Very high efficiency DIY drivers- Mtbr.com
Results 1 to 45 of 45
  1. #1

    Reputation:
    Join Date
    Jun 2009
    Posts
    82

    Very high efficiency DIY drivers

    Are there any electronic enthusiasts out there interested in the design and building of very-high-efficiency constant current led drivers?

    It could be argued that the longer run time afforded by a small improvement in driver efficiency could be equally well achieved by a small increase in the capacity of the battery. That is true, but in practice we tend to make a decision in advance about how many cells of a particular size and, once that decision is made, an "x" percent improvement in driver efficiency leads to an "x" percent increase in run time, which is nice to have. A higher efficiency driver also means less thermal loading on the housing, and makes the driver cooler running, more reliable and less needful of heatsinking. In addition, there is just something beautiful about very high efficiency that cannot be defined or logically justified. OK, I'm an efficiency nut.

    I'm interested to hear what others achieve and, FWIW, here is how my new DIY driver is going.

    Firstly, I prefer to use a buck controller with "N" cells in series to run"N" leds in series, as this minimises the difference between battery voltage and led voltage, which in turn improves driver efficiency, so all my discussions refer to that arrangement. In my case, 4 Li-ion cells and 4 leds in series, namely 2MC-Es in 4S2P. In some cases that leads to the controller dropping out of regulation before the batteries are fully discharged, but I attack that 2 ways. Firstly, by using MC-E leds which operate at lower current per led (700mA) compared to XR-E (1000mA), and so have a lower forward voltage - mine are only 3.15V each at full current. Secondly, by building a controller with an unusually low voltage drop. Most Taskled controllers require the total battery voltage to be around 1V higher than the total led voltage, which is likely to result in the controller dropping out of regulation near the end of the discharge cycle. The excellent, brawny Taskled "Hiplex" uses only 0.5V, but I aim for much lower. I shall refer to this Vout-Vin requirement as the voltage ovehead.

    Some history. I began by building a very simple linear regulator, a single op-amp driving an N-chanel mosfet, which requires a voltage overhead of only 0.014 volts at my full current of 1.4A, virtually elimating the problem of regulator dropout. However, the efficiency is (for me) unacceptably low during the first part of the discharge, when the battery voltage is considerable higher than the led voltage. The average discharge voltage of the batteries is around 3.7V per cell, and the measured led voltage is 3.15V, so the average efficiency is only 3.15/3.7 = 85%. Considering the very low cost, small size and simplicity of the regulator, that's not too bad, but I knew from the minute it was built that I would never use it - efficiency too low for my liking.

    Current project is therefore a very-high-efficiency switchmode buck driver, with correspondingly low "on resistance" so the voltage overhead will be only a fraction of a volt. I have the prototype running on the bench, with the following measured results :-

    Total ON resistance: 0.067 ohms (total resistance of mosfet+inductor+sense resistor)

    Voltage overhead: 0.095 Volts (at my max current of 1.4A)

    Peak efficiency: >99% (as Vin approaches Vout)

    Average efficiency over battery discharge cycle: 96.5%


    Measuring very high efficiency with high accuracy is potentially difficult, as both the input power and output power have to be measured very accurately. All measurements are taken with professional quality equipment, with the result that the error in the measured efficiency is about 0.1%

    Naturally, I'll be taking more efficiency measurements over a wide range of operating conditions, for now my priority is to move from a prototype, to a compactly packaged unit that can fit in the back of my dual MC-E "flatpack" light.

    The "average efficiency" was measured under the following conditions.

    Battery voltage: 15.004 V (typical average over discharge)
    Battery current: 1.0829 A
    Input power: 16.248 W

    Led voltage: 12.497 V
    Led current: 1.2541 A
    Led power: 15.672 W

    Efficiency: 96.5% (average over discharge)

    The switching mosfet is presently overkill, capable of switching 60V worth of leds at 25A, that's 1500W of leds!!, but it was the only one vaguely suitable in my storage drawers so in it went... Of course, the inductor would need to be upsized slightly for a 1500W load. Challenge to Troutie - you build a mega light with 1500W of leds, and I'll build you a controller to drive them.

    The inductor is of my own design (I'm a designer of magnetic devices) and, like the mosfet, was based on what I had lying around. Hand wound, and the ferrite core DIY cut, ground, lapped and gapped. It's bigger than needed, but fits OK in the "flatpack" housing, so I may not bother to order some smaller ferrite for a physically smaller inductor.

    If I bought the optimum components rather than using what I had, I might just squeeze average efficiency up to 97% and the circuit could be physically smaller, but IMHO 97% is close to the practical limit. I have designed much higher powered (5kW) switchers before with similar conclusions. You start out with a very ordinary design and easily get 90%. Then it's hard work getting to 95%, and above that you fight like hell to gain every 0.1%.

    But it's all good fun.

  2. #2
    www.hahntronix.com
    Reputation: mhahn@hvc.rr.com's Avatar
    Join Date
    Mar 2007
    Posts
    252
    So what do you use as a driver chip? Or do you roll your drivers out of discrete components.

    I'm kind of taken with the national semi led driver parts. The LM3406, with an internal mosfet, would do the 1.4 amps you need, but the sense voltage is around .2 volts, so that will kill your efficiency.

    I have an older lm3404hv 1 amp design, driving a pair of Cree XREs, running on a friends 44 volt electric bike (stokemonkey type). He puts 20 miles a day on it all year long, and it's still hanging in there after a year and a half.

    The national parts seem pretty hard to kill. I've given up on the cheap Chinese driver chips. I've had several blow up just connecting the battery to the light.

    Mark

  3. #3

    Reputation:
    Join Date
    Jun 2009
    Posts
    82
    Quote Originally Posted by [email protected]
    So what do you use as a driver chip? Or do you roll your drivers out of discrete components.

    I'm kind of taken with the national semi led driver parts. The LM3406, with an internal mosfet, would do the 1.4 amps you need, but the sense voltage is around .2 volts, so that will kill your efficiency.

    I have an older lm3404hv 1 amp design, driving a pair of Cree XREs, running on a friends 44 volt electric bike (stokemonkey type). He puts 20 miles a day on it all year long, and it's still hanging in there after a year and a half.

    The national parts seem pretty hard to kill. I've given up on the cheap Chinese driver chips. I've had several blow up just connecting the battery to the light.

    Mark
    Hi Mark,

    I'm not using a dedicated led driver chip, though I should thoroughly investigate the many led driver chips that are out there, even if I do use an external mosfet for the lowest possible on resistance and voltage drop..

    The prototype uses one of the earliest (1970's !) general purpose PWM controller chips - the TL494, but I'll upgrade to the TL598, which is identical except for a proper totem-pole output designed for driving mosfets.

    I find the later generations of PWM power supply chips (not dedicated led drivers) to be almost useless for anything except the narrow range of power supply applications for which they were designed, which does not include driving leds.

    It's quite easy to roll your own PWM controller from a few op-amps to get around some of the annoying limitations of the dedicated chips, but because I need to keep this thing as small and simple as possible to fit in the housing, I went with the TL494/TL598 because it is a single chip solution, and I have managed to make it do what I want.

    I'll take a look at the National led driver chips. Zetex specialize in led driver chips and may have something I like, or could adapt, for my need of very high efficiency and low inout-to-output voltage.

    Colin

  4. #4
    www.hahntronix.com
    Reputation: mhahn@hvc.rr.com's Avatar
    Join Date
    Mar 2007
    Posts
    252
    Do you have any reference designs for the TL598? Either ones available on-line or ones of your own you'd like to share?

    The only power supply designs and app notes I've been able to find that use the TL598 tend to have way more discrete components that I'll be able to solder by hand on a tiny surface mount PCB.

    Mark

  5. #5
    mtbr member
    Reputation:
    Join Date
    Feb 2009
    Posts
    169
    I pretty much gave up on N lithiums and N LEDs, unless going linear. Most of the current LED drivers chips have too high a sense voltage. Many are .25 volts like the AX2002. There is just not enough headroom for that. There are some that are less. Kind of gets me in the mode of N batterys and N-1 Leds. The Zetex 300 and 310 have only 25mV sense voltages, but the device can only take 9 volts. From what I can tell most of the LED drives are meant to minimize cost, not maximize performance. I have yet to find the optimum device for an application like this. Doesn't mean it's not there though. Some are hard to find. The Asian parts like the AX2002 ,or the PT4115 can't easily be purchased. Getting them on a board for DX is the only source I know of. These driver chips are interesting and they are worth a look. I'm sure that you can get to 95% with some work.

    Don't forget temperature effects. As the temperature goes down the battery voltage drops while the LED forward voltage rises. What works well at room temp, might not do so well at freezing. It cuts into your voltage margins.

    With a slight increase in circuit design you can go with a SEPIC topology(aka Single Ended Primary Inductor Converter). No more problems with input vs output voltage that way.

    The simplicity or using the 7135 linear regulators is attractive to me. I'll take the 85 to 90% efficiency. It's just a tradeoff like just about every engineering decision. There is one big advantage of going with a single cell, or several in parallel. Many lithium battery issues go away when you are charging a single cell. No problems with cell balance and a lot less of a chance of a battery failure. A vent with fire failure is something to behold. I'll take less efficiency to make it a bit safer. Don't want to burn the house down when charging.

    I really want other added features. Something that turns the light down to 25% when the battery gets to 90% discharge is something I might want. Same for high temp protection.

  6. #6
    mtbr member
    Reputation:
    Join Date
    Oct 2004
    Posts
    3,061
    Microcontrollers like the Atmel ATtiny85 and the Microchip PIC12F683 work really well for generating a PWM signal. They each have an A/D converter which can be used to monitor a sense resistor. They are both 8-pin devices and are available in a DIP package (which makes it handy to experiment with them using a solderless breadboard).

    It's kind of a toss-up which to use for an application like this. The Atmel device is a faster processor, but the Microchip device has a significantly faster A/D converter. I didn't find one device to be appreciably harder or easier to program than the other.

  7. #7
    mtbr member
    Reputation:
    Join Date
    Apr 2006
    Posts
    596
    I pretty much agree with VR; except I feel that you have to be proactice with li batteries. Realisic voltage margins are very important. How you implement this, your concern. A/D's pick any 2; cheap, fast, small. lol James

  8. #8

    Reputation:
    Join Date
    Jun 2009
    Posts
    82
    Do you have any reference designs for the TL598? Either ones available on-line or ones of your own you'd like to share?

    The only power supply designs and app notes I've been able to find that use the TL598 tend to have way more discrete components that I'll be able to solder by hand on a tiny surface mount PCB.
    After trying the TL598, I have rejected it and gone back to the TL494, because the supply current for the TL598 turned out to be a fair bit higher than the TL494, which killed my efficiency, especially at low output currents where any fixed ovehead current becomes significant.

    To be precise, supply current my TL494's is measured as around 7mA, whereas it is 20mA for the TL598. Both these figures are within spec, so I may just have been unlucky with my TL598's, or it may also be that the slightly faster TL598 with faster totem-pole outputs is simply a bit thirstier on supply current. Either way, from where I stand, the TL598 is chewing up an extra 13mA compared to the TL494, and that significants degrades my efficiency.

    By how much? Depends on the output current. I'm running a 15V (approx) battery pack, so that 13mA represents a power wastage of 15x0.013 = 0.2W. If the controller is pushing out my maximum of 1.4A, at around 12.6V, then output power is is 17.6W, and I'm therfore losing 0.2/17.6 = 1.1%. Given how hard I've worked to get up to 97%, I'm not going to throw away 1.1%!! That may be academic, but it looks much worse at lesser output of say 100mA. in this case, output power is around 12V x 0.1A = 1.2W. Now that 0.2W becomes signifiicant, and represents a wastage of 0.2/1.2 = 16.6%. No way will I put up with that.

    So, I've gone back to the TL494, and buffer the output with a mosfet driver chip.


    Nonetheless, with the TL598 I did achieve 95.7% efficiency under my standard test conditions (see original post) of Vin=15, Vo=12.5, and that's still pretty good so you may still be interested in using TL598, which provides a single chip solution without additional buffering to drive the mosfet gate.

    Both the TL494 and TL598 are very simple to use in this application, with a minimum of external components. If you are still interested, I'll provide more details. There's an R and C to set the oscillator frequency. There's a current sense resistor, which can be very small (I use 25 mohm) as common mode range of the inbuilt error amp extends down to ground, and that's one of the reasons I like these chips. There's an external P-channel mosfet, inductor and schottky freewheel diode of course, and the internal 5V reference is used with a potentiometer to produce a variable reference voltage for dimming control. Can't get much simpler, and with appropriate choice of inductor, mosfet switch and schottky you can easily design for as many output amps as you want.

    These chips were originally designed for (now obsolete) push-pull switchmode designs, where the duty cycle for each transistor must not exceed around 45%, or else there is a danger of "shoot through", with both transistors on, resulting in fireworks. For our application the output control is set to single ended, but it is absolutely infuriating that the maximum duty cycle is still limited, now to around 90%. It is infuriating because it is purely historical and relevant only to push-pull designs, and just a pain-in-arse for our application. I get around this by wiring the output stage "active low" resulting in a duty cycle range from 10% to 100%, rather than 0% to 90%. However, it does mean that the second error amp can in effect no longer be used because of the internal ORing of the op-amp outputs, which is a great shame, as that 2nd amp would have been absolutely ideal for regulating the maximum housing temperature, for example. Typical silly chip designers ..... That's why often I prefer to roll my own PWM controllers from scratch, but in this case there is simply not enough space. I will probably add over temp protection simply by other means, but that's another story.

    I have now purchased an optimum mosfet and schottky rather than using what happened to be in my storage drawers, and as a result have tweaked the efficiency up to 97.1%, which represents the average efficiency that I will obtain over a full discharge cycle of the battery pack. I will use 4 Li-ion cells in series with 4 leds in series, and am conservatively assuming that the average battery voltage will be 15V, and that the average led forward voltage will be 12.5V. In practice, those two voltages will probably be somewhat closer together resulting in a slightly higher efficiency, if anything. Here are the exact results:

    Battery voltage: 15.006 V (typical average over discharge)
    Battery current: 1.0765 A
    Input power: 16.154 W

    Led voltage: 12.504 V
    Led current: 1.2550 A
    Led power: 15.692 W

    Efficiency: 97.1% (+-0.05%) (average over discharge)


    In addition, the total controller "ON" resistance has now been reduced to 0.06 ohms, which represents a negligible voltage drop of 0.084V (or 0.021V per cell, if you prefer) at my full current of 1.4 amps. Thus the controller will not contribute to a lack of regulation near the end of the discharge cycle.

    Minimum led current is around 5mA, which is less than ever required. In practice, I'll probably scale the pot to control from 50mA minimum, up to 1400mA maximum, for a very useful 28:1 control range.

    Because I'm totally insane, I plan to build this controller using full-size DIL chips, on a 26x21mm piece of veroboard (!!) and with leaded resistors and caps, so it will fit inside my dual MC-E "flatpack" housing, and still with enough space to fit a pot and two toggle switches. Hmmm, we'll see. I know it won't be easy though.

  9. #9
    www.hahntronix.com
    Reputation: mhahn@hvc.rr.com's Avatar
    Join Date
    Mar 2007
    Posts
    252
    I'd be very interested in seeing your TL598 schematic if you feel like sharing. I've found a few examples of how to make a buck converter using the TL494, but they typically don't have much of a circuit description, and include a few mystery parts that I'm not sure why they are there. I've been reading about the TL598, but there's a lot of stuff to wade thru, and no good buck converter examples. I'm starting to get my head around the concepts, but have never used a pwm controller before. Most of the single chip solutions I've used have been based on hysteretic control of the LED current (the timing diagrams are MUCH easier to follow ).

    I'm sure there are many folks who would like to find a thru-hole design that can handle high currents. And anything over 90% efficiency is pretty good compared to most of the cheap Chinese drivers we've been playing with. You're lucky to get 80% out of those. The biggest power waste for the Chinese drivers is usually the current sense circuitry. They typically run at a 200 to 300 mV level. Which at 2.8 amps costs you over half a watt. I'm pretty impressed that you managed to get your current sense voltage down around 35 mV. You must be using the error amp as an op-amp to amplify your current sense signal. What kind of gain are you using?

    Your vector board dimensions sound like a challenge to build. You might want to consider surface mount. Surface mount is hard to prototype with, unless you are willing to design your own PCBs. Soldering surface mount parts by hand using a soldering iron is not much fun, even with a PCB, and good magnifing glasses.

    I've had pretty good success using a technique I learned from SparkFun electronics (www.sparkfun.com, look in the tutorials section). You hand cut a stencil of your board using a very sharp exacto blade. Then spread paste over the board using the stencil and an old credit card. Hand place your components. Finally you gently place the board in the center of a cheap $20 skillet, and set the skillet on high. After 2 to 3 minutes, you'll see the solder start to flow. Turn the skillet off and wait a few minutes before picking up your board. I've done temperature profiles of a couple runs I did this way and they look pretty close to the soldering temperature profiles manufacturers recommend. If you use a lead based paste, you typically never get the parts too hot, since most are designed these days to survive lead-free solder which melts at a higher temperature.

    The above technique is best for boards with surface mount components on one side only. I do two sided boards by putting all the small, hard to solder, stuff on one side and all the larger parts (inductor, sense resistor, freewheel diode, etc.) on the other. Then I solder that side by hand.

    I've managed to build boards that have a 1.5 amp output, 7 to 40 volt input, buck converter and an Atmel 8 pin microcontroller on a 2 sided board that measures about 1.25 by .75 inches. Not quite as small as some of the Chinese flashlight drivers, but my boards have low voltage detection, thermal management, and a user interface that I can live with. Over the range you are designing your board for, mine would probably be a bit less than 95% efficient, so probably not up to your standards .

    Mark

  10. #10
    mtbr member
    Reputation:
    Join Date
    Apr 2006
    Posts
    596
    Cool idea; lol on the homemade reflow. I've always used the magnifiers (cheaters) and my Weller iron to blunder through. I usually breadboard with throughboard parts and then move to SMD's. Now-a-days though, lots of parts are only SMD. Very sad! Parts get smaller and my vision gets weaker. I personally feel that vector board is the hard way. I use the Dremal and cut my own traces on PC board stock. YMMV James

  11. #11
    mtbr member
    Reputation:
    Join Date
    Oct 2004
    Posts
    3,061
    Quote Originally Posted by OldMTBfreak
    I personally feel that vector board is the hard way. I use the Dremal and cut my own traces on PC board stock.
    I like the One Pas, Inc. Super Boards. Their other (more conventional) boards work well for certain applications too. The Super Boards can handle some SMT components.

  12. #12
    www.hahntronix.com
    Reputation: mhahn@hvc.rr.com's Avatar
    Join Date
    Mar 2007
    Posts
    252
    Never heard of using a dremel create surface mount PCBs. You could do a thermal transfer with an iron and laserjet printout of your circuit on the right kind of paper. Then just dremel away the copper you don't want. For some of the smaller parts I'll probably need to give up my morning coffee.

    It might be a good idea to have a strong fan blowing behind when you do this. I've found out the hard way that inhaling copper dust or fumes can be bad for your lungs. But with the right kind of dremel bit this probably isn't a big worry.

    Mark

  13. #13

    Reputation:
    Join Date
    Jun 2009
    Posts
    82
    Thanks to everyone for info and suggestions. I can see the appeal of using a microcontroller.

    I have virtually finished the final product (as opposed to solderless breadboard prototype) and making it small enough with 2 DIL ICs, one 16-pin and one 8-pin, was a nightmare - NOT recommended - though there would be no problem at all if space is not a critical issue. With surfacemount components it would be easy though,after you made the board, that is. The TL598 would have made life easier as well, as no additional gate driver chip would be required.

    I squeezed a bit more efficiency by using lower ESR caps on the input and output. Here are the final numbers for the finished product, at a number of different operating conditions.

    Vin=15V, Vo=12.5V, Iout=1.25A, Eff=97.40% (my standard test conditions)

    Vin=15V, Vo=12.5V, Iout=1.52A, Eff=97.44% (eff slightly better at higher current O/P)

    Vin=15V,Vo=12.5V, Iout=0.57A, Eff=96.5% (eff still good at lower current)

    Vin=13V, Vo=12.5V, Iout=1.25A, Eff=98.4% (eg battery near end of discharge cycle)


    All of the above measurements relate to having 4 cells and 4 leds in series. Here is what you get with 4 cells and 3 leds in series, for those who worry about dropping out of regulation near end of discharge cycle, though I don't think that will be a problem with the negligible voltage drop of this controller.

    Vin=15V, Vo=10.2V, Iout=1.9A, Eff=96.2%

    Still pretty darned good, though personally I'll stick with the higher efficiency of 4 cells and 4 leds. I'm now satisfied with performance of this driver.

    One cute thing about such high efficiencies is that heatsinking of the driver is completely unecessary, even at power outputs high enough to drive 4 MC-Es at full current. At full current I was able to place my finger on every component on the board, and with difficulty was just be able to detect a warmth above room temperature.

    OK,OK, I'll stop boasting now ....

    Mark, when I get a chance I'll scratch a basic TL598 circuit on a piece of paper, take a photo, and post in this thread.

    Colin

  14. #14

    Reputation:
    Join Date
    Jun 2009
    Posts
    82
    I'm sure there are many folks who would like to find a thru-hole design that can handle high currents. And anything over 90% efficiency is pretty good compared to most of the cheap Chinese drivers we've been playing with. You're lucky to get 80% out of those. The biggest power waste for the Chinese drivers is usually the current sense circuitry. They typically run at a 200 to 300 mV level. Which at 2.8 amps costs you over half a watt. I'm pretty impressed that you managed to get your current sense voltage down around 35 mV. You must be using the error amp as an op-amp to amplify your current sense signal. What kind of gain are you using?
    Keep in mind, that my 35mVcurrent sense voltage is at full current. At the lowest controlled current, which is more than 20 times less, the sense voltgae is only a couple of mV and still works just fine - try that with someof the newer PWM chips! Actually I've dropped the sense resistor even lower from 25 mohms down to 20 mohms - no point in having more than the few mV that I need .... My DIY sense resistor consists of an 80mm length of 0.3mm diameter enamelled copper wire, wrapped around a 1/2W resistor - cheap and easy to adjust. Purists may turn up their noses at the relatively large change in resistance with temperature (+0.39%/DegC), but it could be argued that this is an advantage, having the effect of decreaing the led current with increasing temperature. Indeed, I'll set the maximum current (pot controlled) slightly above max led rating, in the knowledge that this will be automatically reduced if and when the housing (and therefore leds) get hot.

    The error amp does not amplify the sense voltage as such, it just compares it with an equally small reference voltage. The "error signal" thus derived is then amplified by almost the full open-loop gain of the op-amp, with the result that the error is brought to essentially zero, so the quality of regulation is very high. Changes in battery or led voltage have essentially no effect on the controlled current.

  15. #15
    mtbr member
    Reputation:
    Join Date
    Apr 2006
    Posts
    596
    Sounds nice Colin. Plz post a schematic. Kevin, those proto boards look nice. I may try a few. I really like to layout with a Sharpe, then cut the traces on the pc board with a dremal. I use the narrow saw for a cutter. SMD's require slightly more care. lol James

  16. #16
    mtbr member
    Reputation:
    Join Date
    Feb 2009
    Posts
    169
    I think that your sense resistor is if anything better than using a "normal" resistor. Your's will likely be better cooled and will not get as hot as a surface mount resistor. All resistors change with temperature and those little ones can get hot.

    There are many sources of drifts and changes. You might find that the input offset drift in your error amp might be one of the larger sources of changes.

    One reason that the cheapo drivers are lower efficiency are the components. Minimal sized inductor have high resistance and the switches themselves are also higher resistance. I imagine that you are using a somewhat better inductor and a big low Rds on mosfet.

    I'm still thinking of using this device. It's a specific LED driver. The big downside that I see is low gate drive at only 3.5ma. That can switch smaller mosftes, but a big device will need a stronger driver to switch quickly.

  17. #17

    Reputation:
    Join Date
    Jun 2009
    Posts
    82
    There are many sources of drifts and changes. You might find that the input offset drift in your error amp might be one of the larger sources of changes.
    Offest voltage temperature coefficient is not specified for the TL494 but, based on similar op-amps such as the CA3140, is likely to be around 8 uV/DegC, which is so small as to cause no problems.


    One reason that the cheapo drivers are lower efficiency are the components. Minimal sized inductor have high resistance and the switches themselves are also higher resistance. I imagine that you are using a somewhat better inductor and a big low Rds on mosfet.
    Absolutely true. My inductor is 23 mohms, and my external mosfet 13 mohms. I should also mention that my DIY inductor is HUGE compared to what commercial designs use, but I don't care as it still fits inside my housing, and my"flatpack" housing is as small as it it possible to make a dual MCE light and still have adequate cooling.


    I'm still thinking of using this device. It's a specific LED driver. The big downside that I see is low gate drive at only 3.5ma. That can switch smaller mosftes, but a big device will need a stronger driver to switch quickly.
    It will be interesting to see how it goes. I personally can't see much to recommend this IC. As you say, not enough gate drive to do anything useful. Also, the maximum recommended 8V supply severely limits your options. Finally, it is not designed for dimming - surely you would wish to have adjustable led current?

  18. #18
    mtbr member
    Reputation: sdnative's Avatar
    Join Date
    Aug 2008
    Posts
    336
    Good, thread. Great inputs! Bumping it back up to the top again for forum vis. Will try to put together my own work and findings when I get the time. The availability and diversity of simple constant current IC drivers has put drivers into the DIY realm that's for sure!

  19. #19
    mtbr member
    Reputation:
    Join Date
    Feb 2009
    Posts
    169
    For max efficiency eliminating the diode can help a lot. All switching power supplies actually have two switches. Just about every LED supply only has one active component and uses a diode for the other. Even a Schottky diode has about a half a volt of loss. Replacing it with a transistor can reduce that loss to millivolt levels. The penalty id pretty large if you are only using a single LED. The efficiency loss due to the diode is pretty much (V_diode/V_led) * (1/duty_cycle). If the voltage supply is large that can be as much as .5/3.6 or 13%. If you are trying to run a P7 with a two series Li battery it's about 7% loss just for that diode.

  20. #20

    Reputation:
    Join Date
    Jun 2009
    Posts
    82
    For max efficiency eliminating the diode can help a lot. All switching power supplies actually have two switches. Just about every LED supply only has one active component and uses a diode for the other. Even a Schottky diode has about a half a volt of loss. Replacing it with a transistor can reduce that loss to millivolt levels. The penalty id pretty large if you are only using a single LED. The efficiency loss due to the diode is pretty much (V_diode/V_led) * (1/duty_cycle). If the voltage supply is large that can be as much as .5/3.6 or 13%. If you are trying to run a P7 with a two series Li battery it's about 7% loss just for that diode.
    Yes, agree that diode loss is a heinous crime.

    I believe your efficiency formula is slightly wrong,and should be :-

    For buck regulator, efficiency loss due to diode is :-
    (V_diode/V_led) *(1-duty_cycle)

    The better Schottky diodes have around 0.35V loss. The trick here is to choose the diode voltage rating to be no more than necessary.

    However, be warned NOT to scan through the data for every diode and choose the one having the lowest forward voltage, which can be as low as 0.3V. The ones with the very lowest voltage drops also have the highest leakage currents. Problem is, the leakage current increases with temperature, and and our housings are an unfortunate example of where the operating temperature can be high if the bike is stationary. This can easily lead to schottky thermal runaway, where the leakage current heats the diode, which increases the leakage still further, and so on until destruction. Some of the tiny surface mount diodes are quite vulnerable in this respect. So, choose a diode with just enough voltage rating, and look carefully at the leakage current specs at elevated temperature.

    Of course, and as vroom points out, efficiency is also optimised by running your leds in series, for a higher total led voltage. I chose a 4-led-in-series (14.8V) system for that reason. And for the efficiency purists, running a single P7 will always be crud because the voltage is too low.

  21. #21
    Frt Range, CO
    Reputation: pursuiter's Avatar
    Join Date
    May 2008
    Posts
    2,576
    Quote Originally Posted by cdcdcd
    ...running a single P7 will always be crud because the voltage is too low.
    Time for synchronous rectification?

  22. #22
    mtbr member
    Reputation: troutie-mtb's Avatar
    Join Date
    Sep 2007
    Posts
    3,062

    any chance of a laymans terms diy driver .

    I Am sure this thread means something to a few folks gifted in the black art of electronics and who understand the flux capasitator ION drive stuff .

    it would be a good thing if you clever people could do a simple driver that us colour blind non electronic types could have a go at building like a circuit diagram for dummys and a parts list that did not assume any prior knowledge at all
    I am sure I dont just speak for myself here and there must be quite a few diyers who would like a go at their own driver too but dont know a resister from a bottle of milk.

    Sdnative quoted this "availability and diversity of simple constant current IC drivers has put drivers into the DIY realm that's for sure!"

    Well come on guys please show us how , we want to learn this witchcraft .

  23. #23
    mtbr member
    Reputation: znomit's Avatar
    Join Date
    Dec 2007
    Posts
    1,980
    Quote Originally Posted by troutie-mtb
    Well come on guys please show us how , we want to learn this witchcraft .
    Heres a nice one to start with troutie. I use these with a 12v wall wart on test circuits when away from the bench power supply.
    http://users.telenet.be/davshomepage/current-source.htm

  24. #24
    mtbr member
    Reputation:
    Join Date
    Oct 2008
    Posts
    1,322
    +1 to Troutie's request. I like knowing how to build lights and all that.
    A better understanding of drivers would be useful.
    So lets say I want to get the new XPE colors maybe a red and blue set. Put those into a housing and have the blue and red flash alternatively about every 2 to 3 seconds.
    Off road use only of course just to torment my friends.............
    Last edited by odtexas; 08-22-2009 at 11:49 AM.

  25. #25
    mtbr member
    Reputation:
    Join Date
    Feb 2009
    Posts
    169
    Troutie, you nicked me. I've got a degree in electrical engineering.

    One issue is the physical nature of modern components. They are so tiny that it is hard to deal with them. The small size because it makes for tiny boards. For example many of the device that I am considering using are in SOT-23 packages. They are 1.9mm x 2.9mm. The alternative is to use older devices like cdcdcd employed in his design. They are a lot larger than modern integrated circuits, but they also need a lot of external components to work. That makes them a little more complicated to build and what you end up with is much larger. A good "old" chip to use would be the TL598. It is available as both a surface mount and as an old style dual in line packace that is .3 inch by .9 inch (8mm x 20mm). It also will need a bunch of resistors and capacitors to make it work. In comparison there are some current devices that can make a driver with only four external components.

    These factors make it kind of hard to do a DIY type circuit that is easy to assemble, small and efficient. It's worth some thought thoug to see if something could be made.

  26. #26
    mtbr member
    Reputation: troutie-mtb's Avatar
    Join Date
    Sep 2007
    Posts
    3,062
    Hi Vroom9
    not sure what nicked me means

    when we come on here as newbs we know nothing but want to try and have a go
    a bit pointless when a DX bastid light is only 50 but the forum is still strong and new arrivals keep coming .

    We can mostly take a set of components and bolt them together to make a killer light using off the shelf drivers .
    When that has been done we have a light that is as good or better than most main makers lights OK they may look a bit Ghetto but its dark who sees them .

    then along comes a bunch of guys who start talking in strange symbols and language about building their own light drivers this then intrigues the likes of me who has had nothing to do with these strange little devices .

    So I challenge all you leccy wizards to design a driver that is as kick ass as the lights we all build and do it so a numpty like me can build it and learn a bit in the process

  27. #27
    mtbr member
    Reputation:
    Join Date
    Feb 2004
    Posts
    197
    Right on!
    I'm an idiot with a soldering gun.

    Quote Originally Posted by troutie-mtb
    Hi Vroom9
    not sure what nicked me means

    when we come on here as newbs we know nothing but want to try and have a go
    a bit pointless when a DX bastid light is only 50 but the forum is still strong and new arrivals keep coming .

    We can mostly take a set of components and bolt them together to make a killer light using off the shelf drivers .
    When that has been done we have a light that is as good or better than most main makers lights OK they may look a bit Ghetto but its dark who sees them .

    then along comes a bunch of guys who start talking in strange symbols and language about building their own light drivers this then intrigues the likes of me who has had nothing to do with these strange little devices .

    So I challenge all you leccy wizards to design a driver that is as kick ass as the lights we all build and do it so a numpty like me can build it and learn a bit in the process

  28. #28
    mtbr member
    Reputation: tamen00's Avatar
    Join Date
    Mar 2004
    Posts
    1,025
    Awesome!!! I agree 100% troutie... that would add a lot of value to the board and maybe get some things going while everyone is waiting for the next batch of better leds to come out... Things get a little dead around here in between updates to the leds... so lets spend the time learning a thing or two about drivers...

  29. #29
    www.hahntronix.com
    Reputation: mhahn@hvc.rr.com's Avatar
    Join Date
    Mar 2007
    Posts
    252
    Troutie,

    All the drivers I've built (I think I should have version 5 tested by the end of the week) have used off the shelf LED driver chips. Now a days that means playing with very tiny surface mount parts. Unless you're doing electronics for other than hobby stuff, it's probably not a worthwhile skill set to pick up, and it's definitely not cheap to learn. You can't really prototype surface mount circuits, so you need to learn how to lay out a printed circuit board (makes the tiny bits easier to solder, but it's still way harder than thru-hole soldering). And the way most of the drivers work, you can't really test or debug them with a voltmeter, so you need a decent oscilloscope. It gets pretty spendy pretty fast.

    I'd love to see cdcdcd's design that can be built using thru-hole components... hint hint. If it works as well as he says it does, it should be pretty simple to design a printed circuit board for it. The only reason I sound skeptical (no disrespect intended) is I've had my share of driver designs that lit up the LED but on closer examination weren't working as expected at all. His design might require a larger housing, but with the powerful new LEDs coming out you'll likely need one anyway. With a printed circuit board and a parts placement diagram, probably most folks who can solder would be able to build his driver.

    Of course, cdcdcd may choose not to share his design. That's his right. I haven't really shared many details about the ones I've worked on. I've put a lot of time and money into them and then dropped them for one reason or another. I've got a big box of over $500 worth of surface mount parts, sitting next to my desk, as I write this. I might get to salvage some of them for version 5 of my driver. This has become an expensive hobby

    I suppose someday I should at least fess up about the off the shelf chips I tried, but thought weren't worth pursuing. Just to save others some time.

    Mark

    PS: Troutie, if (big IF there) version 5 works as hoped. I might have something for you to try to drive a phlatlight with. Let's see if I burn out the 2 samples I have before I say any more.

  30. #30
    mtbr member
    Reputation: bumphumper's Avatar
    Join Date
    Jan 2009
    Posts
    136

    blinky

    Here is a simple led flasher circuit that might do the job; http://www.elecfree.com/electronic/b...c-555-or-7555/
    Instead of using multiple 5mm leds, just use a couple of xpes.
    I found the link posted in a thread on candlepowerforums entitled "how do I dyi blinking led's?"
    hope this helps

  31. #31
    mtbr member
    Reputation:
    Join Date
    Oct 2008
    Posts
    1,322
    cool...........

  32. #32

    Reputation:
    Join Date
    Jun 2009
    Posts
    82
    Now a days that means playing with very tiny surface mount parts. Unless you're doing electronics for other than hobby stuff, it's probably not a worthwhile skill set to pick up, and it's definitely not cheap to learn. You can't really prototype surface mount circuits, so you need to learn how to lay out a printed circuit board (makes the tiny bits easier to solder, but it's still way harder than thru-hole soldering). And the way most of the drivers work, you can't really test or debug them with a voltmeter, so you need a decent oscilloscope. It gets pretty spendy pretty fast.
    All true. Even with through-hole components, you can't really experiment with or debug a switchmode driver without an oscilloscope, and some electronics background. Quite honestly, by far the cheapest and easiest way to get a good driver is to buy one from Taskled, for example. The only reason you would wish to roll your own is unless you are a perfectionist like me who gets a kick out of building something better than you can buy, and/or if you enjoy the challenge and learning exercise of building your own. Surface mount technology makes incredibly compact designs possible, but is virtually unusable for the casual experimenter.


    I'd love to see cdcdcd's design that can be built using thru-hole components... hint hint. If it works as well as he says it does, it should be pretty simple to design a printed circuit board for it. With a printed circuit board and a parts placement diagram, probably most folks who can solder would be able to build his driver
    It's mostly just laziness that I haven't got around to drawing the circuit and posting it. However, I have already given the broad details that an experienced electronics person such as mhahn would need in order to build this driver. It would be very easy to build a PCB for this design, and I suppose that if a parts placement diagram was available, and plenty of photos, then even a novice would stand a fair chance of getting it working correctly. However, I suspect a novice would stand an equally fair chance of making a minor slip up, and it's virtually impossible to debug switchmode circuits without an oscilloscope and a fair understanding of how the circuit works. Correct PCB layout is important, but if a correctly laid out PCB was available then that gets around that problem. I also have to ask what others hope to achieve with my circuit. For example, part of the reason I can get such high efficiency is because I deliberately operate the circuit at only 40kHz so as to virtually eliminate switching loss, and minimise the power required for gate drive of the very-low-Rds mosfet. The downside to this is a huge (25x20x7) homewound inductor that occupies more volume than an entire Taskled driver! The circuit would still work nicely with a smaller inductor at 100 to 200kHz, but then the efficiency may be no better than Taskled, it would be MUCH bigger if through-hole components were used, may well cost more to make, so for many there would be little point. However, for those that would enjoy to learn, and get a kick out of making their own, the type of circuit I have used would be quite suitable. The TL494/TL598 chips are about the sweetest and easiest general purpose switchmode chip to use, and ideal for experimenting and getting your feet wet, though you will need an oscilloscope.

    They are also a so called "voltage mode" chip, rather than the more modern 'current mode" chips, which are actually inferior for this application, because of the higher required sense voltage, which degrades efficiency. "Current mode" is more modern and trendy, and has a number of advantages such as inherently faster response time. However, the makers of dedicated led driver chips have apparently not figured out that fast response time is irrelvant for most led driver applications. Battery voltage changes on the timescale of minutes or even hours, and the user requires intensity changes on a timescale of a few tenths of a second at fastest, and both these requirements are trivially easy to meet with a voltage-mode controller. Apparently many modern designers are unable to see the wood for the trees.


    The only reason I sound skeptical (no disrespect intended) is I've had my share of driver designs that lit up the LED but on closer examination weren't working as expected at all. His design might require a larger housing, but with the powerful new LEDs coming out you'll likely need one anyway.
    I love healthy scepticism. I am similarly reserved about many claims of very high measured efficiency, unless I know exactly how it was measured, and with what quality of equipment. I have a professional (analog) electronics background, though these days tend to specialize more in magnetic design, which I enjoy. In the physics research environment in which I work, we also have need of very fast switching of high voltage, say 5kV switched in a couple of nanoseconds, and that's kinda fun as well. The waveforms of my driver are all "textbook perfect", and I can assure you it works EXACTLY as intended. I could post pics of some of the key waveforms if you like.

    I also have a very nice linear regulator using a single CA3140 op amp which would be hard to beat in terms of simplicity and performance. I am much more confident that others could build that regulator and be confident it would work, and it can be debugged and understood with no more than a digital multimeter.


    I Am sure this thread means something to a few folks gifted in the black art of electronics and who understand the flux capasitator ION drive stuff .
    Now tell me,Troutie how the heck do you know about "ION drives"! An ion thruster is a type of rocket engine which fires high energy ions out the back of the spacecraft (or satellite) to provide thrust, and has the advantage of requiring less fuel to be carried. We design and build ion thrusters in one of the Labs here. Lots of high voltage electronics required to accelerate the ions.
    Last edited by cdcdcd; 08-24-2009 at 02:23 AM.

  33. #33
    mtbr member
    Reputation: troutie-mtb's Avatar
    Join Date
    Sep 2007
    Posts
    3,062
    Nice Explainations guys and thanks for taking the time to do it

    that is me convinced it is ways over my head and I am far too old a dog to learn new tricks like those .and will stick to the off the peg stuff

    And CDcdcd I guess I watch too many discovery programmes

  34. #34
    mtbr member
    Reputation: yetibetty's Avatar
    Join Date
    Dec 2007
    Posts
    1,781
    Blinkys "simple"LED flasher curcuit lost me.

  35. #35
    www.hahntronix.com
    Reputation: mhahn@hvc.rr.com's Avatar
    Join Date
    Mar 2007
    Posts
    252
    Blinky's simple circuit will probably not work for higher power LEDs, or at least it won't be very bright. A LM555 can sink a couple hundred mA (thousandths of an ampere). A decent high power LED takes most of an ampere or more to drive it. With the addition of a switching transistor, Blinky's circuit could work, but it won't be efficient.

    Well I can see that Cdcdcd has thrown the gauntlet:
    However, I have already given the broad details that an experienced electronics person such as mhahn would need in order to build this driver.
    I'll try to draw a schematic of what I think he's built, and he can belittle me over what I've got wrong . His resume, with respect to power supply design, is much more impressive than mine. I've never switched more than a few hundred volts, and nobody in my lab is working on Ion Drives (we do have the odd Flux Capacitor lying around though , maybe I should just fire up the DeLorean, drive to tomorrow and find the answers before I draw the schematic). I'm just an embedded systems guy (we mostly do firmware), who knows enuff analog theory to be very dangerous.

    But I have a couple of questions first (if the master will answer a few humble questions from the student):

    Do you do cycle by cycle current sensing? By that, I mean do you make sure that the current thru the LED never exceeds some maximum during every cycle the mosfet switch is turned on.

    The error amp in a TL598 has a unity gain bandwidth of about 800kHz. Won't it make a pretty crappy comparator? You implied you were using it as a comparator in an earlier post. Of course if you run your switcher at 20 kHz, that may not be an issue, but you talk about being able to run at 200 kHz.

    How do you handle noise? Most inductor switching I've done (not much actually) causes a hideous amount of ringing on the current sense resistor differential voltage. I figure the TL598 deadband helps with turning the mosfet on. How do you handle turning it off? Do you add hysteresis to the op amp? Or do you do some kind of snubber circuit?

    I'm sure I'll have more questions as I stumble ahead.

    Mark

    PS: If the student may offer an observation to the master:

    Most recent US driver chips seem to designed for the automotive market, hence the desire to control current thru the LED very tightly. Automobile battery voltages jump all over the place. Of course the master is right, that if you only wish to run off a battery connected to a single light, a simpler driver will suffice.

    Most Chinese drivers are designed for flashlights. They typically don't have decent thermal regulation, and are not (IMHO) designed to last longer than necessary to get a customer (chump) to buy the light. But hey, I'm a very cynical person.

  36. #36

    Reputation:
    Join Date
    Jun 2009
    Posts
    82
    G'day Mark,

    Now let me make it clear, I ain't no "analog master", but having you stir me up is just what I deserve. BTW, the "modern designers" that I suggested could not see the wood for the trees were not tintended to include you. I'm not sure I have the circuit written down anywhere, as I build directly from my head.

    I should take the trouble to draw the circuit, but just for now I'll try to answer your questions because you are a bit off the track as to how this "old fashioned" type of PWM controller works. You are trying to analyse and understand it in terms of the cycle-by-cycle hysteresis based controllers, but that is not how "voltage mode" chips such as the TL494/TL598 work at all.

    The heart of these chips is the pulse-width-modulator block. How this block works internally is not important. It is a simply a fixed frequency oscillator, where the duty cycle of the output is proportional to a control voltage input.

    The current is NOT sensed cycle-by-cycle. Usually there will be a filter capacitor across the leds(s), so that the voltage across the leds is is essentially smooth, and therefore so is the led current. The current sense resistor monitors the led current, which is essentially a smooth DC value, rather than the mosfet current which is discontinuous, or the inductor current which has a large sawtooth ripple component. Thus all the issues of ringing and other BS and artifacts on the current waveform are almost non existent, and it becomes practical to use a very small value of sense resistor, in my case 20 milliohms.

    So now to the op amp. It is NOT used as a digital comparator, and AFAIK I never said it was. The op amp is used as a high gain LINEAR amplifier. On one input is the DC reference voltage, and on the other input the (essentially DC) current-sense voltage. The output of the op amp is a DC voltage that represents the ERROR, being the difference between the desired current and the actual current, therefore this amp known as the error amp. The error voltage is wired to the control input of the pulse width modulator. This is a classic analog feedback control loop. By the action of the feedback loop, the duty cycle is maintained at exactly the correct value required to keep the desired current (= reference voltage) the same as the measured current (= voltage across sense resistor).

    That's all there is to it. Am I making any sense? This type of switchmode circuit is very easy to get up and running, and works well with much lower values of sense resistor than would be possible with "hysteresis style" current-mode controllers. Is this all too good to be true? Yes and no. Precisely because this type of circuit does not operate on a cycle-by-cycle basis, it cannot respond really fast on a cycle-by-cycle basis, either. The frequency at which it can respond MUST be less than the switching frequency, maybe 10 times less. But here is where it is important to see the wood from the trees. If the circuit operates at 100 kHz, then it can still respond to reference/line/load changes at 10 kHz, which is way faster than ever needed for portable LED driving applications. Furthermore, it is possible to achieve at least a 30:1 range of current control, without the lumen efficiency loss associated with "low frequency PWM chopping", and still use a low value of sense resistor. IMHO it's a case of win, win and win. Bugger, I should have been a salesman, not a scientist.

    Colin
    Last edited by cdcdcd; 08-25-2009 at 10:42 AM.

  37. #37
    mtbr member
    Reputation:
    Join Date
    Feb 2009
    Posts
    169
    I spent quite a few hours last weekend surfing the various semiconductor manufacturer websites. There are literally hundreds of current devices. They are all for specific purposes that don't match too well to what I was looking for.

    First the majority are designed for boost operation only, with no simple means to use them as a buck converter. I pretty much crossed them off the list. Just for runtime we tend to have at least one cell for each led so we don't really need boost. Boost requires all the current to go through a diode so there is loss there. It also requires higher currents.

    Second, I'm considering on using the P7 so being able to run 2.8 amps is a requirement. That disqualified almost all the chips with internal switches.

    Almost all the chips used sense voltages of at least 200mV. Some were 500mV. That is a problem at 2.8 amps. A 250mV sense voltage dissipates 700 mW at 2,8 amps.

    About the only chips left were the Diodes Inc (Zetex) 300,310, and 400 chips. They do cycle by cycle current sensing and have a sense voltage of around 20mV. The 300 and 310 suffer from a killer characteristic. The sense voltage increases with temperature and it changes by a pretty large factor. That is not a good thing as the current goes up the hotter they get. We want the opposite. The 400 chip has a much better temp characteristic. The 400 also has two sense inputs. The same 20mV switch current sense and a second 300mV voltage sense input.

    I also took a look at using a switching voltage reg chip. There are literally thousands of them. The modern ones all seem to suffer from the same problem, high sense voltage of 500mV to 800mV or more. That leaves only the old chips like the TL598 with the uncommitted sense amps.

    For me I'm just at the research and design stage for a switching driver. I been using cheapo DX drivers for my first testing. I think I'll be using a linear op-amp and mosfet regulator with about a 50mv sense for my first helmet light.

  38. #38
    www.hahntronix.com
    Reputation: mhahn@hvc.rr.com's Avatar
    Join Date
    Mar 2007
    Posts
    252
    OK Colin,

    I finally put together a schematic of sorts. I took a TL494 app note and cobbled up this:



    Anyway this is a quick and dirty photoshop hack of the schematic in the app note. The parts values are probably way off. And let's just agree to ignore R6, R7, and C2 for now (I'm pretty sure they just implement a soft-start using the DTC input).

    So you probably have a capacitor paralleled with R7 to cut the high frequency gain of the error amp. That would help eliminate the effect of the high frequency ringing you get when you switch an inductor on or off. A snubber across the inductor might help as well, but designing them seems like voodoo to me.

    Using a TL598 (with the error amp inputs swapped) you could replace Q1 and Q2 with a P channel mosfet. I'd imagine you'd want a low resistance resistor to drive the gate and some kind of pullup resistor. Your mosfet will need to have a max Vgs of whatever your input voltage is. There's lot's of P cahnnel mosfets with a +/- 20 volt Vgs, so that should be easy. The only problem I see with the '598, is it has an undervoltage lockout that will turn the mosfet on hard when the Input voltage drops below 6 volts. It'd probably work fine for a couple of P7's in series, or a MCE wired in 2S2P mode, but would blow up a P7 connected directly to it when the input voltage fell below 6 volts. And I really haven't figured out how'd you'd use the dead time circuit (DTC) to do some kind of inverted soft-start.

    This circuit sets the inverting input on the error amp at about .25 volts. With R11 at .1 ohms, you'd get an average drive current of around 2.5 amps. To drop the voltage at the inverting input, change the resistor divider made up of R3 and R4. That would let you boost the LED current, or else lower the value of R11 to get a more efficient driver.

    I'm curious to hear what kind of mosfet driver chip you used. There are a number that implement synchronous N channel mosfet driving. That would argue for using a '494. The biggest disadvantage I see to the '494 is it needs 7 volts in to keep the reference stable. If you can live with a gradually dimming light, once your battery goes south of 7 volts, that's not really an issue.

    Thanks for the introduction to a PWM alternative to dedicated LED drivers. The '494 and '598 look like interesting chips, but for now I'm going back to my surface mount custom designed LED driver chips. They seem a bit more robust, especially for applications that don't use a dedicated battery, and use less and smaller parts.

    Your latest posting makes you sound like you're from way south of the equator. What are a bunch of Foster's drinking blokes doing with Ion Drives? I apologize if Foster's is the Oz equivalent of Bear Wizz Beer, it's the only beer I've ever had form down under.

    Mark
    Last edited by [email protected]; 08-30-2009 at 01:41 PM.

  39. #39

    Reputation:
    Join Date
    Jun 2009
    Posts
    82
    Hi Mark,

    You have done a brilliant job of coming up with a general purpose TL494 schematic, though as drawn it's not real good for leds because it can't get to 100% duty cycle, as discussed in an earlier posting. I'll get back to you about my exact circuit, but for now will answer your easiest question first.

    Yes, I'm from way down south in Australia, home of Foster's beer, kangaroos, and Physics research. We do some pretty interesting stuff where I work, including plasma confinement for fusion research, ultracold Bose-Einstein condensates, and ion thruster rocket engines. Our main research area are listed at :-

    http://physics.anu.edu.au/Areas/

    For ion thrusters, click on "Plasma Applications and Technology", then on "Space Engines of the Future", for a description of our Helicon Double Layer Thruster (HLDT).

    I'm presently trying to attract funding for further research into advanced led bike lights and ultra-efficient drivers.

    Cheers, Colin

  40. #40
    mtbr member
    Reputation:
    Join Date
    Mar 2006
    Posts
    110
    Quote Originally Posted by cdcdcd
    I'll get back to you about my exact circuit, but for now will answer your easiest question first.
    Anytime soon?

    How about high efficiency with parallel cell and parallel leds so you could dim them individually? Any recommendation?

  41. #41
    www.hahntronix.com
    Reputation: mhahn@hvc.rr.com's Avatar
    Join Date
    Mar 2007
    Posts
    252
    Colin,

    Yes, I'm from way down south in Australia, home of Foster's beer, kangaroos, and Physics research. We do some pretty interesting stuff where I work, including plasma confinement for fusion research, ultracold Bose-Einstein condensates, and ion thruster rocket engines.
    Yikes, "Bose-Einstein condensates"! So then you guys can build some pretty serious beer coolers to keep your Foster's cold! And even launch it into orbit for future generations to enjoy ( just kidding, I know ION drives aren't that powerful, unless you've made some big breakthru recently).

    Thanks for introducing me to the '494, and '598. I might look at designing something with them, but they look really hard to get into a tight small PCB, at least in the thru-hole version.

    Mark

  42. #42
    mtbr member
    Reputation:
    Join Date
    Oct 2004
    Posts
    3,061
    Quote Originally Posted by Ekke
    How about high efficiency with parallel cell and parallel leds so you could dim them individually?
    As I understand it, the difficulty with driving LEDs in parallel is that the slight variations in the forward voltages will (of course) cause differing amounts of current to flow through the LEDs. This might not be a problem if you're staying well below the rated maximum current for the LEDs combined, but if you're trying to drive them at their maximum rated current, you'll end up over-driving some of them and under-driving the others. (I'm assuming that you don't place a current limiting resistor on each LED. Doing so is perfectly acceptable, recommended even, but it won't result in high efficiency.)

    As for dimming LEDs individually, that would suggest that you have a way of controlling the current to each individual LED. If you could solve this problem, the problem that I mentioned above would be solved too. But that suggests to me that you have independent drivers for each LED and that they're not really being driven in parallel.

    If you look at an LED data sheet, there's a value specified for maximum continuous forward current. There's often another, significantly larger value, specified for when the LED is pulsed. If you had an array of LEDs that you wanted to be able to control individually, you could multiplex them, and control the intensity of each LED individually by controlling the length of the pulse applied to each LED. And, of course, you can turn some of them off just by omitting the pulse. I have played around with controlling LEDs in precisely this fashion; I wasn't interested in dimming them individually, but I was interested in turning some on and some off at certain times and dimming those that were on as a group.

  43. #43
    mtbr member
    Reputation:
    Join Date
    Mar 2006
    Posts
    110
    Quote Originally Posted by KevinB
    But that suggests to me that you have independent drivers for each LED and that they're not really being driven in parallel.
    Yeah, that was what I was thinking.. But said something else.

    Currently I think I want 3 leds, one spot and two wides, maybe one extra wide. Would be nice to have even light so I think individual dimming would be nice. Not sure though, don't have those lenses yet.. Or even leds.

  44. #44
    mtbr member
    Reputation:
    Join Date
    Sep 2006
    Posts
    251
    Extra wide is useless for a bikelight, allmost 50% of the light will be waisted. It will illuminate the trees above you

  45. #45
    mtbr member
    Reputation:
    Join Date
    Mar 2006
    Posts
    110
    Quote Originally Posted by super-fast
    Extra wide is useless for a bikelight, allmost 50% of the light will be waisted. It will illuminate the trees above you
    I was thinking lens with 120 x 40 beam or so.. And it won't be just bikelight, in slow speeds (walking/running) wide beam would be nice. Just needs adjustable brightness spot with it that doesn't make you blind.

Members who have read this thread: 0

Posting Permissions

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

THE SITE

ABOUT MTBR

VISIT US AT

© Copyright 2020 VerticalScope Inc. All rights reserved.