Difference between STM32H743 and STM32H747?

[eeeeee]

Hi,

I've made a custom board where I followed the schematics of the STM32H747I-DISCO board. Instead of the stm32h747 I use the stm32h743 microcontroller, because of the availability.

I'm working on the 'FoodReco' project and modified it to my needs (changed the pinouts for the SDRAM/DCMI etc). When I cover the camera with my finger, the camera frame buffer on my custom PCB looks really similar as the camera frame buffer on the disco board:

For the DISCO board:I believe the camera frame buffer is used to perform the image classification model on. The output however, is not the same at all. When it is pitch black it should output 'floor', when it isn't pitchblack it should output 'ceil'. The DISCO board, as expected, outputs 'floor 100%'. My PCB however reports a random value 'ceil 96 %'..

I haven't changed the functions APP_FramePreprocess, APP_NetworkInference and APP_Postprocess.

How is it possible that the result of the model is not the same? Am I misconfiguring something?

Thanks in advance!

[TDK]

H743 has a single M7 core. The H747 is dual-core and has an M7 core and an M4 core. There are many differences, although the peripherals should be the same.

I would never expect two pictures to be identical even if they look "really similar".

[eeeeee]

Yeah I noticed that the H747 is dual core, but for this application the CM4 core is left unused. That shouldn't be the problem.

[Uwe Bonnes]

H743/747 have the same die, with some function disabled, e.g. access to the M4. Comparing the framebuffer output is probably not a good idea, as camera properities and environment may impact the result.

[eeeeee]

I'm testing in the same environment and with the same camera, so that shouldn't be the problem I believe. The reason why I'm just looking at the buffers is because the stm32h743zi cannot have an LCD and Camera at the same time, so I can't verify if I see the correct output on a display...

[Tesla DeLorean]

As I recall the H747 supports DSI, and there are a handful of pins (balls) which are assigned slightly differently. Perhaps be more precise about the fully qualified part number and packages we're talking about.

Printout the footprint and play pin bingo

[Uwe Bonnes]

"Cover the camera with my finger" means that infrared still get through. Depending on ambiant IR, results will differ substantial.

[eeeeee]

I will try an IR blocking object to see if that makes a difference.

[TDK]

You realize that camera pictures are subject to considerable noise, right? So covering with your finger, or an IR block, or anything, will not produce a 100% black image.

[eeeeee]

Yes I'm aware of that, but that does not pose a problem for the neural network that is used on the image. I figured out what the problem was though, it was in initializing the SDRAM. I use a 8 data bits wide SDRAM, so I thought connecting the FMC_NBL0 bit to GND was okay. Turns out this was not the case. The output now is to my satisfaction.

Thanks for your time and input!