Page 1 of 2 1 2 LastLast
Results 1 to 25 of 29
  1. #1
    mtbr member
    Reputation:
    Join Date
    Mar 2013
    Posts
    28

    Bricked my triple XM-L Chinese lamp, need regulator advice

    Dear Forum

    I assume you know what I mean by "triple XM-L Chinese lamp", i.e. the ones on eBay, etc. Long story cut short: I blew the regulator.

    I set about repairing it and discovered that the original regulator was probably de-tuned (presumably in order to constrain the heat generated by the LEDs). I figured I would improve the heatsink and run the LEDs at the full 3A @ 3.35V and get the max output from them.

    I've opted for three LM2596 regulators but I noticed that in the 3A @ 3V range the efficiency is around 75% and heatsinking is required.
    My question is, would a regulator like this one be a better choice as it doesn't seem to need a heatsink so presumably it is more efficient and just a better choice for my application.

    Thanks in advance for an pointers you can give me. Love from UK

  2. #2
    mtbr member
    Reputation:
    Join Date
    May 2005
    Posts
    2,248
    The LM2596 you linked is a voltage regulator. LEDs really need current regulation. I would not recommend it.

    The DX driver would work if you ran the LEDs in series and used a 14.8V battery. DX shows that item as sold out so you may be waiting a while to get one.
    Last edited by Vancbiker; 03-20-2013 at 04:37 PM. Reason: add missing word

  3. #3
    mtbr member
    Reputation:
    Join Date
    Mar 2013
    Posts
    28
    Quote Originally Posted by Vancbiker View Post
    The LM2596 you linked is a voltage regulator. LEDs really current regulation. I would not recommend it.

    The DX driver would work if you ran the LEDs in series and used a 14.8V battery. DX shows that item as sold out so you may be waiting a while to get one.
    That explains why they only quote one voltage output for most of the regulators, I did wonder at that. I have tried the LED with a bench power supply and the current seems to increase in a linear relationship with the voltage which indicates to me that these regulators will work in parallel. This is also confirmed by the XM-L spec. I think I will give it a go and let you guys know how I get on.

  4. #4
    mtbr member
    Reputation:
    Join Date
    Apr 2008
    Posts
    153
    The problem with driving LEDs with constant voltage sources is twofold.
    1) LEDs have very low dynamic resistance. A tiny ~0.15V change can alter the XML current by ~1 amp.

    2) LEDs have negative temperature-to-voltage coefficients, -2.1V/C in the case of the XML. This is the bigger challenge than #1 above. Even if you have a perfect 100% voltage regulation, the shift in the I-V curve due to heat will cause the current to drastically change. Let's say you're driving the LED at a perfectly regulated 3.0V. According to the datasheet it will draw 1000mA at 25C. However, the LED will quickly heats up. 100C junction temps are quite normal, so let's take that as an example. This will shift the I-V curve by 0.21V, effectively placing you at the 3.21V point on the 25C plot, which means a current of 2100mA. This increased current will cause the LED to heat up even more, causing more current, which will cause more heat, and on and on.

    The current regulator you linked to is the way to go.

  5. #5
    mtbr member
    Reputation:
    Join Date
    Mar 2013
    Posts
    28
    Quote Originally Posted by desolder View Post
    The problem with driving LEDs with constant voltage sources is twofold.
    Thank you so much for taking the time to explain that. I wish I had come here sooner. I will try the voltage regulator as I already have the parts, just to satisfy my curiosity, but I'm currently looking at making something like this

  6. #6
    mtbr member
    Reputation:
    Join Date
    Aug 2010
    Posts
    96
    Quote Originally Posted by Carlos Fandango View Post
    Thank you so much for taking the time to explain that. I wish I had come here sooner. I will try the voltage regulator as I already have the parts, just to satisfy my curiosity, but I'm currently looking at making something like this
    A circuit like that one you listed is a linear LED driver. True, efficiency "can" be high, but only at the point where the input voltage and the vf or the LED are basically the same. As the input voltage raises above the vf of the LED, the efficiency of the circuit drops dramatically. The only way to have a true, highly efficient LED driver that will work through a range of input voltages is to have a DC to DC regulator (also called a switching regulator), where you have a Capacitor and Inductor to store energy during operation.

  7. #7
    mtbr member
    Reputation:
    Join Date
    Mar 2013
    Posts
    28
    I tested the voltage regulator and as predicted, the current started to run away almost immediately. I shut it down at 3150mA and went back to the drawing board.

    The linear driver does seem inefficient when used with batteries because of the voltage variation. I could minimise that effect by running my 4 batteries in parallel at 3.7V. My LED power consumption is 3.35V*3A*3=30W
    The nominal circuit power would be 3.7V*3A*3=33W
    By the time I discharge the cells it will be something like 2.7V*3A*3=24W

    Only 3W lost at 3.7V and at the low end of the voltage spectrum it is still putting out 24W. Does this make sense to you?

    The most appealing thing for me is the simplicity of the circuit.
    Last edited by Carlos Fandango; 03-23-2013 at 12:59 PM. Reason: Clarification

  8. #8
    mtbr member
    Reputation:
    Join Date
    Aug 2010
    Posts
    96
    Quote Originally Posted by Carlos Fandango View Post
    I tested the voltage regulator and as predicted, the current started to run away almost immediately. I shut it down at 3150mA and went back to the drawing board.

    The linear driver does seem inefficient when used with batteries because of the voltage variation. I could minimise that effect by running my 4 batteries in parallel at 3.7V. My LED power consumption is 3.35V*3A*3=30W
    The nominal circuit power would be 3.7V*3A*3=33W
    By the time I discharge the cells it will be something like 2.7V*3A*3=24W

    Only 3W lost at 3.7V and at the low end of the voltage spectrum it is still putting out 24W. Does this make sense to you?

    The most appealing thing for me is the simplicity of the circuit.
    Yes, that sounds about right. Putting the cells in parallel will allow higher efficiency as the battery voltage will not be much higher than the vf, but if the cells are not protected, then putting them in parallel could be dangerous - please be careful.

    Another alternative to the linear circuit is to use switching current regulator, like those offered (and used by many here in the forums) from TaskLED. You can always send George (owner/designer) and email to ask for recommendations given your specific project.

  9. #9
    mtbr member
    Reputation:
    Join Date
    Mar 2013
    Posts
    28
    Those drivers look great but for this application it seems overkill. The light itself only cost me 30 and I would need a h6flex driver at $40 +postage and still not get the full 9A I need. I will ask him what he suggests. Thanks for the link.

  10. #10
    mtbr member
    Reputation:
    Join Date
    Dec 2012
    Posts
    16
    With a TaskLed driver you can run the 3 leds in series and use a 4S battery, with plenty of voltage above the LEDs drop. The linear regulators have poor efficiency, they take extra drop linearly. So at full charge the Li-Ion element has 4.2V and the driver will take 1V for each LED (if the number of leds is equal to number of Li-Ion elements), dissipating 3A*1V=3Watt per LED. Go with a switching inductive driver that takes care of the current, has protection cycle-by-cycle and so on. Efficiency is always in the range of 85-95% for such driver. For LED drivers, when it comes to efficiency, reliability and constant output lumens, there is no overkill. Otherwise everybody will use linear regulators for simplicity. Look for LED drivers at LT semiconductor or Zetex and see, they have products that require only an inductor, few capacitors, a shunt resistor to set the current limit per cycle and a schottky diode. They have bucks, boosts, and buck-boosts. So you can design the driver the way you want:
    1) with leds in series (10.2V drop on the leds), using a 2s-3s-4s battery and a boost or buck-boost.
    2) with leds in parallel using a 2-3-4s battery using a buck.

    I almost forgot:
    For a linear driver, when the battery voltage drops bellow the Vf of the combined LEDs configuration, the LEDs will be driven with voltage, not current. They will take the current according to the Vf/If characteristic at the temperature they have (it's an equilibrium somewhere), bellow the one desired, so the lumen output will drop to whatever the Vf/If shows. Don't count on just lower output, but tremendous lower. At 2.5V the leds just start behaving like diodes, only light dimly.

  11. #11
    mtbr member
    Reputation:
    Join Date
    Mar 2013
    Posts
    28
    Quote Originally Posted by uiflorin View Post
    With a TaskLed driver you can run the 3 leds in series and use a 4S battery, with plenty of voltage above the LEDs drop. The linear regulators have poor efficiency, they take extra drop linearly. So at full charge the Li-Ion element has 4.2V and the driver will take 1V for each LED (if the number of leds is equal to number of Li-Ion elements), dissipating 3A*1V=3Watt per LED.

    I almost forgot:
    For a linear driver, when the battery voltage drops bellow the Vf of the combined LEDs configuration, the LEDs will be driven with voltage, not current. They will take the current according to the Vf/If characteristic at the temperature they have (it's an equilibrium somewhere), bellow the one desired, so the lumen output will drop to whatever the Vf/If shows. Don't count on just lower output, but tremendous lower. At 2.5V the leds just start behaving like diodes, only light dimly.
    OK but if I look at the Panasonic battery spec the battery charge does not stay at 4.2V for very long at all. Typically they roll off to a linear region starting at 3.7V almost straight away.

    In this region, between 3V and 3.6V, at a discharge rate of 2.5A the linear regulator is =>90% efficient after which the LED starts to significantly dim from lack of voltage. From the XM-L spec this will only happen beyond ~3V, at which point I am reaching the last part of the battery charge. The key to this strategy is that the supply voltage can be selected in increments of 3.7V and extra capacity added easily. In addition I will be able to see when my battery is low without pushing the cells into the 2.5V protective shutdown mode without warning.

    I'm going to try it as a starter for ten and think about implementing a buck regulator so I have something to compare it against. I had a look on LTs website and there weren't so many converters that would give me 9A and of those that would the ancillary component list is quite extensive, maybe 15 devices or more. Certainly much more involved than this simple linear regulator for not a huge gain in functionality or performance, especially considering I'm not constrained in my choice of battery configuration.

    Can you give me an example of a very simple buck regulator with minimal components that will deliver the full 9A?
    Last edited by Carlos Fandango; 03-26-2013 at 09:59 AM. Reason: clarification

  12. #12
    mtbr member
    Reputation:
    Join Date
    May 2005
    Posts
    2,248
    If it were me, I'd get away from the parallel LED setup because of the need for 9A. Using parallel cells and a linear regulator is fine at lower currents as M8 connectors and 22AWG or 24AWG wires are OK at 3A. Wires and connectors to handle 9A without significant voltage drop get much larger.

  13. #13
    mtbr member
    Reputation:
    Join Date
    Dec 2012
    Posts
    16
    Look on the XM-L Vf/If graph. To drive the LED full currrent the Vf is 3.3V at 25C. Probably at 85C the Vf is 3.15V. The driver needs 150-200mV drop. Except the case the driver needs only few mV drop, I doubt you can drive the leds full power bellow 3.4V per cell. This is the reason a high frequency inductive switching driver works better, assuring always desired current, has a larger input voltage to acomodate the desired output, recycles the energy with inductive elements, reaching always 85-90% efficiency and has protections. Use a buck that can drive 3 leds in series, or a boost or buck-boost, so the leds will be in series and the current only 3A max.

    Why do you think the chinese do not use those linear drivers even for their cheap lights? It doesn't work for high power.

  14. #14
    mtbr member
    Reputation:
    Join Date
    Mar 2013
    Posts
    28
    Quote Originally Posted by uiflorin View Post
    I doubt you can drive the leds full power bellow 3.4V per cell.

    Why do you think the chinese do not use those linear drivers even for their cheap lights? It doesn't work for high power.
    I aim to find out. I don't doubt that the switching regulator is a better overall solution, I'm just groping around for a solution that is practical for me to implement and I thank you for your input, it is much appreciated.

  15. #15
    mtbr member
    Reputation:
    Join Date
    Mar 2013
    Posts
    28
    OK a little update on my experiment. It doesn't look like it's feasible to use a 3.7V voltage source with the MOSFET linear regulator after all. The problem seems to be that there isn't enough gate to source voltage drop (Vgs) to turn the MOSFET on at low voltages. This is probably due to the ~3V voltage drop caused by the LEDs, there's just not enough to go around, particularly as the cells get down to ~2.7V.

    I'm going to reconfigure it for a 7.4V supply but I expect the efficiency to be poor and lots of heat to get rid of.

  16. #16
    mtbr member
    Reputation:
    Join Date
    Apr 2008
    Posts
    153
    Efficiency will be a tad under 50% with a 7.4V supply. (3.45V / 7.4V = 49%). At 9 amps, the MOSFET will be dissipating about 30 watts. That's a lot of heat! In order for that MOSFET to survive, it's going to need a gigantic heatsink. For a passive heatsink, you generally want a minimum of 1 square inch of heatsink per watt, or 30 square inches total. A 6"x6" square of thick aluminum plate should suffice. An old Pentium III class CPU fan+heatsink should also work, since they need to dissipate similar power levels. Old CPU heatsinks are handy to have around for the purpose of experimenting with power semiconductors (transistors or LEDs) which lots of need heatsinking.

    At this point, you might want to consider a simple resistor. The whole point of transistors is to introduce dynamic resistance, in the order of hundreds of thousands of ohms. This resistance creates a stable current source. But with such a high source voltage over the load voltage, you can use a simple resistor and it will introduce enough resistance into the circuit to give you a reasonably stable current. Let say you design for the peak voltage of 8.4V. This gives you a required resistance of 0.55 ohms to limit the LED current to 9 amps [ (8.4V-3.45V)/9A = 0.55ohms ]. When the cells are exhausted at around 3.0V, the current flowing will still be about 5A [ (6.0V - ~3.15V)/0.55ohm = 5.2A ].

    Another issue - how are you going to cool those LEDs? They were under-driven for a reason. The housing is too small to dissipate the heat at the full power an XML is capable of.

  17. #17
    mtbr member
    Reputation:
    Join Date
    Mar 2013
    Posts
    28
    It's funny you should mention the P III CPU heatsink because that is exactly what I've got! The slot style connector arrangement makes for a really handy heatsink for the light unit, I've cut the heatsink to size and sanded the paint off the back of the light.

    I don't think I'm going to bother at all with the MOSFET solution, it won't work at 3.7V and with 7.4V it's tremendously wasteful. The trouble with the resistor method is the question of power dissipation. I^2 x R gives me 81A x 0.55 = 45W.....ouch.

    Still on the linear regulators, I considered the AMC7135, which are self contained linear regs outputting 380mA but able to run in parallel. The trouble is I would need about 20 of them! This problem just seems intractable without resorting to a buck converter. Sod it.

  18. #18
    mtbr member
    Reputation:
    Join Date
    Mar 2013
    Posts
    28
    OK an update on the linear regulator. I took Vancbikers advice and tried series connected LED and the results are much better. By about 9.2V I am getting 2.3A (my bench PSU limit) which is plenty bright and this would mean if I connected three 3.7V cells in series I would have 11.1V - 8.1V supply range.

    This indicates that when the circuit was configured for parallel LEDs and 3.7V the MOSFET wasn't turning on enough (probably due to the Vgs requirement).

    Should I try a different MOSFET with lower gate input resistance? Rds for this device (IRFI5210) is 0.06 ohms and I haven't seen many other devices with lower resistance in my simulation package.

    What about a series parallel arrangement of LEDs? Does it make sense to have 2 in series and one in parallel to give me 6.7V voltage requirement?

    By the way I noticed the phenomenon of thermal runaway while trying to take current readings. The current seems to climb with the temperature. I measured about 80 C at 2.3A. Does this mean that as the current increases I am getting more light out of the LED or is it releasing energy in the form of heat?
    Last edited by Carlos Fandango; 03-30-2013 at 08:00 AM. Reason: spelling

  19. #19
    mtbr member
    Reputation:
    Join Date
    Dec 2012
    Posts
    16
    The LEDs have negative slope for light power vs temperature. The Vf drops with temperature, like for any diode, so the electric power Vf*If drops too, even if the current is steady. So expect some drop in light with temperature, and lifetime of diode too, above 85C.

  20. #20
    mtbr member
    Reputation:
    Join Date
    Apr 2008
    Posts
    153
    You've already got a pretty good MOSFET. A better one won't help you much. At 3A, the minimum power dissipation will be I^2*R = 3^2*0.06 = 0.54W, with a voltage overhead of V = I*R = 3*.06 = 0.18V. The actual power dissipation will depend on the amount of voltage that needs to be dropped. Your full design power (3A) won't happen till your source voltage exceeds the LEDs' voltages (3*3.45V @ 3A @ 25C) plus all the overheads (0.18V for the mosfet, plus ~0.7V for the sense resistor).

    You can't series-parallel 3 LEDs. The voltage needed to light up the two LEDs in series would blow the single LED in parallel. Paralleled LED strings need to have the same number of LEDs, so the voltages will be the same. e.g. you could series-parallel 4 LEDs in a 2-series + 2-parallel matrix, or 9 LEDs in a 3-series + 3 parallel matrix.

  21. #21
    mtbr member
    Reputation:
    Join Date
    May 2005
    Posts
    2,248
    You can do a series parallel configuration with 3 LEDs. 2 leds in parallel then 1 in series with parallel pair. The parallel pair will each see half the drive current and the Vf will be the same as a 2S configuration. This has worked using 2 XPGs and a single XML. XPGs in parallel in series with the XML. Drive current at 2.5A and Vf at ~6.6V.

  22. #22
    mtbr member
    Reputation: Khrystyan27's Avatar
    Join Date
    Jul 2011
    Posts
    141
    When you manage to repair it, can you please post some pictures with this light turned on, and aimed straight up (high beam).

    I want to buy the same light, but from DX. Thanks!

  23. #23
    mtbr member
    Reputation:
    Join Date
    Mar 2013
    Posts
    28
    Quote Originally Posted by Khrystyan27 View Post
    When you manage to repair it, can you please post some pictures with this light turned on, and aimed straight up (high beam).

    I want to buy the same light, but from DX. Thanks!
    Sure but I don't know what you mean by "aimed straight up". Surely horizontal at some object would be better? The result does rather depend on a lot of factors like ambient light and distance of objects.

  24. #24
    mtbr member
    Reputation:
    Join Date
    Mar 2013
    Posts
    28
    OK I've talked to a few people and it seems like parallel LEDs might not be such a good idea because you cannot guarantee the current will be equal in all branches due to the variability in the thermal properties of the LEDs and housing. At least with series LEDs you can guarantee the current is equal and in addition to that there is a better choice of regulators for lower current configurations. Plus smaller wire sizes/reduced resistive losses.

    Looks like it will have to be a 6 cell battery to get me to 11.1V and have enough capacity.

    I just need to put the components into a PCB and I will ready for a road test.

  25. #25
    mtbr member
    Reputation:
    Join Date
    May 2005
    Posts
    2,248
    Quote Originally Posted by Carlos Fandango View Post
    OK I've talked to a few people and it seems like parallel LEDs might not be such a good idea because you cannot guarantee the current will be equal in all branches due to the variability in the thermal properties of the LEDs and housing.
    While in theory parallel LEDs potentially have problems, there have been countless bike lights made using parallel circuits. I have done several with as many as 3 parallel strings and never yet seen a problem.

Page 1 of 2 1 2 LastLast

Similar Threads

  1. Regulator or Resistor?
    By thebronze in forum Lights DIY - Do It Yourself
    Replies: 3
    Last Post: 12-17-2012, 10:42 AM
  2. Mukluk triple crankset and other questions! Advice?
    By clayman29 in forum Fat Bikes
    Replies: 6
    Last Post: 04-14-2012, 06:36 PM
  3. looking for Magicshine style lamp, only lamp
    By chollaball in forum Arizona
    Replies: 12
    Last Post: 02-02-2012, 12:35 PM
  4. Two batteries & one lamp advice
    By thatdrewguy in forum Lights and Night Riding
    Replies: 9
    Last Post: 08-22-2011, 12:49 PM
  5. Advice on triple XML build for the road
    By Magnum9 in forum Lights DIY - Do It Yourself
    Replies: 31
    Last Post: 07-24-2011, 01:32 AM

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •