LPSPI sampling configuration API #135
Conversation
Expose a public method to allow users to change sampling point of an LPSPI peripheral. I've experienced RX FIFO overruns when the SAMPLE: SAMPLE_1 (delayed edge) is used with non-hardware controlled chip select pins (ie, manually driving a CS pin with GPIO embedded_hal traits). Changing this configuration setting to SAMPLE_0 fixed the issue. Tested on a Teensy 4.0 and 4.1.
remove text
fix formatting
There was a problem hiding this comment.
New API and docs LGTM. However, the method set_sample_point should be attached to Disabled. From the 1060 reference manual (rev 3 section 48.5.1.9.2),
CFGR1 should only be written when LPSPI is disabled.
We've included the "delayed edge by default" behavior since this driver was originally integrated. Now that the user can change the setting themselves, we could remove that implicit behavior and keep the hardware default. That's a future change for a breaking release.
When using CS0 (pin 10 on Teensy 4.X), the SAMPLE_1 setting works fine. When using pin 4, the SAMPLE_0 must be used to avoid overruns. I'm curious if anyone else has experienced this issue.
Sorry, not sure why this sampling point setting would be based on chip select. But if this is the difference between hardware- and software-controlled chip selects, I'm curious if there's a timing issue between CS changes and the LPSPI transactions.
Co-authored-by: Ian McIntyre <ianpmcintyre@gmail.com>
|
Moved method to |
|
Whoops - accidentally closed when responding - Ian, can you reopen? |
|
I squashed these commits, added a changelog entry, and merged. If there's any follow-on work between the close and now, we can discuss in a new PR. |
Expose a public method to allow users to change the sampling point of an LPSPI peripheral.
I've experienced RX FIFO overruns when the SAMPLE: SAMPLE_1 (delayed edge) is used with non-hardware controlled chip select pins (ie, manually driving a CS pin with GPIO
embedded_haltraits). Changing this configuration setting to SAMPLE_0 fixed the issue.Tested on a Teensy 4.0 and 4.1 with a Trinamic TMC5130 stepper driver IC.
When using CS0 (pin 10 on Teensy 4.X), the
SAMPLE_1setting works fine. When using pin 4, theSAMPLE_0must be used to avoid overruns. I'm curious if anyone else has experienced this issue.