Drivers: linear vs. switching vs. PWM

Hello

I may be bringing something up which has been covered before, but I am trying to refresh my memory from electronic training I had many years ago (and even then, I only went to the Associate’s level, and I haven’t worked with electronics in a while), and maybe my questions will also help someone else new to working on drivers and may have the same questions.

Some of the recent discussions on constant current is where my questions originate from. Some of the information I have found on buck and boost drivers comes from here (buck) and here (boost) as well as various threads here and the flashlightwiki.

The way I understand it, from a power source perspective, is that a linear driver is only useful for low current applications, say below 200mA, because of the inefficiencies of “burning off” the extra power in current regulation. This is where a switching regulator came in as they are more efficient at higher currents, although they can be much more complex. Recently, I have noticed an increased interest in “constant current” drivers. Am I incorrect in assuming that “constant current” is referring to linear regulation? If constant current does refer to linear regulation, why is it desirable in a high current LED driver seeing as it is inefficient? I understand that running an LED with a PWM signal is not very efficient at low levels, but at higher levels is the efficiency of a PWM driven LED less then a LED run with switching regulation? I understand that PWM signals and linear/switching regulation are separate topics, but both are items used in drivers which I am seeking to expand my knowledge of.

When people refer to constant current they are not specifically referring to linear power supplies. Those things are two different topics entirely.

  • Switching (such as buck, boost, or SEPIC) is always going to have some losses. For the more basic (current controlled) buck drivers expect efficiency as low as 75-85%. Those losses happen all over inside a buck driver: the FET (switching losses), the inductor, and the diode. There will also be small losses in the sense resistor bank.
  • Linear drivers will have some losses two. They’ll reduce the voltage until the current matches their set current, so you multiply the excess voltage by the set current to get the amount of power lost. There are also small losses in the sense resistor bank (even a 7135 likely has an internal sense resistor).

It should be easily to see that if your input and output voltages are close enough… the linear driver becomes more efficient than a mediocre switching driver! In our applications input and output voltages are often fairly close together, so sometimes a linear driver makes the most sense. The downside is that if they are too close and the battery has too wide a useful voltage range you may not be able to maintain regulation (input drops below output). That’s where a buck driver would shine even if it’s less efficient. Using an input voltage of approximately double the output voltage would yield a tightly regulated output until the batteries were completely drained.

PWM output to an LED is always less efficient than a smooth constant current output. The LED becomes less efficient at higher currents. An LED operating at an average of 1 amp using 50% duty cycle PWM is actually operating at 2 amps half the time and 0 amps the other half of the time. Let’s say that the LED is 60% efficient at 1 amp and 45% efficient at 2A. You’re getting 45% efficiency now with the PWM when you could be getting 60% if you had constant current!

From your post it sounds like you are still mixing up PWM, linear, and switching… again, one of these 3 things is not like the others. PWM can be used with either linear or switching power supplies to further control the output of a fixed power supply. If the power supply is configured for 3 amps and not adjustable, my only option is to switch it on and off very rapidly (PWM) in order to simulate lower currents (1 amp would be a 33% duty cycle for example).

Thank you for taking the time to write such an informative reply. I am sorry if my questions sounds like it is confusing PWM and current regulation. I have used PWM signals with servos in the past and know it is a control signal vs. a power source such as linear and switching supplies. Your reply has helped to expand my knowledge as I guess what I am really after is trying to understand the most efficient ways to drive high powered LEDs as well as the reasoning behind using different drive methods.

You’re welcome, but… unfortunately your reply makes it sound as if you may still be confused. We use PWM which is visible on the driver board’s output, eg we switch things like 6vots at 10amps on and off at ~20khz.

When a PWM control signal results in a flat, DC line on the output from the board… that’s still a constant current driver. This behavior is seen in the LDCH LD29 for example. The MCU outputs PWM but after that PWM is handled by other sections of the driver the actual output to the LED is a steady, constant DC current with no more ripple in it than you’d expect. See the scope image above the words “Again a very smooth output, no traces of any pwm or driver regulation.” over here at HKJ’s review of the LD-29 to see what I’m talking about.

I seem to have muddied things up by bringing up PWM. Based on your above information, are the following statements true?

Linear drivers are useful when the load voltage is near the input voltage. So if a led have a forward voltage of 3.3v a linear driver would be used when running the LED off of one 18650 (3.7V nominal) or possibly even two in series (7.4V nominal).

Switching drivers are used when the input voltage is different (either higher or lower) than the output. So, a switching driver would be good if we were to use the LED from the example above with a 12V source (buck) or a 3V source (boost).

PWM is a modulated signal which controls the output of one of the above regulators by switching it on/off at a high rate of speed which is determined by the duty cycle of the PWM signal.

If the above are true, then for applications using li-ion battery cells for power (1 or 2 in series) a linear regulator would typically be used.

not quite, you can not use a linear driver for 2s Li-ion to a single 3v LED. You’re going to be getting 8.4v (7.4 nominal) from the 2s Li-ion, to drive a 3v LED (such as a CREE with a vF from 3.3-4.2 or so) you’ll require a buck driver, 2s Li-ion will instantly kill it.

Note you can obviously run a 6v LED (such as the cree MT-G2) from 2s in a flashlight application and if ran at high enough current’s (well over 10A) you can even run an MT-G2 (or multiple MT-G2’s in parallel) from 3s Li-ion because of the ever rising vF as current [and heat] increases as well as voltage sag from the cells however this trick will never work to run a 3v emitter from 2s. The MT-G2 can handle abuse better than any emitter I’ve ever seen.

To have the ability to run from 1 or 2 Li cells you would probably want to choose a SEPIC however many chineese companies will simply go with a buck driver causing single cell operation to suffer greatly. Buck/boost would also work but really the best efficiency will be from a SEPIC.

I was surprised to learn that bibihang’s friend is running an XP-E2 from 2s using a Nanjg driver w/ Zener mod… apparently at 2-3A. It doesn’t work well (which is not surprising). See post #164 here. Clearly it’s a bad idea and not one which should be pursued!

It’s quite possible to run low current devices from a linear converter, even when output is massively lower than input. LDOs are linear and they are commonly used to supply from a few mA up to 100mA. 3.3v LDOs running on 10v+ input is normal. Building an entire buck section to run a 5mA or 20mA status LED wouldn’t be necessary for example.

ADV - your current understanding, expressed in post #4, seems generally correct. As Cereal_killer pointed out, running a ~3v emitter on 2s with a linear driver is bad. Otherwise everything sounds fine.

… now on to the things that haven’t been addressed yet: dirty power in buck circuits. Linear power supplies are known in all sectors for their clean power. Switching? Not so much. Switching always brings in some ripple or something. In the case of the cheap, hacked up, high current buck circuits we use the bad behavior can be extreme. Very tall voltage spikes in the output are possible, enough to quickly destroy an LED. So that’s another aspect to watch out for. There is some discussion of that behavior in this thread: ImA4Wheelr - Driver Mod: HX-1175b & HX-1175B1 (Pic Heavy)