# Driving series parallel combination of LEDs with a constant current source

I came across some products (fairly reputed) that use a series-parallel combination of LEDs and drive it directly using a CC source.

Here is the configuration of LEDs:

A series string consisting of 12 LEDs. 16 such strings in parallel. A total of 192 LEDs are laid out on a 30 cm X 30 cm metal core PCB. Thermals are good. After running for over an hour, the MCPCB back is roughly 60 degree C (140 degree F).

LED used - Samsung LM301H (Product page) (PDF datasheet here)

Vf of LED is roughly 2.7 V.

Sorting current - 65 mA but max current mentioned as 200 mA.

Power supply used is a meanwell 36 V 2500 mA CV-CC PSU (PDF datasheet here). I feel that this PSU runs in CC mode because PSU can deliver enough voltage (12*2.7 = 32.4 V). This design is relying on the wattage limit of meanwell driver so that driver hits the current limit and starts running into CC mode, in my opinion. When this happens, each LED will be getting 2500 mA / 16 = 156.25 mA (well under 200 mA limit).

LED datasheet mentions the bin voltage as a range with 0.1 V gap. (for ex: 2.7 to 2.8 V).

Is it okay to wire up such LEDs in the discussed series-parallel combination? Will the current distribution happen evenly in such case? (quite likely the string's collective Vf will be close to each other because of randomness but I want to understand how much Vf difference is acceptable for such designs to work reliably)

Datasheet says:

Samsung maintains measurement tolerance of: forward voltage = ±0.1V

These LEDs are sorted (binned) by Vf by the manufacturer (also color temperature, tint, efficiency, etc). So if you order a reel of them, all the LEDs will be well enough matched that you can get away with wiring them in series parallel strings without resistors.

Of course Vf depends on temperature, so the heat sink should be designed to keep them more or less at the same temperature, or the series strings should span the thermal gradient so they each include the same number of hot and cold LEDs, for example.

If you want to calculate some probabilities, let's assume the voltages are uniformly distributed in [0,1] interval for normalization, and plot the probability density function of the sum of 10 samples:

If you use long LED strings, the law of large numbers is on your side.

Note the pdf of the sum of two variables is the convolution of both pdf's, so just run:

from numpy import *
p = ones(11)
q=convolve(p,p) # 9 times
plt.plot(q); plt.show()

• Nice overview. I'd add that it still could be (with bad luck) that one string is -1.2 V and another string is +1.2 V, relative to the average. That would possibly impress 0.2 V per LED on a string which may impact current sharing enough to be a bother even with the binning. That's a very unlikely situation, though I've not tried to quantify just how unlikely. I'm sure that the "well-reputed" manufacturer also tests the assembled units before packaging just in case that rare event happens. Anyway, I think you hit the main two points right: binning and thermal arrangement vs edge effects.
– jonk
Apr 5 '21 at 18:33
• @jonk - Given that there is a probability of +- 1.2 V mismatch, what would be the corrective action if I come across such a situation? Desoldering and soldering new set of LEDs doesn't seem like a practical solution for bulk manufacturing? What is the Vf difference (between various strings at worse temps while running) that I can get away with without risking the product quality? Apr 6 '21 at 6:51
• The law of large numbers is on your side. It's a very low probability. Apr 6 '21 at 7:35