Related
Recommended

Keithley 2450: LED characterization

neuromodulator

 

 

3.4 LEDs

 

Diodes are two-terminal electronic devices that facilitate current conduction in one direction and block it in the opposite direction. Their electrical characteristics can be described by the Shockley diode equation which defines the relationship between current and voltage across a diode:

 

 

Where:

I is the diode current.

Is is the reverse bias saturation current.

V is the voltage across the diode.

n is the quality factor.

VT is the thermal voltage (~25.7 mV at 25 ºC).

 

I-V characterization of diodes can be very challenging as voltage affects current exponentially (according to the equation). At low currents, unless shielding and guarding is used, measurements begin to get dominated by noise. As I do not have a test fixture, I used inexpensive alternative approaches to reduce the noise: keeping leads short, and filtering (averaging multiple measurements).

 

 

 

Of course this can not compete with a proper shielded and guarded test fixture, but it still noticeably reduces the noise level.

 

Another difficulty that I had to solve was the photoelectric effect caused by external light hitting the LED, I reduced the effect by keeping ambient light low and covering the LED with a small metal box.

 

 

3.4.1 Steady state LED I-V curve variability

 

Electronic component manufacturing is an imperfect process that produces components with different electrical characteristics. To study the variability of LEDs I-V curves, I measured the curves of 6 groups of 8 LEDs each. Voltage sweeps were made with KickStart and saved on separate CSV files for each measured LED. I found KickStart particularly useful for this task as it allowed me to graphically verify that the LED was making good contact to the terminals and that I did not reverse bias the LED by mistake (keep in mind that I measured 48 LEDs, so the chances of making mistakes are high). I used Python (LedColors.py) to parse the CSV files and plotted the results.

 

 

 

Interestingly some LED I-V curves match each other perfectly (ie: yellow LED), while others have a lot of variability (ie: white). Its also interesting how some curves spread at low currents but converge at higher currents, while others behave the opposite way.

 

 

3.4.2 LED step-stress and I-V characterization

 

Overdriving an LED can accelerate its degradation rate, but not necessarily if it is done for very short periods or if the overdriving current is not absurdly high. I programmed the SMU to run a loop where at each iteration it progressively current overdrives an LED and then measures the I-V curve to see if damage can be detected as a shift of the curve. Each loop iteration caused more damage to the LED until it completely failed. The loop iterates through 3 phases:

  1. Overcurrent phase: Generates a 60 s pulse at a constant current that begins at 0 A in the first iteration, and increases 10 mA on each new iteration. A custom trigger model was created to tightly control the overdrive timing and measure the voltage.
  2. Cooling phase: It gives the LED 120 s to cool from the overcurrent phase, so that the I-V characterization is performed at ambient temperature. This phase was implemented with the delay() function.
  3. I-V characterization phase: A voltage sweep with a 10 s delay per point, at increments of 0.1 V with a current limit of 20 mA is performed. smu.source.sweeplinearstep() was used to perform the sweep.

 

Python was used to control the instrument, save the data, and then plot the results.

 

LedOverdrive.py

LedOverdrive.tsp

LedOverdrivePlot.py

 

The next plot shows the voltage time series during during the overcurrent phase.

 

 

From 10 mA up to 100 mA the voltage trace behaves in a very stereotypical way, as the junction heats, the voltage drops. From 110 mA up to 140 mA the curve looks noisier and at 140 mA some kind of thermal runaway effect appears to show up. When the LED was overdriven with 150 mA it began to fail after 30 s, at this point the LED was flickering. At 180 mA the LED failed open.

 

Next plots show the I-V curve with the current axis in linear and in log scale.

 

 

 

There are 3 curves that are completely different to the rest, the 150, 160 and 170 mA curves, during the overdrive phase the LEDs began to flicker, so its quite evident that the LED got seriously damaged at 150 mA. Judging by the shape of these curves, they look like a resistor in parallel to a diode. The 130 and 140 mA curves also deviate from the rest, and that could be a sign of LED damage. At low currents the the curves also spread, but its hard to know if the spread was caused by LED damage.

  • in reply to dubbie

    Well you can go as high as you want, but its likely impact on the lifetime of the component. Still it may be long enough that it doesn't matter for your application. For instance maybe for an emergency alerting LED, you may care more about making it really bright than extending its lifetime.

  • in reply to Gough Lui

    Manufacturer sometimes publish recommendations on how to overdrive the LEDs at low duty cycles, but in the end you can overdrive them anyway you like, and of course it will impact the lifetime, and how quick they degrade. In these test I was just trying to show how the instrument could be programmed to perform a simple step-stress test. LIV characterization is something I thought it would be interesting to try, but it would have taken me far more effort to do.

  • in reply to dubbie

    Overdriving LEDs safely really depends on the duty cycle and the lifetime you expect from the LED. On the whole, efficiency usually drops off in terms of light output produced per unit of current once you go above a threshold (sometimes this is even less than the rated current for larger LEDs), so you're going to be paying a price. This price is "doubled" as heat production which further shortens the life of the LED. In the case of plastic-potted style LEDs (e.g. the ubiquitous 3mm/5mm types), the heat cannot escape quickly enough and often the die will overheat - the flickering that is commonly seen can be a result of even the bond wires "crackling" off the die as it melts in place. While some LEDs will survive overcurrent for more than just a transient moment, they usually won't operate anywhere near as long as their datasheet ratings of 20,000+hours. Sensible overdriving tends to focus on a duty cycle - say in the case of LED strobes, it may be possible to go 2-3x overcurrent as long as the duty cycle and pulse lengths are kept short enough to minimise any adverse effects.

     

    It's a good illustration of how much abuse is necessary to destroy an LED ... I did the "reverse" (literally) in my RoadTest where I looked at LED damage in reverse bias and that definitely is quite variable. Each component is unique and sometimes differences in design and manufacture can make all the difference. Other times, some "leakiness" can just be down to semiconductor purity issues or ESD damage.

     

    - Gough

  • A very interesting set of results. It is interesting to see components being pushed to their very limits as this is not something that you would normally do. When over driving LEDs I would normally only go to double the maximum current but your data seems to indicate that much higher over driving might be perfectly OK.

     

    Dubbie