Power dissipation in PWM-dimmed LED?

[Artlav]

An LED driver, ZXLD1360, have a dim pin, to be driven by PWM.
To get full 1:100 dimming range, it should be driven at under 500Hz.
As a result, the LED is essentially driven with a PWM waveform, with peak at it's full current's voltage, and zero in-between.

Question is - how does that affect the power dissipation?
Obvious part is that the heat dissipation will be based on LED's full power efficacy point, which is less than what it would be at the lower one.

Less obvious is - how does the fact of PWM affect it?

Let's say the LED have Vf of 15V at 1A current. Normally, it would dissipate 15W (ignoring the *0.75 factor for light output).
If it was to be driven at 0.25A, it would have Vf of about 13V, and power dissipation of 3.25W.

However, what values to use for driving it with a 0.25 duty cycle PWM?
Is it RMS current, I=Ip*sqrt(D) = 0.5A?

What about Vf?
Assuming Vf of 1A, it would be 7.5W.
Assuming Vf of 0.5A, it would be 6.75W
Assuming RMS Vf, it would be 3.75W
Which one is it?

Basically, how to estimate the power dissipation of an LED driven by PWM?

[Cside]

why is the Vf dropping from 15V ro 13V ?

also "Assuming Vf of 1A..." Vf (foreward voltage) is not a current.

As a first pass the power should be the foreward voltage (supply voltage in your case?)  x  the LED current at given voltage x duty cycle.

[Marco]

You are mixing up a lot of stuff. As a first approximation power simply goes down linearly with duty ratio.

In principle due to efficiency droop a greater percentage of the power will go to heat, but cooling design generally isn't precise enough to care about such small differences. Exactly how much is hard to say, efficiency droop is not given parameters in datasheets.

[Artlav]

why is the Vf dropping from 15V ro 13V ?

Because Vf of an LED is current dependent?

also "Assuming Vf of 1A..." Vf (foreward voltage) is not a current.

Meaning "assuming Vf at current of 1A".

You are mixing up a lot of stuff. As a first approximation power simply goes down linearly with duty ratio.

The origin of the question is that some components produce waste heat by RMS current, so simply V*A*D would give quite a bit less than expected.

The difference can be more than two times, which matters a lot when selecting a heatsink.

[owiecc]

Maybe the attached graph will help. Start with DC on the right. When you lower the duty cycle, the average power you dissipate in the LED lowers, the Tj also lowers. This causes the Vf to increase a little. You can even see heating and cooling of the die during PWM dimming. For your calculations of power start with DC and assume linear change with duty cycle. This will be good enough. Either way you need to size the heatsink for D=100%.

For more you can check out my dissertation.

[Artlav]

Some testing later, constant current seems to be better.

I had two LEDs with the same build and driver running:
-One with a 256Hz PWM, duty cycle of 20%, from 500mA current source.
-Other with a constant current of 100mA.

The PWM one appears noticeably dimmer, but it's heat sink is 2-3 degrees hotter (thermocouple at the base of the star).

I tried to measure the brightness with a luxmeter, however it appears to have failed - the PWM one shows 15000 lux (at 1 cm away), the CC one - 80000 lux, which i find suspicious.

Bottom line is - i'm getting more light for less heat with CC dimming than with PWM.
And i'm curious how that works and how to estimate it.

For more you can check out my dissertation.

That's quite interesting, thank you.

[mikerj]

Maybe the attached graph will help. Start with DC on the right. When you lower the duty cycle, the average power you dissipate in the LED lowers, the Tj also lowers.

The problem is that even though the average current is lower, the peak current remains the same, as does the peak Vf.  This means you get higher total Pdiss in the LED compared with driving it from a constant DC current.

[DanielS]

Bottom line is - i'm getting more light for less heat with CC dimming than with PWM.

The things that break peak-current PWM are lower LED efficiency at high power, higher Vf, higher I2R losses, higher losses in the power supply (if using switchers) and likely a bunch of other parameters.

You can have the best of both worlds by using a switching constant-current source - dimmable by changing the switching regulator's reference voltage to change the output current.

[owiecc]

Some testing later, constant current seems to be better.

I had two LEDs with the same build and driver running:
-One with a 256Hz PWM, duty cycle of 20%, from 500mA current source.
-Other with a constant current of 100mA.

CC will always be better due to higher efficacy at lower current levels. You can compare the difference on the attached figure. Red: CC, grey: PWM. E.g. for that particular green diode you get twice the lumens at 0.2*Ipeak.

Bottom line is - i'm getting more light for less heat with CC dimming than with PWM.
And i'm curious how that works and how to estimate it.

It is a bit hard as you will need some radiometric data. It requires either some expensive HW or a tricky calorimetric setup. Maybe this would be nice for a student project: find some cheap, easy and practical way in which these things can be estimated