Product longevity of STM32F76xx

[SR NA.2]

Kindly provide Product longevity of STM32F76xx series MCU

[Peter BENSCH]

Welcome, @SR NA.2​ , to the community!

for all STM32, including the STM32F76xx, there has so far been a rolling commitment for guaranteed production over the next 10 years, currently guaranteed until at least 2032.

However, one must not confuse longevity with scarce current availability.

Does it answer your question?

Regards

/Peter

[Tesla DeLorean]

Perhaps convey to management that ST needs to start getting a lot more transparent in what it's plans are moving forward here.

With 12, 18 or 24 months of runway at lot of us are likely to decamp and find alternatives in the mean-time, and ST's going to be out of the design refresh.

Most companies dealing with consumer products don't work on 10 year availability expectations, or don't have staff retention to sustain that. All the institutional knowledge leaves or gets fired.

Vendors like Micron have a limited subset of parts they guarantee availability on, and one of their strategies is to die-shrink to reduce cost, and expand availability, via equivalent form and function.

Is there some prospect that the F1 or L1 families will be die shrunk onto tighter geometries?

[Peter BENSCH]

@Community member​ 

Very interesting question!

Several points have to be taken into account:

• While memory devices are high-density logic, they are much less complex in terms of different functions than general-purpose microcontrollers, and they are not mixed-signal designs. If you were to shrink a microcontroller, this would also result in different parameters of the analogue functions, which would then no longer be in accordance with the original data sheet, resulting in a completely new component that would have to be completely re-qualified.
• In addition, the STM32F1 and STM32L1 are based on the Cortex-M3, which is no longer used by ST in new MCUs.
• Another point: the so-called pad limitation. If you make the core logic smaller, at some point the area required for the pads of the bond wires outweighs the space, which makes the process no longer economically viable.

So does it really make sense to think about smaller structures in MCUs?

Regards

/Peter

[Tesla DeLorean]

>>So does it really make sense to think about smaller structures in MCUs?

Well if you make the die smaller the yield per wafer increases geometrically, aided by how square and rectangular die tessellate onto a circle.

A lot of the F1 and L1 parts are relatively low pin count,and the bulk of the die size comes from FLASH, RAM and peripheral logic. Yes I can imagine the pad ring, and packaging can be an issue, I just don't see the will to do anything as the biggest obstacle.

There should be some serious reflection into the RP2040 approach, rather than the kitchen sink strategy ST has embarked on, especially if they expect to retain any of the low end / simple General Purpose MCU market share.

Most of the original STM32 market share came from ATMEL being completely dysfunctional 15 years ago. They and MICROCHIP are a different adversary today.

>>In addition, the STM32F1 and STM32L1 are based on the Cortex-M3, which is no longer used by ST in new MCUs.

But that's a choice. uBlox is using the same CM3 core they used a decade ago, in a current TSMC process that's clocking at 384 MHz, they didn't go to the CM4F because single precision floats are completely worthless for mathematical problems, and none of the ARM FPU support transcendental.

Then again the F1, L1 still have significant demands

[waclawek.jan]

Clive,

note, that some of it is just glitter, not gold. The 45nm node comes with drawbacks. Good for gadgets, bad for the slightly better stuff.

JW