[DEV] STM32G4xx support

[Giovanni]

Support for DAC3 and DAC4 has been added but it is untested.

Giovanni

[JMifune]

Thanks, I've ported DAC1 testhal example, but the DAC streaming callback does not fire.

It is still on DACD1 as in original example, the only changes I made was to DAC_TRG bits - 0x7 for Tim6 on G474.
From the reference manual DAC_TRG_MASK is 4 bits long, but the top bit is used only for HRTIM.

[Giovanni]

There must be some problem in DMA-related changes.

Giovanni

[JMifune]

Looks like it. DAC itself works, output is stuck on initial value.

I've tried to route DAC3 and DAC4 outputs through opamps configured as followers.
Always get ~0.6V on outputs, no matter the DACs initial values.

[Giovanni]

There were 2 problems:

1) DMAMUX was not activated on G4 and this prevented triggering.

2) DACv1 was accessing DAC registers using half word accesses, this is forbidden, accesses must be performed using full words. This used to work on previous devices (violating RM specification) and stopped working on G4.

Now I reworked the driver to use 32 bits accesses everywhere and it seems to work but I cannot test it fully, I would be grateful if everybody could give it a try now. If it is OK then I will also backport it to 18.2 and 19.1. I also cleaned some parts, the driver was a bit old.

Note, there is a G4 builder in the /testhal/STM32/multi/DAC demo now.

Giovanni

[JMifune]

Yes, I have tested it DMA and DAC1 output is working now.

Still can not figure out why I can not get signal out of DAC3/4.
DMA is working on DACD5-8 and all opamps enabled as followers and working properly when VP_SELs are set to connect to external pins.
But when I set OPAMPx_CSR VP_SEL to 11 to connect to DAC output, then opamp outputs are at contstant ~0.6V.

[JMifune]

Tried the same opamp6 config + DAC3 CH1 in CubeMX.
It works.

[JMifune]

I've figured it out.
Before enabling DAC3/4, DAC_MCR register MODEx bits must be set to 011 (DAC channelx is connected to on chip peripherals with Buffer disabled).
Default MODE bits are 000 (external pin with Buffer enabled), witch is not an option for DAC3/4.
Also HFSEL bits must be managed according to different AHB speeds.

[julester23]

I have been working on getting the FDCAN peripheral to work on the STM32G434, and have some working internal loopback demos working. I tried submitting to ChibiOS-Contrib, however the pull request I created depends on many of the additions in Chibios-RT master (not in stable_19.1.X). Would it be possible to backport the STM32G support into stable 19? or is 20 coming out soon? or would you be interested in getting some of this work into the mainline ChibiOS-RT? Please let me know if there's anything I can do to help get CAN support in the STM32G series.

I admit it is not complete, but I tried to follow the naming and style as much as possible.

One way or the other (merging both would be redundant):
ChibiOS-Contrib Branch which depends on master branch of ChibiOS-RT and so it is not merge-able.
ChibiOS Branch with a "CANv2" driver.

Thanks for your help.

[Giovanni]

I have been working on getting the FDCAN peripheral to work on the STM32G434, and have some working internal loopback demos working. I tried submitting to ChibiOS-Contrib, however the pull request I created depends on many of the additions in Chibios-RT master (not in stable_19.1.X). Would it be possible to backport the STM32G support into stable 19? or is 20 coming out soon? or would you be interested in getting some of this work into the mainline ChibiOS-RT? Please let me know if there's anything I can do to help get CAN support in the STM32G series.

Hi,

Good work, I examined it briefly and there are some things I observed:
- It should be named FDCANv1 as the current rule is that LLDs take the name of the peripheral (CANv1 is historically wrong, it should be BXCANv1).
- The number of RX and TX mailboxes is set to 1, I believe there are more.
- IO macros are not very efficient, it forces the compile to write and/or read the register each time because the volatile attribute, the preferred way is to prepare a variable with all values the do a single write (unless things need to be written is sequence separately which is sometime required). In general I prefer to read/modify/write explicitly.

Overall it needs some more work.

Giovanni