## Is an LED driver practical?

36 posts / 0 new
Author
Message

Hi everyone,

I'm developing a circuit that will run some LEDs (for indoor lighting) powered off some batteries. What I don't want to do is just hook up the batteries to the LEDs, because this will deplete the battery charge too quickly.

To solve this problem I thought of driving the LEDs with a current that I could regulate. To do this, I thought of using a zener diode and an NPN BJT. However, when I thought about it, I'm not sure this is such a good idea.

The total battery voltage is going to vary between about 3-4 V. The LEDs are 3.2 V. If I have a BJT, won't the voltage drop across it make the LEDs dim?

Thanks!

Eric

Batteries are not going to last very long with 'power' LED's.

My cycle headlight works pretty well with 4 AA cells. The rear light is very effective. It flashes. Hence you get a very good battery life.

You will probably get the best performance with a specific LED driver chip. But a simple compromise is to PWM the LED with a varying duty. You get the MOSFET or junction transistor to saturate. Pass a very high current to the LED for a short duty cycle. This avoids wasting power in resistors or linear regulators.

David.

Eric,

Think of LEDs as current regulated devices. As a diode it does have a certain voltage across it when it is forward biased, (conducting, lighting up), but you don't control or regulate the forward voltage, Vf. You control the current.

The easiest way to control the current is with a series resistor. The voltage across the resistor is equal to the supply voltage minus the forward voltage across the LED. One can then select a resistor for the amount of current they want to flow through both the LED and the resistor, (in series). R=V/I R = (5-1.2)/10mA = 380 ohms, for example, for an LED with Vf = 1.2 V, 10 mA current, 5V supply.

The "problem" with this is that the resistor just gets warm, burning up energy. The energy spent in the resistor does not go towards light generation.

An alternative method is to use an active LED driver circuit which controls the current through the LED, without "wasting" energy in the series resistor. One typically uses a constant current source to do this. With higher energy LEDs the energy spent in the electronics driving them is negligible. This is part of what David mentioned above.

If you are good with analog circuitry, and like op-amps, (or transistors...), you can design your own. Otherwise there are many LED driver chips on the market that will do this for you. There are also some Application Notes on several of the supplier's sites describing the theory.

Lastly, batteries for room lighting? Again, as David mentioned, there is only so much energy stored in a battery(s). Walk around at night in your house with an LED flashlight to get an idea of how much light vs time one gets...

JC

Some pros of using a constant current boost led driver:
You can nearly deplete the batteries before the led stops emitting light.

You can combine it with pwm to vary the intensity without changing the hue.

LEDs have an optimum current where they are most efficient. You can make sure the led is at this point as long as possible.

Cons:
It's a lot more expensive than a resistor in series.

If you're going for high power LEDs like 350mA, you should think about getting a heatsink. And you should consider using a lens or diffuser, or it will be very unpleasant to be around.

Hi all,

1. I ran some tests and, according to some back-of-the-envelope calculations, I want about 40 mA to pass through each LED. At this current they're pretty bright.
2. Resistors in series are an option but I would rather not do this due to the heat dissipation.

Here are my questions:
1. Is it bad practice to just hook up LEDs to a battery because in this design the current is unregulated? This was also my thought in obtaining a current regulator.
2. The NPN BJT that I mentioned above would amplify a current that I provided from the zener and a resistor to the 40 mA I calculated. Would this work?

Eric

Just operate the LEDs in pulsed mode as has been suggested. There are plenty of circuits around.

Leon Heller G1HSM

Ah, yeah ... about that, and why this thread is in the Gen. Electronics forum ... we're actually doing some redesigns of the system which may or may not include a microcontroller, depending on some external factors.

So I'm considering the case where there is no microcontroller. What would you suggest in this case?

lagger, how do you find the point of optimum efficiency?

Thanks!

Eric

Quote:
The total battery voltage is going to vary between about 3-4 V.

It would help if you are more precise on this. Battery technology, how many cells in series, etc.

Quote:
2. Resistors in series are an option but I would rather not do this due to the heat dissipation.

This is the easiest and cheapest solution. But you'll need a reasonably constant supply voltage. Heat dissipation at 40mA is probably not relevant.

Quote:
1. Is it bad practice to just hook up LEDs to a battery because in this design the current is unregulated? This was also my thought in obtaining a current regulator.

This is done in many cheap flashlights because of cost and because the used battery has a large enough internal resistance so the current does not gets way to big.

Quote:
2. The NPN BJT that I mentioned above would amplify a current that I provided from the zener and a resistor to the 40 mA I calculated. Would this work?

Although one of the BJT parameters is the current multiplication you can not count on it for your application because it can vary a lot from device to device and it depends a lot on the BJT temperature. The second issue is worse as the power dissipated in the device will drive the temperature up which will drive the current up in turn. This until the BJT calls for help using smoke signals :-).

You best option for efficiency is to build/buy a constant current device, switched if you like, and place multiple LED in series. Like 12V supply, three LED and your constant current circuit in serie.

Markus

In the datasheet, you usually have graphs for relative luminous intensity as a function of current and as a function of temperature.

Google for "joule thief" It can suck power from the battery up to the very last drop.

Nachus

Hey guys,

Thanks a lot for the feedback. I have some updates.

1. The batteries are three NiMH cells connected in series. The nominal voltage on each is about 1.2 V, and the capacity of each cell is about 2300 mAh.
2. The LEDs are white, 3.2 V. I don't actually have a data sheet because they're a supply from China that I inherited so I don't know where they came from. Because of this and my battery cells, I can't really put them in series.
3. I connected the batteries to the LED array I'm trying to drive and the current draw was about 760 mA, but wavered a lot if you touched or moved the wires. At this draw, some of the LEDs got quite hot and one even turned blue.
4. I am currently trying to use a p-type MOSFET in saturation mode and tweaking the gate voltage as a means to limit the total current (I_D) passing through the array. Does this sound plausible? The MOSFET also has a temperature dependence, and no doubt with a current draw of about 200 mA it's going to warm up, but it seems like it's sturdier than the BJT.
5. As I may not be able to use a microcontroller, a switched supply is currently out of the question ... but thanks for the tip, the joule thief looks like a handy little circuit! :)

emoney wrote:
1. The batteries are three NiMH cells connected in series. The nominal voltage on each is about 1.2 V, and the capacity of each cell is about 2300 mAh.

This complicates your circuit a lot, could you change to four NiMH in series ?
- The NiMHs give you 3.6V to 3V, you need 3.2V right in the middle of it. Needs complex electronics !
- Direct connecting does not work, as the inner resistance of NiMHs is too low. They happily supply enough current to fry your LEDs as you have seen.
emoney wrote:
2. The LEDs are white, 3.2 V. I don't actually have a data sheet because they're a supply from China that I inherited so I don't know where they came from. Because of this and my battery cells, I can't really put them in series.

Fair enough, you just need to know how much current you want to give them. The Battery configuration prevents a simple series configuration, but it complicates stuff enough that I'd change it.
emoney wrote:
3. I connected the batteries to the LED array I'm trying to drive and the current draw was about 760 mA, but wavered a lot if you touched or moved the wires. At this draw, some of the LEDs got quite hot and one even turned blue.

Smoke signals :-).
emoney wrote:
4. I am currently trying to use a p-type MOSFET in saturation mode and tweaking the gate voltage as a means to limit the total current (I_D) passing through the array. Does this sound plausible? The MOSFET also has a temperature dependence, and no doubt with a current draw of about 600 mA it's going to warm up, but it seems like it's sturdier than the BJT.

A simple constant current source works better with a BJT. See the circuit here as an example:
http://www.ecircuitcenter.com/Circuits_Audio_Amp/BJT%20Current_Source/BJT_Current_Source.htm
But you need 1-2V for it to regulate correctly.
emoney wrote:
5. As I may not be able to use a microcontroller, a switched supply is currently out of the question ... but thanks for the tip, the joule thief looks like a handy little circuit! :)

You don't need a microcontroller for a switched circuit. The joule-thief is a switched circuit and is quite simple.
On the other hand many microcontroller are quite happy with your 3-3.6V supply voltage, if you want to go down that route.

In the end you have the choice:
1) Keep your 3-cell battery and build a complex and error-prone circuit to supply a single LED
2) Keep your 3-cell battery, add a boost converter to some higher voltage and supply a couple of LED in series
3) Change the battery configuration:
3a) to 4 cells (or more) and build a current limiter with a BJT, use a simple resistor or a switching mode current regulator.
3b) to 2 or 1 cells and build a joule-thief style switching regulator

3a with a current limiting resistor is the simplest circuit.

Markus

You can probably find a 'similar' device and read its data sheet.

You should never do a 'constant' drive without a series resistor. If they were drawing 760mA, this is 2.7W. Your batteries may give you 60 minutes or so. You have to read those data sheets too!

Even if you do not have a controller, you can drive a 37% duty cycle. This will give you 1W with fully charged batteries. By the time you have a multivibrator and an o/p transistor, you might just as well have used a proper LED driver.

The punter is going to pay serious money for the batteries. The cost of driver chips is considerably less. Unless you are in the market for 'sell it quick and run' products.

David.

emoney wrote:

5. As I may not be able to use a microcontroller, a switched supply is currently out of the question ... but thanks for the tip, the joule thief looks like a handy little circuit! :)

YEs, but if you use NIMH you are at the risk of deplete the cells to zero, which usually ends damaging the battery. If you're going to use a joule thief (originally used with primary batteries), then you need a battery cutoff (more circuitry) that stops the switching circuit when there's a certain voltage between the battery terminals (usually 1V per cell, altough some manufacturers suggest 0.9V/cell in packs of less than 5 cells)

Quote:
I am currently trying to use a p-type MOSFET in saturation mode and tweaking the gate voltage as a means to limit the total current (I_D) passing through the array. Does this sound plausible? The MOSFET also has a temperature dependence, and no doubt with a current draw of about 200 mA it's going to warm up, but it seems like it's sturdier than the BJT.

What you can do is, to do a N-MOSFET + opamp current regulator low side, and you won't have to "tweak" the gate by hand (with what that implies) This way you'll end up with a continously variable and repetable current control. The only drawback AFAIK is the heat dissipated, but, on the other hand there are plenty of NMOS power fets around. It's easier to find the perfect fit for the task.

Nachus

The most usual way to go, is to have the three AA NiMH batteries in series, and a boost LED driver that controls up to many white LEDs in series. Plenty of IC's out there that do that for you (check out Maxim, ST, Linear, TI, etc). Not cheap, but more efficient and simple, and SAFE (for your LEDs).

Guillem.
"Common sense is the least common of the senses" Anonymous.

Quote:
2. Resistors in series are an option but I would rather not do this due to the heat dissipation.

You run your white chineese LEDs so they turned blue but heat on a shunt resistor worries you?

Quote:
LEDs (for indoor lighting) powered off some batteries

Could you please tell us why do you want to use NiMH cells to drive LEDs indoor? Neither LEDs are suited for indoor lighting (they give very focused light which needs to be additionally dissipated to be tolerable), nor is battery powered light source suited for indoor lighting..

No RSTDISBL, no fun!

David:

Quote:
Batteries are not going to last very long with 'power' LED's.

Me:
Quote:
there is only so much energy stored in a battery(s). Walk around at night in your house with an LED flashlight to get an idea of how much light vs time one gets...

Brutte:
Quote:
nor is battery powered light source suited for indoor lighting..

I guess some things are only learned the hard way...

In the USA public buildings, schools, etc. have the exits marked with "Emergency Exit Signs", these are illuminated with batteries and LEDs if the Mains power fails. Old ones used filiment light bulbs.

Routine lighting of the interior, via LEDs and batteries, doesn't seem to be a very practical idea.

JC

## Attachment(s):

Quote:
Routine lighting of the interior, via LEDs and batteries, doesn't seem to be a very practical idea.

The lighting isn't for homes in the US. Grid power is prohibitively expensive or simply not available. The rooms to be illuminated are rather small and clients are interested in affordable, lightweight home lighting systems. There are some definite drawbacks to using LEDs and batteries, but I think it could get the job done in this case.

Also, just want to thank everyone again for their feedback! I really appreciate all of it.

Eric

Quote:
Grid power is prohibitively expensive

Eric,
Recharging your batteries with solar power? If you recharge from the grid you are still paying for the power to light the room.

Not using rechargable batteries? Now THAT would be expensive! (And environmentally unfriendly).

JC

Quote:
I guess some things are only learned the hard way...
In the USA public buildings, schools, etc. have(...)

Do you know what "an indoor lighting" is? Your torch lying on a chair or a flash in your camera is not an indoor lighting, although it is indoor actually.

Quote:
Grid power is prohibitively expensive or simply not available

Are you thinking about a commercial indoor light source? How much does a bucket of energy cost at your place? Unit of 1MWh is frequently used, this one costs max about 130USD, or 420PLN (including tax) in Poland. And you want to make it with batteries:
-cheaper when it is expensive or -more attractive than other light sources when not available?
Think it over emoney because I assure you it is impossible to even get close to that price with any reasonable technical solution (not stealing it).

No RSTDISBL, no fun!

emoney wrote:
3. I connected the batteries to the LED array I'm trying to drive and the current draw was about 760 mA, but wavered a lot if you touched or moved the wires. At this draw, some of the LEDs got quite hot and one even turned blue.

You've gotten some good advice so I'm going to go slightly OT here. When soldering an 0603 blue LED a while ago I got it too hot (I was experimenting with hot air for the first time). The LED worked just fine afterward... except that it was no longer blue. It was green.

So white LED ---heat--> blue LED
blue LED ---heat--> green LED

I wonder if you kept on heating up that formerly white LED if it'd turn green? I can't say I am a master of semiconductor physics or optoelectronics - but I for one find this bizarre!

White leds are not white actually. Most of them uses luminophore (just like fluorescent bulbs).

No RSTDISBL, no fun!

White LED's are blue in fact, plus a yellow-ish resin that absorbs part of the blue light-or any energy and turns it yellow. The addition of both colours seem white to our eyes. The 'frequency domain' or wavelength energy distribution found on some white LED's datasheets is quite enlightening.

Guillem.
"Common sense is the least common of the senses" Anonymous.

With regards to

JC:

Quote:
Recharging your batteries with solar power? If you recharge from the grid you are still paying for the power to light the room.

Not using rechargable batteries? Now THAT would be expensive! (And environmentally unfriendly).

Yes, the batteries are being recharged from a solar cell. We're working in villages in Africa, actually, so the grid is simply not an option (read: it doesn't exist). And we are using rechargeables! 2300 mAh NiMH ... if people use our system, then maybe we can combat others that use primary cells only and last for only two weeks until the batteries need to be thrown away ... on the ground ... because there is no solid waste management system, really.

Brutte:

Quote:
Do you know what "an indoor lighting" is? Your torch lying on a chair or a flash in your camera is not an indoor lighting, although it is indoor actually.

Our system consists of three lamps hung from the ceiling. Two have five LEDs and are suited for a kitchen or den; the third has four LEDs and is suited for a bedroom. This isn't a lantern.

Quote:
Are you thinking about a commercial indoor light source?

No, this is meant for a two- or three-room home.

~~~

To anyone that was curious, after all this input and considering the circumstances I've decided to use ballast resistors with a value of around 22 Ohm. I don't have the resources right now to create a voltage/current source because I think I would need a boost DC-DC converter. My chosen solution is a compromise that at least promises protection of the LEDs and batteries, so the system should behave in a predictable way for its predicted lifespan (about 1.5 - 2 years, which is the lifetime of the battery).

The resistors will deliver about 20 mA of current to the LEDs while the NiMH cells are around their nominal discharge voltage. There will be some variation when the batteries are fully charged and reaching zero capacity, but it's a sacrifice I'm willing to make. The battery charge cycle and discharge cycle is also being regulated to prevent overcharging and overdischarging.

Thanks again everyone! I'll take all your suggestions into designing "version 2.0"! :D

Eric

Just a side observation: NiMH capacity degrade really fast over 30ÂºC. For hot environments other chemistries will work better, probably Pb-Acid batteries. Li-pol would be too expensive.

It would be much better if you can check the real environment values before you get started, since there are many things that can go wrong if one doesn't check for everything, even in such 'simple' project. IP would be an issue also, specially if used inside a kitchen.

Guillem.
"Common sense is the least common of the senses" Anonymous.

Guillem,

Thanks for the suggestion. It actually turns out that the temperature where the systems will be installed is around 21-25 deg C (high altitude). If I could order the components to suit the purpose of the system, I would probably use lead-acid because of their robustness. However, I'm just coming into this job and the stock is there ...

... either way, I will be sure to try to introduce heat sinks because both the batteries and the LEDs are pretty temperature-sensitive.

Appreciate it,

Eric

Quote:
The resistors will deliver about 20 mA of current to the LEDs

Take a knife and a carrot and try to slice it under 3x20mA white leds (5mm I suppose?) in your kitchen and you will know it is not reasonable. Your idea is to make a stationary lighting device using some most expensive and not suited technical solutions.
1. Why do you use NiMH cells when these are second after LiPolymer most expensive cells in terms of energy storage? I am sure something powered by solar panel does not need to be light. Use low current Pb-Acid, possibly those not maintenance free (hey are much cheaper tham SLA and your system is not a Mars Pathfinder - maintenance is not a problem since it is for kitchen). Add a 1L demineralized water container and it will last for 5 years of service.
2. LEDs are some most expensive light sources (perhaps only HIDs are more). LEDs have the advantage of being very endure (100.000h) and your cell will not last for even 10% of that time. LEDs are small and are wonderful for cell phones, but it is a disadvantage for kitchen light. What efficacy do you expect(with a diffusor of course)? 20 - 30lm/W?? I doubt it.
Use fluorescent bulbs which have the efficacy of about 100lm/W for a fraction of the LEDs price (in terms of lumen/watt). Your cells will be 3 times smaller and you will not cut your fingers.

No RSTDISBL, no fun!

AFAIK, white LED's are about 100l/W if choosen properly. Two minutes ago, I've received an OSLON white LED. 1W, 115l. Too expensive. 2.5 by 2.5 mm (yes, 2.5x2.5mm).

LED's are in a rage for efficiency, size, cost, power and endurance. Don't underestimate the power of them. The diffuser and light delivery system is really tricky, though.

My recommendation would be to use a bunch of 20mA white LED's with high efficiency (up to 150 lumen/W now, but more in two years) in a 'big' area, behind a white translucid diffuser. The enclosure may be be the worst part and the one that would need more work. Optics is the driving headache here, while electronics is on a second term.

Guillem.
"Common sense is the least common of the senses" Anonymous.

Brutte,

Unfortunately we're committed for the time being to the NiMH cells and the white LEDs. It was beyond my control, but different battery chemistries and lighting fixtures are definitely being considered for the next design iteration. I think the motivation for using NiMH was their energy density -- so that they would last longer on a single charge.

We never did any optics calculations -- just demo'ed it and people seemed to be pleased -- they paid for it, at least. Will definitely take your fluorescent bulb idea into consideration, however.

Guillem:

For the time being I don't think we'll be able to have those optics. But like I said with Brutte ... version 2.0 ... very valuable input.

Either way, the current system beats out kerosene in terms of light given and cost over their lifetime, and it promises to be more adaptable than kerosene as well. And that is what makes it worth it to the customer. :)

Quote:
energy density -- so that they would last longer on a single charge

Energy density and energy capacity are two different parameters. You can have 1000J from lemons and 1J from a LiPolymer cells. One has nothing to do with the other.

Quote:
I've received an OSLON white LED. 1W, 115l.

Quote:
up to 150 lumen/W now,

Could you Guillem Planisi please provide a datasheet of that LED?.
If you would like to know a fluorescent bulb with 100lm/W, I can recommend OSRAM Lumilux.

And lumen is [lm].

No RSTDISBL, no fun!

Oslon and other LED's I'm using are from Osram Opto Semiconductor also. Regular new 'power' LEDs from Osram are now 100lm/W, but other manufacturers had announced (but there is still no public datasheet for them) white LED's with 150lm/W @ 20mA (power LED's run from 100mA upwards, and usually have lower efficiency).

http://catalog.osram-os.com/cata...

http://www.ledsmagazine.com/news...

Nichia tends to be the cutting edge for low power LED's, while Osram works on the car manufacturing industry where output power is more demanding. Even when Osram seems expensive when compared to others, it is the main supplier for car manufacturers over Europe. Being there, doing that.

Guillem.
"Common sense is the least common of the senses" Anonymous.

Oh, I almost forget: fluorescent tubes are harder to drive than the simplier LED's, so any (theoretical?) gain obtained by those, is totally outweighted by the complex and relatively high voltage electronics involved to supply it from batteries. In comparison, even switched mode LED drivers are simple, while obtaining better efficiency and results.

Regarding the optics, the problem with LED's is that they emit their relatively high amount of light from a damn really small area, thus they have a pretty bright hot spot, while fluorescent tubes emite their light from a long and big cilindrical surface. That is the reason why some diffuser must be used when using LED's, otherwise, you may wear sunglasses...

Guillem.
"Common sense is the least common of the senses" Anonymous.

Quote:
fluorescent tubes are harder to drive than the simplier LED's

With the same technique - just put a shunt resistor and off you go. It is as simple as that, although cold startup is more complex. With high voltage this is true - tubes operate at 50-80V while a single white led only at about 3.4V. But with such photovoltaics and battery operation it is the flux for \$ meeting the criteria that matters, not a voltage. Even with LEDs - running them in parallel without shunt resistors is impossible - they must be connected in series to have one constant current source.

No RSTDISBL, no fun!

Brutte, can you post some link to those fluorescent lamps you comment? What about the cold start? Cost per unit? Datasheets? I'm intrigued about them. My experience was for Uninterrupted Lighting with standard 230VAC tubes. And I have also experience with CCFL, but those are, ehm, small and harder to drive (>1000V).

Guillem.
"Common sense is the least common of the senses" Anonymous.