# How long can an LED withstand its rated peak current?

I am working with several 7 segment displays HDSP-F103. According to the datasheet, each segment has a peak current of 45 mA, but it doesn't specify for how long in time or in PWM duty cycle.

It does mention 'average power' (37 mW). So I wonder if it's OK to use this parameter to work out the peak current time. It that case, I calculated the limit of 45 mA at 44.4 % of duty. Based on this, I assume safe to run the segments at 40 mA, 20%, which is my desired configuration.

Do you find any problem with this? Thanks.

• Until it breaks :-) – vicatcu Dec 23 '18 at 17:15
• Peak mA is an absolute. If you want to hold 45mA for some time then you need to consider junction temperature. The 15mA DC current allows you to calculate a base PWM with a maximum duty cycle for 45mA at 33%. Notice that note(1)&(2) tells you to consider junction temperature. – Jack Creasey Dec 23 '18 at 18:28

## Conclusion

20% @ 40mA is not safe. Proof at the end.

The AlGaAs Red a.k.a. ultrabright Red operates at a higher voltage than GaAs.

These parts have the following TYP. specs @ 25'C:

• 1.6 V @ 1 mA (1.6 mW)
• 1.7 V @ 5 mA (8.5 mW)
• 1.8 V to 2.2 Vmax @ 20 mA (36 to 44 mW)
• Derate above 91°C at 0.53 mA/°C
• 255 to 430 °C/W/Seg depending on device
• Average power /seg 37 mW is an ABSOLUTE MAXIMUM

You asked is 20% duty cycle safe for 40mA?

So let's extrapolate the Pd at 40mA. Using linear regression, the Vf=1.7V(@ 5mA)+ΔV/ΔI = 2.2Vmax (@ 20mA) thus Vf= 1.7V + 30Ω * If = 1.7+ 30* 0.04A= 2.8V (112mW)

Thus 20% duty cycle = 22.4 mW which is less than the ABS MAX of 37mW. OK so far.

## the gotcha

But that's only for 1 segment.

So if we activate 7 segments + dot = 8 x 22.4mW=179mW .

We know the current derating is above 91'C yet the max operating temp is 100'C.

They also spec the thermal resistance from junction to pin as 255 to 430 °C/W/Seg

Let's pick 345'C/W. So with 179mW, all 8 junctions rise 345'/W*0.179W=62'C above 25'C pin temp (big assumption) which makes the junction pretty close to 100'C and then the pin/pads start to heat up.

(boiling hot is bad for popcorn failures from common moisture ingress may results in LED chip delamination, as clear epoxy has poorer moisture seal on the interface compared to black epoxy.)

## Recommendation:

Verify Vf max on every lot is closer to typ and not max @ 40mA and/or don't exceed 30mA at 20% duty cycle or don't use 20% PWM @40mA. using 50% at 15mA for the same brightness. Use fat thermal tracks for each segment to radiate some heat.

## Other info

Stanley and HP pioneered the GaAs RED LED's so that others could multiplex them in the 1st LED digital watch in the early 70's. Since they had Vf= 1.6V vs AlGas of 2V they could handle more current but were far less bright. ( My buddy Kirk did the 1st LED watch at Litronix bought by Siemens A.G. of Germany.)

The problem was to make the gold whisker wire bond small enough to be almost invisible yet carry enough peak current not to fuse. Meanwhile the chip itself is thermal insulator is tiny so it cannot handle much power owing to the high tempco of >250/C/W.

The result is your average 5mm LED has an Abs. Max. to rated current ratio of 30/20=1.5 These parts have a ratio of 45/15=3 which permits MUXing to an absolute max of 3 digits as reasonable intensity but only 2 at max intensity.

• I am amazed for the completeness of your answer. Thank you. Since I am limited at 20% duty by my driving circuitry, I think I'll be using a more conservative current figure, like 25 or 30 mA. – user162889 Dec 23 '18 at 23:27
• Actually, I am thinking about it and I don't understand why you say 'So with 179mW, all 8 junctions rise 345'/W*0.179W=62'C'. Why do you use the total 8 segment power to work out the temp rise? Why aren't the temp increases calculated independently for each LEDs as the rest of the calculations? I mean, why not 345º/W*0.0224W (or worst case, 430 * 0.0224). Why are they supposed to be 'thermaly coupled'? – user162889 Dec 24 '18 at 0:43
• If one segment rises, it adds to the ambient for the next segment. So you have to use all power sources in close proximity to compute temp rise. but for PCB temp rise, you need the packaging design spec for thermal resistance to outside ambient. A precise analytical answer takes more space. – Tony Stewart EE75 Dec 24 '18 at 0:46
• The thermal resistance between segment junctions is assumed to be less than to the outside ambient. – Tony Stewart EE75 Dec 24 '18 at 0:54
• But that is a good thermodynamics question on heat transfer and it is not precise without all details. I did the worst case. – Tony Stewart EE75 Dec 24 '18 at 1:12

Your approach is correct, but you should design based on the power specification from the "operating parameters" not "maximum parameters" section.

That's because power levels above the operating recommendations, while they don't cause permanent damage if within maximum ratings, may still result in unwanted temperature rise and possible wavelength shift as well as deleterious effects on efficiency, beam pattern, etc.

• Thanks, I usually design according the operating parameters but in this case there is no relevant data for this calculation in that section. – user162889 Dec 23 '18 at 17:32

I think you'll be fine with 40mA at 20% duty cycle, assuming you are multiplexing fast enough that the segments appear to be solidly on (perhaps a few hundred Hz to 1kHz).

Forward voltage of the AlGaAs red at 40mA is about 1.9V (see figure), so the average power dissipation is about 15mW at 20% duty cycle. The dissipation at 15mA DC is about 1.7V * 0.015 = 25mW. Those are typical figures and you have to allow that the forward voltage may be higher than typical, but you're well within the limits.

This part appears to be limited by the maximum peak current rather than the dissipation, when you have a 20% duty cycle.