Sharing Current Between Parallel COB LEDs. Transistors? Circuit Design Help!

Hi CPF! This is my first post so I hope I'm in the right place. I wasn't sure if this would do better in one of the custom forums so feel free to move this thread or whatever.

So I've got some high powered COB LEDs. They're rated at 100W which is made up by about 3A at something like 35V. They came with drivers to match which I suppose are some kind of constant current source and are tuned to run one LED each at full power.

My plan however is to run the LEDs at approximately half power by running two in parallel from one driver. My hope is that this will make cooling easier by reducing heat per an LED, increase the lighting efficiency and prolong the life of the LEDs. Further more this will allow me to run two different spectrums from the same driver - one warm white, one cool.

I also want to be able to control how the current is shared between the two LEDs. If I could conjure up a circuit with a fully analogue control to adjust the balance of current between them that would be fantastic. I suspect however that that it would be easier and cheaper to design a circuit with say three balance settings. For example these may split the 3A by approximately:

• 1A to the cool white LED and 2A to the warm
• 1.5A each
• 2A cool, 1A warm

Of course the easiest way to achieve this would be to use resistors in series, however efficiency is a primary concern of mine so I'd rather not go down this route. I am not too worried about precise amperages but it needs to be fairly cheap. So I've been reading about current controlling circuits using transistors. This is all quite new to me however and I'm struggling to comprehend it all well enough to come up with a solution to my problem. Perhaps to someone more experienced it is quite straight forward?

If anyone has any ideas for either an analogue current balance control or a number of discreet settings I'd really appreciate they're input.

Many thanks,
Gully.

gully.moy said:

Hi CPF!

Click to expand...

Hello!

gully.moy said:

My plan however is to run the LEDs at approximately half power by running two in parallel from one driver.

Click to expand...

I need some more coffee, but wouldn't two LED in series from one driver would do what you're saying?

gully.moy said:

My hope is that this will make cooling easier by reducing heat per an LED, increase the lighting efficiency and prolong the life of the LEDs.

Click to expand...

Reduced power would do this, yes.

gully.moy said:

Further more this will allow me to run two different spectrums from the same driver - one warm white, one cool.

Click to expand...

Maybe so, maybe no. Different LED types have a different Vf, aka "forward voltage." Not saying it won't work, just one will be brighter than the other. Led response to current varies among the various led types out there.

gully.moy said:

I also want to be able to control how the current is shared between the two LEDs. If I could conjure up a circuit with a fully analogue control to adjust the balance of current between them that would be fantastic.

Click to expand...

You do that with two drivers. Been done before. Not so bad. This will also allow you to run each one properly if you're going to try color/tint mix.

gully.moy said:

I am not too worried about precise amperages but it needs to be fairly cheap.

Click to expand...

Drivers for regular flashlight applications of LED are only a couple of bucks. $2-$4 or so.

gully.moy said:

So I've been reading about current controlling circuits using transistors. This is all quite new to me however and I'm struggling to comprehend it all well enough to come up with a solution to my problem. Perhaps to someone more experienced it is quite straight forward?

Click to expand...

No problem. We had a warm happy thread all about it here: http://www.candlepowerforums.com/vb/showthread.php?394834-battery-efficiency-increased-runtime-mod

You could come back and ask questions in this thread, if you like.

gully.moy said:

If anyone has any ideas for either an analogue current balance control or a number of discreet settings I'd really appreciate they're input.

Click to expand...

That'd be based of the two driver boards you selected and the modes that they offer. No big deal.

more_vampires said:

I need some more coffee, but wouldn't two LED in series from one driver would do what you're saying?

Click to expand...

It's my understanding that if I connected them in series they would have to share the voltage. Since my drivers are only rated at something like 30-36V they wouldn't supply enough voltage to meet the voltage drop of two of these LED modules in series. Does that make sense?

more_vampires said:

Maybe so, maybe no. Different LED types have a different Vf, aka "forward voltage." Not saying it won't work, just one will be brighter than the other. Led response to current varies among the various led types out there.

Click to expand...

I believe that the Vf listed on the sellers page is the same for the warm and cool versions. I'd have to check it, but as long as it's similar I'm not too concerned with the exact brightness.

more_vampires said:

You do that with two drivers. Been done before. Not so bad. This will also allow you to run each one properly if you're going to try color/tint mix.

Click to expand...

My very limited understanding of transistors would suggest that it should be possible with one driver given the right circuit.

more_vampires said:

Drivers for regular flashlight applications of LED are only a couple of bucks. $2-$4 or so.

Click to expand...

Remember I'm not talking about flashlight LEDs here. These are high powered COB LEDs running 100w per a module at around 35V. The drivers are not so cheap and it's hard to find ones with the correct voltage/current combinations at the right price.

more_vampires said:

No problem. We had a warm happy thread all about it here: http://www.candlepowerforums.com/vb/showthread.php?394834-battery-efficiency-increased-runtime-mod

Click to expand...

Thanks for this, I'll check it out in a minute.

And thanks for getting back to me :thumbsup:

more_vampires said:

I need some more coffee, but wouldn't two LED in series from one driver would do what you're saying?

Click to expand...

I suspect the required voltage for running the two in series would exceed the maximum voltage range of the constant current driver, given that these drivers were sold along with the LEDs. But yes, normally a driver will increase its voltage until the desired current flows through the circuit.

more_vampires said:

Not saying it won't work, just one will be brighter than the other.

Click to expand...

Really hard to say. One might be a little brighter than the other. The only way to know is to try it.
There's no risk to the LEDs here.

Since the LED chip is 35v, it is comprised of ~9 diodes, so I would think that any voltage bin differences would mostly average out.

It can be problematic to run LEDs in parallel for several reasons, but it can work in some situations.

You're right that running them in parallel is the correct way to achieve what you want.

You're also correct that switching in and out resistors is one way to do it, and that it can also be done with transistors. But using current control techniques with transistors is no more efficient than using resistors.

If the Vf of the two leds is the same (or at least pretty close) then it would be reasonable to run them in parallel. Even if the specs are the same, there's no guarantee that any two individuals will be at close to the same point in the range.

Assuming you have two good candidates, the most efficient way to run them would be to use PWM. Basically you put a transistor in line with each LED, and turn the transistors on and off at high frequency. Say you want slightly warm white. So leave the warm one on all the time, and turn the cool one on and off. Say it's on half the time and off half the time. When it's on, the two LEDs share current. When it's off, the warm one gets all the current.

So the warm one would be on 100% for half the time and 50% half the time, while the cool one would be on 50% half the time. Want it a little cooler? Leave the cool one on 60% of the time, or 70%. Want it really cool? Reverse the positions and turn the warm one off.

You could do this with a single PWM module, and a switch to select which LED gets controlled. Then you'd have warm to neutral on one switch setting, and neutral to cool on the other.

Anders, DIW, thanks for setting us straight guys.

This is my absolute favorite kind of thread. The experts showed up!

Thanks DIW, that seems like just the kind of thing I'm looking for. Although I'm surprised to hear that transistor circuits are as inefficient as resistors since I've read otherwise providing the components are chosen carefully. But I'm not too worried about that because I like the sound of Pulse Width Modulation anyway for it's efficiency and [virtual] analogue control.

Reading about it though I'm still a bit lost in new concepts, and therefore struggling to understand exactly what components I need. Most of the PWM modules I can see for sale list output values as a fixed current, and a varying voltage. I want the opposite.

If I understand the theory correctly, the actual current and voltage would actually both be unaltered by the module when on, it is just the ratio of on time to off time that varies. So I guess this produces an average voltage decrease. But what happens to the current. Would there also be an average current drop? Or do I need a different kind of PWM module for that?

Thanks again.

gully.moy said:

it is just the ratio of on time to off time that varies. So I guess this produces an average voltage decrease. But what happens to the current.

Click to expand...

Actually an average current decrease. Current is the amount. Voltage is like the "pressure". Basic electronics.
To go through two LEDs in series, it takes twice the voltage. The current is the same because it is the same current going through both of them. To go through two LEDs in parallel, it takes twice the current. The voltage is the same because it is not any harder for the current to go through either of the LEDs. The voltage potential does not fall just because the current splits apart and flows through two separate wires*.

It gets more complicated when discussing AC voltage, but I will not go into that here.

*Assuming there is plenty of reserve current available, so this general rule may not apply to static electricity .
Actually, since we are discussing a constant current supply here, the voltage probably actually will be a little lower. Having two resistive loads in parallel (the two LED chips) creates a lower resistance, meaning it will take a slightly lower voltage to drive the rated current through. Nothing you need to worry about though.

Yeah sorry, I was confused by motor speed controllers like this one and LED drivers like this one which I found on eBay when searching 'PWM'. I saw the analogue dial and thought they were what I was after but didn't get why they were offering a range of voltage outputs when I want variable current.

Then I confused myself more thinking that when the PWM module was turned off that the voltage would be 0, so the average voltage whilst the PWM was 50% on 50% off would be half of the voltage when on. But of course even when turned off the voltage across the terminals would still be full. If I was thinking of potential difference rather than voltage I might have clicked sooner.

Of course the voltage and current couldn't both half when the PWM was turned down half way because then the power would have to quarter for only 50% less 'on' time.

I'm largely just talking through my thoughts in case anyone else gets stuck in the same rut as I did.

So what I want is something more like this PWM chip and learn how to connect it to my LEDs and a potentiometer for control! Have I got it?

Sorry I'm a bit dense with electrics and need things spelling out sometimes.

Anders Hoveland said:

I suspect the required voltage for running the two in series would exceed the maximum voltage range of the constant current driver, given that these drivers were sold along with the LEDs. But yes, normally a driver will increase its voltage until the desired current flows through the circuit.




Really hard to say. One might be a little brighter than the other. The only way to know is to try it.
There's no risk to the LEDs here.

Since the LED chip is 35v, it is comprised of ~9 diodes, so I would think that any voltage bin differences would mostly average out.

It can be problematic to run LEDs in parallel for several reasons, but it can work in some situations.

Click to expand...

With typical VF off even on a China FAB, unlikely. 10-12. 12 is the most common for COB.

Yeah, when using PWM, concepts like 'average' can be very useful or very misleading. Average current is pretty reasonable when calculating battery life, decent when estimating brighness, not so good when looking at tint, and pretty rotten when calculating power.

For power you should use RMS values. The RMS value of a PWM signal at 50% duty cycle is 0.7071 times the full value. Since it applies to both voltage and current in this case, P = 0.7071V * 0.7071I. It's no coincidence that 0.7071*0.7071 is 0.5, so the equation simplifies to P = 0.5 * V * I, which is exactly what you intuitively knew it should be.

I find it's generally best to consider what's happening when the device is on (and at full power, at least for the moment) for calculating voltage drops, max current, tint etc. Then multiply by the duty cycle to get 'average' power, brightness, battery load.

The TLC5940 is designed to run a number of small LEDs, not a single large one, and it has current regulators for each channel. While you could adapt it to your use, it's kind of overkill. What you need is a small portion of that chip with bigger output transistors. Also, you need to create a stream of data to control that chip. That means you need a microprocessor or special output from a PC or similar (it used to be that you could use the parallel port on a PC for this, but those are pretty non-existent today).

There are some PWM modules around that are purely PWM, with no driver circuitry, which is what you want. Taskled used to make the D2Flex, but has discontinued it. I was sure that led-tech.de had one, but I can't find it now. I'm sure you could roll your own with any of the DIY microprocessor units like Arduino, Raspberry Pi, BeagleBone, etc.

DIWdiver said:

For power you should use RMS values. The RMS value of a PWM signal at 50% duty cycle is 0.7071 times the full value. Since it applies to both voltage and current in this case, P = 0.7071V * 0.7071I. It's no coincidence that 0.7071*0.7071 is 0.5,

Click to expand...

for a sinusoidal waveform. but what happens when the output is being alternately shunted between two different outputs? for the average current to be a factor of 0.7071, that would (most likely) require current to be flowing through the two outputs simultaneously, seems a little counterintuitive

Anders Hoveland said:

for a sinusoidal waveform. but what happens when the output is being alternately shunted between two different outputs? for the average current to be a factor of 0.7071, that would (most likely) require current to be flowing through the two outputs simultaneously, seems a little counterintuitive

Click to expand...

I think you missed the point entirely. The RMS value can be very different from the average value. Average values are useful for some things, not for others. RMS values are useful for some things, not others. And they are pretty much mutually exclusive.

I didn't say that the average was 0.7071 times the max. I said that the RMS value was 0.7071 times the max. Quite different statements.

In order to make useful calculations in circuits that are not DC, with 'DC' meaning that the current and voltage are fixed over a short term, one must understand not only what average and RMS are, but also where they are to be used. To use either where the other is called for is to make a significant error (usually). The average value of a sinusoidal waveform is zero. It's pretty obvious that won't produce useful results in most calculations.

RMS means literally "square Root of the Mean of the Square". That means that you take the square of the signal, average over one cycle (or some other useful period), then take the square root of the average. It turns out that for a sine wave, the RMS value is 0.7071 times the peak value. For a square wave (by definition, 50% duty cycle), it's also 0.7071 times the peak value. But for any other duty cycle, the PWM signal has different RMS value.

In general, the RMS value of a PWM signal is the square root of the duty cycle. If you have a duty cycle of 0.1, or 10% (on 10%, off 90%) the RMS value is 0.32, because the square root of 0.1 is 0.32. Likewise, if you have a duty cycle of 80%, the RMS value is 0.89, since the square root of 0.8 is 0.89.

Anders Hoveland said:

for a sinusoidal waveform. but what happens when the output is being alternately shunted between two different outputs? for the average current to be a factor of 0.7071, that would (most likely) require current to be flowing through the two outputs simultaneously, seems a little counterintuitive

Click to expand...

No, the RMS voltage is sqrt(DutyCycle)*peak value. If the duty cycle is 0.5, peak is 1, then RMS = sqrt(0.5) = 0.7071

For your example of going between two values: say 0.25 peak for 25% of the duty, and 0.5 peak for remaining 75% of the duty, you could do RMS =

sqrt((0.25 * sqrt(0.25))^2 + (0.5 * sqrt(0.75))^2) = 0.452. I.e. the RMS voltage is 0.451V. The power into a 1 ohm resistor is the square of this or 0.203W


You could also get that from (0.25^2*.25+0.5^2*0.75) = 0.203


Semiman

DIWdiver said:

I think you missed the point entirely. The RMS value can be very different from the average value. Average values are useful for some things, not for others. RMS values are useful for some things, not others. And they are pretty much mutually exclusive.

RMS means literally "square Root of the Mean of the Square". That means that you take the square of the signal, average over one cycle (or some other useful period), then take the square root of the average. It turns out that for a sine wave, the RMS value is 0.7071 times the peak value. For a square wave (by definition, 50% duty cycle), it's also 0.7071 times the peak value. But for any other duty cycle, the PWM signal has different RMS value.

In general, the RMS value of a PWM signal is the square root of the duty cycle. If you have a duty cycle of 0.1, or 10% (on 10%, off 90%) the RMS value is 0.32, because the square root of 0.1 is 0.32. Likewise, if you have a duty cycle of 80%, the RMS value is 0.89, since the square root of 0.8 is 0.89.

Click to expand...

When talking about PWM, occasionally I mention mechanical voltage regulators such as used by Bosch in cars and motorcycles in the 1900s.

There's the battery, there's the regulator. It's just a switch that slams off and on quickly. Under an oscilloscope, it looks like big nasty square waves. The only reason that it works is that over time, it equals out. It tones down the voltage from the charging system. It's really crappy regulation. It doesn't care about current.

Modern current control is smooth (usually.) In PWM, there may be a visible flicker that can induce headaches, nausea and vertigo in some people (like me.) Not a fan of PWM, smooth current control is best.

Those type of regulators work by adjusting the field current in the alternator. Since the field winding is a big inductor, the PWM duty cycle of the voltage ends up adjusting the average field current. The field current ends up being a sum of DC and AC components. If it's designed properly, the AC component is much smaller than the DC component. The field current and engine speed control the output current of the alternator. The battery soaks up most of the AC component of the alternator output. So the 'big nasty square waves' at the output of the voltage regulator aren't simply averaged. They are filtered by a number of other components in the system. The field coil should do a decent job of averaging the PWM signal from the regulator, but there's a lot more to it than that.

I'm sure someone will correct me if I'm wrong, but I'd bet big that modern electronic regulators do exactly the same thing, just using transistors instead of contacts, and higher PWM frequency.

If there is visible flicker in a PWM system, it's because the frequency is low, or being modulated by the scanning frequency of a camera. If we're talking direct to the human eye, 200 Hz. is plenty to make it imperceptible to even the most sensitive eyes. And it's usually quite easy to go much faster than that.

The problem with PWM in illumination is with cameras. If your PWM frequency is 3000 Hz, and your camera scans an image in 1/1000th of a second, you get three cycles of on and off during the scan, which look like light and dark bars in the image. Talk about a headache!

For more insight into the PWM problem inducing epilepsy and other factors, please consider this wiki page:
http://en.wikipedia.org/wiki/Flicker_vertigo
I mention this because I, myself, am vulnerable. I've been knocked out with a blackjack on two occasions breaking teeth. This is what will happen, similar to flicker vertigo/the Bucha effect. It's basically the same effect. Avoid it.

If you want to talk safety, please don't subject a percentage of the population to a medical condition by viewing your display.

You don't want people hitting their heads upon pavement or tile. These are not good materials for this.

PWM over 1 KHz, problem solved, at least where people are concerned.

Agree, SM. Bad pwm on most lights is just a headache. It's slow strobe like at night club/dance halls that really gets me. Haven't been to one in probably 14 years. Most didn't run them long enough to get me.

Also forgot to say that PWM could be a step towards smooth current regulation, then full wave rectifier plus filtering.

If you PWM at 1KHz plus, there is no need to post filter. As well, assuming you are PWMing DC, no need for the rectifier either.