[ABANDONED] Trace with ozone

Brice_38_FR · Oct 3rd 2018, 9:34am

HI

I have more time to investigate why my trace isn't working as expected.

So I have created a CubeMX project for my stm32g767bi and generated an Atollic project.
I configure ozone to use the 4 pin trace
The µC is running @96MHz after the call to SystemClock_Config()

So on this screenshot, we can see that the trace is working

but when I step into the SystemClock_Config function, the trace no longer works after the call to HAL_RCC_ClockConfig on line 195

Any idea ?
I can provide cubemx project, atollic projet and elf file if required

SEGGER - Nino · Oct 4th 2018, 1:31pm

Hello,

Thank you for your inquiry.
When changing CPU clock and thus trace clock during execution trace can get out of sync depending on the trace pin design on your board.
To counter this the trace sampling delay can be adjusted: segger.com/products/debug-prob…hnology/setting-up-trace/
Do the troubleshooting steps from the website help in that issue?
You would need to find a setup that works continuously with default clock (16 MHz for STM32F7 usually) and the 96 MHz setting.

Best regards,
Nino

Brice_38_FR · Oct 4th 2018, 4:40pm

Hello,

I don't mind if the trace doesn't work before the clock configuration, it's more important for me that it works after

I read the troubelshooting but I only have a USB j-trace so I don't have/find a webserver.

I saw that I can overide timing in the ozone trace settings, but there is way too much combination to try them all !

Do I need to probe the traces pins to get the right timing ?

SEGGER - Nino · Oct 4th 2018, 5:04pm

Hello,

Brice_38_FR wrote:

I read the troubelshooting but I only have a USB j-trace so I don't have/find a webserver.

Oh I see now. The trace sampling adjustment is not supported with the old model.
So any timing issues you might have on the board must be countered with your hardware design.
How do the trace signals look like on a scope? Are they comparable to the ones from our Wiki? wiki.segger.com/Tracing_on_ST_STM32F769
Are there any components on the trace lanes that might interfere with high speed trace signals e.g. capacitors, diodes or inductors?

Generally we recommend upgrading to a current J-Trace PRO model. The benefits are listed on our Wiki page: wiki.segger.com/J-Trace_vs._J-Trace_PRO

Best regards,
Nino

Brice_38_FR · Oct 5th 2018, 3:32pm

It's seems more complicated than expected ....

Trace signals are directly connect between the µC and the MIPI connector. The layout of the signals haven't been made perfect (it was a test to use the trace) but seems pretty straight forward.
Here is a "dirty" capture of the layout. PCB height is around 65mm.
(MIPI connector is on the bottom)

I will try to probe them to see if there is any synchronisation problem.
About the upgrade, I would love to get the last model but my boss won't understand and allow to buy a new one ...

SEGGER - Nino · Oct 11th 2018, 1:35pm

Hello,

Thank you for the additional information.
The PCB layout looks fine on first glance.

Brice_38_FR wrote:

About the upgrade, I would love to get the last model but my boss won't understand and allow to buy a new one ...

That is very unfortunate. The old J-Trace CM lacks any options to counter trace timing issues.
You could try reducing 4-bit trace to 1 or 2 bit to see if signal integrity improves.
Does that make a difference?
Could you try setting a BP right after the PLL init and resume execution from there. Does the same issue appear or is the trace stream in sync?

Best regards,
Nino

[ABANDONED] Trace with ozone

[ABANDONED] Trace with ozone

Brice_38_FR wrote:

Brice_38_FR wrote:

Share

Tags