     
  17. Patrick Meijers

    Thread Starter New Member

    Apr 16, 2017
    13
    1
    Thanks for the suggestions. I found an adjustable power supply with range 6-9V for only ~10 euro extra compared to 5V supply. Hence I think I will go for that for the green.

    PS, I cannot put the led in series because I want to be able to control each individually (change the color and intensity for each RGB led).
     
  18. ebeowulf17

    Well-Known Member

    Aug 12, 2014
    2,441
    437
    That sounds like a good plan to me. I hope I'm not leading you astray and wasting your money, but it seems like a good plan to me. The extra voltage should make your circuit easier to spec, but keeping below 12V and reducing heat dissipation will hopefully pay for itself in reduced heat sinking requirements (which might have necessitated bigger MOSFETs and/or external heat sinks.)
    That's what I thought. I just wanted to make sure before steering you away from 12V.

    I'm no expert (really can't emphasize that point enough!) but I believe you're on a good track now. You'll still want to think through all the tolerances as discussed earlier, but with more voltage overhead, it should be much easier now to find suitable parts. Cheers!
     
  19. RichardO

    Late Member

    May 4, 2013
    2,274
    890
    Some quick thoughts...

    The voltage from the PWM source must be 5 volts not 3.3 volts minimum. Either that or the T1 threshold voltage has to be 1 volt maximum.

    Plan on a high ambient temperature. It can be 50 degC on a hot day in Vegas. :eek: The inside of your case could easily get much higher than that. Plan on needing a fan.

    Aim for no more than 100 degC _junction_ max for your LED's. Brightness and LED life decrease with elevated temperatures. The LED's and PWM transistors will need lots of heat sinking!!!

    Consider a 3.3 volt power supply for the red LED's. Two reds in series with might work with a 5 volt power supply but then you may have a similar problem to green LED.

    Consider using a PC power supply for 3.3 volts and some 5 volts.

    There is going to be a lot of heat in PWM FET (T1). This is because it is being used as a switched current source. The FET is never hard on and has a voltage across it when it in on.

    Run different color LED's at different currents to better match brightnesses.

    Data sheets for the LED's is a must. Not Chineese "data sheets". If the data sheet does not have information on brightness at high temperatures than choose a different LED.
     
  20. Patrick Meijers

    Thread Starter New Member

    Apr 16, 2017
    13
    1
    @RichardO, thanks for the thinking.

    I have Arduino Mini as PWM source. This should be (close to) 5V.

    The leds are to be used for a project in a church in December. Its typically quite chilly at that time of a year. So I am lucky in that sense.

    I am using a heat sink calculator to test my heat sink design. I will check max junction temp as you indicated.

    I did not account for a heat sink for the FET.
    I saw that power supplies have some adjustable output range. 4.5-5.5V for the red and blue. 6-9V for the green.
    Thus I expect to have ~< 2V @ 700mA drop over the FET. Do you think a heat sink is still needed then?
    Fans are not easy to be added.

    Brightness calibration between the different leds I planned to do via an led specific PWM offset programmed into the Arduinos.

    I already bought a batch of 100 leds for ~ $30 (chinese site, brand: Genesis). Trying these first and if not succesful will follow your advice and switch to e.g. Cree leds with proper data sheets.

    Grt, Patrick
     
Loading...