Working on a project recently I ran into an issue with a lack of on hand components and a desire to simplify my project so that you didn’t need to look at mounting complex components.
The project I am playing with is pretty simple, a 5V USB powered lamp to be used off of a USB power pack that everyone seems to carry around nowadays or driven from a USB power source like the Artesyn DCH5. I know that I can run a single LED easily as the forward voltage drop is well within the 5V range, but I wanted to add a bit of a challenge and I wanted enough light to read a book by or navigate a tent while out camping so I set up my designs for four LEDs.
LEDs in Series vs Parallel
Usually when adding multiple LEDs into a system you want to run them in series, this allows for consistent current across each LED so they light more uniformly and it prevents an LED from hogging current and burning out, but this setup also causes the forward voltage drop of the LEDs to add up and quickly pass the 5V source I have available. Lower Vf LEDs can be used, something like 1.2Vf would work, but finding this low of a forward voltage in a visible spectrum LED, much less a white one, is very difficult. The LEDs I have on hand are around 3.2Vf meaning that running four of them like I was planning will not work in series.
You can run with a higher Vf than your source will support if you are willing to utilize an LED boost driver like the Linear Tech LT3491. There are quite a few that will take a 5V source and push out more than 20V but I was hoping to avoid adding an IC to this very simple project. The LEDs I am using are very similar to the CREE C513A family of parts but I have a 3mm diameter variant.
Parallel vs. Series
Connecting LEDs in Parallel
Not wanting to use a boost driver I was left with driving the LEDs in parallel. Driving LEDs in parallel allows you to maintain a lower forward voltage since the voltages do not stack in parallel but you do increase your current source draw to maintain a similar overall wattage. When driving LEDs in parallel you can run into situations where the internal resistances of the LEDs are not matched. Mismatched internal resistance will lead to one LED hogging the current from the power source since the power will look to take the path of least resistance. By hogging the power, the LED will have a shorter lifespan and may burn out very quickly or immediately.
One way to combat this is to check your LEDs beforehand and bin them by internal values. To bin your parts or match them up with ones most similar to them, you utilize a constant current power source and a decent multimeter. For the LEDs on hand I knew I planned to run them at their rated 20mA so I fixed my current source there and input an over voltage limit at 3.6V since my parts should not go beyond it. Hook up the multimeter at the same positive and negative points on the led as your power supply to measure the actual voltage with some accuracy and resolution.
Once you power on the constant current power supply the LED should light and the power supply will adjust the voltage across the LED as needed. The LED is a diode device meaning that its characteristics will settle a bit as it warms up, so I waited for the output voltage to stabilize around a consistent value. On a sheet of paper, I taped down components I had tested and wrote the resulting values from my testing. I selected the 4 LEDs with the closest stable voltages at 20mA. To simplify this whole process some suppliers, Cree for example, will actually offer to bin parts for you by forward voltage and color temperature so you get the most consistent results possible.
Running a quick test.
From my measurements I found some interesting details. I was using a batch of LEDs from the same manufacturer and out of the same box, they may have come from different production runs but that is not easy to discern from what I had on hand. After measuring the LEDs which were spec’d for a 3.2V forward voltage drop I ended up with a range from 2.991V up to 3.127V at 20mA meaning I had over a tenth of a volt variation among my parts and none of them had the expected 3.2V drop at 20mA. If i boosted the current to 30mA I did see drops closer to the expected spec meaning I can probably drive these harder than I planned to but for the moment my design does not require it. With this range of voltage drops I would see internal resistances varying from 156ohms to 149ohms. This range is pretty small but as current increases the differences are more pronounced and could lead to system issues.
Highly scientific documentation in action.
When using LEDs in your projects and systems make sure to characterize them as best as possible especially when driving them in parallel without some sort of intelligent driver IC. By binning your parts, you can create higher quality systems that are more durable and stable.