This is an old revision of the document!
For best results put panels of different specs on separate MPPT controllers.
You are blindfolded and placed in a field that is flat except that there is an anthill somewhere in it.1) Your job is to find the highest point in the field.
The simplest approach would be to walk until you start going up and stop walking when you start to go down again. You might shuffle a bit to make sure you are on the very top of the anthill. It will take a while but it's not complicated.
This is like having one panel or multiple identical panels on the same MPPT. There is one anthill (one power curve, the blue line in the image) and one highest point on that anthill (Maximum Power Point on the curve).
Unbeknownst to you there are really multiple anthills. You could perform the walking algorithm perfectly and still still be on a low anthill with no way to know it.
This is like having mixed panels on the same MPPT controller. There are multiple I/V curves laid on top of each other making, in effect, a lumpy and unpredictable curve. Normal MPPT have no way to know they are on a lump (local maxima) rather the actual top (maximum non-local power point). You might luck into the max but don't count on it.
If they tell you in advance there may be many different anthills you can explore the entire field and remember each hill you find. Eventually you will find the tallest one.
This is what happens when you use an MPPT with an advanced algorithm that makes full sweeps of the I/V curve instead of looking only for local high spots. The downside is it cost.
During these extended sweeps harvest falls off dramatically because the algo has to check the whole I/V curve, including points that make little-to-no power.
This algorithm will find the MPP of the mismatched array but cannot make as much power as different panels/arrays on their own MPPT.