Really strange timing. What could be causing it?

[Nick Alexeev]

Colleagues,

Here’s a code snippet which changes the state of two GPIO lines with delays in between.
[My actual firmware has more code, and I’ve reduced it to this condensed example which still has the issue.]

HAL_GPIO_WritePin(LED_DRIVER_ENABLE_GPIO_Port, LED_DRIVER_ENABLE_Pin, GPIO_PIN_SET);
HAL_Delay(1);

HAL_GPIO_WritePin(LED_STATUS_GREEN_GPIO_Port, LED_STATUS_GREEN_Pin, GPIO_PIN_SET);
HAL_Delay(1);

HAL_GPIO_WritePin(LED_DRIVER_ENABLE_GPIO_Port, LED_DRIVER_ENABLE_Pin, GPIO_PIN_RESET);
HAL_Delay(1);

HAL_GPIO_WritePin(LED_STATUS_GREEN_GPIO_Port, LED_STATUS_GREEN_Pin, GPIO_PIN_RESET);

The resulting picture on the oscilloscope looks quite wrong. 

oscilloscope screenshot of GPIO outputs

Legend for the oscilloscope plot:
Channel A is LED_DRIVER_ENABLE (PA3)
Channel B is LED_STATUS_GREEN (PB0)

The most troubling part is that the LED_STATUS_GREEN goes up before LED_DIVER_ENABLE on the oscilloscope.  In the code LED_DIVER_ENABLE precedes LED_STATUS_GREEN.  [It's more visible in my update below.]

In addition, the first call to HAL_Delay(1) produced a 0.1ms delay instead of expected 1ms.  The second and third calls to HAL_Delay(1) produced 4ms and 2ms delays.

What could be causing these issues?

Any suggestion, insight or reference is really appreciated!

STM32G474. I ran the code on two copies of the board, and the microcontrollers are from the same reel.  I see the same anomaly on both of them.

The compiler is CubeIDE v1.15.1 . The optimization is -O0 (no optimization, which is the default setting).  Let me know if I should provide more information.

Sincerely,
- Nick

[Pavel A.]

(you know of course that HAL_Delay(1) can take any time from 1 to 2 ms ? )

[Nick Alexeev]

update:

Here's an even more condensed code for this issue.  I removed the delays, but the GPIO outputs are still out of order.

HAL_GPIO_WritePin(LED_DRIVER_ENABLE_GPIO_Port, LED_DRIVER_ENABLE_Pin, GPIO_PIN_SET);
HAL_GPIO_WritePin(LED_STATUS_GREEN_GPIO_Port, LED_STATUS_GREEN_Pin, GPIO_PIN_SET);
HAL_GPIO_WritePin(LED_DRIVER_ENABLE_GPIO_Port, LED_DRIVER_ENABLE_Pin, GPIO_PIN_RESET);
HAL_GPIO_WritePin(LED_STATUS_GREEN_GPIO_Port, LED_STATUS_GREEN_Pin, GPIO_PIN_RESET);

oscilloscope screenshot of GPIO outputs

Legend for the oscilloscope plot:
Channel A is LED_DRIVER_ENABLE (pin PA3)
Channel B is LED_STATUS_GREEN (pin PB0)

time axis in [microseconds]

edit:  Here's the disassembly.  I'll have to read-up before I can understand it.

222           HAL_GPIO_WritePin(LED_DRIVER_ENABLE_GPIO_Port, LED_DRIVER_ENABLE_Pin, GPIO_PIN_SET);
0800077c:   movs    r2, #1
0800077e:   movs    r1, #8
08000780:   mov.w   r0, #1207959552 @ 0x48000000
08000784:   bl      0x8003828 <HAL_GPIO_WritePin>

226           HAL_GPIO_WritePin(LED_STATUS_GREEN_GPIO_Port,  LED_STATUS_GREEN_Pin,  GPIO_PIN_SET);
08000788:   movs    r2, #1
0800078a:   movs    r1, #1
0800078c:   ldr     r0, [pc, #328]  @ (0x80008d8 <main+808>)
0800078e:   bl      0x8003828 <HAL_GPIO_WritePin>

229           HAL_GPIO_WritePin(LED_DRIVER_ENABLE_GPIO_Port, LED_DRIVER_ENABLE_Pin, GPIO_PIN_RESET);
08000792:   movs    r2, #0
08000794:   movs    r1, #8
08000796:   mov.w   r0, #1207959552 @ 0x48000000
0800079a:   bl      0x8003828 <HAL_GPIO_WritePin>

232           HAL_GPIO_WritePin(LED_STATUS_GREEN_GPIO_Port,  LED_STATUS_GREEN_Pin,  GPIO_PIN_RESET);
0800079e:   movs    r2, #0
080007a0:   movs    r1, #1
080007a2:   ldr     r0, [pc, #308]  @ (0x80008d8 <main+808>)
080007a4:   bl      0x8003828 <HAL_GPIO_WritePin>

[Pavel A.]

Does one of the GPIOs "gate" the other? So that one cannot raise until the other is in high state?

Bring more water.

[waclawek.jan]

As @Pavel A. said above, a minimum-1ms-time-wasting-delay with 1ms granularity can take 2ms, and it does, if it is invoked two times in a row.

> The most troubling part is that the LED_STATUS_GREEN goes up before LED_DIVER_ENABLE on the oscilloscope. In the code LED_DIVER_ENABLE precedes LED_STATUS_GREEN.

My guess is that you set LED_STATUS_GREEN in the code you haven't shown us, before the snippet you've shown us.

Place a breakpoint - or infinite loop - just before that code and observe LED_STATUS_GREEN.

JW

[gbm]

Make sure to set the scope to CHOP mode (not ALT).