Constant Current Theory Help

Discussion in 'The Projects Forum' started by adikted, Oct 7, 2008.

  1. adikted

    Thread Starter New Member

    Oct 7, 2008
    9
    0
    Hi All,
    I am enlarging my original design (test_circuit.1.jpg) to power another set of 3 LEDs. In doing this on the proto board I saw the redundant parts in making 2 identical circuits running side by side.
    Anyway, though of a test design (test_circuit.2.jpg) but am not sure about running it like this, 2 N-FETs [Q3,Q4] off of 1 NPN[Q1] + 1 current set R [R2] + overload R [R1]; instead of each N-FET having its own set.

    Is the second schematic possible? I want to run up to 8 N-FETs with 1 set of components. Will make my footprint a hell of a lot smaller.

    Thanks any input you might have!!!
     
  2. Audioguru

    New Member

    Dec 20, 2007
    9,411
    896
    Where can you buy matched Mosfets? Nowhere.
     
  3. adikted

    Thread Starter New Member

    Oct 7, 2008
    9
    0
    So your saying because each mosfet is a bit different it will though the whole circuit off?

    Thanks!
     
  4. adikted

    Thread Starter New Member

    Oct 7, 2008
    9
    0
    Anybody with non-sarcastic comments to this project?

    Why can't I power multiple similar gates, no identical, with one lead?

    Why do the MOSFETs have to be 100% identical in the second circuit?
     
  5. beenthere

    Retired Moderator

    Apr 20, 2004
    15,815
    282
    If you simply select a FET that will handle that current, you won't need two. Some are made to handle currents of over 100 amps. Since it will be operating in the ohmic region, you will probably need a heat sink.

    Your proposed circuit would have let most of the current pass through the FET with the best conduction for a given gate voltage. For a pair to run as in your circuit, matching would be necessary, as well as thermal balancing. More trouble than it's worth.
     
  6. neon9

    Member

    Oct 8, 2008
    15
    0
    What is the deal with mosfets trans. if you are not switching why the extra hardware all you need is a simple resistor. you indicate a lot of I for those guys you better have a way to keep them below power dissipation 700ma 3.6v what is W. and the npn current limit is there to do what blow itself up if ever turn on. is the source I limited ? if not puff.
     
    Last edited: Oct 8, 2008
  7. adikted

    Thread Starter New Member

    Oct 7, 2008
    9
    0
    Hi There,
    Thanks for the comments! to answer the questions:
    Beenthere: I can push more amps through it the FET but I do not want to bother with a boost circuit to power more LEDs. 12v input lets me do 3x 3.6v LEDs in series, with a CC on each series. Thank you for the information about balancing the 2 FETs. Do you have any suggestions for powering a lot of 3.9w LEDs that would be more practical than this?

    Neon9: A resistor is a simple solution that is not always reliable and can be grossly inefficient. This circuit will be hooked up to a MCU PWM timer eventually. This constant current circuit creates a feedback loop, trans is used to turn the FET variable on or off depending on excess current. Since running a battery, I will not always have 12v in; will range from 10.4v to 15v, depending on state. The FET burns off the extra energy:
    ~12v(input)-.5v(dropout)-[3.6v*3]=0.7v; thus ~0.7v extra to burn at 0.7a=0.5w heat
    ~15v(input)-.5v(dropout)-[3.6v*3]=3.7v; thus -3.7v @ 0.7a=2.59w heat

    A resistor would not compensate for the over or under drive of input, its output would be variable as well.

    Thanks for the feedback!
     
  8. Wendy

    Moderator

    Mar 24, 2008
    20,766
    2,536
    Why are you focused on MOSFETs? A regular BJT makes a good constant current source by itself. You can use also variations like current mirrors. Parallel paths only need a common reference.

    *********************

    Here is what I'm talking about. I've never built these, they are concepts only.

    [​IMG]

    [​IMG]
     
    Last edited: Oct 8, 2008
  9. neon9

    Member

    Oct 8, 2008
    15
    0
    Then if a resistor is eneficient how about an LM117 as current regulator for thre LEDS I should be 15ma not .7a.
    some people will never work for me simplicity is the spice of life. and to sugest that a passive component is more reliable then an active that will never pass mustard.
     
  10. adikted

    Thread Starter New Member

    Oct 7, 2008
    9
    0
    Bill_Marsden: I never thought about doing something like that. This design is the simplest and most effective, as far as I knew. I have never seen either of those two built before, but will look into it for my application.

    Neon9: I use Luxeon K2s and LEDengin LEDs, they run from .35a to 1.5a. These aren't your radioshack 3mm .015a LEDs, these are the latest and greatest and very expensive; hence the use of more reliable and accurate circuitry. I have done an LM317 Voltage regulation circuit trial, but found this circuit delivers more consistent results with varying input ranges.
     
  11. Wendy

    Moderator

    Mar 24, 2008
    20,766
    2,536
    With that much current you probably don't want to use the current mirror, though it would work if you make sure they are all the same temperature (sinked together).

    The other variation could probably use a Darlington transistor, with the appropriate mods to adjust. It would have the advantage of being temperature compensated somewhat, unlike the first design.

    Another way to go about it is PWM, which would be pretty efficient. You might even be able to mix the second design and a PWM circuit.
     
  12. Wendy

    Moderator

    Mar 24, 2008
    20,766
    2,536
    I've just noticed a problem with your second schematic, both current paths go through the 0.75Ω resistor. The comments about balance MOSFETs not withstanding (and they are valid, you can't share the sense/drive circuit like that), you will end up with ½ the original current in each path.

    [​IMG]
     
  13. adikted

    Thread Starter New Member

    Oct 7, 2008
    9
    0
    Thanks for the ideas Bill. I did some research into BJT circuits for this application and came to the same conclusion that you did. Too much current and variance in needed current based on LED temp.

    Very good point on the second circuit!!! Thank you for pointing that out!!! I was really thinking that something like that would happen in the second design, though not sure where.

    The darlington pair is a good idea but I believe that I will lose too much voltage over the transistors, min is what 1.2v for a darlington. The first design loses only 0.5v.

    Thank you for the ideas!!! Really appreciate your creative input!

    Nick
     
  14. Ron H

    AAC Fanatic!

    Apr 14, 2005
    7,050
    657
    Adikted, don't get your hopes up about using a BJT current mirror. It is not practical using unmatched discrete transistors without the use of emitter ballast resistors, and it is not practical for high-current (i.e., 700mA) outputs, because the reference current (through the resistor) also has to be 700mA.
    In your original circuit, you are wasting ≈100mA through the NPN. Since the gate current is zero, you should be able to use a much larger value of collector resistor (I would try 10k), although you will probably have to change the value of your 0.75Ω current sense resistor.
     
  15. Wendy

    Moderator

    Mar 24, 2008
    20,766
    2,536

    BJTs don't work the way you're thinking, the voltage on the base has nothing to do with the collector voltage. The collector can get very close to what the emitter is voltage wise.

    This design will be slightly over an amp by my reckoning (never built it though), and can be adjusted down from there. You can add limited temperature compensation by thermally linking 2 of the diodes (say CR2 and CR3) to one of the transistors.

    [​IMG]


    I thought about using a LM317 for the voltage reference (it uses few parts), but decided to stick with the diodes for the temperature compensation, which works but isn't very great. I figure the LEDs will soak up 3X3.6V, or 10.8V, and the emitter resistor will eat another .7V. This gives the transistors .5V to play with, which means they'll disappate somewhere around .5W, not too bad. You might not even need temperature compensation with that little wattage disappation.

    Going through the math the max current you could get (if you adjusted the base bias circuit accordingly) is 1.6A
     
    Last edited: Oct 10, 2008
  16. Wendy

    Moderator

    Mar 24, 2008
    20,766
    2,536
    Heck with it, here is the LM317 variation I was thinking about...

    [​IMG]
     
  17. adikted

    Thread Starter New Member

    Oct 7, 2008
    9
    0
    Thank you for your continued input on this project!! I really appreciate the great feedback and ideas.

    Ron_H: I made a mistake in my original diagram, the collector is supposed to be much higher but didn't think that it would cause too much variance in circuit output since I had already put my order in at Digikey.

    Bill: Great diagrams!! I was thinking about BJT the wrong way. That parallel LM317 looks very promising. I had thought of running parallel series off of the regulator circuit, but could not figure out how it would be possible to ensure each series had its own way removing excess energy (each series has been paired with LEDs that are closest in current requirement; since I did not order LEDs from the same bin :( ). There would be little wasted energy with a 12v input but the input will sometimes go as high as 16v. I would also have to ensure the input doesn't go below 12v or things stop working properly. Driving that much current through a 317 will heat it really fast, changing the output, correct? I really need to build that sucker and give it a try. Will hook her up to the scope here in the next couple days let you know how it works!

    I am now working on a circuit for charging large batteries from solar. I have two 12v batteries powering a ~150w application, and one 200w input panel array. Each battery had a A/D sensor for MSP voltage reading, when one battery gets to low, I need the MSP to switch which battery is getting charged and which battery is being used, without losing power. No schematic yet, still figuring out what is needed.

    Got to run, thanks again for all your help with this!!! It has been fun getting good input!

    Nick
     
  18. Wendy

    Moderator

    Mar 24, 2008
    20,766
    2,536
    The LM317 in my last circuit is only a voltage regulator that is lightly loaded, it should not get hot at all. The transistors are another story, they could get quite toasty indeed, so plan on good heatsinking.

    Basically the transistors impose a voltage on the emitter resistors. This sets up a current, which is transferred to the collector. On the collector side it becomes a constant current source.

    As the transistors heat up the BE voltage changes a little, which is the temperature drift I was talking about. It's not major, but it exists. Diodes exhibit the same drift with temperature, which is how the diode circuit can compensate.

    The tools I'm drawing these schematics with can be found in my blogs, which are linked on all of my posts in the upper right hand corner.
     
  19. Wendy

    Moderator

    Mar 24, 2008
    20,766
    2,536
    I put some thought into it, this should be pretty stable if the diodes CR1 and CR2 are thermally linked to one of the transistors. It works by sending a constant current (17ma) to the variable resistor R2. As the resistor varies in value it will change the voltage drop across it, the diodes however will drop a constant voltage, varying only with temperature. I figure each darlington stage will take 0.5ma or less input current. The LM317 will disappate between 0.16 to 0.25 watts given you're stated voltage range and varability, and the transistors can disappate up to 5W each, worst case, best case being less than ½W.

    [​IMG]
     
  20. adikted

    Thread Starter New Member

    Oct 7, 2008
    9
    0
    Sorry for the delay in writing back.

    Bill that is a brilliant circuit! I designed a proto and sent it off, should be about 2 weeks for it to come in. I am guessing that R2 can be adjusted to compensate to the heat of the transistors as well. The proto has both CRs, 317 and transistors sinked together. The output can be watched as they all heat up, and with varying input voltages, to determine what is the best R2 to keep the output consistent. Thank you so much for your help with this project! You are a very kind and intelligent designer, your input has been greatly appreciated.

    I will post a schematic for the charger/source circuit here shortly, got some p and n channel mosfets to play around with to determine the best setup.
     
Loading...