I am planning to measure the voltage of an rgb LED pulsed with an ADC. The pulse is around 20ms long. The pulse is necessary so the self-heating is within reason so it won´t affect the any characteristics. My Question is now what properties my ADC should have like Sampeling Rate, Bits, etc. so my measurement will be good.

• The first thing you need to do is define what "good" means to you. That will be necessary to define all of the ADC specifications. Mar 14 '19 at 14:02
• Why are you applying a voltage pulse instead of a current pulse? If you have a good model for Vf=Vt+If*Rs you only need to know Vt for each colour and measure Rs which may have a tolerance of +25%/-10% for good parts from 10%*If to max for calibrated color mixing Mar 14 '19 at 14:40
• Using a test validation of the Cauer model and tempco, ΔVf/ΔT of -2 to -5 mV/'C (?) you choose an error voltage ΔV from the desired ΔT for a pulse of Δt then measure the Vt at 100uA or 0.1% Pmax to know that your assumption of ΔT~0 is valid. Mar 14 '19 at 14:49
• What does an "analog to digital converter" have to do with pulsing an LED? How would those two things interact? Mar 14 '19 at 20:50
• If you want to measure a voltage peak, a voltmeter with peak-hold feature is an ADC with display and range controls. Mar 18 '19 at 22:14

Regarding your use of the pulse: good idea. Lets examine how long that pulse should be, if you want to measure the LED properties.

A cubic meter of silicon has 11,400 seconds thermal time constant. This is measured by uniformly heating one face, letting the heat flow out the opposite face, and insulating the other 4 faces.

A cubic 0.1 meter (10cm) is 100X faster, at 114 seconds thermal time constant.

A cubic cm is another 100X faster, at 1.14 seconds thermal time constant.

Notice all these sizes --- meter, 0.1meter, 1cm --- are much larger than the size of the LED junction depths.

A cubic millimeter is another 100X faster, at 0.0114 seconds (11.4 milliSeconds). This may be the size of the LED silicon, but what about the junction depth, right below that bond wire?

A cubic 0.1mm ---- 100 microns in size, or 4 mils, or 0.004 inches --- is another 100X faster, at 0.000114 seconds, or 114 microSeconds.

A cubic 10 microns is size is thermal time constant of 1.14 microSeconds.

The 10 micron size may be about the accurate sampling window, if you wish to avoid thermal heating effects.

• Well, i already tried to create a thermal model with ltspice ( Cauer Type). The self heating is within reason. But i am not sure what sampeling rate etc my adc should have in the range of 20 ms Mar 14 '19 at 14:32
• @ValentinHuppertz you should consider taking several fast readings and then you will know how many milliseconds might be "too many". In my experience on power MOSFETs suffering from catastrophic thermal runaway, the device can be "dead" between 100 us and 10 ms after applying power. Mar 14 '19 at 14:43
• @ Andy Some years back I used a grid simulation, in SPICE, to show how an OVER TEMPERATURE detector needs to be within 10 microns of the Power Transistor being protected. And 10 microns is1.14 microSeconds.; the 10 micron size is about the DEPTH of the collector junction for bipolars, and perhaps the WELL depth for MOSFETS. That 10 micron distance is to PROTECT THE DOPING, to shut off the transistor before the temperature is melting the metal and migrating the doping. Mar 14 '19 at 15:10