How Can I Lower Voltage More Efficiently?

[wingman358]

Hi all, I'm working on an automotive LED lighting project in which the supply voltage is from the battery/alternator and so varies 12 to 14V. I'm trying to find a way to drop the LED circuit supply voltage that would be better than using resistors. Here's the current design:

R1 = 100Ω
R2 = 3.9kΩ
Vsupply = 14V

LEDs: 20mA @ 2V

The problem I see with the above circuit is R1 must dissipate

P = I*I*R ≈ 0.120A * 0.120A * 12V ≈ 1.5 W

which is rather large considering I would prefer to use 1/4 or 1/8 W rated resistors.

So that got me thinking in a more general sense about the efficiency of voltage dividing circuits that use resistors. There must be some other circuit that I can use that would be more efficient and not that much more complex. Something like a transformer but for DC. I read a bit about linear regulators and SMPS but I'm quite sure how to use those. Any tips or ideas?

I'm a newb with electrons so Thanks in advance for any help!

 

[berkeman]

Hi all, I'm working on an automotive LED lighting project in which the supply voltage is from the battery/alternator and so varies 12 to 14V. I'm trying to find a way to drop the LED circuit supply voltage that would be better than using resistors. Here's the current design:

View attachment 69804

R1 = 100Ω
R2 = 3.9kΩ
Vsupply = 14V

LEDs: 20mA @ 2V

The problem I see with the above circuit is R1 must dissipate

P = I*I*R ≈ 0.120A * 0.120A * 12V ≈ 1.5 W

which is rather large considering I would prefer to use 1/4 or 1/8 W rated resistors.

So that got me thinking in a more general sense about the efficiency of voltage dividing circuits that use resistors. There must be some other circuit that I can use that would be more efficient and not that much more complex. Something like a transformer but for DC. I read a bit about linear regulators and SMPS but I'm quite sure how to use those. Any tips or ideas?

I'm a newb with electrons so Thanks in advance for any help!

Yeah, you don't use resistors for LED lighting projects. You use constant-current buck regulators, with low-side current sensing. Maybe try doing some searching with those terms, and have a look at the typical circuits used. Let us know if you have questions about those circuits.

 

[pixelpuffin]

couldn't you just put the six LEDs in series?

 

[berkeman]

couldn't you just put the six LEDs in series?

If they are well-matched, you can put LEDs in series and get uniform brightness. That is used in LED lighting applications, but you still would use a current-sensing buck regulator usually to maximize the efficiency. When you buy a bag of LEDs, they are generally not well-matched, though. If you buy them on a reel or Cut Tape, they should be matched fairly well.

 

[wingman358]

If they are well-matched, you can put LEDs in series and get uniform brightness. That is used in LED lighting applications, but you still would use a current-sensing buck regulator usually to maximize the efficiency. When you buy a bag of LEDs, they are generally not well-matched, though. If you buy them on a reel or Cut Tape, they should be matched fairly well.

Thanks for the ideas, I'm reading up on these buck babies now.

I actually was originally designing for series wiring, hence the six LEDs (12V / 6 gives 2V over each, in theory), but an EE friend advised against it for the reasons mentioned above. Just curious though, how does one "match" them? What characteristics exactly are matched?

And in all honestly the higher priority for me is just getting the project done. I don't need 99% efficiency, I'm just looking for something better than wasting 1.5W to heat. Although that would be acceptable especially if it is much simpler and easier for a newb to do. That being said, what do you think about max power dissipation ratings for resistors? What goes into that? I mean what does that assume in terms of convection, radiation, conduction, etc.?

Thanks for the ideas and thought-provoking discussion; that's why I love these forums!

 

[berkeman]

Thanks for the ideas, I'm reading up on these buck babies now.

I actually was originally designing for series wiring, hence the six LEDs (12V / 6 gives 2V over each, in theory), but an EE friend advised against it for the reasons mentioned above. Just curious though, how does one "match" them? What characteristics exactly are matched?

And in all honestly the higher priority for me is just getting the project done. I don't need 99% efficiency, I'm just looking for something better than wasting 1.5W to heat. Although that would be acceptable especially if it is much simpler and easier for a newb to do. That being said, what do you think about max power dissipation ratings for resistors? What goes into that? I mean what does that assume in terms of convection, radiation, conduction, etc.?

Thanks for the ideas and thought-provoking discussion; that's why I love these forums!

I think that if you use the "Cut Tape" option when you buy the LEDs from Digikey, you will get LEDs that are from the same batch, so they will be matched fairly well. There is another level of matching that I think LED Lighting high-power LED manufacturers use (like tighter tolerances on the efficiencies), but if you are doing a fast one-off project, that may not matter.

And a good compromise on complexity versus efficiency is to buck the 12V down to around 5V, and use a single resistor for each 2xLED string. Place as many 2xLED strings in parallel as needed to get you to your desired lighting level.

 

[berkeman]

I think that if you use the "Cut Tape" option when you buy the LEDs from Digikey, you will get LEDs that are from the same batch, so they will be matched fairly well. There is another level of matching that I think LED Lighting high-power LED manufacturers use (like tighter tolerances on the efficiencies), but if you are doing a fast one-off project, that may not matter.

And a good compromise on complexity versus efficiency is to buck the 12V down to around 5V, and use a single resistor for each 2xLED string. Place as many 2xLED strings in parallel as needed to get you to your desired lighting level.

Or adjust the number of LEDs and the buck output voltage as needed -- you need to look at the tolerance on Vforward at your target current to decide what combination of Vout and Rseries you will use.

 

[meBigGuy]

Basically, if you have 3V across the LED's (depends on the LED's spec, based on color, etc) you have 14-3 = 11V across 100 ohms which is 110ma. If the LED's are matched that's 110/6 = 18ma each.
The 3.9K does nothing except draw about 1.3ma.

I don't know whether matching is harder for parallel or series connections, or if there is any difference. Intuitively I think parallel matching may be more of an issue since 1 LED can hog a lot of current.

Regarding matching when LED's are in series: all you are trying to match is the brightness at a given current.

Try it and see if they are uniform enough, and swap out the ones that don't match until you get acceptable results. Nothing to lose.

 

[berkeman]

I don't know whether matching is harder for parallel or series connections, or if there is any difference. Intuitively I think parallel matching may be more of an issue since 1 LED can hog a lot of current.

Regarding matching when LED's are in series: all you are trying to match is the brightness at a given current.

Try it and see if they are uniform enough, and swap out the ones that don't match until you get acceptable results. Nothing to lose.

It's pretty much impossible to match them in parallel as shown in the OP. Series matching is a much better bet, and is done routinely in LED lighting applications.

 

[wingman358]

OK so an alternative circuit (assuming 12V supply) would be to just wire the 6 LEDs in series, and swap outliers? Would that distribute 2V across each LED evenly? or close enough to an even distribution?

 

[berkeman]

OK so an alternative circuit (assuming 12V supply) would be to just wire the 6 LEDs in series, and swap outliers? Would that distribute 2V across each LED evenly? or close enough to an even distribution?

You still need a current-defining resistor. Drop about 2V across the resistor, and put 5 LEDs in series. Use the value of the resistor to set the series current.

What current are your LEDs rated for?

 

[wingman358]

You still need a current-defining resistor. Drop about 2V across the resistor, and put 5 LEDs in series. Use the value of the resistor to set the series current.

What current are your LEDs rated for?

They're 20mA at 2V and the max anticipated supply voltage is 14V. So I can keep the 6 LEDs, drop in a current limiting resistor of R = 2V / 0.120A = 17 Ohms and I should be all good, no?

I've read its not a good idea to run LEDs at their max rated continuous current, but I figure I'll be fine as this circuit is on a momentary switch and will be relatively low duty.

Thanks for all your help berkeman!

 

[berkeman]

They're 20mA at 2V and the max anticipated supply voltage is 14V. So I can keep the 6 LEDs, drop in a current limiting resistor of R = 2V / 0.120A = 17 Ohms and I should be all good, no?

I've read its not a good idea to run LEDs at their max rated continuous current, but I figure I'll be fine as this circuit is on a momentary switch and will be relatively low duty.

Thanks for all your help berkeman!

What is the minimum anticipated supply voltage...?

 

[wingman358]

It's for a motorcycle brake light, so battery voltage which ranges ~12V to 14V

 

[berkeman]

It's for a motorcycle brake light, so battery voltage which ranges ~12V to 14V

So you're back to 5 LEDs, correct?

 

[wingman358]

6... why?

 

[berkeman]

Because if the LEDs are 2V apiece and the supply is 12V, there is no voltage drop across the resistor and the LED current is undefined. You always want to have a voltage across the current-setting resistor.

And you can see the tradeoff/disadvantage of using just a resistor to set the current when you power supply voltage can vary. To get a fairly well-defined current, you need to drop more of the PS voltage across the resistor (like 6V across the resistor and the rest across 3 LEDs, so you don't get a 2:1 variation in LED current over the PS range of 12-14V). But doing that wastes a lot more power in the resistor. When the power supply can vary, going with the DC-DC buck with current sensing makes more sense...

 

[wingman358]

Ok so my question of the voltage distribution arises again... intuitively I imagine the battery/supply voltage being divided evenly (or close to that) amongst each of the elements in the circuit, which must be wrong if what you're saying is true.

I suppose what I am assuming about the LED forward voltage must be incorrect. I am thinking that the forward voltage is the "ideal" supply voltage, rather than a fixed value.

So then to reconcile my understanding with what you are saying, the LEDs exhibit a voltage drop equal to the forward voltage no matter what? if that is the case, then you are right and I must go back to 5 LEDs and use a higher resistance...

 

[berkeman]

Ok so my question of the voltage distribution arises again... intuitively I imagine the battery/supply voltage being divided evenly (or close to that) amongst each of the elements in the circuit, which must be wrong if what you're saying is true.

I suppose what I am assuming about the LED forward voltage must be incorrect. I am thinking that the forward voltage is the "ideal" supply voltage, rather than a fixed value.

So then to reconcile my understanding with what you are saying, the LEDs exhibit a voltage drop equal to the forward voltage no matter what? if that is the case, then you are right and I must go back to 5 LEDs and use a higher resistance...

No, please have a look at the datasheet. It will give the V-I characteristic (typical and tolerances). The V-I characteristic is non-linear (not linear like for a resistor). You need some mechanism to set the LED current, so that small variations in the Vf of the LED do not cause large changes in the LED current.

 

[meBigGuy]

LED current increase logarithmically with voltage across the diode. The standard solution is to use a constant current LED driver. A resistor will work also, but it's hard to calculate the correct value for a given string. Also, the voltage varying from 12 to 14V would significantly increase the current if there was 2V across the resistor..

An lm317 voltage regulator makes a very nice current source with 1 sense resistor. You can google the circuits. Nice thing about it is that the current won't vary with voltage, which would be a significant variation with a resistor. The to-220 package is cheap ($0.72)and has good thermal dissipation. Read the data sheet.

There might be other regulators that need less headroom and have a lower reference voltage. There are other LM317 packages also.

 

[Windadct]

There are some online LED calculators that help will seeing the various possibilities - ether way each parallel path should have its own resistor. You need to know the LEDs V and Current - from it's datasheet - for example basic Red LEDs could be 0.7 to 1.2V, but I have some purples that are 3.3V -- made a design with 4 different sets, by collor a real pain - each set had a completely different arrangement. There are also LED driver ICs - that can help - but add cost - these typically send pulses to the LED and make it look brighter than continuous DC will.

 

[donpacino]

I would like to caution that even with 'similar' LEDs you will get slightly different voltage drops with the same current through them. As a result the LEDs will all be different brightness.

I would recommend putting a resistor of high resistance value (~100k) in parallel with each LED. This will help to regulate brightness without using too much power. That is how the problem is solved with typical LED based light fixtures.

 

[meBigGuy]

these typically send pulses to the LED and make it look brighter than continuous DC will.

I ran some side by side experiments that showed that the perceived brightness depends only on average current. I tried ratio's up to 20 to 1 or so. That was back in red-only LED days, but I don't see why that would change. The eye integrates the brightness.

 

[wingman358]

Okay guys just want to follow up on all that was discussed here. I got my project to work without letting any of the magic blue smoke out .

I tried diving into the NP junction theory stuff but that avenue was too complicated, and I really didn't want to have to buy some complicated flux capacitor / linear regulator 9000HD. So instead I spent a few hours going through experiments with a multimeter, breadboard, and some LEDs until I had tested all of my assumptions.

Here's a couple things I noticed doing my experiments:

-My LEDs must have all been from the same batch, as they were all very close in terms of brightness and voltage drop across them (we're talking \pm0.01 V). I tested them in varying arrangements from 2, 3, all the way up to 6 in series, and checked the voltage across each. I also strung all of them together across a 2V parallel arrangement and didn't notice any dim/bright ones.

-The most interesting result from my tests was that each LED received an equal voltage drop (\pm0.01 V) regardless of the number of LEDs whether it was 3, 4, 5, or 6 in series. They all got the same voltage.

-With an 11 V supply and 6 LEDs in series, each LED got 1.75 V and the resistor got 0.08 V. I was surprised this circuit actually worked. It was very dim, but the LEDs were working. (Current was measured at 0.5 mA).

So those results were very interesting and helped me understand a little better how the voltage got divided up through the series string of LEDs. I had originally assumed each LED would automatically just have 2V across it, and if that couldn't be supplied, they wouldn't work. Well that was sort of correct, but its more like they're just very dim if the voltage is too low.

I think the biggest lightbulb that went off (pun fully intended) in my head was when I was falling asleep thinking about how the current in the circuit was determined - whether it was determined by the LEDs or the resistor. So then I realized it's sort of like a step input with some oscillation during the settling time. The voltages between the LEDs and the resistor affect each other, as does the current through the resistor and thus the current through the LEDs. I tried coming up with an equation to describe all these relations, but just decided to solve numerically.

So the first thing I did was I took the current-voltage curve from the data sheet and entered a couple points in excel and used a trendline to find a function for it. Then I was able to use that function to calculate the current drawn by an LED at a certain voltage.

So what I did was input the supply voltage, set a resistance (say 150 Ω) then start my iterations using the voltage across the resistor as 2 V. Then the current across the resistor was calculated using V=IR (ignoring the LEDs for this part). Knowing from my experiments above that the LEDs get the remainder of the voltage divided evenly between them, and having the LED current-voltage function, I calculated the current the LEDs were trying to draw at the calculated voltage. So I then had two current values, one from the resistor calculation and one from the LED calculation, and at first they were way off. So I iterated on the voltage over the resistor (adjusted up or down) until the two values for current were less than 0.1 mA different.

I couldn't think of any reason why that wouldn't be valid. So I did that iteration for each 14 V and 16 V until I got reasonable numbers for current (~20 mA @ 14 V and ~30 mA @ 16 V) using the same resistance value. Then I checked the power dissipated over the resistance and determined how many resistors to put in parallel to get the right total resistance and a low power dissipation over each resistor (~50 mW per resistor).

Those of you paying close attention know I originally was using the voltage range 12-14 V but am now saying 14-16 V. Well I decided to go to the connector and measured voltage using my multimeter with the bike off, but key on, and with the bike idling (12 V then 14 V). Then I revved it and noted it spiked around 16 V. I decided I won't ever really care if the LEDs are dim when the bike is off (but key on). So I designed for 14-16V.

Now on to the next project.