I would definitely skip the resistors if possible– and it certainly doesn’t look to me like they’re needed in the application that you’re describing.
The note on page 12 is based upon how it is designed to work given power dissipation considerations in the IC package. It is not a statement about the most voltage that is allowed there.
But, let’s see where that voltage value comes from, and what it really means. At the device maximum current of 90 mA per channel, with 1 V between LED voltage and power supply voltage per channel, the total power dissipation is P = I*V = .09 A * 1 V = 90 mW per channel. With 16 channels, that’s 16 * 90 mW = 1.44 W. That value is consistent with the maximum power dissipation for some of the smaller packages. The ‘GN (RoHS DIP) package that we use actually allows up to 2 W, so there’s actually some additional safety headroom for us.
If you wanted to stay below that level of power dissipation with only 30 mA per channel (the limit of those RGB LEDs), you would need to have less than 3 V headroom (on average) between the LED voltage and the power rail.
Suppose that V_F is 1.8 V for the red LEDs and 3.6 V for green and blue, and that we have 6 red, 5 green, and 5 blue LEDs. Then with a 5 V power rail, the average drop is ((5 V – 3.6 V)*10 + (5 V – 1.8 V)*6)/16 = 2.075 V, and the total power dissipation is 16 * 2.075 V * 0.03 A = 966 mW. That’s about 2/3 of the power dissipation in the example given on page 12, and well within the safe operating conditions of the device. If you wanted to run a bit cooler yet, pick a slightly lower system voltage, maybe 4 V or 4.5 V, which will make it so that the MBI5026 will run cooler yet.
