Conversation

Notices

1. minute (mntmn@mastodon.social)'s status on Tuesday, 25-Jul-2023 22:17:42 JST minute

   tracked down another issue with BPI/RCM4. i was wondering why the ethernet was working only at 100mbit/s and not reliably. turns out it had to do with the connection/polarity of the LEDs. the bpi-cm4 expects LED1 to be active high, but on the Reform motherboard, it's wired as active low. and: the LED pins also double as config straps for the RTL8211F PHY on the bpi-cm4. so one or two config bits must have been unexpected. cutting the traces on RCM4 made 1gpbs work for the first time.

   • minute (mntmn@mastodon.social)'s status on Tuesday, 25-Jul-2023 22:18:36 JST minute

     in reply to

     so i'm going to add another inverter and resistors to RCM4.

   • minute (mntmn@mastodon.social)'s status on Tuesday, 25-Jul-2023 22:23:02 JST minute

     in reply to

     weird: on the official pi CM4IO board, both LEDs go to cathodes of the ethernet LEDs. so this looks to me like a design error with the bpi-cm4.

   • minute (mntmn@mastodon.social)'s status on Tuesday, 25-Jul-2023 22:30:45 JST minute

     in reply to

     • Giles Goat

     @gilesgoat the screenshot above is from the pi CM4IO. but here is a screenshot from the BPI (banana pi) CM4 IO board, where it's the other way around on one of the LEDs.

   • Giles Goat (gilesgoat@toot.wales)'s status on Tuesday, 25-Jul-2023 22:30:47 JST Giles Goat

     in reply to

     @mntmn Could be not an error , both anodes are at + and the resistor is on the -, which means they are driven "on" by a '0', which means they are probably driven by an open collector transistor, which normally "had more current to ground" than trying to drive a '1'. What is "a bit odd" is that resistor should have been on the anodes not on the cathodes but it probably still works.

   • minute (mntmn@mastodon.social)'s status on Tuesday, 25-Jul-2023 22:38:13 JST minute

     in reply to

     • Giles Goat

     @gilesgoat yes, they do. but that is an incompatibility between the raspberry pi cm4 and the banana pi cm4 that is not documented as such. this kind of stuff creates problems for integrators like me

   • Giles Goat (gilesgoat@toot.wales)'s status on Tuesday, 25-Jul-2023 22:38:15 JST Giles Goat

     in reply to

     @mntmn Don't know those boards but could it be they use different ETH controller chips that drive those leds differently ?

   • minute (mntmn@mastodon.social)'s status on Tuesday, 25-Jul-2023 22:45:49 JST minute

     in reply to

     all ethernet LEDs are specced as active low in the raspberry pi cm4 manual. sigh. pondering how to solve this incompatibility with bpi cm4

   • minute (mntmn@mastodon.social)'s status on Tuesday, 25-Jul-2023 22:49:38 JST minute

     in reply to

     idea: XOR gate driven by a GPIO that is set/unset by software depending on rpi or bpi

   • minute (mntmn@mastodon.social)'s status on Tuesday, 25-Jul-2023 23:36:55 JST minute

     in reply to

     this is what we're gonna do

   • minute (mntmn@mastodon.social)'s status on Tuesday, 25-Jul-2023 23:39:47 JST minute

     in reply to

     • Paul Sutton

     @zleap yes

   • Paul Sutton (zleap@qoto.org)'s status on Tuesday, 25-Jul-2023 23:39:48 JST Paul Sutton

     in reply to

     @mntmn

     Ok without looking it up that is an XOR gate or eXclusive OR ?

   • minute (mntmn@mastodon.social)'s status on Tuesday, 25-Jul-2023 23:47:59 JST minute

     in reply to

     • Paul Sutton

     @zleap

   • Paul Sutton (zleap@qoto.org)'s status on Tuesday, 25-Jul-2023 23:48:00 JST Paul Sutton

     in reply to

     @mntmn

     :0 I can't remember the truth table off hand but I think it is something like

     0 - 0 - 0
     1 - 0 -1
     0 1 1
     1 -1 -1

     so output high when either of the inputs are high AND both inputs are high.