#linuxcnc-devel Logs

May 17 2020

#linuxcnc-devel Calendar

04:14 PM seb_kuzminsky: jepler: i'm looking at your rpi4 images
04:18 PM seb_kuzminsky: back in january there were three: "lite", "", and "full"
04:18 PM seb_kuzminsky: in may there are only two: "lite" and ""
04:18 PM seb_kuzminsky: i think i used "full" for the buildbot rpi4 slave
04:18 PM seb_kuzminsky: what's the difference between them?
04:18 PM jepler: https://github.com/jepler/pi-gen/commit/cf3ce09ca5da252183860c3d4f1e74b5d1fa2393
04:18 PM jepler: the "lite" is without a desktop. I have not tested it
04:18 PM seb_kuzminsky: lol ok
04:18 PM jepler: It also doesn't reach the "install linuxcnc" step which would make it useless to most people
04:20 PM jepler: but kudos to JT for documenting how to build kernel images, and pi foundation for providing the original scripts
04:30 PM JT-Shop: :)
04:38 PM jepler: > This branch is 1009 commits ahead, 6643 commits behind raspberrypi:rpi-4.19.y.
04:38 PM jepler: hmm maybe I should rebuild the kernel
04:39 PM jepler: 1009 commits ahead must mean they rebased it
04:41 PM jepler: anyway reminder that the kernel is also in github: https://github.com/jepler/linux changes mostly represent packaging customization
05:27 PM jepler: braindump(ag) == U-Boot on Raspberry Pi (including RPi 4)
05:27 PM jepler: https://andrei.gherzan.ro/linux/uboot-on-rpi/
05:28 PM jepler: Would be nice to just put uboot on the SD and the rest on a hdd!
05:29 PM pcw_home: I have a minor hm2 encoder patch is this OK for master?
05:29 PM jepler: Still uses the proprietary pi bootloader as a first stage, I think
05:29 PM jepler: pcw_home almost certainly
05:30 PM pcw_home: adding a parameter to enable a high resolution timestamp
05:31 PM pcw_home: (and set the timestamp freq to ~10 MHz rather than ~1 MHz)
05:32 PM pcw_home: 1 MHz is fine for encoders but when the encoder is used for frequency counting from V-Fs
05:32 PM pcw_home: for example, it has only about a 10 bit resolution at a 1 KHz servo thread
05:38 PM jepler: Neat
05:38 PM jepler: The downside is for counting really slow pulses I assume?
05:39 PM jepler: We always feels like there was one more bug in the velocity estimation, I feel like
05:40 PM pcw_home: The downside is I'm pretty sure the wrap logic will fail at a count/servo period > 32768
05:41 PM pcw_home: so 10 MHz is fine of 1 KHz and 500 Hz with 1ms latency but that's as low as you can go
05:41 PM pcw_home: (servo thread waise)
05:46 PM jepler: So if you know you need it, set to fast, otherwise leave at slow?
06:03 PM pcw_home: Yeah
06:06 PM pcw_home: I guess I could bump up the slow rate at the same time, 1 MHz is a bit too conservative
09:21 PM Tom_Rpi: ./usr/lib/linuxcnc/modules/hm2_rpspi.so: undefined symbol: hm2_register
09:47 PM jepler: you need to loadrt hostmot2 first
09:48 PM jepler: then loadrt hm2_rpspi second
09:48 PM skunkworks: I think he figured that out. :)
09:48 PM skunkworks: (on linuxcnc channel)
09:49 PM Tom_Rpi: yeah thanks
09:50 PM Tom_Rpi: both eth and spi work on the rpi4
09:50 PM Tom_Rpi: 7i80 eth & 7i90 spi
09:53 PM skunkworks: awesome
09:53 PM Tom_Rpi: and i flashed a bit file via spi on it
10:34 PM Tom_itx is now known as Tom_L