7135 drivers with higher voltages revisited

Rufus,

Best of luck to you. Once upon a time, I tried exactly what you are after here, and personally came to the conclusion that it could not be done. My light was 5x MCE wired as 4p5s, powered from 16 Eneloop AA in series.

The problem as you have probably noted with using the Download method, where the mcu of the mode board is powered in series with the LEDs, is that on any single mode, you could negotiate the difference between the Vf of the emitters at the desired drive current, and the Vbatt under that load with a resistor of the proper value to burn off the excess and just feed the mcu a voltage that’s within it’s acceptable range. The problem with that design is that under different modes, the Vf of the emitters changes so drastically from one drive current to another that a single valued resistor ends up not leaving enough voltage for the mcu on high and way too much voltage on low. My mode board would either overheat nearly instantly on low or not operate at all.

In an attempt to uncouple Vin on the mode board from the Vf of the LEDs at various currents, I selected much higher valued resistors and fed the MCU directly from the switch instead of from the LED- of the N-1 LED. Among the resistors I had available to me (salvaged from parts discarded while modding) I recall finding that I still encountered the same problem as with the MCU in series with the LEDs… some times the light would light, but I couldn’t get the full range of modes from Low-High to work.

I think I theorized back then that full extent of the voltage sag between essentially open pack voltage on low and a 3A load on high, the 16AA pack sagged ~7V (from 23V open down to 16V under load on high) and that was just too much, what with the operating range of the MCU being only about half that (something like 2.5v-6V).

If only there were an easy way to change the value of the resistor responsible for stepping down input voltage to the MCU, as the modes change. It would require the lowest value on high mode and the highest value on low mode.

If you manage to discover an elegant way to do that, my old LED Ultra Stealth Light might finally get some modes after all.

So far using the 7805 regulator, I’ve been able to run a DX 7612 attiny board at 1.4A and get hi, med, low, strobe, and sos modes. Dr. Jones suggested that chip as a more viable alternative to resistors to control the input voltage to the processor for the very reasons you mentioned. Using that chip allows me to forgo using an led to drop the voltage in front of the board. The led’s are all in series together. V+ to led + to 7135’s to ground. The processor essentially runs on a separate circuit: V+ to 7805 to processor to Vdd 7135’s to ground. Any resistor needed would only affect the voltage to the high current side of the system as the 7805 would continue to feed the processor with the correct voltage. The 7805 chips are available on ebay and at mouser. Mouser even has a LDO version that would allow 2s2p setups with 2 li-ions and maintain voltage regulation down to 6V.
I would love to see this mod done in a flashlight as my motivation come from bike lighting where higher voltage is sometimes preferable to high current. Because of that I lack the specific knowledge that would apply to handheld lights. I expect there are some out there waiting for this to happen in a 4-D Mag but it won’t be me doing it.

Since I was last at home and was able to verify the demise of the 7135 ships on the heat sink(not from over-voltage but from overheating during a second reflow), I did some work on a new heat sink. Of course, not being able to leave well enough alone, I changed it a bit and experimented with adding fins.

It’s a nifty little job but although it should be fine for multi-emitter optics it’s to big to use with an single XML in the lamp heads that I use so I made this one as well.

For scale the circular part of the first sink is the same size as a 17mm board and the two parallel rails on the second are ~ 9mm overall width x ~ 25mm long. Either sink would be connected through the shell to more cu exposed to moving air. The Ledil EVA tir optics I will use are 17mm deep and the aluminum shell is only 35mm long on the longer bottom side. The switch occupies the area above the centerline of the screw holes in the open back of the shell and rests on the aluminum “hook”.
One other thing I should mention is that with the 7135 chips removed, only about half of the board is still needed, allowing for further space economy(similar to the old pwm boards).
I like the way the fins turned out and will likely use that process to shed heat outside the shell. It was done be using a step drill to punch holes in thin(no idea what guage) copper sheet the size of the “pill”, and separating them with hoops of 24guage(?) solid cu wire. I coiled the wire around pipe a bit smaller than the pill and slipped the coil over the pill to measure and cut the hoops. I have access to a high output soldering iron so I may try to solder the chips onto one pill with that and attempt reflow again with the other. A realistic goal for these would be to burn off 2V over the combined Vf of the led’s anything above that would be gravy.

Yes, of course, how could I forget (I was actually nodding off at the keyboard last night when I wrote…), the problem of the voltage range isn’t only with the MCU but with the 7135s as well. It seems that’s what’s needed there is something like 7135s for the 7135s. I wonder if they can be run two in series, with the ground of the first connected to the LED- of the last, and the last’s ground connected to ground as usual (and the MCU connected to the VDD of all of them).

I’m hoping that for a 2s2p mod, with the 7135’s sinked separately from the LEDs they might then be able to handle the extra ~2V. For 3s or more I agree that a resistor would be needed but a 3-position switch(on w/resistor-on w/o resistor-off) Judco makes them(Digikey)you could have your cake and eat it too. The data sheets indicate that with ~1000mm^2 of heat sink area a chip can drop up to 4V excess. Given the lamp size I have to work with I don’t expect to do that well but 2-3V is conceivable. Post #45 shows how the 7805 chip replaces a 7135 to regulate Vin for the mcu. I ordered them from eBay but they’re also available from mouser.

At one time I was a Boy Scout and some vestiges of the motto still cling to my behaviour. I feel compelled to admit that one xml and 8 additional 7135 have given their all for the sake of knowledge. All was going well, I had 8 chips soldered into a piece of cu being controlled by my special controller board(no 7135’s, just atmel and 7805)powering 4 xml’s. With 10 series nimh’s I was getting all modes and 1.4A on high(Vbatt < Vf). I added 2 more nimh’s for a no load voltage of 15.3V and used a 1/2 ohm 2W resistor in series with the leds and measured 2.08A on high and the chips were still cool so I removed the resistor from the circuit. I started on low mode, all fine, chips cool, the med, still good, chips cool, when I went to high first one xml went out then the modes still functioned but at very low levels. I believe when the xml went, the remaining voltage to the 7135 chips jumped up to ~5V and they went poof!
Here is a picture of the driver setup showing the chips loaded with the ground post and Vdd wire in place
And here with the pwm controller added(I cut off the portion of the board that had 2 empty 7135 pads)

So far in all of the tests the controller portion has worked perfectly and has survived each time. On the first sink I added some extra cu after the chips were loaded. In this case I forgot to solder the ground wire to the sink in advance of loading the chips and this may have proven costly. Soldering to the pins after the chips are loaded is no problem but any soldering done on the sink needs to be finished first before the chips go on. In any case, my reflow skills need improvement(as does my memory). I do not know for sure if this has any relavance but some of my “failure analysis” has noted a lack of current control in the 7135 chips afterwards. I have removed the wires and cookies(cooked chips, lol) on the sink and will try to reload them this week. I may just be making excuses as I am reluctant to call this idea “busted” but this is certainly a setback.

The thing that appealed to me most about your end goal was that it could add some flexibility to a simple, affordable solution. At this point, it is no longer simple, and the heatsinking makes it less affordable. I’m compelled to ask, is it just the challenge of it, or is there some reason for a strong preference for a 7135 based solution instead of DC-DC buck circuit?

Probably just stubborn and don’t want to quit yet. Here are the reasons I started investigating these boards: I’ve used the KD super P7 driver but they suck for efficiency. Sharkbucks cost twice as much as the KD driver and don’t work they way I would like and the B-flex and Der Wichtel are cost even more and are too big. The Sharkbucks is the only 3A buck driver close to the size of a 105C but if I can make this work it will be small, close to 90% avg efficiency, still cheap($4/chips, $3/ board, $.50 regulator, $3 resistors, and the sinks are made from scrap cu) so similar to KD but more efficient and in spite of the trials and tribulations I’ve gone through it is still simple to do.
If you are willing to have an on-on-on-off switch with 2 resistors(1.2 ohm, .6ohm), you can use a 3A 7135 board with just the voltage regulator addition to drive 4xml in series. That part is done and proven. I’m being stubborn about the resistors(don’t want them) and my testing is to see how much heat sinking is practical for ME. I’m not yet adept at reflow and this has cost me some chips and aggravation but I’m not quite ready to cash it in yet. Who knows, it may end up that it works fine but I’m just doing it wrong.

If it is the case that these spikes are the/a problem rather than my at present poor reflow skills then the Zener diode might provide a solution. Would a resistor also be needed to limit the current. I don’t really like the idea of a resistor permanently part of the drive circuit this way.

Would a capacitor be able to capture this power and bleed it off through a resistor during the Zener-inactive cycle.?

Could the Zener output be connected to the Vdd trace of a separate 7135 with another 3.3V Zener back to the 7135 output turning wasted power into light?
The duty cycle migh need reprogramming to maintain desired light output levels.

If it seems I am easily diverted you’re right but please remember that I have a definite goal in mind. I just want to learn things along the way.

to be honest, after using both 7135 based and Taskled drivers, if you’re going to be spending $40+ in LEDs and similar on a battery, $3-40 on a decent driver is a pretty good investment. The programmable UI, temp and undervolt warnings and amazing customer service are easily worth the premium over a $5 driver. If you’re really strapped for space, why not an Lflex with the 4 XM-Ls run in 2S2P with the driver set at 3.5A for 1.75A per LED? That’s right in the middle of the XM-Ls efficiency curve and any housing that won’t fit a b3flex also won’t dissipate the 40W plus that 4 XM-Ls @3A will generate.

I can see the appeal in forging your own path, but sometimes it’s worth ponying up the extra and spending that time relaxing or using the light instead :wink:

It’s not really about the money at this point. I have 2 quad xpg lights that have Maxflexs installed, 2 tri xre lights that use the KD super P7 driver, and another with a B3flex partly awaiting LEDs. Lflex is an option even though it has the same heat issues as 7135 based drivers(uses an fet instead I think) but all of Georges’ flex drivers use a momentary switch that requires separate wiring of limited length. The XML example is more likely to end up in a dive light where heat is not a problem. In the end , I’m doing this because I want to and my curiosity has not succumbed to my attention span.

Here are the latest results from this set up:
12V battery(10-sub c nimh cells in series).
3-XML U2 LEDs in series.
1-DX 7612 board with 7135 chips removed and 1-78L05 installed.
8-7135 chips stripped from 7612 board and 1A 1-mode boards and reflowed to custom (lol) heat sink.
1-Judco switch with 2-1ohm resistors in series.
On low using both resistors I measured .102A, 13.45Vbatt,13.43V after resistors, 7.3V across the 3 emitters, and 5.6V across the 7135’s.
On medium with 2 resistors I measured 13.40Vbatt, 12.29 after resistors, .707A, 7.6V across emitters, and 4.1V across 7135’s.
When I switched to high I read 2.37A but the resistors began to smoke so I switched back to low.
It has been running continuously on low since then which is when I began typing this post(~1 hr) so I don’t think pwm voltage spikes are a problem at this output level.
The resistors were both supposed to be .5 ohm resistors but some readings don’t seem to be consistent with that value. Also a cheap DMM. Of necessity I’m getting better at both stripping and reflowing chips but I (and you as well, sorry to say) will be waiting for the full power test until I have higher wattage resistors in place. I’m going to run it on medium for awhile and let you all know the outcome of that after work tomorrow.

interesting results, looking forward to the High results with the beefier resistors. Looks like the 7135s are getting the right voltage on medium but a rather low Vf - is that normal for that current draw (0.7A)?

I hear what you’re saying about the pull of curiosity! You’re clearly not doing this out of necessity :slight_smile:

The goal at the outset was for 4 LEDs powered by 4 Li-ion and though I’m very pleased with how the 7805 is working I’m not overly confident about the 4x4. 2s2p I think is very plausible and 3x3 would be great. What to me makes this idea questionable is if you forget where a switch is set you could blow all the 7135’s by accident. Even if I can make it work, this represents a serious flaw. Oh well, one set of chips a week is less than popcorn money. To some I’m sure this seems a pointless exercise but it has been good practice and has taught me how to make a custom heat sink for the chips allowing installation in odd places. In any case it ran without incident on medium for another 20 minutes by which time Vbatt had dropped to 9.3V(recovered to 11.1V).

Also, the voltage measurements may be suspect since I am measuring a pwm signal which may not correspond to a current vs Vf curve. I use a lot of big words but I really don’t know s**t. Time for a beer.

I’m going to try this mod on 4 LEDs 2s2p with 8.4 V. Since the 78L05 requires ~2V overhead I ordered some LD2981 very low dropout regulators from Mouser which will stay in spec down to full discharge. Also ordered were some smt 0805 1 and 2.2 micro farad ceramic caps.

I think I may have discovered a relavant tidbit. When I tried this with the pack fully charged to 14.3V the chips died(gates lock open and can’t be dimmed). Remember, on low and medium I was measuring 7.3v - 7.6v across the emitters so the fresh pack was exceeding 4V across the chips even with the resistors. Even though the chips don’t even get warm on low and medium, they poof at to high an input V. I went over the 3 nimh/1 Li-ion per led limit. If this is the case, then 4 LEDs on low-med at ~10v and high at ~13.2v would stretch the ability of a resistor to both lower the voltage sufficiently for safe operation on low-med and maintain voltage s above Vf on high. At this point 2s2p might prove more reliable which is the reason for the Mouser order. For my next magic smoke, I’ll reload the sink with fresh chips and run the test again with 9 nimhs. For the record, I still haven’t spent enough on toasted chips to buy even half of 1 quality Taskled or Shark driver.

It may appear that you are talking to yourself, believe me when I say Ive followed this with interest even though I dont understand it. I may learn something if I'm lucky. Keep it up and good luck with the outcome.

Thanks, it does seem that way but I notice that some are reading( or at least looking at) this thread. Watching someone muddle through isn’t how textbooks are written but at least it’s genuine. Posting just the end result would be better as a “How Too” guide and I hope to get to that point but this is more of a journal.

I follow your so-called "muddling through" with scrutiny every day Rufusbduck

And I start to see some relevance to it too so keep it up. You are doing a hell of a job and I hope you will nail it in the end. Sure makes for some exciting reading :-)