This is a good article but it's a few years old, and out of date now. I can't see its date with my current adblock settings. It would be great if someone could find it and put it in the title.
There are stupendously powerful $1 microcontrollers now, including the RP2040 (dual core cortex m0+ with 264KB ram on the chip), ESP32-S2 (Espressif core with 400k of ram and wifi), and ESP32-C3 with wifi and RISC-V core and again something like 400k of ram.
Jay Carlson also has a series about microcontrollers as cheap as THREE CENTS, i.e. the famous Padauk 3 cent MCU:
> It would be great if someone could find it and put it in the title.
There's no publication date -nor anything marking it as updated at any time afterwards- on the article itself, but first comments are from November 2017 and passing references in the article point to "end of 2017" too.
The article about the Padauk is interesting. Especially the description of the FPPA feature. Multiple CPU contexts? This sounds suspiciously like SMT to me which would be a total novelty for a micro controller (afaik).
The ESP32 has two separate cores, not SMT. But look at the Parallax Propeller which goes back quite a long way. Also the Padauk is not exactly SMT afaict. It's more like time slicing. The idea of SMT in eg. an x86 is that one thread can run while another is blocked waiting for memory.
Sounds like a barrel processor [1], used to hide latency of individual instructions to allow code to be written without having to think as much about instruction scheduling / dependencies.
The Xmos chips give the programmer virtual time sliced contexts over one or more hardware core(s). Say you have a 400 Mhz base clock and 4 threads, each one would see the wall clock time advance at 100 Mhz. This is useful for having concurrent realtime operations on each hardware thread w/o having to resort to trickery.
These chips are ideal for DSP, phased arrays and beam forming. Most of their devkits are centered around voice applications.
It has an interesting trade-off: no internal flash. But I've found that I add a SPI-flash chip to the MCU in many designs, so it's a reasonable choice.
NXP has rt1020 (Cortex-M7, no flash), or rt1024, which is same but 4 MB flash in the package- it's connected to the SPI-flash controller of the rt1020 core, so I think it's a multi-die package.
The RP2040 actually has built in 12Mbps USB 1.1, no serial bridge required. But if you do want to use a serial to USB bridge the max baud rate is 921600.
Those are the three micros I use for projects/products, too. A while back I bought a few hundred of the Padauk chips with the intent of doing a small manufacturing run of a side project, but I haven't seen them in stock at the normal places for a while. It looks like the lead time for cheap micros in general is about a year.
There are stupendously powerful $1 microcontrollers now, including the RP2040 (dual core cortex m0+ with 264KB ram on the chip), ESP32-S2 (Espressif core with 400k of ram and wifi), and ESP32-C3 with wifi and RISC-V core and again something like 400k of ram.
Jay Carlson also has a series about microcontrollers as cheap as THREE CENTS, i.e. the famous Padauk 3 cent MCU:
https://jaycarlson.net/2019/09/06/whats-up-with-these-3-cent...