STM32F723 does not trigger HAL_PCD_DataOutStageCallback for USB Audio on Windows

[fxtentacle]

Dear STM team,

When I created a new STM32F723ZETx project (FW v1.17.2) with STM32CubeMx and selected USB audio, then it worked nicely on Linux, but AUDIO_PeriodicTC_HS was never called on Windows. I've verified that Windows is sending isochronous audio packets, which means the problem is caused by the STM32 HAL not triggering the appropriate callbacks despite the hardware receiving the data.

My impression is that this line:

https://github.com/STMicroelectronics/stm32f7xx_hal_driver/blob/52bfa97ba66afc08481f6fd7631322593bd89691/Src/stm32f7xx_hal_pcd.c#L1377

is trying to verify that the even/odd flag inside DOEPCTL is matching with the current frame number's parity. If that is the intention, then instead of

((RegVal & (0x1U << 16)) == (hpcd->FrameNumber & 0x1U)))

it should read

(((RegVal & (0x1U << 16)) != 0) == (hpcd->FrameNumber & 0x1U)))

because we want to check if the EONUM bit is set and not check for its absolute value. And, indeed, that small change repairs the HAL_PCD_DataOutStageCallback on Windows 😃

[ALABSTM]

Hi @fxtentacle,

Thank you for this report. The point looks valid. A question though please. After you changed the code in the hal_pcd.c file, did you try again on Linux to check everything is still OK?

With regards,

[fxtentacle]

Yes, the change had no influence on Linux where the HAL_PCD_DataOutStageCallback was being called both before and afterwards. My subjective impression is that Linux now has a slightly improved timing from when I connect the STM32 until when the audio device changes, but that's really hard to measure objectively.

[ALABSTM]

ST Internal Reference: 185291

[ALABSTM]

Hi @fxtentacle,

I hope you are doing well. Just to let you know that according to our development team, the fix should rather be as below, in order to check for both odd and even frames.

(((RegVal & (0x1U << 16)) >> 16) == (hpcd->FrameNumber & 0x1U)))

With regards,

[ALABSTM]

Fixed in commit c393a68