Hi again,
Yes, mounting them on the same heatsink with thermal epoxy connecting the two would help to some degree because that would make the two diodes look like one single die thermally, so the only difference then would be the two bulk resistances and curve.
The following is assuming the diodes are thermally insulated from each other the way most diodes are used...
I wrote up a few equations and used them to try to make some sense of it, and i found out that it looks like when one diode starts out heated more than the other still no thermal runaway occurs, but rather the cooler diode heats up little by little and eventually the main difference is in the diode curve or the bulk resistance. If both diodes have the same exact specs, the two reach the same temperature. I think that's encouraging too.
One of the problems would be if one of the diodes got too much of the bulk of the current and that's probably where the real problem lies. If the diodes are over sized (relative to 50 percent rating) by say 150 percent it may work out ok in most cases. That would mean that for say a 4 amp diode two 3 amp diodes are used for example.
With even slightly different characteristics such as one diode changing by 2.2mv/degC and the other by 2.3mv/degC (internally of course) i was seeing about a 3 deg C difference in the final equilibrium temperatures. With one diode with 0.16 bulk resistance and the other 0.20 bulk i was seeing around 13 deg C difference, but the current ratios were as far off as 64 percent in one diode and 36 percent in the other. That would mean that if both diodes were rated for 75 percent of full load they would survive.
If one diode had it's curve off more than the other plus an unfavorable bulk resistance the situation became worse however, but interestingly, if one diode had it's curve off in one direction and the other diode had it's resistance off in a favorable direction, the currents could be evened out again!
A second way the currents would be evened out again turned out to be if the one diode had it's curve off and it's bulk resistance off in a favorable direction.
The bottom line here is that depending on any of the possible combinations of curve and bulk resistance between two diodes some combinations will work much better than others, and of course some may fail.
So it looks like it's a bit of pot luck, if we get two diodes that work well together we may get nearly even current sharing, but the other extreme is we get two that work very unfavorably together and we get serious current differences.
The other interesting thing this brings out is if we have two diodes that work well together and then try to use one of them with a third unknown diode the new set may fail even though the first set worked fine. This also means that the two diodes dont necessarily have to be 'exactly the same', but they simply have to draw the same currents when in parallel.
Can you say, "Current Probe"