Posted on July 03, 2015 at 17:44The Arduino connector does present a fair few signals, again perhaps end-running the need for a Nucleo offering.
To your other question, they design the part to be of as general use as possible. The different markets for the part use different subsets of the functionality, rather than all at once, and in some cases use very low pin count devices, again restricting the external escape of many peripherals.
If your application doesn't require external memory, and display, then that cuts the pin usage significantly.
The different peripheral and pin muxing options allow different combinations to be fit in the part. They can also help board designers escape pins in a way that eases layout. The choices are somewhat limited, but there is at least some flexibility.
The peripherals in the STM32 are more generic and lamer than more custom solutions. The fact that the timers only have a single counting element means you need more of them to generate different frequencies. I can see applications that could eat 5-6 USARTs. The parts are designed to appeal to everyone, not just niche groups.
The semi business is about scale, selling vast numbers of the same thing (die), run through the diffusion process in very large batches. While there's a trade off between die size, and parts-per-wafer, you also don't want to be changing the masks around, or reconfiguring the wafer level tester.
Putting the same die in different packages is a lot easier, and then marking them based on how much time you want to spend on the tester vs the sell price. All the F405/407 chips have 1MB of FLASH, some are sold as having 256KB, testing the flash takes a long time, you're going to be able to deliver 4x the quantity of the 256KB parts out of the tester in the same time. As this is a high cost bottleneck, they can price the parts accordingly.
