This is why it is important to use a driver that regulates the current through the diode, rather than the voltage across the diode. Vf actually
drops as temperature increases, so in an extreme case a process called thermal runaway can happen -- At a fixed voltage, the diode will draw some current and start to heat up. This will cause the Vf to drop, which will cause the current to increase even more, causing the temp to increase even further in a cycle.
Not only must your driver regulate current, ideally, you should have a separate current regulator for
each diode (or string of diodes) in parallel:
[ driver ] -+----+
. . . . . .[LED].[LED]
. . . . . . . .| . . . |
. . . . . . . .[Ground]
is BAD
[ driver ] -+[LED]- [Ground]
[ driver ] -+[LED]- [Ground]
is good
In the first case, one diode can heat up more than the other, causing it to take a larger fraction of the current (say 75/25 instead of 50/50). If you have many diodes in parallel, driven by a current source, you can burn one out due to thermal runaway. Or worse, if for some reason one of your two LEDs in parallel (driven by the same current source) becomes an open circuit due to bad wiring or something, the current source will overdrive the other one by double, as it's only trying to control
total current.
The one exception where running multiple LEDs in parallel is acceptable is with multi-die LED, such as the P7 or the MC-E. In that case, the dice are all on the same substrate so they stay the same temperature as each other. Also, the dice are fabricated at the same time, so their characteristics should be very well matched. The problem is putting separate LEDs (which can have different temps than each other, and different properties than each other) in parallel.
