maximizing cap energy discharge over LED string

Discussion in 'The Projects Forum' started by auvie, Oct 31, 2013.

  1. auvie

    Thread Starter New Member

    Oct 8, 2013
    25
    0
    Hi guys,

    I'm trying to maximize the usage of energy stored in a capacitor that is to be discharged over a string of LEDs connected in series. Because the cap's output voltage drops rapidly during discharge, as soon as it falls below the combined forward voltage of the LEDs, current stops flowing and the LEDs turn off, even though there's still a significant amount of energy remaining in the capacitor.

    Rather than try and hook up some sort of boost converter, I think I need to take advantage of the fact that any number of LEDs may be on at any one time (all that matters is total overall light output). I would like to implement some sort of cascading LED bypass circuitry. That is, as soon as the cap voltage drops below the necessary forward voltage and current stops flowing, the last LED on the string could be bypassed, reducing the necessary forward voltage and squeezing a little more energy (and hence, light output) out of the capacitor; as the cap voltage continued to drop, this would continue in a cascading fashion, until there was only one LED left in operation.

    I've attached a diagram to try and illustrate this concept. The real circuit would have a lot more LEDs, with a possible need to bypass two or three at a time, however at this point I'm just trying to get help with general ideas on how it could be implemented.

    Looking at it from a high-level, it should be able to be achieved through some sort of switching functionality, i.e. cap voltage drops too far, current stops flowing through LED string, causing bypass1 to switch on, and also enables bypass2 circuit ---> cap voltage continues to drop, current then stops flowing through LED string again, switches on bypass2, and enables bypass3 ---> and so on.

    With that in mind, perhaps it could be implemented with BJTs? I also found an STMicro part, LBP01, that is a bypass switch for when an LED fails to open-circuit -- maybe that could be used, in conjunction with BJTs for the cascading/enable circuitry.

    Any comments/suggestions greatly appreciated!
     
  2. wayneh

    Expert

    Sep 9, 2010
    12,100
    3,034
    The boost converter - joule thief - is a better idea for efficiency.

    But I think you could use an LM3914 to give you a falling bargraph.
     
  3. auvie

    Thread Starter New Member

    Oct 8, 2013
    25
    0
    Thanks wayneh for your reply, both parts of your answer are interesting. I had originally dismissed the idea of a boost converter, because I couldn't find anything off-the-shelf to suit my needs, namely a high voltage input/output capability in a small package . It would need to accommodate a Vin range from 360V to 50V or lower, and Vout to remain around 360V -- it would also need to handle a peak current output somewhere between 1A and 3A, but only for short pulses (duration in the range of 1ms - 10ms), repeating every 1-2 seconds.

    I just did a bit of reading on the joule thief, looks interesting, though I'm not well versed in DC/DC converters in general. Given the above ballpark specs, do you think a joule thief could be designed fairly easily? Could you point me to any relavant tutorials?
     
  4. wayneh

    Expert

    Sep 9, 2010
    12,100
    3,034
    OK, your needs are very different - much higher voltage - than what I expected. None of the simple joule thief circuits you see out there are meant for 360V, or even 50V for that matter.

    So you need a 360V, 2±1amp square wave with a 1% or lower duty cycle?

    And you're powering a string of LEDs? Do you care if the LEDs dim during the flash, as long as they're all lit?

    Why are you powering this with a capacitor - where does its charge come from? For that level of current, the capacitor needs to be huge. Is it?

    Sorry about all the questions but without details it's hard to speculate about what your requirements really are.
     
  5. auvie

    Thread Starter New Member

    Oct 8, 2013
    25
    0
    Doesn't need to be a square wave (i.e. dimming is fine, as per the cap's discharge curve). They all don't even have to be lit by the end of the pulse, like with the bypass idea I mentioned in the first post. What's important is total light output over the pulse duration.

    I'm driving the LEDs with a capacitor because the device's power source will be low voltage DC, 24V max. I will have a transformer-based charging circuit to charge the capacitor, and re-charge it in between flashes. The capacitor will be at least 1000uF, driving several LED strings in parallel for 1-10ms. However, the first step is to just get a single LED string working, similar to my diagram above. In that diagram, I forgot to put in a small series resistor to control max current draw.
     
  6. #12

    Expert

    Nov 30, 2010
    16,261
    6,770
    On the surface, this seems absurd. I can't imagine what is forcing you into this configuration.

    That said, an inductor will produce whatever voltage is needed to get rid of its energy when it is suddenly switched off. I have made a drawing of the basic concept. The critical part is the inductor and the current you run through it before switching off. You need to calculate how much energy is needed to fire the LEDs and have that much energy flowing in the inductor. When you switch off, the inductor will get rid of its energy through the LEDs even if it has to develop a thousand volts and suck the capacitor dry. I have not accounted for enough transistors to get a 4 amp current flow. I have not accounted for how to spread the time out to .01 seconds

    Warning: if the LEDs are not connected, the inductor will splatter the pass transistor like an AK-47 into a stick of butter.
     
    Last edited: Oct 31, 2013
  7. crutschow

    Expert

    Mar 14, 2008
    13,000
    3,229
    An inductor in series with the LEDs, all by itself, will aid in the efficient transfer of energy from the capacitor to the LEDs. Basically you create a resonant circuit which operates for 1/2 cycle to transfer essentially all the capacitor energy to the load. The LC time-constant determines the resonant frequency and thus the time of the half cycle. You may need a reverse connected diode in parallel with the LEDs to carry any current from the second half of the resonant cycle.
     
    #12 likes this.
  8. #12

    Expert

    Nov 30, 2010
    16,261
    6,770
    Do you mean like this?
    (I'm really not all that good with inductors.)
     
    Last edited: Oct 31, 2013
  9. auvie

    Thread Starter New Member

    Oct 8, 2013
    25
    0
    Thanks #12 and crutschow for your input, since first reading your posts I've been reading up on inductors to try and understand your ideas.

    At max power, I would like the device to provide the LED array with 54Ws worth of energy over 5ms. The power would be delivered by 360V @ 30A, spread over 10 parallel strings of LEDs (so 3A per string).

    So, 54Ws @ 30A would require a 0.12H inductor, and along with that, say I used a 360V, 1000uF capacitor, which can store 64.8Ws. That would put the the LC resonant frequency at 14.52Hz, putting the length of a half-cycle at ~34ms.

    Does that mean it will take 34ms for the capacitor to charge the inductor, making that my minimum trigger delay? I was hoping for a trigger delay on the order of µs, but I may be able to work around that requirement.

    Am I on the right track here?
     
  10. #12

    Expert

    Nov 30, 2010
    16,261
    6,770
    Charge Time = 30 amps times .12 Henrys over 360 volts.
    10 milliseconds
    The capacitor doesn't charge the inductor, the volts do.

    Discharge time = a half cycle of {2 Pi [square root of (LC)]}
    I get 34.4 milliseconds

    I don't see how this interferes with flashing the LEDs every 1 second.

    I also don't see why you want to generate 360 volts to charge the inductor when 24 volts will do it in .15 seconds. I also grow weary of you changing the requirements so I can do all the math again. There are other customers here. I'm going to help them for a while.
     
  11. auvie

    Thread Starter New Member

    Oct 8, 2013
    25
    0
    My apologies if it looks like I keep changing the requirements, they're sort of incremental. I don't need a final circuit now so much as I need to discuss concepts and ideas. At this point, since the discussion has gravitated towards inductors, I'm mainly just trying to understand how I might use an inductor to flash these LEDs.

    Let's spec the LED array to require 360V @ 30A, where the LEDs have a total forward voltage of 350V, with a total series resistance of 1/3ohm ([360-350]/30) to limit the current to 30A (remember that it is 10 LED strings in parallel, so it would actually be 10 3.33ohm resistors in parallel).

    I can't use the 24V source to charge the inductor directly because it is lithium-based, and the current requirements are too great. It's why I am using a charging circuit to charge a capacitor over a longer period of time (e.g. 1 second), storing 54Ws of energy.

    So, regardless of how the inductor is charged, say we do charge one to 54Ws @ 30A, then suddenly throw a switch that puts it in series with that LED array. It is here that I don't quite understand the decay characteristics of the inductor. Because it will resist the change in current coming from that abrupt switch, I can understand how it will throw out whatever voltage is necessary to keep that 30A flowing (i.e. 360V). What I don't understand is two-fold:

    1) how can we model the decay curve with this LED-based load? A normal RL decay would see both the voltage and current decay exponentially, but the voltage isn't able to decay below 350V in this case. I guess this means the current would have to decay more sharply, but I'm not exactly sure what the curve would be, given the LEDs' dynamic resistance and what-not. Because the LEDs' efficiency depends on the amount of current running through them, I need to be able to approximate the current decay to estimate the light output over a specified pulse duration.

    2) how can we control the pulse duration? Say we want to cut the power to the LEDs after 1ms, or 5ms. The inductor needs to dump the remaining energy somewhere, and not just throw it away (too inefficient). Would it be possible to use a micro/timer to switch the inductor back to the capacitor circuit, so it dumps the energy back into the capacitor? Once the inductor is empty, rather than let the circuit resonate as per an RL circuit, perhaps the inductor could then be disconnected from the cap, allowing the charging circuit to take over, and top up the capacitor to 54Ws again. Does this sound feasible?
     
  12. ronv

    AAC Fanatic!

    Nov 12, 2008
    3,282
    1,236
    I'm trying to make sure I understand your circuit. Each string runs on 360 volts @ 3 amps. So about 100+ 10 watt leds in series? Then 10 sets in parallel. So about 1000 10 watt leds in total?
     
  13. auvie

    Thread Starter New Member

    Oct 8, 2013
    25
    0
    Hi ronv, yes you are correct.
     
  14. wayneh

    Expert

    Sep 9, 2010
    12,100
    3,034
    I'm still lost on the WHY of this. The circuit flashes 10,800W of LEDs for a few ms every second or so? I'm not really sure stepping up the voltage so far is the best strategy to accomplish that. I need to think about the opposite approach - all LEDs in parallel. That might require 3600A at 3V. Hmmm, sounds like a lot.

    Sorry, just trying to understand what's going on.
     
  15. auvie

    Thread Starter New Member

    Oct 8, 2013
    25
    0
    Hi wayneh, indeed, it comes down to the current. For instance, say we had a 24V, 225mF capacitor(s), which has the equivalent storage capacity of a 360V, 1mF capacitor. It could be charged directly by the 24V source.

    Say we had 7 3V LEDs in series, and 150 strings in parallel. Each string would require a 1ohm series resistor ([24-21]/3) to limit the current to 3A. The total current needed would be 450A, with a total resistance of 6.66mohm (1/150). The problem here is the ESR of the capacitor — we might find one that has an ESR under that, but I’m not sure it would leave enough room to be able to properly limit current in each LED string. However, maybe I am wrong?

    Thinking about this some more, I originally went with the 360V cap because it is familiar too me, and maybe I'm being too stubborn about it. For instance, perhaps I could implement a simpler charging circuit on a 48V capacitor, using a voltage doubler.

    No matter what capacitor voltage and corresponding LED array I go with, at the end of the day, the conceptual design can be the same, in that it seems like an inductor could do a good job of pulsing the LED array. That still leaves the question of decay behaviour, and pulse duration control.
     
  16. wayneh

    Expert

    Sep 9, 2010
    12,100
    3,034
    Is end-to-end efficiency a major concern, or do you just need something to work?

    One problem with long series strings is reliability. When 1 LED fails, they all go down. Or it shorts and the voltage on the others goes up, which could be a bad thing.

    Interesting challenge.
     
  17. auvie

    Thread Starter New Member

    Oct 8, 2013
    25
    0
    efficiency is definitely a concern (running on batteries). That's a good point about reliability, and I'll certainly keep it in mind for the final array design. I would hope, at least, that because they're only being pulsed, they're not getting taxed too much and so would have a higher degree of reliability over the device's lifetime.
     
  18. ronv

    AAC Fanatic!

    Nov 12, 2008
    3,282
    1,236
    This kind of follows your original idea. The TL431 acts as a floating comparator and shorts out 1 led when the voltage drop is large enough over the 3 leds raising the led current back up. It would need to be duplicated for each set of 3 leds. It could be 3 or 4 or 5 leds in the voltage sample. Might take some tweeks. :rolleyes:
     
    auvie likes this.
  19. crutschow

    Expert

    Mar 14, 2008
    13,000
    3,229
    Attached is my concept to generate an inductive pulse through the LEDs, which extracts all the energy from the capacitor.

    R2 was added to stop high frequency ringing in the simulation. It may not be needed in the real circuit.

    For the simulation I charge the capacitor through a 10kΩ resistor, which gives a very large time constant to charge the capacitor. Your charging circuit will determine how long that takes (without the resistor).

    LED Pulse.GIF
    View attachment LED Inductive Pulse.asc
     
    Last edited: Nov 1, 2013
    auvie likes this.
  20. auvie

    Thread Starter New Member

    Oct 8, 2013
    25
    0
    Hey thanks ronv and crutschow, great circuits. It looks like you both are using the same software, LTSpice? I'm thinking I should probably invest in that as well. Can you specify real-world parameters of capacitors and inductors, such as ESR?
     
Loading...