davids5 commented on pull request #1544:
URL: https://github.com/apache/incubator-nuttx/pull/1544#issuecomment-672816427
>@davids5 I think that the default slew rate is for 50Mhz. This is not
excessive for SPI.
I here you, I original thought the same thing, hence the defaults*.
> This seems like a hardware design problem (lines termination etc), and I
don't thing we should make a change in NuttX for this.
This is the point. The line termination and characteristic of an interface
is hardware dependent, really board dependent.
1) Signal integrity is effected. (cross talk) but over driving the lines.
2) FCC certifications is effected by slew-rate (drive strength).
Our EE group (amazing Wizards) have suggested we want to set the minimum
drive strength that achieves quality signal integrity and no more.
Contrast a board 3 x 3 cm PCB (controlled impedance) with an SoC and SPI to
devices on it and one with connectors and cables to it devices. The impedance
will vary greatly. This really requires looking at the signals (with proper
equipment) and adjusting drive strength correctly.
>It make break other user's code. Maybe we should add a new configuration in
Kconfig, so the user can select the slew rate for every SPI, while keeping the
default at 50Hz.
Yes I 100% agree we should not make a change that is breaking, and I would
choose to default it as is.
> (As a side note, ringing usually does not have enough energy to cause
latch-ups in other chips, although max. VIH is violated. But of course you know
better, as you tested the actual hardware...)
The failure was on an H7 using SPI DMA on a mems sensor 7mm from the SoC.
When the clock was continuous because of the DMA the device failed at the high
drive strength. Lower it and it works. Probe it with a 10X 14pf Probe and it
works. Probe it with .9 pf active Probe and you see .7V over VDD on every
rising edge.
My suspicion is that the current injection to the substrate via the body
diodes, caused the ADC to or loose its bias and clamp as the data was a FIFO
full of 00s. Lower the drive strength and once the over shoot was gone it works
fine.
The code need to fix the overcome the problem is [ugly and does not scale
well](https://github.com/PX4/Firmware/blob/master/boards/px4/fmu-v6x/nuttx-config/include/board.h#L411)
but moreover we
[state](https://github.com/apache/incubator-nuttx/blob/master/arch/arm/src/stm32h7/hardware/stm32h7x3xx_pinmap.h#L67-L70)
the setting should be done in board.h. We (I) broke that contract based on
*above.
This opens another heated can of worms. There are defaults. where the pin
mux has only one appearance. These will need to be have numbers added (I
suggest _0 to make it obvious). To allow proper usage. Again massive ripple.
I think I will propagate this to an Issue....
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
