PWM of high Constant Current vs setting lower Constant Current?

RoGuE_StreaK

Enlightened
Joined
Apr 15, 2010
Messages
615
Location
Brisbane, Australia
As per title, what are the pros and cons of running a higher CC and PWMing it, vs running a lower CC and keeping it on? Efficiencies, tint shifts, ??
I'm looking at running XM-Ls at up to 4Amp short bursts, but the majority of the time running it at say 1.5Amp (depending on my test results); I realise there are a lot of ways to skin this feline, just wondering what peoples experienced views are, should I just set the CC to 4Amp and do a fast PWM to reduce it down, or re-set the CC each time I want to incease/decrease by a large amount?
With my planned setup I can have 32 hardware CC steps between 0Amp and 4Amp (actually slightly less 4Amp, but anyway...). I need to vary the light output at all times, so will be using PWM to some degree anyway, should I set to 1.5Amp (about 12/32 output) and PWM it slightly, needing to change the settings before boosting up to 4Amp (32/32), or should I run it "full on" and greatly reduce the duty cycle of the PWM (ie 37.5% duty cycle of 4Amp ~= 1.5Amp)
Any thoughts? PWM frequency will be at least 160Hz if I use the software method, or may well be in the 39kHz range if I use hardware PWM.
 

qwertyydude

Flashlight Enthusiast
Joined
Aug 10, 2008
Messages
1,115
The biggest differences of course will be the tint shift and efficiency. High currents generally let the blue light get too intense for the phosphors and therefore you generally end up a little cooler in tint.

But also there is efficiency. At high instantaneous current, you'll dissipate more power in the internal resistances of the things other than your led. So what ends up happening is you waste power in the internal resistance of your battery and in the electronics.

One other thing is very short pulses in the same frequency tend to make for more noticeable PWM. It tends to make things in motion freeze instantaneously rather than have a natural motion blur.
 

Lux-RC

Banned
Joined
Sep 10, 2011
Messages
55
it may sound strange but freq does matter. At freq above 10Khz pwm method is getting almost equivalent to CC.
 

Th232

Flashlight Enthusiast
Joined
Dec 25, 2008
Messages
1,064
Location
Sydney, Australia
it may sound strange but freq does matter. At freq above 10Khz pwm method is getting almost equivalent to CC.

If you've seen this on an oscilloscope, at those frequencies is the resulting output still a square wave of sorts, or is it dampened (bypass caps or similar) to be more of a flat line with some ripple (like you'd see with CC)?
 

The_Driver

Flashlight Enthusiast
Joined
Jan 6, 2010
Messages
1,177
Location
Germany
You can either have the highest possible high-mode with a high brightness, but with a low efficieny and a lot of heat or a lower high-mode with less total brightness but a lot more efficient. The thing is that if your current is very high the lower pwm modes will also have the same bad efficiency since you are still using the same current. This is why current controlled lights are almost always more efficient (more runtime) compared to similar pwm-driven lights when using the lower levels. SOme of selfbuilt's reviews reflect this (take a look at runtime comparison graphs of similar lights).
 

Lux-RC

Banned
Joined
Sep 10, 2011
Messages
55
...The thing is that if your current is very high the lower pwm modes will also have the same bad efficiency since you are still using the same current....

it's not completely true, the good news that it's easy to test. In fact a very short pulse efficiency IS NOT the same as continuous operation at the peak pulse current.

Another important aspect is the regulator efficiency. I can't agree that most constant current drivers are as good in efficiency at low modes as at 100%. Moreover, most constant current typologies are far less efficient at very low currents than when they run on full. A PWM regulator has a flat regulation efficiency which normally doesn't noticeably depend on the dimming ratio.
 

RoGuE_StreaK

Enlightened
Joined
Apr 15, 2010
Messages
615
Location
Brisbane, Australia
Getting closer to finalising the PCB for the first test batch, I've decided to keep both sets of functionality and can experiment, the Output Enable is connected to a remappable pin on the MCU so can run it on hardware PWM if I want to, it looks like the hardware PWM frequency will be in the order of 39kHz. Current selection is done via SPI (Serial Peripheral Interface), will have to experiment with speed of change, looks like roughly speaking I'll have hardware CC levels of 0.125Amps so may just do the lot using SPI changes, may have enough resolution and speed to not require PWM at all for variation of brightness.

Another benefit is that I could do a rolling change of the output pins being used, to decrease heat build-up in the drivers; eg., with two drivers both using 6 of their 16 pins, that gives 1.5Amps, can constantly roll from pin to pin on each SPI change, so they all get the chance to cool down if need be.
 

The_Driver

Flashlight Enthusiast
Joined
Jan 6, 2010
Messages
1,177
Location
Germany
it's not completely true, the good news that it's easy to test. In fact a very short pulse efficiency IS NOT the same as continuous operation at the peak pulse current.

Another important aspect is the regulator efficiency. I can't agree that most constant current drivers are as good in efficiency at low modes as at 100%. Moreover, most constant current typologies are far less efficient at very low currents than when they run on full. A PWM regulator has a flat regulation efficiency which normally doesn't noticeably depend on the dimming ratio.

Ok, thanks for the correction. I am defnitely not the big expert on this subject. If you make a dedicated monnlight constant current circuit though you should be able to achieve a much higher efficiency than most pwm lights. Take a look at the older HDS edc lights (pre 2008) for example.
 

RoGuE_StreaK

Enlightened
Joined
Apr 15, 2010
Messages
615
Location
Brisbane, Australia
Well I don't know if it's the best solution, but looks to be right for my particular application. It's actually a constant current sink, like the oft-used AMC7135, but needs a bit of supporting infrastructure, usually an MCU, but a clock of some sort at minimum.

The first one I was going to use is the Texas Instruments TLC5917, which has 8x sinks which, depending on the R-Ext resistor value, can sink up to 120mA each; 8x 120mA = 960mA when tied together.
Then I recently stumbled across it's brother, the TLC5926, which is exactly the same, except each chip has 16 sinks, so when set to 120mA it can sink a total of 1920mA (or near-enough 2Amps). Also has the advantage of being a smaller package than the one I was using for the 5917, and a "powerpad" version that isn't available for the 5917; an exposed metal pad on the bottom for thermal management. Pretty sweet.

There's a doc available that shows how you can use them without any MCU controller, just a 555 timer, which even allows external PWM control. Doc's "SLVA346", aka "TLC5916 One-Wire Control - Eliminating Microprocessor Control for LED Driver". I've picked up a few SMD 555s to maybe make up some stand-alone drivers with the unused 5917's I have lying around. Gotta love free samples! (TI are really good in this regard)

Oh, and they're cheap! ~$0.70each for larger orders, and the 5926 is the same price as the 5917, so you can get twice the current for the same price, but with better thermal management!
 
Last edited:
Top