- due to their silly habit of placing them in parallal with no balancing resistors.
Very silly indeed,
- the reason
all diodes and LEDs of one type have different Vf tolerances at max power is
only due to the internal bulk resistance,
Rs and the temperature shift of the negative temperature coefficient (NTC) of the semiconductor threshold voltage which controls the junction temperature with thermal resistance.
Anyone can use linear regression with Rs and Vth to get
Vf=Vth+Rs*If from the power rating of any diode type using Rs= k/P_max then choosing k for some value 0.25<k <=1 (if you know the Vth or knee threshold Vk) ( the more complex formula includes Temp and NTC)
The % tolerance is a manufacturing process limitation and capability of the epitaxial making equipment and any subsequent binning in the process. This is not published and requires a component engineer's skill to understand how it affects the design when doing cost-reductions.
The other major factor in power rating is the thermal resistance or conductance as parallel diodes heat up by thermal resistance * power, the ability of tracking each other vs independent maximum air flow vs , no thought at all about air flow and ambient temperature inside the set may cause this positive feedback effect called "Thermal Runaway" where the hotter diode drops to a lower voltage from the NTC effect which tends to draw more power. There is a formula for the critical threshold of thermal runway, but I forget at this moment. The "balancing resistors" are about the same value of Rs of the diode in question, in parallel which reduces the rate of change of temperature with current by 50% by doubling the total resistance and thus reducing the deviation or effective tolerance of mismatched series resistance for each diode.
If semiconductor or foundry datasheets actually put tolerances back in the datasheets, it would help, but they tend to simplify specs now and
only include maximum values for Vf @ I to reduce cost and not tell you the real reasons why. It may be they outsource the production to a cheaper foundry with wider tolerances or reduced testing or buy from a different epitaxial wafer source.
When I could rely on diode tolerances, I could support designs that put them all in parallel at max current. I did this for one client who used a million of my different LEDs over a 10 year span and put them into products to light up sidewalks, tunnels and roads. But then the LEDs were 100% tested and binned in 0.1V bins and often tested within millivolts of each other , better than any discrete Zener. I only bought binned brightness +/-10% and +/-0.05V voltage at some premium to guarantee performance and also got reverse ESD protection diodes added to each LED due to sloppy customer handling processes. The premium wasn't too much as I chose the bins with the largest population that met my criteria and had close communication with my supplier for orders in the 20k ~50k pc per type range.
This " tangent " report is brought to you by Nescafe.