Re: [Linuxptp-devel] connecting two devices clock via GPIO

[Frantisek R. <Fra...@po...>]

Just a short update along the lines of this thread:

On top of the previous work on PPS input to i210,
I've cobbled together a simple PLL synth with a 25MHz VCXO,
good enough to take an external 10 MHz reference 
and "influence" the crystal on an i210 to run in step with the PPS.
(I could as well just desolder the original crystal, but ahh well.)
I'm attaching the resulting output of "my" i210 PPS proggie :-)
(again based heavily on Mr. Cochran's example.)

As for the capturing side of things, both TCPDUMP (PCAP file format)
and T-Shark (PCAP-NG file format) work for nanosecond-level 
capturing. I'm working on a crude analyzer of PTP traffic. 
I've given up tapping into the Wireshark dissector framework or using 
raw libpcap (and writing my own PTP dissector from scratch). 
Instead, I tend to run tshark -V and parse what I need from its 
textual output. I got through the text parsing and transaction 
recombination/inference, I have yet to write the output side (dump 
per-transaction timing data into CSV files suitable for gnuplot). 

The key bit I'm still looking forward to is the decoding of various 
timing deltas. I have a "time of capture" timestamp, and the PTP's 
own payload timestamps, and the correction value... If I understand 
correctly, the slave will tell me in PDelay Response (+FollowUp, if 
2-step) its own timestamp, which should give me a clue about *the 
slave's* performance...

I got this far using a DIY powered tap for metallic 100Mb Ethernet.
And then there's the hope that I'd be able to convert (reconfigure) 
the DeLock gigabit i210 with SFP from SERDES mode into SGMII mode...
That's coming up. The 100Mb SGMII SFP's seem quite exotic.

If you care, keep your fingers crossed :-)

Frank Rysanek