The basics of LED drivers

JoakimFlorence

Newly Enlightened
Joined
Jun 4, 2016
Messages
137
There have been so many repeated questions about this that I thought I would start a definitive thread explaining it once and for all.

You have your LED, how do you power it?

The first thing you need to know about LEDs are that they are not like regular (incandescent) light bulbs. LEDs, while they have a very long lifespan, are also very sensitive. It only takes a split second of too much power to overload them and cause a burnout. You can't just hook up an LED to a battery. If there's nothing to limit the current, an LED will allow more power through than it can handle.

That's where a constant current driver comes in.

How do constant current drivers work? Well actually they regulate the voltage, but that ends up regulating the amount of current indirectly. A constant current driver will increase the voltage until a certain amount of current flows through the circuit. We normally don't think of an LED as having any resistance, but actually it does (within its narrow operational range of voltages). If the voltage potential across the LED falls just a little bit, the amount of current able to flow through will decrease. A little more voltage and more current will be able to flow through. It would be very easy for too much current to flow through and the LED to suddenly burn out.

Of course drivers can't regulate the voltage to just any level. They're usually custom designed with a certain voltage range. The driver might say 8-12v. That means the minimum amount it can drop the voltage down to is 8 volts, while it also can't go above 12 volts. It's important to select the right driver voltage range depending on how many LEDs you plan to be running in a string (in series in the circuit). Oftentimes the driver will only indicate a single voltage rather than a range. In that case it's usually safe to assume the driver cannot drop the voltage more than 3 or 4 volts below the indicated voltage.

You usually do not want to go above the rated voltage on the LED. Remember, when LEDs are connected in series (in a string) the voltages all add up. Connect them in parallel and it's the current (mA) that add up, not the voltage.

Of course, if you do not have enough voltage, no current will be able to flow through the LEDs at all. (This is another way that LEDs are different from incandescent bulbs)
Being a semiconductor, an LED does need a certain minimum voltage threshold to conduct current.

So if you have three 3.2v LEDs wired in a string, a driver rated for 8-12v would be an appropriate choice.
---O---O---O---


The second thing is the amount of current that the driver is rated for. You usually want the driver current (mA) to match up with the rated LED current (mA). It's okay to use a little less current. For example, if the LED is rated for 350mA it's okay to use a 300mA driver, but be aware that the power will be a little bit lower and the LEDs might not be quite as bright.

Another thing that needs to be covered, when connecting two or more LEDs in parallel it is usually good to supply a little lower current that the sum of what all the LEDs are rated for.
You see, unless the voltage being supplied is at the lower end of the LEDs operational range, there is a tendency for all the current to start flowing through just one of the LEDs, rather than being spread out evenly between the LEDs, and then there will be a burnout. A 3.2v 350mA rated LED may be able to handle 3.4v 371mA, but try connecting three of these LEDs in parallel, naturally with a driver supply designed for 3 times the current which would be 1050mA, and there is a fair chance there could be a burnout. Lowering the power a little bit could avoid this. (The reason for this problem is because once the voltage gets on the high end of the operational voltage range of the LED, the amount of current that can flow through the individual LED quickly increases exponentially. The driver dropping the voltage at a given amount of current through the circuit isn't going to do any good because the circuit in this case is designed for more current than any single LED can handle.) To summarize, when powering multiple LEDs in parallel (or several different strings of LEDs in parallel) be conservative in estimating the amount of power they can handle. If you have two strings of 350mA, using a 600mA driver instead of 700mA wouldn't be a bad idea.

Oftentimes the LED specifications might not indicate an operational current (mA). In that case it's easy. Blue and white color LEDs are generally 3.2 volts. Take the power rating (Watts) and divide by 3.2 to find the current. For example, 1 Watt divided by 3.2 = 0.312 Amps, or 312 mA. (In this case most of these "1 Watt" LEDs are actually rated for 350mA)

The amazing thing about constant current drivers, if you had a driver with a wide enough functional voltage range, you could string 3 LEDs into a string or 5 LEDs into a string, and the amount of power supplied by the driver would be entirely proportional to how many LEDs there are. Remember, the driver is going to change the voltage until a certain current amount is flowing through.

So why don't we just use a resistor with an LED to limit current? This works very well for very low power indicator light LEDs, but when we're talking about more power the inefficiency of using a resistor starts becoming a significant factor. (It would basically require twice as much power)
 
Last edited:

JoeRodge

Newly Enlightened
Joined
Oct 8, 2017
Messages
133
I might be asking a stupid question and the answer is most likely out on the web. But I'm here for the community aspect of things!

What application would you use series or parallel? Is one more suitable for a certain application? Curious...
 

DIWdiver

Flashlight Enthusiast
Joined
Jan 27, 2010
Messages
2,725
Location
Connecticut, USA
I'll give you a real-life example of series and parallel that I worked on.

In those days, 1W LEDs were the most economical way to get illumination (at least from LEDs). But the customer wanted 30 times as much light as one would give, so we needed 30 of them.

The device was to be portable, powered by a lead-acid battery (later LiIon). If we strung all the chips in series, we'd need around 100V at 350 mA. Since we didn't want to have the safety issues with such high voltage, this was ruled out. We could have connected them all in parallel, which would need about 3.5V at 10.5A. This high current would have caused high resistive losses in the circuit, thus reducing the efficiency, and/or requiring large, heavy, expensive components. We eventually settled on a 12V battery and 10 strings of 3 LEDs. This was further broken into a group of 6 strings and a group of 4. Each group was given its own driver, nominally 10.5V at 2.1A and 1.4A, respectively.

These particular choices yield some excellent characteristics:
- 30 LEDs were arranged in a 5x6 array, which was aesthetically pleasing.
- The string voltage being below the battery voltage allowed us to use a buck converter, which is inherently the most efficient.
- The low voltage and low current allowed us to build an efficient driver without using expensive components.
- The closeness of the battery and string voltages allow very high efficiency.

In fact, as the battery neared discharge, the voltages were nearly equal, and we measured efficiencies as high as 99%! With a full charge (largest voltage difference) was the lowest efficiency, around 93%.

Running strings of 6 LEDs and a 24V battery might have yielded slightly higher efficiency. I forget now why that was ruled out. Perhaps because of the voltage limitations of large-value ceramic caps of the day. If I were to do this design today, I'd probably push for this configuration. Well, except there are better LEDs today...
 

DIWdiver

Flashlight Enthusiast
Joined
Jan 27, 2010
Messages
2,725
Location
Connecticut, USA
Another thing that needs to be covered, when connecting two or more LEDs in parallel it is usually good to supply a little lower current that the sum of what all the LEDs are rated for.
You see, unless the voltage being supplied is at the lower end of the LEDs operational range, there is a tendency for all the current to start flowing through just one of the LEDs, rather than being spread out evenly between the LEDs, and then there will be a burnout. A 3.2v 350mA rated LED may be able to handle 3.4v 371mA, but try connecting three of these LEDs in parallel, naturally with a driver supply designed for 3 times the current which would be 1050mA, and there is a fair chance there could be a burnout. Lowering the power a little bit could avoid this. (The reason for this problem is because once the voltage gets on the high end of the operational voltage range of the LED, the amount of current that can flow through the individual LED quickly increases exponentially. The driver dropping the voltage at a given amount of current through the circuit isn't going to do any good because the circuit in this case is designed for more current than any single LED can handle.) To summarize, when powering multiple LEDs in parallel (or several different strings of LEDs in parallel) be conservative in estimating the amount of power they can handle. If you have two strings of 350mA, using a 600mA driver instead of 700mA wouldn't be a bad idea.


This is a fairly poor description of a phenomenon known as thermal runaway. It's been around since before the invention of the LED. It happens because in some semiconductors (including LEDs), as the temperature rises, their resistance decreases. If you have multiple units in parallel initially sharing current equally, but one of them has a heatsink problem, it will get hotter than the others. The hotter one now has lower resistance and draws more current. This causes it to get hotter, which causes it to draw more current. In the situation of thermal runaway, one device eventually reaches a high enough temperature that it fails. In some cases this causes increased current in the remaining devices, rapidly resulting in the destruction of the entire array.

Thermal runaway is much more likely if the original devices are not identical.

With modern LEDs (even back when 1W was considered modern) it is generally considered safe to run LEDs in parallel at full power, with a few caveats.
- Make sure all LEDs are identical, preferably from the same manufacturing lot.
- Use quality LEDs.
- Make sure all LEDs have sufficient and similar heatsinking.
- If you can anticipate one or more 'hotspots', any thermal protection should focus there.
- If you break the 'identical' rule, add some resistance to each string to make them share better.

Of course, cutting back on the drive current can help too, but who wants to do that?

In fact this particular customer (for the design mentioned in the previous post) was so happy with the reliability of the product that they eventually asked for a 'turbo' mode where the LEDs were overdriven by 25% for a few minutes.

Oh, and before someone calls me out for it, I know it's not purely 'resistance' that changes, but the more complicated description wouldn't add to the discussion at this point.​
 
Last edited:

ssanasisredna

Enlightened
Joined
Oct 19, 2016
Messages
457


This is a fairly poor description of a phenomenon known as thermal runaway. It's been around since before the invention of the LED. It happens because in some semiconductors (including LEDs), as the temperature rises, their resistance decreases. If you have multiple units in parallel initially sharing current equally, but one of them has a heatsink problem, it will get hotter than the others. The hotter one now has lower resistance and draws more current. This causes it to get hotter, which causes it to draw more current. In the situation of thermal runaway, one device eventually reaches a high enough temperature that it fails. In some cases this causes increased current in the remaining devices, rapidly resulting in the destruction of the entire array.


While thermal runaway is certainly a potential issue, I don't think that was the over-riding concern. Even when they come from the same voltage bin, there will be small variations in Vf between the LEDs.

With small LEDs, which are the ones most likely to be paralleled, Reffective is big enough that small differences in Vf will be swamped by Reffective. With larger LEDs with small Reffective, difference in Vf can lead to noticeable difference in current between paralleled LEDs. To ensure all are below their rated current, it is necessary to run the total current less than N x rated current.

Today's LEDs tend to have better voltage matching, but they also have much lower Reffective. It can be harder to parallel modern high current LEDs.

On the other hand, low current LEDs, 2835, 3030, 3014, etc. are almost always paralleled (10-12 in parallel is not uncommon), and then those parallel strings connected in series. Every LED tube, almost every LED troffer, many low/high LED high bays are made this way.
 

rayman

Flashlight Enthusiast
Joined
May 6, 2008
Messages
1,219
Location
Germany
Nice summary for beginners :twothumbs. I learned a bit even too.
 

Elieaj

Newly Enlightened
Joined
Jun 15, 2018
Messages
1
Re: The basics of LED drivers - Frequency and Voltage

Thanks for the introduction about LED drivers, and I hope you can answer a question I was wondering about.

Most of the drivers that I found online had a voltage range of 120-277 Volts AC and a frequency range of 47-63.
Is that a trait of the circuitry inside the driver or is this on done on purpose?

For example in a 50HZ grid, grid failure would be usually above 48.5 HZ. (I know in developed countries the margin is much less than 1.5Hz, but some countries have bad grids.) So what is the point of having a driver that can handle 47 HZ? or even 55Hz?

Thank You!
Elie
 

easilyled

Flashaholic
Joined
Jun 25, 2004
Messages
7,252
Location
Middlesex, UK
While thermal runaway is certainly a potential issue, I don't think that was the over-riding concern. Even when they come from the same voltage bin, there will be small variations in Vf between the LEDs.

With small LEDs, which are the ones most likely to be paralleled, Reffective is big enough that small differences in Vf will be swamped by Reffective. With larger LEDs with small Reffective, difference in Vf can lead to noticeable difference in current between paralleled LEDs. To ensure all are below their rated current, it is necessary to run the total current less than N x rated current.

Today's LEDs tend to have better voltage matching, but they also have much lower Reffective. It can be harder to parallel modern high current LEDs.

On the other hand, low current LEDs, 2835, 3030, 3014, etc. are almost always paralleled (10-12 in parallel is not uncommon), and then those parallel strings connected in series. Every LED tube, almost every LED troffer, many low/high LED high bays are made this way.

Do you know whether the triple and quadruple emitters bought as a unit on an MCPCB are usually connected in series or parallel? The Emisar D4 for example which has a quad board setup and uses TIR optics.

I have two Emisar D4s. On the lowest (firefly) level, in one of my Emisar D4s, all 4 emitters are very similar in brightness. In the other D4 at the lowest level, 2 emitters are noticably brighter than the other two.

They are both setups using quadruple XPL-His
 

DIWdiver

Flashlight Enthusiast
Joined
Jan 27, 2010
Messages
2,725
Location
Connecticut, USA
Re: The basics of LED drivers - Frequency and Voltage

Thanks for the introduction about LED drivers, and I hope you can answer a question I was wondering about.

Most of the drivers that I found online had a voltage range of 120-277 Volts AC and a frequency range of 47-63.
Is that a trait of the circuitry inside the driver or is this on done on purpose?

For example in a 50HZ grid, grid failure would be usually above 48.5 HZ. (I know in developed countries the margin is much less than 1.5Hz, but some countries have bad grids.) So what is the point of having a driver that can handle 47 HZ? or even 55Hz?

Thank You!
Elie

In a nutshell, because they can.

Modern electronics have finally allowed for line-powered devices to be made with this broad input capability with little or no increase in price compared to more limited devices. Many manufacturers want to sell their product worldwide, and they now can make a single product that can handle the line voltages of almost any customer in the world. So they do. The reduction in model numbers more than warrants the small cost increase.

Also, they want every customer worldwide to be confident the device will work on HIS particular grid. That's why such a wide range is specified.
 

DIWdiver

Flashlight Enthusiast
Joined
Jan 27, 2010
Messages
2,725
Location
Connecticut, USA
Do you know whether the triple and quadruple emitters bought as a unit on an MCPCB are usually connected in series or parallel? The Emisar D4 for example which has a quad board setup and uses TIR optics.

I have two Emisar D4s. On the lowest (firefly) level, in one of my Emisar D4s, all 4 emitters are very similar in brightness. In the other D4 at the lowest level, 2 emitters are noticably brighter than the other two.

They are both setups using quadruple XPL-His

I would guess that the vast majority are either wired in series or user-configurable, with few hard-wired in parallel. There are several reasons for this:

1. In general, a higher-voltage, lower-current systems can be build with lower losses, and thus higher efficiency. There are of course many exceptions to this, but the majority of products are built to satisfy the rule, not the exception.

2. There is a fear in the industry (not entirely unfounded) of running LEDs in parallel. It's possible to run illumination-grade LEDs in parallel with little issue. It's also possible to have real issues.

As to your particular case, there are several possibilities. I present them in decreasing order of likelihood:
1. The emitters are run in parallel or 2S2P configuration, and differences in Vf between the LEDs is causing different currents and thus different illumination.
2. At very low currents, the emitters may have substantially different efficacies.
3. You could have emitters from different batches, which makes #2 much more likely.
4. Space aliens are influencing your observations.

I'm no expert in operating LEDs at super low currents, but I know a thing or two.

For one, there is a current below which an emitter will stop producing light. In general, this is really low, like turn out the lights and hold the emitters close to your face to tell which ones are emitting and which ones are not. I've seen strings of 3 'identical' emitters where one or two were lit and the others were not. But it's not necessarily this low. The SST-90 data sheet used to say 1-9 A drive current. I thought this was odd and questioned it. Eventually an engineer from Luminus assured me that they were saying that below 1A, they would not guarantee that the SST-90 would produce ANY light. While this was clearly absurd, it suggests that the operation of LEDs at very low currents is not as well understood as we would like.
 

ssanasisredna

Enlightened
Joined
Oct 19, 2016
Messages
457
Re: The basics of LED drivers - Frequency and Voltage

Thanks for the introduction about LED drivers, and I hope you can answer a question I was wondering about.

Most of the drivers that I found online had a voltage range of 120-277 Volts AC and a frequency range of 47-63.
Is that a trait of the circuitry inside the driver or is this on done on purpose?

For example in a 50HZ grid, grid failure would be usually above 48.5 HZ. (I know in developed countries the margin is much less than 1.5Hz, but some countries have bad grids.) So what is the point of having a driver that can handle 47 HZ? or even 55Hz?

Thank You!
Elie

What you see sold is going to depend on where you live. If you live in North America, you will see a lot of drivers advertised as 120-277 or a wider 100-277. 100 is used in Japan, 277 is industrial/commercial in the United States. If you are in Europe, most drivers are 220-240, with some Asian models at the wider range. The larger driver vendors will typically have a 220-240 model because it is common in many parts of the world, and then a wide range model for the rest of the world. Single voltage models can be design less expensively and with better performance at that voltage than a wide range model.

w.r.t. frequency, the grid is not what is always going to power those lights. Buildings will often have backup generators under which the lights need to work properly, not to mention not being susceptible to grid events. Then again, a lot of it is just industry enertia and other than a reduction in performance at 47Hz, (or 57), drivers would rarely care what frequency the line is at, though they will often have circuitry internal that is synchronized to the line frequency.
 

easilyled

Flashaholic
Joined
Jun 25, 2004
Messages
7,252
Location
Middlesex, UK
I would guess that the vast majority are either wired in series or user-configurable, with few hard-wired in parallel. There are several reasons for this:

1. In general, a higher-voltage, lower-current systems can be build with lower losses, and thus higher efficiency. There are of course many exceptions to this, but the majority of products are built to satisfy the rule, not the exception.

2. There is a fear in the industry (not entirely unfounded) of running LEDs in parallel. It's possible to run illumination-grade LEDs in parallel with little issue. It's also possible to have real issues.

As to your particular case, there are several possibilities. I present them in decreasing order of likelihood:
1. The emitters are run in parallel or 2S2P configuration, and differences in Vf between the LEDs is causing different currents and thus different illumination.
2. At very low currents, the emitters may have substantially different efficacies.
3. You could have emitters from different batches, which makes #2 much more likely.
4. Space aliens are influencing your observations.

I'm no expert in operating LEDs at super low currents, but I know a thing or two.

For one, there is a current below which an emitter will stop producing light. In general, this is really low, like turn out the lights and hold the emitters close to your face to tell which ones are emitting and which ones are not. I've seen strings of 3 'identical' emitters where one or two were lit and the others were not. But it's not necessarily this low. The SST-90 data sheet used to say 1-9 A drive current. I thought this was odd and questioned it. Eventually an engineer from Luminus assured me that they were saying that below 1A, they would not guarantee that the SST-90 would produce ANY light. While this was clearly absurd, it suggests that the operation of LEDs at very low currents is not as well understood as we would like.

Thank you for your response.

Its definitely reason 4! :nana:
 

Kiwi123

Newly Enlightened
Joined
Aug 11, 2018
Messages
3
This is just amazing work, the first thing you need to know about LED are that they are not like regular light bulbs, it only takes a split second of too much power to overload them and cause a burnout, this is not also for regular use...
 
Top