Which you do by not using a 16-bit or 32-bit driver but settle for a 8-bit driver if you intend to build a multiplexed display capable of use outdoors or in strong light. If the design need only support +30°C ambient temperatures in a shaded environment (shaded from heat and strong light) then you can run the LEDs with reduced current (averaging way below the datasheet limit) which both reduces the heat from the diodes and the heat from the drivers.


It isn't a constraint of the LED that it will give dimmer light if it isn't lit for 100% of the time. With a constant efficiency figure, common natural laws would suggest that for every reduction in the pulse quota, you must compensate with corresponding power increase in the pulses to keep the average power through the LED the same.

But a driver chip that only has to drive one LED with 0.1W for each output will be way cooler than a driver chip that has to handle 8 LED for each output and has to light each LED with 0.8W for 1/8 of the time to make sure that each LED on average gets 0.1W when lit.

So in the end, you as a designer have to think about:
- Required sign intensity.
- LED intensity at max average current.
- Max pulse quota for the LED, to figure out how much extra current you can add (or loss of intensity) if multiplexing
- Max current the driver chip can handle.
- Max power dissipation of the driver chip.
- Available board space for your hot driver chips.
- Max ambient temeprature.
- Max internal heat rise from diodes and driver chips.
- Worst possible convection cooling.
- ...

But since everything is physics, it really is natural laws. The 16- and 32-bit driver chips can not be used in some designs unless you use active cooling or accepts a dim sign. Using heatsink + fan or water cooling are possible choices. But not choices that are likely selected. Finding a 16-bit driver chip in a more than twice-as-large capsule with half the thermal coefficient would keep the status-quo but then you wouldn't get any PCB space by switching from a 8-bit to a 16-bit chip.

If the LED manufacturers can manage to squeeze out more light from the diodes without increasing their costs (and without dropping the ratio between max pulse and max average current at the same time) then signs can be built that will shine at the required intensity but at lower drive currents.

In the same way, product improvements may allow new drivers to be designed that can work at higher temperatures without the component price going sky-high at the same time.

So while this is a question of physics, it is definitely possible that future 16-bit or 32-bit drivers with current form factors can be used in signs capable of working in direct sunlink (surviving heat and having enough intensity to be visible). But we are not there yet, so Eriks note about problems with the bigger drive chips is very relevant.