lower the pwn frequency the higher the runtime?

[cheapo]

is this true? the led is being run for a shorter period of time right?

 

[LukeA]

is this true? the led is being run for a shorter period of time right?

If you can stand the strobing, yeah.

 

[greg_in_canada]

Nope. 50% on at 100Hz is the same as 50% on at 2000Hz. In both cases the LED is getting full current 50% of the time and no current the other 50% of the time. There may be some minor efficiency differences due to the transistor switching time, inductor losses, etc. but basically they are the same.

Greg

 

[LukeA]

I was saying 50% @, say, 60Hz, and 25% @30Hz, using equal pulses and shifting the frequency, not that that would be especially usable.

 

[RV7]

Usually the PWM frequency is fixed. What changes is the proportion the LED is on in each cycle (mine works this way, for example). In such cases, greg is completely right, every word in his post.

 

[2xTrinity]

Using PWM to drive an LED at 1amp, 50% duty cycle will generally be more efficient than 1A at 100% duty cycle (due to less heat buildup, and other factors) but far less efficient than running the LED at 500mA constant current. In the case of something like a Cree, I believe efficiency peaks at around 80mA. Below that, reducing current would decrease the efficiency. Therefore, the absolutely most efficient way to extract runtime would be to find the most efficient current for the LED, then feed it low duty-cycle pulses at that current. I would love to see someone implement a continously variable brightness light implementing this technique (current limiting at the high end, PWM at the low end), naturally using as high a frequency as possible (if the frequency is high enough, flickering will not be noticeable)

Also, one of the advantages of PWM as opposed to current limiting is the fact that the LED's color does not shift at all. I have found that most white LEDs shift toward being more bluish at high current, and more greenish at low current. Varying brightness using PWM as opposed to current-limiting prevents tint variations.

 

[RV7]

Cree's output at 500mA is about 59% of that at 1A. So it's less efficient at 1A, but I'm not sure if I should call it "far less". :thinking: Good point though Trinity! I almost forgot this factor.

Cree's efficiency max out below 100mA. But feeding such a low current to PWM you can only get even lower outputs. :shrug:

Yeah I like the fact that PWM retains the same LED color at all levels. :twothumbs

 

[IMSabbel]

I think in reality its not really as clear-cut as in theory (a la "x% duty cyle" is the only thing that counts).

Seeing that low duty cycles at high frequencies mean that very short rise and fall times are needed, and the circuits are as cheap as possible, there could be capacitive and inductive losses in the electronics at high frequences and low duty cycles.

 

[RV7]

I think in reality its not really as clear-cut as in theory (a la "x% duty cyle" is the only thing that counts).

Seeing that low duty cycles at high frequencies mean that very short rise and fall times are needed, and the circuits are as cheap as possible, there could be capacitive and inductive losses in the electronics at high frequences and low duty cycles.

I agree. Greg mentioned it too. It's usually negligible though, unless the PWM frequency is extremely high.

 

[Minjin]

I've gotta think from a pure physics standpoint that turning it on and off more often MUST use more energy. If you went from 100hz to 200hz, you should have twice the switching losses. Is this not correct or are you guys just saying that the switching losses are negligible?

 

[RV7]

It's normally negligible because modern electronic switches are extremely efficient (milli Ohm level) and fast (well beyond MHz level so the transition period is hardly measureable).

 

[BB]

Actually, capacitive switching losses in FETs (and control logic) can be fairly substantial in switching supplies--losses going up with frequency and the voltage levels being switched (charging and discharging parasitic capacitance in the FETs).

-Bill

 

[macforsale]

*

 

[2xTrinity]

FYI
HDS method of PWM.



http://www.candlepowerforums.com/vb/showthread.php?t=161920

Wow, very interesting technique. I've never seen a PWM system operate like that (although, I suppose PFM or pulse frequency modulation is more appropriate). Switching rate on that is also very fast. 11kHz at the highest level, 300 Hz at the lowest. One advantage I can think of by doing that, is that flickering should only be even noticeable at the very lowest levels.

Actually, capacitive switching losses in FETs (and control logic) can be fairly substantial in switching supplies--losses going up with frequency and the voltage levels being switched (charging and discharging parasitic capacitance in the FETs).

Interestingly, by switching the frequency of the pulse, that ensures that the switching loss (assuming every "switch" costs the same, regardless of factors like pulse-width) is always going to be the same, relative to total power consumption (the lower the ouptut, the lower the frequency, and therefore the lower the switching losses). In a normal PWM setup, switching losses would be expected to be fixed, and therefore at low levels, they would make up a larger fraction of total power use.

 

[TORCH_BOY]

is this true? the led is being run for a shorter period of time right?

Its also has a lot to do with the Mark to space ratio.
time on against the time off

 

[RV7]

Actually, capacitive switching losses in FETs (and control logic) can be fairly substantial in switching supplies--losses going up with frequency and the voltage levels being switched (charging and discharging parasitic capacitance in the FETs).

-Bill

Yeah, I've seen switching supplies consuming more than 20% of the power fed to them. But with careful design, the efficiency of a switching supply can reach 95% or better.

 

[RV7]

FYI
HDS method of PWM.

http://www.candlepowerforums.com/vb/showthread.php?t=161920

Thanks macforsale for the link. Never noticed that before.

So this one generates fixed width pulses at different rates to achieve different levels? Since PWM stands for pulse width modulation, it seems that we need a new name for it. PFM suggested by Trinity is a good candidate.

Do you know the benefits of this kind of design? I can't seem to see myself. If we plan to minimize switching loss, we can find out the minimum frequency that doesn't cause noticeable flickering, and set it as the fixed PWM frequency. Thus no matter the level, the switching loss is kept at the minimum (at the 100% level switching loss is actually totally eliminated). With PFM, if you ensure there's no flickering at the lowest level, then at higher levels the switching loss would be tens of times higher than a PWM.

 

[orbital]

+

If I have 'Light A' that's ON for 550 flashes a minute,
and 'Light B' that's ON for 650 flashes a minute,
is it safe to say the light that's ON 100 less/minute would be using less energy and ultimately run longer, if mA is the same?

 

[TMorita]

+

If I have 'Light A' that's ON for 550 flashes a minute,
and 'Light B' that's ON for 650 flashes a minute,
is it safe to say the light that's ON 100 less/minute would be using less energy and ultimately run longer, if mA is the same?

Let's say there's a blackout, and you have two identical flashlights.

Let's say you used flashlight A once, and fliashlight B twice.

Now, which flashlight used more power?

Toshi

 

[Morelite]

+

If I have 'Light A' that's ON for 550 flashes a minute,
and 'Light B' that's ON for 650 flashes a minute,
is it safe to say the light that's ON 100 less/minute would be using less energy and ultimately run longer, if mA is the same?

If the duty cycle is the same then light "B" wound be using less energy since it is actually on for less time in that minute.