I have years of experience in asm programming with PIC processors, and my job is embedded software development. I got an STM32 development board a couple years ago to get my feet wet in ARM embedded development. I got completely bogged-down in setting up the development environment. It took me forever to get a blinking LED, and I spent an ungodly amount of time unsuccessfully trying to figure out how to change the rate based on a potentiometer input.
It was a very frustrating experience, but articles like this make me want to dive in and try again!
Do you mean in assembly? If you just mean C, then if you’re already in embedded then I’d say a few years ago it was easy. These days it’s super easy - tools like STM32CubeMX will allow you to get even custom hardware up and running in no time.
It will even spit out multiple project types - everything from ARM Keil/MDK/uVision to just a simple makefile.
I disagree that GUI tools like STM32CubeMX are useful. They end up being a crutch that end up adding increasingly complex wizard driven autogeneration, rather than just exposing the configuration in easy to consume ways. Sure, you can get something running fast, but it’s incredibly difficult to move from “blink a led via this template project” to “build a useful project integrating multiple peripherals”.
I think there’s some actually interesting work being done with standardization like SVD files, but too many vendors treat them like second class citizens compared to their bulky code gen solutions.
I think it’s decent - configuring a clock tree and initialising the peripherals by hand might be a good way to learn the chip but in terms of rapid prototyping I’ll take the GUI driven leg up any day.
This is true! If I was writing something to be optimal speed- and size-wise, I'd do that. I think I just liked the consistency with the other parts of the code where using both was necessary :)
I did something similar with CMake and a lot of the details are incredibly challenging to piece together. Everyone just blindly copies existing linker/startup scripts. Very low level articles like this are rare and invaluable
I spent a month and got things working with a modern toolchain C/CMake/GCC/OpenOCD but gave up on further work bc I couldn't nail down a lot of details. micro programming outside arduino is painful and backward. it's stuck in the 90s
This was a really interesting article, and I learned a lot. I'm very much lacking in chip/assembly level knowledge, and this helped me to understand a little bit more. Thank you, Lochsh, for writing it!
Fantastic article, thanks for taking the time to write it up! I'm trying to learn about embedded systems at the moment and it feela like a super steep learning curve, really appreciate your post breaking everything down.
A little typo I think I spotted:
' So, to turn on our LED we want to set the BR8 field, and to turn it off, we want to set the BS8 field.'
Should this be: ' So, to turn on our LED we want to clear the BR8 field, and to turn it off, we want to set the BS8 field. '
Ahh I see now, thank you for teaching me. My mistake!
Do you happen to know of any great learning resources, basically more posts like you've written that go into lots of details and explain why things are done?
They use IP from synopsis, but <rumor>synopsis refuses to allow their docs to be published so every manufacturer has to read them and regurgitate them into their own docs in their own words</rumor>. In any case official STM docs are incomplete. See their driver source - it accesses undocumented registers and sets undocumented bits in documented regs. Without them the usb core will not start or run.
Your options are to carefully rewrite STM's libs (which are based on real synopsis docs and DO work) or use them as is. Both solutions IMHO leave you subject to STM's EULA and such
this is the #1 issue in stm32
a close #2 is the their so-smart-it-is-useless i2c controller. I know of no project that uses it. Everyone bitbangs i2c master on stm32. The hardware unit is very easy to wedge to a point where only a power cycle unwedges it.
I think it's worth noting that there are multiple I2C peripherals by ST -- one for STM32F1, F4, and L1; and a different one used for STM32F0, F3, F7, L0, and L4.
I’ve recently used the “HAL_I2C_IsDeviceReady” function to scan the whole address range to find devices on the bus with no issues. I’m not near my PC to double check but I’m pretty sure that does what you’re talking about and seems to work ok for me - unless you’re suggesting it’s intermittent?
(I've chosen the version for STM32F4 arbitrarily; if there's something similar in one of their other USB support libraries, feel free to point to that instead.)
Avoid ST at all costs in my experience. Their documentation is terrible. They never acknowledge hardware errata even after clear evidence of it. They’re not a company with a solid engineering culture imo
I'm really surprised to hear that. ST has always been one of my favorite manufacturers. I find their datasheets to be clear and well laid out, though I will note that I'm usually more interested in the hardware side than the software side. Their chips and boards are also high quality and convenient, e.g. GPIOs are often five-volt tolerant, Nucleo boards come with well-made breakout headers and several peripherals (at seemingly impossibly low cost), chips have wide supply voltages, etc. I can't say I know anything about their errata (I've never run into any), but even if the situation is as bad as you say, that's quite a harsh condemnation, especially considering how widely-used and successful they are in industry. Do you have other complaints?
I’ve used their accelerometer (LIS2D). Performance and specs are great - it was just a pain to develop the embedded interface with it because the documentation had a bunch of mistatements and was missing critical information (a weak pullup on a certain line caused a bunch of current consumption - took days to figure out). Also they had an off by one bug in their LIFO queue that took a long time to figure out. Once you getting it working its great. Last time I checked they still hadnt updated their documentation with what I reported to them...
And they actually acknowledge hardware errata in the documentation.
More so for Microchip than Atmel --- it's pretty common to find on some of their newer and more complex parts a ton of errata, among which there are some extremely "WTF!?" ones like "feature X does not work at all".
Yeah I love those. They highlight features xyz on first page of datasheet. Then on the last few pages its like btw features xyz dont work on rev A-F. And your supplier doesnt know what revs they have...
The STM32 USB libraries are provided under the oddly named "Ultimate Liberty License" [1], which is actually incredibly restrictive -- it requires the code to only be used on ST hardware, and prohibits it from being used "in any manner that would subject this software to any Open Source Terms".
I don't have experience with this, but from talking to people who do, there are indeed open source implementations. They say the documentation is a bit confusing, but it does describe everything.
That's what "clean room reverse engineering" is for. If you document what the code does, then code written based only on those docs wouldn't need to be under the same license.
Is clean-room reverse engineering even viable? That is, would anyone agree that an independent discovery of what bit does what is even possible, and this is not lifted from known drivers?
Clean-room reverse engineering doesn't require that. If someone were to read the existing drivers and write documentation explaining how the hardware appears to work, that documentation would be unencumbered, and a freely licensed driver could be written based on that documentation. The only methodology that would be questionable would be a developer writing a new driver while referring directly to the vendor driver.
To put it another way -- the IP protections on the current library are a matter of copyright law, and the protections are on the code itself. The facts of which bits do what -- which are embodied in that code -- are not protected (or protectable!), and can be reused freely.
On ARM Cortex M4 devices like the STM32 there is a simple timer called SysTick used for exactly this type of purpose! You only need to set up a clock divider and you start it up- now you have a regular interrupt at whatever frequency you need!
The other approach to delay's (keep in mind, that there is no other process to yield the CPU too) is to use a timer (a peripheral of the micro controller).
This would take a fair bit more code to configure the timer, and setup an interrupt to handle the timer.
Microcontrollers (assuming no RTOS in use), typically don't have multiple processes, they have the main thread of execution (which in the article is the entire program) and have interrupt routines that a run in response to an external event (in this case, the timer has expired).
Using the interrupt driven approach, can led to better performance both in CPU time (async communication with slower periphials etc) and battery life (sleep states).
by either having the timer fire regularly and have run code to figure out what needs to be done now, or having code to set it to the time to the next needed interrupt
Jokes aside one of the easiest ways (for developer) to do so is just run JS on it since it have sleepy event loop built in. Works wonderfully with Espruino. So sad that it is not as widespread as should be.
I think you missed the point, which in this case would be "how does that js engine does it". If it does it the same way then it doesn't answer him at all.
You don't need an OS to do more advanced or efficient sleeping. You could have one of the timer peripherals generate an interrupt on a regular basis, and then put the chip in a lower power sleep mode in between interrupts.
You can set up a wakeup interrupt using a timer by moving a few values into a few registers. It doesn't require anything like an OS. Even tiny 8-bit microcontrollers can do this.
Sure, but putting the chip to sleep is one instruction, and programming the timer interval is only about a dozen instructions. So you don't really need an OS in this context.
It was a very frustrating experience, but articles like this make me want to dive in and try again!