Conversation

Notices

1. benjojo (benjojo@benjojo.co.uk)'s status on Wednesday, 15-Jan-2025 00:40:55 JST benjojo

   in reply to

   • jaseg
   • Andrew Zonenberg
   • F4GRX Sébastien

   @f4grx @jaseg @azonenberg :)

   • F4GRX Sébastien (f4grx@chaos.social)'s status on Wednesday, 15-Jan-2025 00:40:56 JST F4GRX Sébastien

     in reply to

     • jaseg
     • Andrew Zonenberg

     @jaseg @azonenberg (oh hi, @benjojo )

   • jaseg (jaseg@chaos.social)'s status on Wednesday, 15-Jan-2025 00:40:57 JST jaseg

     in reply to

     • Andrew Zonenberg
     • F4GRX Sébastien

     @f4grx @azonenberg At 38C3, someone presented how they did Toslink over fiber using standard SFPs. The TLDR is that 1G SFPs eat pretty much any signal down to a few hundred kHz, and only in 10G SFPs do refiners get into the way. Also they found some newer 1G SFPs are really just relabeled 10G ones and therefore don’t work.

     https://media.ccc.de/v/38c3-going-long-sending-weird-signals-over-long-haul-optical-networks

   • F4GRX Sébastien (f4grx@chaos.social)'s status on Wednesday, 15-Jan-2025 00:40:58 JST F4GRX Sébastien

     in reply to

     • Andrew Zonenberg

     @azonenberg I think I have a few old 850nm SFPs that were used for fiber channel so it's 4 Gbps max IIRC. I would not spend money on a pair of 10G SFP for such a hack lol

     edit: ah I did not think about AGC. hmm.

     edit2: also retimers will be a problem if they do clock recovery. We'll see. a breakout does not look too hard to build.

     I think i found the one-chip-does-all: https://eu.mouser.com/datasheet/2/268/SY100EPT28L_3_point_3V_LVTTL_to_Diff_LVPECL_and_Di-1891475.pdf

   • Andrew Zonenberg (azonenberg@ioc.exchange)'s status on Wednesday, 15-Jan-2025 00:40:59 JST Andrew Zonenberg

     in reply to

     • F4GRX Sébastien

     @f4grx The bandwidth of the AGC, coupling cap cutoff frequency, etc. will vary with the type of SFP.

     You might have better luck with e.g. 100base-FX SFPs than 10Gbase-SR ones.

   • Andrew Zonenberg (azonenberg@ioc.exchange)'s status on Wednesday, 15-Jan-2025 00:41:00 JST Andrew Zonenberg

     in reply to

     • F4GRX Sébastien

     @f4grx You'd have to make it differential via a buffer, balun, whatever. The SFF spec has the requirements for power and differential amplitude.

     As far as what the lower bandwidth cutoff is, I'm not sure. They are AC coupled and the coupling caps are on the module.

     Probably easy enough to just make a breakout, pick up some cheap SFPs, and characterize?

   • F4GRX Sébastien (f4grx@chaos.social)'s status on Wednesday, 15-Jan-2025 00:41:00 JST F4GRX Sébastien

     in reply to

     • Andrew Zonenberg

     @azonenberg yep, I'll attempt to do that, thanks for confirming it's worth trying!

   • F4GRX Sébastien (f4grx@chaos.social)'s status on Wednesday, 15-Jan-2025 00:41:01 JST F4GRX Sébastien

     • Andrew Zonenberg

     hello @azonenberg random crazy question I think you know about, could a normal cheap 1G/100M SFP easily eat a 10 MHz signal source for fiber distribution of a clock without too much work? Like, I dont know, a LVDS driver chip right on the signal pair of the SFP? Would that work for reception too? with the SFP pair to a LVDS receiver feeding more CMOS circuits? I have no pretention to keep compatibility with anything ethernet.

     the challenge is the slow speed here I guess. thanks!