-
Notifications
You must be signed in to change notification settings - Fork 208
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Allow setting external 32khz quartz for slow rtc clock #2032
base: main
Are you sure you want to change the base?
Conversation
We might be able to place this into |
I'm gonna adjust this draft once the other pull request is merged then |
@bugadani I have questions on how to exactly do it Lines 726 to 742 in 08d406e
Here, I should probably add the enum |
I'm not very familiar with the rtc API, but I think you should add |
Oh, this will not work at all because post_init still configures the RTC clock with the default RtcSlowClock, there is no way to pass to it the config |
@MabezDev can we get rid of post_init in favour of esp_hal::init? |
Why does it exist in the first place? |
To run code before main but after initialisation. Useful for stuff like psram and until now, watchdog. Rather than get rid of the post_init , maybe just move the watchdog disabling code to the init function and have that be configurable. |
Okay, I did it for the esp32c6 for now. Does it look fine? If yes, I can do it for the rest of the cpu's. Also I tested it on esp32c6-devkit-m-1.0 with a soldered in xtal 32khz, tested if quartz turns on with an oscilloscope, it did. It required a special bootloader as I said, otherwise it won't output hello world. |
Co-authored-by: Dániel Buga <[email protected]>
Silence means it's fine, right 😃 ? |
I think so. Please rebase :) |
@@ -0,0 +1,50 @@ | |||
//! This turns on the 32 khz xtal and then just prints hello world to UART. Requires a special bootloader |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I don't see the value in this test. For one, what's "a special bootloader" that is required for it? Next, what does this example show, that changing the clock source doesn't break the whole system?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The bootloader needs to be build with CONFIG_RTC_CLK_SRC_EXT_CRYS=y
Well yes, also allows to measure the sine wave with an osciloscope
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Can we select the clock source here, instead of relying on the bootloader?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Not without losing some optimization https://github.com/espressif/esp-idf/blob/59e18382702b6986be3d3f55e9ac7763c1397cf7/components/bootloader_support/src/bootloader_clock_init.c#L82-L90
And maybe something more is behind it, I don't know
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
if (!rtc_clk_32k_enabled()) {
Unless that takes long to decide, this is safe to do. If the bootloader enabled the clock source, and the user selects in in esp-hap, we can skip initialization. Otherwise, in the future we might allow other bootloaders which may or may not set up the clock source, so I'd say let's enable the clock ourselves, if it's not done already.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Please note it's inside the macro statement #if CONFIG_ESP_SYSTEM_RTC_EXT_XTAL
Well maybe it is safe but we lose some optimization, I don't think we should lose any performance because at default we don't control the bootloader fully. In the future we probably will by using a rust based bootloader so it will work the same just like now, but without the hassle of recompiling it
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I don't think we should lose any performance because at default we don't control the bootloader fully
Let's prefer the option that works by default. If the user has problems with the startup time, they can use a custom bootloader.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Making it work only in the application would require changing the code for it to wait to start up, not only is it not efficient but also I don't know how to do it, I looked at a lot of esp-idf code and haven't really found anything that waits for it, I ques booting the bootloader is enough time for it to start after all. Using a fixed time is also bad, would require cpu dependent testing to optimize it.
This is simply not the way to do it...
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Fine. At least please print a warning or panic if the bootloader forgot to enable the clock source.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Ok, I will try to find a way to detect it (probably just failing the calibration step)
Does this look fine? It's able to return to the default clock and everything seems to work fine, but I can't guarantee it Also how do I enable logs from esp-hal, the log feature is already enabled when running examples |
To enable logs via |
Shouldn't there be an option / feature for esp-hal to enable it by itself? The logs I want are being executed before user code. Basically show esp-hal logs if the user wants it, because otherwise it's not possible to enable them directly from the user code and I have no idea how early can I put the log init for it to work Well anyway Does the esp32c6 code for setting the 32 khz oscillator look fine? If yes, can I implement it for other cpu's. I don't want to rework it later for all of them again |
Enable it before calling |
I think this PR should contain the code to initialize the external quartz before we can merge it, depending on a custom bootloader to do it isn't acceptable IMO. Is this something you're willing to work on @Szybet or shall we close this? |
I definitely want to do something with it, but working on a custom rust bootloader? that's out of my league. Maybe just accept the custom bootloader? both probe-rs and espflash support a custom bootloader binary argument |
Hello
This is just a working draft for setting the external 32khz quartz for slow rtc clock. I would like to discuss some potential fixes for issues I encountered. The final goal is to get this merged some day, some how.
So to the issues:
post_init
at esp-hal/src/soc/esp32c6/mod.rs turns the RTC on, I don't know why it is the case, but this requires choosing the rtc clock as the default variable in the code, as there is no config system yet. The obvious solution is to not do that? I don't know the reason behind it, but it messes things up. Now, if the user wants the rtc variable, it will be initialized / calibrated 2 times for no reasonRtc::new(LPWR::steal(), RtcSlowClock::default());
this breaks all current examples and programs. In my experience, this will not be an issue 😃