Page 1 of 2 12 LastLast
Results 1 to 100 of 158
  1. #1
    mtbr member
    Reputation:
    Join Date
    Jan 2010
    Posts
    291

    LFlex for XML and Li-ion

    Hi folks, I've been working on new flex driver prototype and since I've kicked off a run of boards I figure I may as well let you know what's going to be available in a few weeks...

    LFlex is a Linear Regulator style driver. It is NOT a stack of AMC7135 regulators but instead uses a high performance FET along with a uController to provide standard Flex features with constant current output care of the FET.

    The whole idea is to provide a solution for a single XML run for a single Li-ion cell where a buck driver would not be able to run effectively.

    The LFlex will have a few current tables to choose from with the highest at 3A for the XML. With an XML and its low Vf a single Li-ion cell will give good runtime and regulation throughout the useful battery voltage range. Dimming will be via adjustable current output with the lowest levels likely to be a PWM of a low current output - yet to be determined once I have production boards and port the flex UI-UNI2 firmware to them.

    LFlex will have battery monitoring (but no STAT output), it can warn via the main LED just like my other flex drivers. It'll have all the medium/low/cutoff settings as well. Given I'm thinking this would be a good helmet light solution, not having a separate STAT LED wouldn't be too much of a loss. The reason I can't provide the STAT is a) lack of pins on the uC and b) even if I use a higher pin count uC there is a lack of board real estate to fit it all while still keeping it a single sided board.

    LFlex will have the standard temperature monitoring of the other Flex drivers.

    Basically you can think of the LFlex as a D2Flex except the LFlex is regulated and provides voltage monitoring.

    LFlex will operate to at least 9V (likely higher after more testing of the current regulation scheme), so running 2 li-ion with 2 XML is also feasible.

    Size will be a hair under 20mm and components only on one side, bottom side should be thermally mounted to the housing/heatsink.

    Target cost is close to d2flex pricing since this doesn't have the expense of the switcher regulator core.

    I'll post up pictures when I get the production boards. I may post a picture of a kludged up d2flex mule - but it just looks like a d2flex with some bits and wires hanging all over it

    cheers,
    george.

  2. #2
    Light freak
    Reputation: scar's Avatar
    Join Date
    Jun 2005
    Posts
    2,818
    Hey George, thanks for offering up yet another great driver option.

    20mm is still quite large to some people You ever think about making something smaller with maybe just two modes - 100% / 50%??

    Thanks again for all of the different options



    ****

  3. #3
    Dirt Deviant
    Reputation: savagemann's Avatar
    Join Date
    Jan 2008
    Posts
    3,693
    This is great news George.
    I'm glad you posted as I was about to order a bunch of 7135 boards.
    I'll hold out for the LFlex for sure.
    Put me down for at least a few.
    Look, whatever happens, don't fight the mountain.

  4. #4
    mtbr member
    Reputation:
    Join Date
    Apr 2010
    Posts
    489
    Now I'm happy! I'll take a few too. Can't wait to get my hands on them.

  5. #5
    mtbr member
    Reputation: troutie-mtb's Avatar
    Join Date
    Sep 2007
    Posts
    3,062
    Quote Originally Posted by scar
    20mm is still quite large to some people You ever think about making something smaller with maybe just two modes - 100% / 50%??
    ****

    20 mm is pretty good size wise bearing in mind most usefull optics for th XML will be 20 mm or larger .

    and the heat handling properties of a case will determine the size of the light if you want to go to the max .


    Marvellous I see this being a cool driver for some pretty ace lights and single sided for ease of mounting .
    Thanks George

  6. #6
    mtbr member
    Reputation: Goldigger's Avatar
    Join Date
    Nov 2009
    Posts
    1,669
    sounds bloody good to me... my search is over..
    I can run 2s1p with this new board george?

    If is a helmet light, who needs a stat led..
    cant see on top of my head

  7. #7
    Bigger is better!
    Reputation:
    Join Date
    May 2005
    Posts
    699
    Quote Originally Posted by troutie-mtb
    20 mm is pretty good size wise bearing in mind most usefull optics for th XML will be 20 mm or larger .
    Agree 100%

    Very good news, George!!

    Can you say anything about efficiency when running it from 1S vs 2S li-ion?

  8. #8
    mtbr member
    Reputation:
    Join Date
    Jan 2007
    Posts
    752
    It's a linear regulator so the input current will be the same as the output current, except for the low modes where it gets more complicated due to the PWM. So efficiency will be Vout / Vin. Vout will be the Vf of the LED in most cases.

    For 1 li-ion cell with an XML at 3A

    3.3 / 3.7 = 89% efficient.

    Obviously it'll be less efficient in the first part of discharge of the battery when the voltage is above 3.7, and more efficient near the end when voltage is below 3.7. You also wouldn't want to run a single LED off a 2s pack. That would be terribly inefficient. Running 3 or 4 LEDs off 3 or 4 li-ion cells would start to get quite a bit less efficient as well because of the bigger difference between Vin and Vout.

  9. #9
    mtbr member
    Reputation: Goldigger's Avatar
    Join Date
    Nov 2009
    Posts
    1,669
    1s2p?

  10. #10
    mtbr member
    Reputation:
    Join Date
    Jan 2004
    Posts
    1,436
    George, is there a prepay list? in for 2

  11. #11
    mtbr member
    Reputation:
    Join Date
    May 2005
    Posts
    3,340
    Microlight v3.0 just got a bit smaller!

  12. #12
    mtbr member
    Reputation:
    Join Date
    Jan 2010
    Posts
    291
    MtbMac has the efficiency info correct. This is a Linear (hence LFlex) driver and is a great single LED / single Li-ion solution. Efficiency is basically output voltage/ input voltage.

    Dimming will of course extend runtime since it will lower output current.

    Anyhow, it's not the perfect driver (haven't figured out how to make an arbitrary 100% efficient buck/boost that's 12mm in diameter yet ), but it ain't half bad either.

    2 LEDs and 2 Li-ion is likely the most I'd recommend from an efficiency/heat loss perspective, beyond that a true buck or boost driver will do a better job. It would be hard to beat the efficiency of a h6flex pushing 3A to 3 XML's from 4 li-ion cells.

    The key to this design is small size and single sided for ease of mounting to a heatsink to dump excess heat from the current regulating FET versus a bucket of 7135 chips. The other benefit is dimming is via lower output current so more efficient than PWMing a bunch of 7135 chips.

    Total headroom to regulate at 3A output is about 0.15V (less at lower currents), so it is what I would consider a very low drop out regulator design.

    No prepay list - I've NEVER done prepays and never will. If you are interested, email me and I'll add you to a first come/first served contact list that I'll email to before offering the drivers for general sale.

    It'll be around 3 weeks before they are available, I'll have boards in about 1 1/2 weeks and then I need to port the production code and test and characterize the current tables etc etc.

    Thanks for your interest,
    cheers,
    george.

  13. #13
    mtbr member
    Reputation: bravellir's Avatar
    Join Date
    Nov 2008
    Posts
    199
    nice.. two xmls with to cells will be a good combination

  14. #14
    mtbr member
    Reputation:
    Join Date
    Oct 2009
    Posts
    179
    Sounds good George. Will it have reverse polarity protection?

  15. #15
    mtbr member
    Reputation:
    Join Date
    Jan 2010
    Posts
    291
    Quote Originally Posted by JezV
    Sounds good George. Will it have reverse polarity protection?
    Yes, ALL my drivers have had reverse polarity protection for several years now.

    cheers,
    george.

  16. #16
    mtbr member
    Reputation: yetibetty's Avatar
    Join Date
    Dec 2007
    Posts
    1,781
    I'll join the queue and I bet it will be long one.

  17. #17
    mtbr member
    Reputation: mattthemuppet's Avatar
    Join Date
    Jul 2004
    Posts
    2,342
    wow, this is awesome! Simply awesome! I could knock out single XM-L + Regina helmet lamps like nobody's business with this driver using a simple 1in.sq "sled" design. I was going to try a Quazzle board, but this should be as efficient, throw better and be ~1/2 the price.

    could you list the differences between the h6flex and the Lflex? Largely so I can justify to myself that getting the h6flex was a good idea but also so I can think about future dual XM-L builds as this will seriously simplify housing design over the h6flex (and even the b2flex).

  18. #18
    mtbr member
    Reputation:
    Join Date
    Nov 2006
    Posts
    484
    I just sent a PM but basically put me in for two when available.
    thanks

  19. #19
    mtbr member
    Reputation:
    Join Date
    Jan 2010
    Posts
    291
    H6Flex is a damn fine driver - it is very robust. It is a buck driver, it will do DC/DC step down conversion and is very efficient. It can output up to 6.7A fully current regulated. It CAN'T operate from a single l-ion cell.

    The LFlex is a Linear regulator style driver - it BURNS excess voltage as heat, i.e. it does not do DC/DC conversion. Where it will excel is when input voltage is reasonably well matched with output voltage. It will also operate to low voltages before dropping out. I doubt you'll find a switcher based LED driver that can run from a single li-ion and drive a single white LED at 3A and do it efficiently. Such a switcher based driver would be very challenging to design and then to try and fit it into a 20mm envelope etc etc....

    So, the LFlex will fill in nicely where you have a single LED and single li-ion or 2 LEDs and 2 li-ion running at up to 3A.

    I don't see the LFlex stealing the H6Flex thunder, rather it filling in a gap at the single/double LED/cell end of the lighting spectrum. It will allow for a fully regulated and dimmable compact helmet type light where weight/size is an issue. It will provide voltage monitoring/warning/protection and thermal monitoring/protection while still having the same identical user interface as the rest of the flex drivers. So, you could have a bar mounted light and a helmet mounted light and operating the two would not require thinking what interface is on what light...

    The real comparison is the LFlex versus the D2Flex.

    cheers,
    george.

  20. #20
    mtbr member
    Reputation: mattthemuppet's Avatar
    Join Date
    Jul 2004
    Posts
    2,342
    Quote Originally Posted by georges80
    H6Flex is a damn fine driver - it is very robust. It is a buck driver, it will do DC/DC step down conversion and is very efficient. It can output up to 6.7A fully current regulated. It CAN'T operate from a single l-ion cell.

    The LFlex is a Linear regulator style driver - it BURNS excess voltage as heat, i.e. it does not do DC/DC conversion. Where it will excel is when input voltage is reasonably well matched with output voltage. It will also operate to low voltages before dropping out. I doubt you'll find a switcher based LED driver that can run from a single li-ion and drive a single white LED at 3A and do it efficiently. Such a switcher based driver would be very challenging to design and then to try and fit it into a 20mm envelope etc etc....

    So, the LFlex will fill in nicely where you have a single LED and single li-ion or 2 LEDs and 2 li-ion running at up to 3A.

    I don't see the LFlex stealing the H6Flex thunder, rather it filling in a gap at the single/double LED/cell end of the lighting spectrum. It will allow for a fully regulated and dimmable compact helmet type light where weight/size is an issue. It will provide voltage monitoring/warning/protection and thermal monitoring/protection while still having the same identical user interface as the rest of the flex drivers. So, you could have a bar mounted light and a helmet mounted light and operating the two would not require thinking what interface is on what light...

    The real comparison is the LFlex versus the D2Flex.

    cheers,
    george.
    never doubted the pedigree of the h6flex, I don't regret buying it at all. I do now understand the whole linear regulator thing better though - they key, as you say, is matching LED and cell no., i.e. 1 LED, 1S; 2 LED, 2S. I'm guessing that you can't do 3 LED/ 3S as the power burned as the Vin/Vout difference at that LED no. would be too great to safely dissipate, correct?

    I think this will make designing single LED lights (commuter or helmet) so much easier - I wish this was around when I started ordering parts for a bunch of commuter lights I'm building Still, there's finally a reason for people not to use those somewhat suspect DX/KD 7135 drivers

  21. #21
    Barrington Womble
    Reputation:
    Join Date
    Aug 2010
    Posts
    31
    George

    this is the good news my new year needed. Just started a new xml project but the most suitable driver I could find was from Shining Beam, and this was the best of a bad bunch!! I now have no excuse not to stand at the lathe in the cold workshop and get the basic housing sorted. Cheers.

  22. #22
    mtbr member
    Reputation: znomit's Avatar
    Join Date
    Dec 2007
    Posts
    1,911
    Now all we need is a source for a cheap 4p li ion battery pack

  23. #23
    mtbr member
    Reputation:
    Join Date
    Sep 2007
    Posts
    453
    IMHO the diameter is too big. It should be matched with the cell - 18mm
    CNC LED light housing for DIY projects

  24. #24
    mtbr member
    Reputation:
    Join Date
    Jan 2010
    Posts
    291
    Quote Originally Posted by zemike
    IMHO the diameter is too big. It should be matched with the cell - 18mm

    The diameter is what it is to fit the electronics on board. If you want 18mm, then I can leave off the bits that make it work and make the solder pads tiny or I can put components on both sides of the board and you can kiss away a good thermal path. Then you would end up with a kludge like a multi AMC7135 board with components on both sides and no good way to get rid of the heat.

    I already squeezed to make things fit well on this board AND to provide a good thermal path AND to allow the average person a chance of soldering the leads on that can carry 3A in/out of the driver. A round board is already a challenge from a layout perspective since you basically have square/rectangular components.

    Given the size of the optics needed and that the driver would have to be in that housing anyway if you want thermal protection and to minimize voltage drops in the system - i.e. a remote battery pack would not be a good idea either unless you are running THICK wires to deal with the up to 3A current draw... That light housing itself can't be too tiny - you will be dealing with upwards of 10 - 12W of heat you need to dissipate with a single XML at 3A....

    Here's a CAD layout of the board. Take it into photoshop etc and scale it to 0.78" and print it one to one and then see how tight everything is. Then draw an 18mm circle on it and that's what would have to be removed from the design



    cheers,
    george.

  25. #25
    mtbr member
    Reputation:
    Join Date
    Apr 2010
    Posts
    205
    georges80
    George, am I right thinking that LFlex will work properly with 2 serial XML from 2s (7.4 v) LiIons?

  26. #26
    mtbr member
    Reputation:
    Join Date
    Jan 2010
    Posts
    291
    Quote Originally Posted by klynk
    georges80
    George, am I right thinking that LFlex will work properly with 2 serial XML from 2s (7.4 v) LiIons?
    Yes, it will work fine.

    cheers,
    george.

  27. #27
    mtbr member
    Reputation:
    Join Date
    Jan 2010
    Posts
    291
    What current tables do you folk think I need to implement in the LFlex?

    To let the cat out of the bag, with the existing current sense resistor I can support 4A output (the FET can do way more than that - but heat losses are IxIxR so I really don't want to go stupid with the current).

    I'm thinking the following tables:

    4A overdriven XML or underdriven SST50 etc
    3A XML at spec (or a P7 or MCE)
    2A XML conservative or XPG overdriven
    1.5A XPG at spec
    1A XPG conservative or for any of the other "3W LEDS".

    Obviously all the dimming stuff will be there as per the standard Flex UI-UNI2 firmware, I'm just wondering what/how many current tables to provide - without going stupid

    thoughts?

    cheers,
    george.

  28. #28
    mtbr member
    Reputation: mattthemuppet's Avatar
    Join Date
    Jul 2004
    Posts
    2,342
    sounds good to me. Are you limited to 5 entries? As a 350/500mA level might be useful as a single mode XP-G/E commuter light. Then again, why wouldn't anybody set the Duomode to L5=1A and choose L2 or 3 for low if they were going to do that. It's what I'll be doing with the b(2)flex powered commuter lights I'm making at the moment.

  29. #29
    mtbr member
    Reputation:
    Join Date
    Apr 2010
    Posts
    205
    + 1 for 500 mA "commuting" level.

  30. #30
    mtbr member
    Reputation:
    Join Date
    Apr 2010
    Posts
    489
    Sounds right to me to. 4A should do the trick with both XML and MCE.

  31. #31
    Bigger is better!
    Reputation:
    Join Date
    May 2005
    Posts
    699
    Quote Originally Posted by georges80
    I'm thinking the following tables:

    4A overdriven XML or underdriven SST50 etc
    3A XML at spec (or a P7 or MCE)
    2A XML conservative or XPG overdriven
    1.5A XPG at spec
    1A XPG conservative or for any of the other "3W LEDS".
    How about skipping 1A and adding 2,5A?

    My impression is that the "3W" LEDs with 1A max are already made obsolete by XPG (except maybe XPE), so the 1A table would therefore sort of be made to suit old tech.

    And I said above: The size is great!!
    Optic size will be the driving factor of making XML lights as small as possible, and I would very much like to see sub-20mm optics that work with the XML....

  32. #32
    mtbr member
    Reputation:
    Join Date
    Jan 2004
    Posts
    213
    I know some people like the round profile boards, but I wouldn't care if the boards were rectangular... I really wouldn't care if the boards were rectangular and had the option of an external temperature sensor, something like a thermocouple or DS18B20.

  33. #33
    mtbr member
    Reputation: ortelius's Avatar
    Join Date
    Dec 2007
    Posts
    190
    Quote Originally Posted by klynk
    + 1 for 500 mA "commuting" level.
    IMHO, having 500 mA as one of available highest current level for a driver ment primarily for XML and similar high power leds, seem plain silly to me.What would you then do with other 5 (or 8) lower levels with 500 mA top current? I really wouldn't sacrifice one of the available driving levels for such an unsuitable one, as the number of drive levels is limited and must not be unrealistically high.

    For 500 mA commuting level, You you can simply choose setting with 1 A or 1.5 A as top driving current and the next lower level with this setting will be right around 500 mA, as mattthemuppet allready mentioned.

    I'd rather support langen's proposal to even skip 1 A and use 2.5 A instead.

  34. #34
    mtbr member
    Reputation:
    Join Date
    Oct 2009
    Posts
    179
    Quote Originally Posted by ortelius
    IMHO, having 500 mA as one of available highest current level for a driver ment primarily for XML and similar high power leds, seem plain silly to me.What would you then do with other 5 (or 8) lower levels with 500 mA top current? I really wouldn't sacrifice one of the available driving levels for such an unsuitable one, as the number of drive levels is limited and must not be unrealistically high.

    For 500 mA commuting level, You you can simply choose setting with 1 A or 1.5 A as top driving current and the next lower level with this setting will be right around 500 mA, as mattthemuppet allready mentioned.

    I'd rather support langen's proposal to even skip 1 A and use 2.5 A instead.
    +1 on that

    And 20mm round, single sided board sounds perfect to me. Shame we'll lose the stat connection but can't argue with physics

  35. #35
    mtbr member
    Reputation:
    Join Date
    Jan 2010
    Posts
    291
    The number of current tables is not limited to 5. Note, the h6flex has a choice of 9 current tables.

    I'm just looking for some feedback on what tables you think are USEFUL without creating so many that it becomes a pain to make a choice...

    Thanks to folk that have posted their thoughts so far. I won't need to make a decision for about a week, so plenty of time for comments and justifications to be made.

    The LFlex firmware will be the SAME C source code that generates all the other Flex drivers - I just need to add the LFlex specific 'stuff' with appropriate #ifdef's etc to generate the LFlex specific programming files. This means that creating the LFlex firmware isn't a huge task and very little testing needs to be done except for the LFlex specific changes.

    cheers,
    george.

  36. #36
    mtbr member
    Reputation: Goldigger's Avatar
    Join Date
    Nov 2009
    Posts
    1,669
    1.5, 2, 2.5, 3, 3.5
    I thought the idea of this small board was aimed at helmet lights
    I don't think 4 is needed.. that's going to be overkill for a helmet light, especially from a battery perspective and heat sinking..
    3.5 is probably overkill too..but nice to have that boost to show off

    With regard to sst-50 there £18 each, not a great selection of optics... its not a cheap option.. I still haven't seen many people use it, apart from in dive lights.

  37. #37
    mtbr member
    Reputation:
    Join Date
    Apr 2010
    Posts
    489
    I would go even for 5A possibility if the board can take it GD you don't need to set your board to full power if you don't need it, but I would like to drive my XML at 4A and even 4P MCE's at the same current. For a helmet light you can take Lflex+XPG+regina set to 2A and single or double XML at 3-4A for hadlebars.

  38. #38
    mtbr member
    Reputation: Goldigger's Avatar
    Join Date
    Nov 2009
    Posts
    1,669
    Quote Originally Posted by Toaster79
    I would go even for 5A possibility if the board can take it GD you don't need to set your board to full power if you don't need it, but I would like to drive my XML at 4A and even 4P MCE's at the same current. For a helmet light you can take Lflex+XPG+regina set to 2A and single or double XML at 3-4A for hadlebars.
    If you have 4P MCE surely your best off with a H6flex??

    Thanks for clarifying that i dont need to set the board to max

  39. #39
    mtbr member
    Reputation:
    Join Date
    Apr 2010
    Posts
    489
    Lflex will be cheaper and it will suit my needs. 4P I ment as 4 parallel dies on one chip.

  40. #40
    mtbr member
    Reputation:
    Join Date
    Apr 2010
    Posts
    205
    Quote Originally Posted by georges80
    I'm just looking for some feedback on what tables you think are USEFUL without creating so many that it becomes a pain to make a choice...
    IMO, for XML - 1A, 2A, 3A (3 modes is enough).
    for XPG - 0,6A, 1A, 1,5A (again, 3 modes).
    4A for XML would be TOO hot, IMO. I don't think that having such a heater on helmet is a good idea.

  41. #41
    mtbr member
    Reputation: Goldigger's Avatar
    Join Date
    Nov 2009
    Posts
    1,669
    Quote Originally Posted by Toaster79
    Lflex will be cheaper and it will suit my needs. 4P I ment as 4 parallel dies on one chip.
    You want to put 4 amp's through each die? thought they could only handle 700ma each?

  42. #42
    mtbr member
    Reputation:
    Join Date
    Apr 2010
    Posts
    489
    No, I want to put 1 amp through each die, so that's 4 amps for four dies in parallel wiring.

  43. #43
    mtbr member
    Reputation: Goldigger's Avatar
    Join Date
    Nov 2009
    Posts
    1,669
    Quote Originally Posted by Toaster79
    No, I want to put 1 amp through each die, so that's 4 amps for four dies in parallel wiring.
    Im with you, wasn't paying attention..to the parallel part.

  44. #44
    mtbr member
    Reputation: mattthemuppet's Avatar
    Join Date
    Jul 2004
    Posts
    2,342
    Quote Originally Posted by ortelius
    IMHO, having 500 mA as one of available highest current level for a driver ment primarily for XML and similar high power leds, seem plain silly to me.What would you then do with other 5 (or 8) lower levels with 500 mA top current? I really wouldn't sacrifice one of the available driving levels for such an unsuitable one, as the number of drive levels is limited and must not be unrealistically high.
    well, George did list different levels and their matching LEDs, so this driver doesn't have to be used with high powered LEDs. 500mA top current then L3 or L4 as low would make a neat XP-G commuting light. ~150lm on high, 50-75 on low - double the run time of 1A and most likely more than enough light for the street and bike paths.

  45. #45
    mtbr member
    Reputation: troutie-mtb's Avatar
    Join Date
    Sep 2007
    Posts
    3,062
    I am with George`s suggestion I thing it covers the bases real well

    Dont see a need for 500 ma as that is covered in the 1 amp setting there is enough sub settings to get the commute setting low enough
    and why build a light that is only for commuting it may as well be useable both on and off road .

  46. #46
    Dirt Deviant
    Reputation: savagemann's Avatar
    Join Date
    Jan 2008
    Posts
    3,693
    Agreed Troutie.
    But what if we had half steps.....
    4
    3.5
    3
    2.5
    2
    1.5
    1

    I dunno if it really is needed, but a 2.5 setting would be nice for running it slightly underdriven.
    Look, whatever happens, don't fight the mountain.

  47. #47
    mtbr member
    Reputation: znomit's Avatar
    Join Date
    Dec 2007
    Posts
    1,911
    Quote Originally Posted by troutie-mtb
    I am with George`s suggestion I thing it covers the bases real well

    Dont see a need for 500 ma as that is covered in the 1 amp setting there is enough sub settings to get the commute setting low enough
    and why build a light that is only for commuting it may as well be useable both on and off road .
    500mA is a good level for endurance adventure riding where people probably just want an on/off solution (some guys go all night on 50lm petzl headlamps).

    0.5
    1
    1.5
    2.3
    3
    4

    not sure about the 2.3 but 2 isn't much of a bump from 1.5 and 3 is tecnically out of spec for MCE & P7

  48. #48
    mtbr member
    Reputation: ortelius's Avatar
    Join Date
    Dec 2007
    Posts
    190
    Quote Originally Posted by klynk
    IMO, for XML - 1A, 2A, 3A (3 modes is enough).
    for XPG - 0,6A, 1A, 1,5A (again, 3 modes).
    4A for XML would be TOO hot, IMO. I don't think that having such a heater on helmet is a good idea.
    You are not talking about the same thing George is asking. We are not discussing about different levels (modes) that you can have on your light, once you program it. We are talking about available *max current settings* that the board can offer. You then choose one of this and this is the max current your light can have. The number of modes you have is another choice when you program it, but you do not explicitly set their currents in absolute numbers, they are predefined in firmware, based on choice of your max current and number of levels (UI choice).

  49. #49
    mtbr member
    Reputation:
    Join Date
    Apr 2010
    Posts
    205
    Quote Originally Posted by ortelius
    You are not talking about the same thing George is asking. We are not discussing about different levels (modes) that you can have on your light, once you program it. We are talking about available *max current settings* that the board can offer. You then choose one of this and this is the max current your light can have. The number of modes you have is another choice when you program it, but you do not explicitly set their currents in absolute numbers, they are predefined in firmware, based on choice of your max current and number of levels (UI choice).
    Thanks, now I get the point.

  50. #50
    mtbr member
    Reputation: mattthemuppet's Avatar
    Join Date
    Jul 2004
    Posts
    2,342
    Quote Originally Posted by troutie-mtb
    Dont see a need for 500 ma as that is covered in the 1 amp setting there is enough sub settings to get the commute setting low enough
    and why build a light that is only for commuting it may as well be useable both on and off road .
    as they say, horses for courses troutie. The commuter light I'm building for myself won't be used off road - that's what my 2x XP-G helmet light and soon to be built 2x XM-L bar light is for. So for a cheap and simple commuter light, which this driver would suit perfectly, a <1A maximum drive level would be very useful. An XP-G @1A is seriously bright on the road, possibly too much so, and I'd rather have the low level for a "be seen" level when the lights come on but before it gets dark. Plus, battery life becomes more important than absolute light output when commuting - I really don't want to have to recharge my light ever night (or even every other night).

    Anyway, it's up to George. If he can fit a <1A max drive level in without buggering anything else up, I don't really see the harm in it. You might not use it, but that doesn't mean others won't.

  51. #51
    mtbr member
    Reputation: troutie-mtb's Avatar
    Join Date
    Sep 2007
    Posts
    3,062
    Quote Originally Posted by mattthemuppet
    as they say, horses for courses troutie. The commuter light I'm building for myself won't be used off road - that's what my 2x XP-G helmet light and soon to be built 2x XM-L bar light is for. So for a cheap and simple commuter light, which this driver would suit perfectly, a <1A maximum drive level would be very useful. An XP-G @1A is seriously bright on the road, possibly too much so, and I'd rather have the low level for a "be seen" level when the lights come on but before it gets dark. Plus, battery life becomes more important than absolute light output when commuting - I really don't want to have to recharge my light ever night (or even every other night).

    Anyway, it's up to George. If he can fit a <1A max drive level in without buggering anything else up, I don't really see the harm in it. You might not use it, but that doesn't mean others won't.


    Hey up Mattthemuppet what I was meaning is in the sub settings you will get any level you wish so max 1 amp steps down to 50 ma drive .

    if you use the UIP setting then you can set it at what ever level you want at the time you switch it on and click on click off Simples and short press goes to max for the warning flash to the knob head driver

  52. #52
    mtbr member
    Reputation: Ofroad'bent's Avatar
    Join Date
    Jul 2010
    Posts
    1,362
    OK, asbestos suit on- what would happen if you wanted to drive 1 XM-L using a 7.4v battery?

    I know there are other -flex drivers, but not this small, and I'd like to use my nice collection of waterproof 7.4v Lithium battery packs.

  53. #53
    mtbr member
    Reputation:
    Join Date
    Jan 2010
    Posts
    291
    Quote Originally Posted by Ofroad'bent
    OK, asbestos suit on- what would happen if you wanted to drive 1 XM-L using a 7.4v battery?

    I know there are other -flex drivers, but not this small, and I'd like to use my nice collection of waterproof 7.4v Lithium battery packs.
    You would burn the excess voltage in heat - totally useless waste of energy...

    The LFlex is a LINEAR regulator style driver, any excess voltage goes to heat.

    If you set the drive level to 3A you would (fresh off the charger) be pulling 3A from 8.4V to drive a 3.3V LED.

    It would be the equivalent of having a 1 litre container and pouring 2 litres of beer into it - the other litre would spill on the floor and that is CRIMINAL unless we're talking Fosters or Budweiser or ...

    cheers,
    george.

  54. #54
    mtbr member
    Reputation: Ofroad'bent's Avatar
    Join Date
    Jul 2010
    Posts
    1,362
    Quote Originally Posted by georges80
    You would burn the excess voltage in heat - totally useless waste of energy...

    The LFlex is a LINEAR regulator style driver, any excess voltage goes to heat.

    If you set the drive level to 3A you would (fresh off the charger) be pulling 3A from 8.4V to drive a 3.3V LED.

    It would be the equivalent of having a 1 litre container and pouring 2 litres of beer into it - the other litre would spill on the floor and that is CRIMINAL unless we're talking Fosters or Budweiser or ...

    cheers,
    george.
    Damn, that's what I thought but I'd hoped I got it wrong. What's Dr George's prescription fro something as small that would run a single XM-L from 6.0-7.4v?

  55. #55
    mtbr member
    Reputation: mattthemuppet's Avatar
    Join Date
    Jul 2004
    Posts
    2,342
    Quote Originally Posted by troutie-mtb
    Hey up Mattthemuppet what I was meaning is in the sub settings you will get any level you wish so max 1 amp steps down to 50 ma drive .

    if you use the UIP setting then you can set it at what ever level you want at the time you switch it on and click on click off Simples and short press goes to max for the warning flash to the knob head driver
    ah, I get you now. I wasn't familiar with the UIP setting - I was working my approach off UIB, where you can't do that.

  56. #56
    mtbr member
    Reputation:
    Join Date
    Jan 2010
    Posts
    291
    Quote Originally Posted by Ofroad'bent
    Damn, that's what I thought but I'd hoped I got it wrong. What's Dr George's prescription fro something as small that would run a single XM-L from 6.0-7.4v?
    It doesn't currently exist and likely won't. To run 2 li-ion would require a buck driver and that means an inductor, switcher IC and the rest of the control baggage.

    It could be feasible to fit into a 20mm board but would need components on both sides and that becomes somewhat of a mounting challenge because a good thermal path would still be essential. To make it all fit it would need to use the smallest packages (QFN's etc) and that means assembly house setup costs etc etc and only makes sense if the driver sells in decent (1000's) volumes.

    The LFlex is my attempt to address single XML in a compact form factor and the tradeoff is that you run ONE li-ion cell or 2P (or 3P etc for longer runtime).

    cheers,
    george.

  57. #57
    mtbr member
    Reputation:
    Join Date
    Oct 2008
    Posts
    1,310
    Quote Originally Posted by georges80

    It would be the equivalent of having a 1 litre container and pouring 2 litres of beer into it - the other litre would spill on the floor and that is CRIMINAL unless we're talking Fosters or Budweiser or ...

    cheers,
    george.
    Harsh.........

  58. #58
    mtbr member
    Reputation: troutie-mtb's Avatar
    Join Date
    Sep 2007
    Posts
    3,062
    Quote Originally Posted by mattthemuppet
    ah, I get you now. I wasn't familiar with the UIP setting - I was working my approach off UIB, where you can't do that.

    Its worth having a play with the UIP I have been using it for a while now and liking it for the easy use it gives .

    currently using it for the stupid bright quad xml

    from off you hold the button until your desired light level and then let go
    then click and it is off click again and it is on at that level

    press and it goes to max press and it goes back to the preset level .

    perfect for a commuter light

  59. #59
    mtbr member
    Reputation: mattthemuppet's Avatar
    Join Date
    Jul 2004
    Posts
    2,342
    Quote Originally Posted by troutie-mtb
    Its worth having a play with the UIP I have been using it for a while now and liking it for the easy use it gives .

    currently using it for the stupid bright quad xml

    from off you hold the button until your desired light level and then let go
    then click and it is off click again and it is on at that level

    press and it goes to max press and it goes back to the preset level .

    perfect for a commuter light
    thanks for the headsup troutie, I'll look into that for definite. The ones I'm making for friends/family I'll probably stick with UIB+Duomode, just because I don't have to explain anything to them

  60. #60
    mtbr member
    Reputation: mattthemuppet's Avatar
    Join Date
    Jul 2004
    Posts
    2,342
    Quote Originally Posted by odtexas
    Harsh.........
    considering that Fosters is crap beer that Australia exports because no self-respecting Australian would drink it, I think it's pretty fair. I'd say the same about anything with Miller or Coors in the name..

  61. #61
    mtbr member
    Reputation:
    Join Date
    Feb 2009
    Posts
    14
    May be George can build module with leds, similar Quazzle L33 , but for XM-L and 35мм CUTE-3 optics. It would be fine with simple DX housing.

  62. #62
    mtbr member
    Reputation:
    Join Date
    Jan 2010
    Posts
    291
    Quote Originally Posted by hecs_ru
    May be George can build module with leds, similar Quazzle L33 , but for XM-L and 35мм CUTE-3 optics. It would be fine with simple DX housing.
    Sorry, but I don't plan to make modules. That's something other folk can do if they wish.

    The problem with modules, for a lot of folk:

    1) whatever LED you have on there is already obsolete when you release the module or has the wrong tint or lumens/watt.2) whatever number of LEDs you have uses is too many or not enough
    3) whatever size you have made the module is too big or too small or the wrong shape
    4) you've locked yourself into a small choice of optics due to the module/board size.

    From a business perspective, if you are targeting a savvy customer base you are at risk building something that the key component is in a technology improvement cycle. I've seen it happen at CPF for years, a guy buys a reel of the latest and greatest LEDs for $N, before he sells 1/2 the reel at a profit a new LED hits the market and his remaining 1/2 reel is yesterdays flavour of the month. Now he has to sell the remaining 1/2 at a loss for a net $0 if he's lucky.

    Now compound that with it being a finished module with 3 LEDs and driver electronics and you now have a board that no one wants to spend premium money on. To make financial sense you probably ran 500 modules (or more) and you now have to sell the remainder at a discount or over a year or two versus a month or two...

    If one was to make a complete module then you may as well go the whole way and put it into a housing and supply a battery pack and call it done.

    Anyhow, that's my position on a complete module and why it's not something I plan to provide.

    cheers,
    george.

  63. #63
    Dirt Deviant
    Reputation: savagemann's Avatar
    Join Date
    Jan 2008
    Posts
    3,693
    Just keep doing what you're doing George, cause you are doing it right.
    I can't wait to get my hands on the LFlex boards.
    It's exactly what I've been needing.
    And at the projected price, I think I'm gonna get at least 5.
    Lots of my buddies need lights......= )
    Look, whatever happens, don't fight the mountain.

  64. #64
    mtbr member
    Reputation:
    Join Date
    Jan 2010
    Posts
    291
    Just a quick update. My PCB house has shipped boards to me and they should arrive Monday (based on current tracking). I just received the fancy power FETs today, so at this point things are looking good to have some boards done next week so I can port the UI-UNI2 firmware and start testing various current tables etc.

    cheers,
    george.

  65. #65
    mtbr member
    Reputation:
    Join Date
    Jun 2009
    Posts
    10
    Quote Originally Posted by georges80
    You would burn the excess voltage in heat - totally useless waste of energy...

    The LFlex is a LINEAR regulator style driver, any excess voltage goes to heat.
    I think cooling is very essential with these drivers?

    With a fully charged cell the initial heat will be aprox (4.2V-3.35V) x 3A = 2.55W.

    I was planning to make a "clean" LED-lamp without integrated driver. With an external driver my plan was to make it easier to upgrade parts in the future. Maybe I need to put the driver together with the led cooler to handle the excess heat?

  66. #66
    mtbr member
    Reputation:
    Join Date
    Jan 2011
    Posts
    270
    Another vote for a 4A maximum current option. Perfect for my XM-L twin with beam cutoff for commuting. Unless I go triple ...

  67. #67
    mtbr member
    Reputation:
    Join Date
    Jan 2010
    Posts
    291
    Quote Originally Posted by DanneRe
    I think cooling is very essential with these drivers?

    With a fully charged cell the initial heat will be aprox (4.2V-3.35V) x 3A = 2.55W.

    I was planning to make a "clean" LED-lamp without integrated driver. With an external driver my plan was to make it easier to upgrade parts in the future. Maybe I need to put the driver together with the led cooler to handle the excess heat?
    Yes, cooling is important, that's why the board has no components on the bottom side and has an excellent thermal path from the FET to the bottom of the PCB for mounting via the supplied double sided adhesive thermal tape to the housing.

    Do note that the 4.2V will only last a few minutes at most at 3A. It will drop down to 3.7V very quickly, though obviously one must design for the worst case dissipation.

    It makes sense to put the driver in the housing so that it can sense overall 'system' temperature and protect itself AND the LED...

    cheers,
    george.

  68. #68
    mtbr member
    Reputation:
    Join Date
    Jan 2010
    Posts
    291
    lFlex production PCB's arrived today. Here's what it'll look like finished. Now I need to work on porting the ui-uni2 firmware to it. I'll keep you folk updated as I get the code running on a test mule.



    cheers,
    george.

  69. #69
    mtbr member
    Reputation: mattthemuppet's Avatar
    Join Date
    Jul 2004
    Posts
    2,342
    as the Welsh say, "tidy"

  70. #70
    banned
    Reputation: random walk's Avatar
    Join Date
    Jan 2010
    Posts
    2,666
    Still flirting with 20mm?

  71. #71
    Dirt Deviant
    Reputation: savagemann's Avatar
    Join Date
    Jan 2008
    Posts
    3,693
    Nice, can't wait!!!
    Looks good.
    Look, whatever happens, don't fight the mountain.

  72. #72
    Bigger is better!
    Reputation:
    Join Date
    May 2005
    Posts
    699
    Quote Originally Posted by georges80
    lFlex production PCB's arrived today. Here's what it'll look like finished. Now I need to work on porting the ui-uni2 firmware to it. I'll keep you folk updated as I get the code running on a test mule.

    cheers,
    george.
    Nice!

    Do you have a confirmed diameter of the boards?

  73. #73
    mtbr member
    Reputation:
    Join Date
    Jan 2010
    Posts
    291
    Diameter is per the initial spec in this thread - 0.780" or just a hair under 20mm.

    I have a fair bit of the firmware source code modified so likely today I'll download a trial and see if some of the basic UIP code functions (easier to test than UIB).

    For historical reasons the UIP/UIF core is totally separate to the UIB core, so testing the UIP first makes sense.

    cheers,
    george.

  74. #74
    mtbr member
    Reputation:
    Join Date
    Jan 2010
    Posts
    291
    Ok, making some progress with the firmware porting. A bunch more stuff has to be added/changed, but I do at least have the basic firmware running and the current regulation core functional.

    I've only tested the 20mA to 400mA range of output current at this point, but don't foresee any issues going higher - it's the lower end that is more problematic from a regulation perspective.

    I'm pleased to see that the current regulation is stable even at 20mA. I was originally thinking I may have to run a PWM of 200mA for the lower ranges, though it appears that won't be necessary.

    Anyhow, early times still and a lot more work to do.

    cheers,
    george.

  75. #75
    mtbr member
    Reputation:
    Join Date
    Jan 2010
    Posts
    291
    Been making some good progress with the firmware port. I have UIP & UIB2 functional and tested. I still need to test UIF, but it should just 'fall out' of the UIP testing.

    I have the voltage monitoring/warning working and tested. Still need to test the thermal monitoring but no reason it shouldn't work.

    What I've spent quite a bit of time on is the current regulation algorithm at the low end (<200mA). With sense voltage in the 0.05 x 0.015 = 0.75mV range, things get a little tricky. Fortunately the uC I'm using has an internal 20x gain stage that I'm using to get things in the 15mV range and higher.

    Still at those levels, even one PWM count is "visible" to the human eye. Unfortunately (for me), the human eye is very sensitive to small intensity changes at the low end of the range...

    Anyhow, I've got a scheme in place that has pretty well eliminated flicker at the low end by dithering the PWM output that feeds the poor mans D/A converter that drives the gate of the power FET.

    I now have to just finish adding in the rest of the code to deal with autocalibration of the voltage reference etc.

    Given most pictures on this site are folk doing machining with a variety of tools from dremels to CNC's I figured I'd put up a picture of my 'work area' and some of my toys (I mean tools)



    cheers,
    george.

  76. #76
    mtbr member
    Reputation:
    Join Date
    Jan 2011
    Posts
    66
    Looks awesome! (Both the board and the workspace.)

    I'm really liking the idea of this board, and I am very glad that you are working to smooth out flicker on the PWM in the ultra-low end. Reading or working up close with an LED set to low with visible flicker is a major annoyance.

    Looking forward to getting mine when you are done with your work. This board will be an integral part in reaching my goal of a home-brewed, sub-100-gram, headlamp (including the strap and battery). It's going to be a near-run thing, but I think it's doable.

  77. #77
    Rep Power Infinity +2
    Reputation: brad72's Avatar
    Join Date
    Jun 2009
    Posts
    1,538
    Quote Originally Posted by georges80
    the other litre would spill on the floor and that is CRIMINAL unless we're talking Fosters or Budweiser or ..
    As long as it's not the afore mentioned beers I'm happy to take one for the team and lie on the floor with my mouth open to collect the overspill

    Love the workbench. You can never have enough meters etc.

    Brad

  78. #78
    mtbr member
    Reputation:
    Join Date
    Jan 2011
    Posts
    270
    Very interested to see how this progresses - looks very interesting.

  79. #79
    mtbr member
    Reputation: troutie-mtb's Avatar
    Join Date
    Sep 2007
    Posts
    3,062
    ( voice from the jaws film ) I think you need a bigger desk

  80. #80
    mtbr member
    Reputation:
    Join Date
    Jan 2010
    Posts
    291
    Quote Originally Posted by troutie-mtb
    ( voice from the jaws film ) I think you need a bigger desk
    It's surprising to a lot of folk that for the board sizes I work with, I really only need a clear area about 3' wide and 8" deep to do all my work. With all the test equipment arranged around the work 'area' it becomes a very productive environment.

    Now for another status update:

    I've been tuning the current table entries for the past little while.

    There's 5 entries per current table for UIP and UIB and 8 entries per current table for UIF, so well over 100 entries. I run an excel spreadsheet to place the individual points on a curve and then take those values and put them into the firmware. I use the same curve for all the flex drivers, so the relative step changes appear consistent.

    At this stage I'm settling on 7 current tables for the initial LFlex release:

    500mA, 1000mA, 1500mA, 2000mA, 2500mA, 3000mA and 3500mA.

    The L1 (low) on all of the tables will be 50mA.

    These are nominal values and as I test more LFlex (later) I'll get an idea of the +/- % spread over many boards.

    The 4A has been dropped for now since it requires a sense resistor value change (and that would then impact all the other current table entries). So, if the 4A is a big deal for some folk, that will be rolled out in a future run of boards.

    I've done some temperature versus time testing with the 3.5A range, running 4.2V in and 3.2V out that mimics a fully charge li-ion cell driving a low Vf XML. That means 1V x 3.5A = 3.5W dissipation in the FET.

    Using an inferior thermal pad material (non-adhesive for ease of testing), the FET case reaches about 45C with 20C heatsink temperature.This is well within the capabilities of the FET. With the supplied Bergquist adhesive thermal tape, temperature will be lower since the Bergquist material has over 2x the thermal conductivity of the non-adhesive material I'm using to test with.

    UIF is the last part of the code that still needs to be tested, but I'm not expecting any issues to crop up since most of the UIF code 'co-exists' with the UIP code that has already been tested.

    I've tested the temperature monitoring/auto-dimming code and it is also functional.

    So, we're getting close. I'm expecting to release LFlex drivers later next week and will do so by first contacting the folk that emailed me showing interest. After that I'll do a general release.

    Anyhow, a VERY nice sunny day here - so I need to go and make a couple of wheels spin...

    cheers,
    george.

  81. #81
    Dirt Deviant
    Reputation: savagemann's Avatar
    Join Date
    Jan 2008
    Posts
    3,693
    Awesome news George!!!!
    I'm thrilled with the current table options....Perfect!
    Sounds like everything is going to plan. Good stuff.
    Can't wait to get that email from you stating they are ready for purchase.
    Look, whatever happens, don't fight the mountain.

  82. #82
    mtbr member
    Reputation:
    Join Date
    Jan 2010
    Posts
    291
    Running some more tests at 3A and 3.5A to characterize the FET thermal path. Figured I'd measure the headroom needed for the LFlex to remain in regulation.

    I'm using quite an impressive little FET on the LFlex and it has some pretty amazing specifications for its size. But then it is quite an expensive component relative to the rest of the board - it is the singularly most expensive part of the drive.

    Anyhow, at 3A I adjusted the input voltage until it transitioned to direct drive at which point the FET is turned on 100% and measured 0.08V drop across the driver. That's a pretty impressively low dropout at 3A. That 0.08V implies a total driver resistance of 0.08/3 = 0.027 ohms which is reasonable given that there is a 0.015 ohm sense resistor in series with the FET (which has an RDSon of 0.006 ohms with 2.5V gate drive) for a total of 0.021 ohms. Factor in a few milliohms for the copper traces and it all makes sense. I have a nice Fluke 6 1/2 digit bench meter, so measuring such low voltages accurately it quite trivial.

    I'm using the same FET on the Hallsw board - or at least its non-identical twin. It's a TI FET and I noticed that the one on the LFlex had very similar specs to the one on the Hallsw board... when I ordered the LFlex FET I received notice that it's a factory sort of the other FET (for guaranteed and documented performance at 2.5V gate voltage). So, the over-achievers get their own part number

    If anyone is chasing a pretty amazing FET (especially for its size) and can deal with surface mount etc, it's a TI CSD16340Q3

    cheers,
    george.

  83. #83
    mtbr member
    Reputation:
    Join Date
    Oct 2009
    Posts
    179
    George, I haven't a clue what a FET is but if you're happy then I'm happy Like everyone else on this thread I'm looking forward to the lflex with anticipation and your excitement makes me excited without really knowing why, like mass hysteria

  84. #84
    mtbr member
    Reputation: troutie-mtb's Avatar
    Join Date
    Sep 2007
    Posts
    3,062
    Quote Originally Posted by JezV
    George, I haven't a clue what a FET is but if you're happy then I'm happy Like everyone else on this thread I'm looking forward to the lflex with anticipation and your excitement makes me excited without really knowing why, like mass hysteria


    Me too not a clue what George is talking about and happy to let it stay that way .
    what I could glean from the bits I did understand was things are going well and we can look forward to the release of a good product . and by the sound of it at a good price too .

    I have just been back over the thread to recap .

    am I right here

    it will drive 1 xml from 1 or more parallel cells but not 2 series .cells

    it will drive 2 xmls from 2 series cells but not 3 or 4

    it will will be hotter in the early running from fully charged cells .

    does the light then dim when it drops out of reg


    will it be heating up more if using low settings from full cells

  85. #85
    mtbr member
    Reputation:
    Join Date
    Apr 2010
    Posts
    489
    George is talking about really high quality part of the driver (FET= Field Effec Transistor) which is also the heart of the driver. It's resistance in direct drive is much lower than resistance of the copper traces on the board.

    Quote Originally Posted by troutie-mtb

    does the light then dim when it drops out of reg

    will it be heating up more if using low settings from full cells
    The light will eventually get dimmer as it goes in direct drive, or maybe not, as long as you set the cut off voltage to 3V which is actually around the Vf of XM-L.

    I think it won't heat up more at lower settings, because the driver burns excess voltage, and it doesn't behave like resistor, which burns excess power. But someone who is more in to electronics should correct me if I'm wrong.

    Hope I get an email soon, saying "Order shipped"

  86. #86
    mtbr member
    Reputation:
    Join Date
    Jan 2011
    Posts
    270
    Quote Originally Posted by JezV
    George, I haven't a clue what a FET is but if you're happy then I'm happy
    *chuckles* For my sake it's dragging back memories from a long past degree in electronic engineering. Is one reason I'm interested in all this stuff to be honest!

    Having seen the thread on CPF about the limits of overdriving an XM-L I'm now more than happy with the current tables proposed for the LFlex - thanks George.
    Quote Originally Posted by troutie-mtb
    it will drive 1 xml from 1 or more parallel cells but not 2 series cells
    Correct.

    it will drive 2 xmls from 2 series cells but not 3 or 4
    Also correct.

    it will will be hotter in the early running from fully charged cells
    Indeed it will, as it is dropping more voltage, and power, across itself in order to present the same steady voltage to the LED to operate at the same current.

    does the light then dim when it drops out of reg
    Exactly as Toaster79 says.

    will it be heating up more if using low settings from full cells
    Low settings as opposed to high on the same cells? Shouldn't do. Yes, it is a linear regulator so in effect you're putting a clever resistor in series with the LED. However the decrease in overall current has more of an effect on driver power wastage than the increase in effective resistance value, so the power wasted in the driver should drop at lower settings.

    Michael

  87. #87
    Barrington Womble
    Reputation:
    Join Date
    Aug 2010
    Posts
    31
    [QUOTE=JezV]George, I haven't a clue what a FET is but if you're happy then I'm happy

    Brings back some long faded college memories! A lot of what George is saying makes some sense but I'm glad he's doing all the thinking for me. I'm just glad there is someone out there who is willing to put in the hours to produce a good quality product at a good price who asks & listens to what we want & doesn't take 12 months to get the finished item to market! This is what keeps our interest up & keeps us ahead of the big light manufacturers.
    Last edited by cagliari; 02-13-2011 at 04:22 PM.

  88. #88
    mtbr member
    Reputation:
    Join Date
    Jan 2011
    Posts
    66
    MFJ, I believe you to be right on all counts.

    Your right on the last one, too, but I think you overlooked one small issue: the low settings are regulated by pulse wave modulation (PWM), not resistance. So technically, the heat issue should all but go away at this setting (save and except the FET, but even this will not be working hard at this point either).

    I'm self taught in the study of electrical theory, so please take my ramblings with a grain of salt.

  89. #89
    mtbr member
    Reputation:
    Join Date
    Jan 2011
    Posts
    270
    True, true, good point P220C.

  90. #90
    mtbr member
    Reputation:
    Join Date
    Jan 2010
    Posts
    291
    Glad that some of my posts are entertaining you guys - some more than others

    I suppose I need to explain a few things to try and set straight how the LFlex regulates current.

    Firstly, it uses the PWM hardware to create a D/A. Basically it runs very high speed PWM (>100KHz) and then varies the on/off time (duty cycle) of that 100KHz pulse train. That pulse train is the run through an R/C filter to convert the pulse train into an analog voltage.

    Secondly, the FET never sees PWM, what it sees is the analog voltage (from 0V to 2.5V). As that voltage varies (fed into the Gate of the FET) the FET can be turned from fully off to fully on and many levels in between.

    The FET is in series with the LED. So, now we can treat the FET as a resistor that we can vary from 0 ohms to infinite (or close to it) ohms.

    So, one of the questions in this thread is power dissipation and how it varies from high current to low current.

    It's pretty easy, consider worst case, we have a freshly charged li-ion cell at 4.2V and an XML with a Vf of say 3.2V at 3A and lets say 2.7V at 1A.

    So, at 3A drive we have 3A running through the LED and through the FET. The FET drops the battery voltage from 4.2V to 3.2V. So, we have 1V at 3A which means 3W being dissipated in the FET and 3.2Vx3 = 9.6W being dissipated in the LED.

    Now at 1A we have 1A running through the LED and through the FET. The FET drops the battery voltage from 4.2V to 2.7V. So, we have 1.5V at 1A which means 1.5W being dissipated in the FET and 2.7 x 1 = 2.7W being dissipated in the LED.

    Obviously with the battery (li-ion) dropping to 3.7V pretty quickly, the power losses in the FET will be less, since it will have to only drop 3.7V to the LED Vf (at the specific LED current).

    Anyhow, hope this helps shed some light (tee hee) on how the LFlex works. As mentioned above, think of the FET as a variable power resistor. The firmware adjusts the resistance in real time to maintain a constant current flowing through the resistor and hence the LED as the battery voltage varies and the LED Vf varies.

    cheers,
    george.

  91. #91
    mtbr member
    Reputation: troutie-mtb's Avatar
    Join Date
    Sep 2007
    Posts
    3,062
    Thanks George Its nice to know how stuff works

    though the fact that it does work is more important
    and looking forward to some arriving in the UK

  92. #92
    mtbr member
    Reputation:
    Join Date
    Oct 2009
    Posts
    179
    Quote Originally Posted by georges80
    Glad that some of my posts are entertaining you guys - some more than others

    I suppose I need to explain a few things to try and set straight how the LFlex regulates current.

    Firstly, it uses the PWM hardware to create a D/A. Basically it runs very high speed PWM (>100KHz) and then varies the on/off time (duty cycle) of that 100KHz pulse train. That pulse train is the run through an R/C filter to convert the pulse train into an analog voltage.

    Secondly, the FET never sees PWM, what it sees is the analog voltage (from 0V to 2.5V). As that voltage varies (fed into the Gate of the FET) the FET can be turned from fully off to fully on and many levels in between.

    The FET is in series with the LED. So, now we can treat the FET as a resistor that we can vary from 0 ohms to infinite (or close to it) ohms.

    So, one of the questions in this thread is power dissipation and how it varies from high current to low current.

    It's pretty easy, consider worst case, we have a freshly charged li-ion cell at 4.2V and an XML with a Vf of say 3.2V at 3A and lets say 2.7V at 1A.

    So, at 3A drive we have 3A running through the LED and through the FET. The FET drops the battery voltage from 4.2V to 3.2V. So, we have 1V at 3A which means 3W being dissipated in the FET and 3.2Vx3 = 9.6W being dissipated in the LED.

    Now at 1A we have 1A running through the LED and through the FET. The FET drops the battery voltage from 4.2V to 2.7V. So, we have 1.5V at 1A which means 1.5W being dissipated in the FET and 2.7 x 1 = 2.7W being dissipated in the LED.

    Obviously with the battery (li-ion) dropping to 3.7V pretty quickly, the power losses in the FET will be less, since it will have to only drop 3.7V to the LED Vf (at the specific LED current).

    Anyhow, hope this helps shed some light (tee hee) on how the LFlex works. As mentioned above, think of the FET as a variable power resistor. The firmware adjusts the resistance in real time to maintain a constant current flowing through the resistor and hence the LED as the battery voltage varies and the LED Vf varies.

    cheers,
    george.
    Translation for non-EEs:

    lflex + 3.7v Li Ion + XML = lots of light
    lflex + 7.4v Li Ion + (2 x XML) = lots and lots of light

    Stop me if I'm being too technical...

  93. #93
    mtbr member
    Reputation:
    Join Date
    Jan 2011
    Posts
    270
    Quote Originally Posted by JezV
    Translation for non-EEs:

    lflex + 3.7v Li Ion + XML = lots of light
    lflex + 7.4v Li Ion + (2 x XML) = lots and lots of light

    Stop me if I'm being too technical...
    Very good summary!

    George, could the PWM signal be sent direct to the FET and use the FET as a high-speed oscillating switch, switching between on and off to regulate all brightness levels (even full power)? That way you wouldn't have the resistive losses across the FET, although you would of course need to be driving the gate current through all the switching. I don't know what that would do to the LED though - mind you, could you put an R/C filter after the FET? Might be a bit big though as you might need a sizeable capacitor I guess.

  94. #94
    mtbr member
    Reputation:
    Join Date
    Jan 2010
    Posts
    291
    Quote Originally Posted by mfj197

    George, could the PWM signal be sent direct to the FET and use the FET as a high-speed oscillating switch, switching between on and off to regulate all brightness levels (even full power)? That way you wouldn't have the resistive losses across the FET, although you would of course need to be driving the gate current through all the switching. I don't know what that would do to the LED though - mind you, could you put an R/C filter after the FET? Might be a bit big though as you might need a sizeable capacitor I guess.
    Umm... we're talking LEDs here, not hot wire lights....

    With a LED if you turn the FET on 100% then you will have direct battery to LED connection, i.e. fresh 4.2V applied to a LED with say a 3.2V Vf at 3A... With 4.2V applied from a battery capable of driving many amps you would blow the LED (likely >> 10A would flow).

    A sizeable capacitor will do nothing with a battery connected directly (FET 100% on). Talking about R/C filtering or big caps or..... may as well throw an inductor in there too and let's call it a buck driver Hmm - let's call it a b2flex or h6flex

    The simple summary is that if you want a scheme like you outlined - get a d2flex - that is EXACTLY how a d2flex works. It dims by varying PWM duty cycle - but it does NOT regulated the current (only the average current).

    Here's a simple example of what PWM with a 4.2V freshly charged battery (even worse with say 2P) feeding a low Vf XML:

    Pretend your head is the LED, pretend a baseball/cricket bat is the battery going through the PWM FET. By varying the PWM duty cycle you are varying the duration of your head being hit by the bat - but NOT the intensity. The intensity will be VERY high while the battery is at 4.2V. This is the d2flex example.

    Now, instead, we have the lflex, it varies the drive current (by dropping excess voltage across the FET that is acting as a resistor), think of the FET resistance as a pillow on top of your head. Your head will now only see a mild 3A constant pressure being applied versus high impact short pulses of the d2flex.

    You may now go and rest your head

    The lflex functionally works like an AMC7135 that seems to be oh so popular with a lot of folk building single cell/single LED lights. The AMC7135 burns excess voltage resistively across its internal transistor - basically we're talking linear regulator. The lflex has several advantages over the AMC7135: a) the lflex output current is adjustable from 50mA to 3.5A b) lflex uses a FET so the dropout is much smaller c) the lflex FET has an excellent thermal path through the PCB for cooling. To achieve 3A (or 2.8A for the AMC7135), 8 x AMC7135's are needed - so it would be a double stacked board with all sorts of cooling problems due to the stack. But functionally, they act the same - excess voltage -> resistive heat.

    cheers,
    george.

  95. #95
    mtbr member
    Reputation:
    Join Date
    Jan 2011
    Posts
    270
    *chuckles* Knew there had to be a reason! Thanks George.

    *goes and tends sore head after baseball bat experiment*

  96. #96
    mtbr member
    Reputation:
    Join Date
    Apr 2010
    Posts
    143
    Quote Originally Posted by georges80
    Here's a simple example....

    Pretend your head is the LED, pretend a baseball/cricket bat is the battery going through the PWM FET. By varying the PWM duty cycle you are varying the duration of your head being hit by the bat - but NOT the intensity.
    That's the greatest analogy for a PWM circuit I have ever heard. Wow.
    "I applaud your stupid idea because it is genius." - Eric Sovern, Surly

  97. #97
    mtbr member
    Reputation:
    Join Date
    Jan 2010
    Posts
    291
    Very close to releasing the LFlex driver. So, in preparation I've put up the preliminary product/tech/manual onto my website:

    http://www.taskled.com/lflex.html

    and

    http://www.taskled.com/techlflex.html

    It's a work in progress, so more information to add. If you notice any obvious errors I'd like to hear of them.

    cheers,
    george.

  98. #98
    mtbr member
    Reputation:
    Join Date
    Jul 2009
    Posts
    19
    Quote Originally Posted by georges80
    If you notice any obvious errors I'd like to hear of them.
    Hey George,

    I think you have a typo on the product info page...0.780" ~ 19.8mm...you put down 22 by accident.

    Can't wait to order a couple. How would these work in a triple AAA (3.5v ?) Alot of small flashlights use that config...could be a simple mod.

    Bryce

  99. #99
    Dirt Deviant
    Reputation: savagemann's Avatar
    Join Date
    Jan 2008
    Posts
    3,693
    Bryce, I took it as meaning less than 22mm. The way that the boards are manufactured, they all come off of one large sheet of fiberglass, and each board is broken away from the master sheet. Kind of like a saltine cracker.
    Well, where the boards are broken away from eachother, there can be some hair left behind that will make the board measure a big larger. This is easily fixed in about 10 seconds with a small file. Hence the <22mm.
    I'm not sure if I explained that right, but it sounds good to me....= )

    George, looking good. Can't wait.
    Look, whatever happens, don't fight the mountain.

  100. #100
    mtbr member
    Reputation:
    Join Date
    Jan 2011
    Posts
    66
    "Yessss...."

    -Napoleon Dynamite

Page 1 of 2 12 LastLast

Members who have read this thread: 1

Posting Permissions

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