Series or parallel for multiple LED's?

[2xTrinity]

I am interested in applying more than 7 (this is the voltage blackout range of the boost drivers I am aware of (22V+)) SSC/Cree LEDs in a single light, to be run on batteries (read "Limited space"). This could be done in strings of series, one long series, or X*parallel...

I would recommend these drivers from DX. I have used them and absolutely love them. They are buck, or voltage step-down only. They appear to be reasonably efficient and are perfectly well regulated on 2x CR123, or 2xLiIon -- anything just so long as the voltage doesn't drop below 3.6 (then it will start dimming). The best part is, you can buy two four-packs for $13 shipped, meaning it's actually cheaper to buy EIGHT of these drivers than to buy a single similar driver elsewhere!

If space is permitting, wire one driver to each LED, and connect all the drivers in parallel to your batteries. Since each LED has its own regulator, current will be evenly divided between all of them. Also, if you have an LED failure or driver failure for some odd reason, your other 6 LEDs will work undisturbed. The only drawback that I can think of is that you won't be able to do variable brightness using external PWM switchers or potentiometers like you can do with Shark or Fatman drivers. If you want a quick and easy way to vary brightness though, you can simply turn some emitters off and on with separate switches.

 

[MrAl]

Hi there,

Yeah it is interesting to see this old thread again.

The problem with connecting the LEDs in parallel is that if the characteristics
of each LED is not matched serious overcurrent could occur for some LEDs
and not others. It is always a good idea to use some resistance in series
with each LED and make sure the current isnt too high to each LED.
It's a pain to have to connected even one resistor in series with each
LED but it is worthwhile in the long run. Many times you can get by with
a tiny little 1/8 watt resistor too for each LED, which keeps things small.

In the EE course i talked about earlier in this thread i had done an analysis
on LEDs in parallel based on a theoretical model of a 20ma LED, and allowed
the characteristic of one LED to be drastically different than the others but
still something that could actually occur in real life.

Another technique that could help is to match the LEDs by voltage.
To test the voltage, run 20ma (or whatever you plan to operate each LED
at) through all the LEDs and measure the voltage. You can then bin the
LEDs according to what voltage you measure say within 0.05 volts.
LEDs that measure within 0.05 volts of each other might run ok in parallel,
but then again the long term stability is difficult to predict because it is
known that the characteristic voltage of an LED changes over time. If all
the LEDs change the same you get lucky, but if not perhaps one LED will
burn out. If it goes open then the rest of the pack have to absorb the extra
current. Just some things to think about.

Oh yeah, another nice benefit to connecting a small resistor in series with
each LED is that you can then easily measure the current through each LED
after you run the whole bunch up. Simply measure the voltage across
each resistor and divide by the resistors value, and this gives you the
current. For example, say you use 10 ohm resistors for each LED, and you
are connecting 10 LEDs total, and you measure 0.2 volts across the resistor
going to LED #1 and 0.15 volts across the resistor going to LED #2... this would
mean LED #1 has 20ma going through it and LED #2 has 15ma going through it.
You would repeat this for all 10 LEDs.
No need to insert a meter, which could cause a large error in measurement anyway.

 

[Bimmerboy]

This discussion has got me thinking about how hard I'm pushing some of these LED's, and that I'd prefer to not burn them out. I'll likely do some current measurements tonight so there's at least some idea of how much torture I'm putting my mods through. Could be scary.

Over time, I'll be considering some form of regulation for most, or all of these DD'd and paralleled mods, so as to ensure long and healthy lives for them... you never know when one might actually need to be used for any length of time. Each one will have some different parameters to work within in terms of space, number of LED's, and power, so it should be a great learning experience.

Xiorcal - Your post reminded me of a quote someone I knew used to say... *with British accent* "It's all a question of ratios!". I forget where he got the quote from. As applies to your question, it's a balance of design parameters.

For example... What kind of host are you planning to use, and how creative can you get with stuffing different power options into it? Starting from that point should help many other decisions fall into place, or at least eliminate what's not possible.

altis - Gee, thanks for sending me off into an abyss of endless Wikipedia distraction. I barely made it back alive. It all started with "ring counter". :laughing: However, your idea sounds really interesting. Interesting enough to try on some future mod, if I get to the point of understanding that kind of circuit. Could you give some more detail on how you'd go about doing this?

2xT - While those buck boards in particular won't work for the mods I'm currently (pun intended) looking to convert to regulated, it's a great idea, and will definitely look into the other DX drivers. Also, there was a recent discussion somewhere around here, where it was said that two of them can be connected (in parallel, I think) to create a buck/boost. Could give rise to some effective, and cheap possibilities.

MrAl - Hello there, and good to see you again! Although the already built mods might not be able to take advantage of the great info in your reply, one of the upcoming ones certainly can! The design will easily allow for it. Also, I recently came across your old LDO-CC thread, and would like to make use of that circuit for something.

BTW, awesome tip about measuring V across a resistor to solve for current. Too cool!

 

[altis]

Bimmerboy, sorry for distracting you!

I, too, have been having some more thoughts about this. Obviously, the simplest way to do this is to have one regulator for each LED - as previously suggested (and they're pretty small and cheap too).

However, this denies the beauty of the problem. For another solution that uses the minimum of components I suggest the following:

Think of a classic buck regulator. From the supply there is a switch which goes to the top of a flywheel diode. This is followed by a coil then a capacitor to ground and the load. Something like this:

By turning this round and introducing a number of switches we can create a simple, multi-channel current regulator. Instead of the load, we could just have a low-value current sensing resistor. We can probably do away with the capacitor too. Then add several switches, commoned together at the junction of the diode and the coil. Then connect each LED in the supply to each switch (near Iin in the picture). Simple - job done.

Apart from controlling the switches of course! You'd need to switch each one on and off in turn round the ring. You could probably use any common switch-mode regulator IC and have some sort of counter to route its output to the apropriate switch. Or simply build your own. It should be a fairly simple job to include a dimmer too.

Alternatively, you may be able to make something out of one of the recent multi-phase ICs that are used to regulate the CPU core voltage on modern motherboards.

 

[TorchBoy]

Interesting how things have changed since this thread was started, and interesting too for the issues that haven't changed in that time. The price of LEDs has dropped so much (and I've found cheaper sources) that it would now cost about the same to connect a dozen Crees in parallel instead of a dozen 5mm LEDs. (OK, that was four years ago, not two.)

The main issue I have now with connecting up lots of stroppy LEDs in parallel (or series) is heat.

 

[xiorcal]

Altis:
I don't suppose this (which I found earlier) http://www.freepatentsonline.com/6621235.html has any resemblance to the "ring counter" method you were talking about.
I don't quite get what you were on about with "If there were N of them (LEDs)then each would experience N * max-current but for only 1 / N of the time", as if there were 9, running at 350mA, that would indicate each would experience >3A for some (1/9th) amount of time, and I thought that was magic smoke territory...

2xTrin:
Those drivers are interesting, but I was thinking dimmable. If you used those, and just turned LEDs off, you wouldn't get those wonderful gains in efficiency from running many LEDs at lower currents.

MrAl:
Thanks, I gathered the resistance variance thing. I think it would be a little expensive to do voltage matching, I wouldn't want to buy that many emitters. Wouldn't using resistors in series with the LEDs (I presume this is to make the resistances "chunkier", and therefore proportionally more current sharing) defeat the energy efficient nature of the LEDs?
Nice trick for current reading.

Bimmer:
I was considering a MagC or MagD, but I'm not sure. I may just make a crude heat-sink light, with a handle. If I used a Mag, I'm not sure whether I would put batteries in it anyhow. Multiple battery configurations seem annoying, and AW's C cells would cost a bomb in the form of adequate charger. I've been looking at Li-ion packs on battery-space, and thinking how much easier it would just be to strap one on my waist. Wouldn't have to worry about heat degrading the Li-Ion then.
Drive-wise, Probably 9 LEDs at <1W (~350mA) variable down.
The Shark would do it in 3 by 3 parallel series.
It would almost be worth just running 6 in series (from max 22V), a nice 3.66V max per LED, and down-dim control, however there is an optic which channels 7 LEDs, and I would centre three LEDs on the middle optic to give flood, rather than leave it blank.

I also had an interesting idea about a CPU fan in the round of the battery compartment, dragging/forcing air through a heat-sink, but that is for a different discussion, and at 1W each, not necessary (?). Wouldn't be Waterproof

-Xiorcal

 

[TorchBoy]

Altis: ...
I don't quite get what you were on about with "If there were N of them (LEDs)then each would experience N * max-current but for only 1 / N of the time", as if there were 9, running at 350mA, that would indicate each would experience >3A for some (1/9th) amount of time, and I thought that was magic smoke territory...

It certainly is, but Altis also mentioned "some beefy current sinks - one for each LED" which I took to be capacitors to even out the current flow through the LEDs. But I presume a resistor would still be needed in series with each LED so most of the current did actually go through the capacitor.

Anyway, driving LEDs at higher currents for less time isn't the most efficient way to drive them.

Has anyone checked how even the operating voltage of Crees are?

 

[altis]

Yes, the circuit I was suggesting would certainly be pulsey. This is not a problem for traditional LEDs but I'm not sure about these new high-power devices. Bear in mind that this circuit will switch at tens if not hundreds of kilohertz and it's the RMS current that heats things up.

You could even out the current by storing it in an inductor - one for each LED. That way you'd end up with something a bit like a Cuk regulator. But if you carry on like this then you'll end up with something very complicated and you may as well just use one regulator for each device - as suggested above.

Some more info here:
http://www.steve-w.dircon.co.uk/fleadh/mphil/history.htm

By the way, a 'current sink' is just a big fat transistor. And that patent is for a multi-output current mirror. This is a linear device so will not be at all efficient.