cancel
Showing results for 
Search instead for 
Did you mean: 
Highlighted
Participant
Participant
365 Views
Registered: ‎02-14-2019

Byte alignment at synchronous Ethernet (SyncE) and GTH transceiver

Jump to solution

Hi everybody,

let me introduce you my test setup:

1st board: Ultrascale+, TSU of ZYNQ is used, Linux running ptp4l as master, GEM0 is connected via PCS/PMA IP core and gtwizard_wrapper to serial GTH transceiver, using 1G optical sfp module

2nd board: almost the same as 1st board, Linux running ptp4l as slave, using fractional PLL and QPLL0 for rx-clock recovery to achieve SyncE

Both boards are directly connected via 1m optical cable.

Observation:

a) TX-Clock (1st board) and RX-Clock (2nd board, PLL recovered clock) are synchronous to each other with very low jitter.

b) Output of ptp4l-logfile (2nd board, ptp slave):

ptp4l[4791.083]: master offset         -8 s2 freq     +0 path delay       720
ptp4l[4792.083]: master offset         -6 s2 freq     +1 path delay       718
ptp4l[4793.083]: master offset         2 s2 freq     +3 path delay       718
ptp4l[4794.083]: master offset         0 s2 freq     +2 path delay       720
ptp4l[4795.083]: master offset         -8 s2 freq     -0 path delay       720
ptp4l[4796.083]: master offset         0 s2 freq     +2 path delay       720

c) I wrote a decoder for the internal GMII interface (125MHz) to give me a pulse when the SFD-Byte of the Ethernet frame is received relative to rxoutclk_bufg_out. rxoutclk_bufg_out is the buffered (BUFG_GT) output RXOUTCLK of the transceiver with 62.5MHz frequency. What I observe is, that the SFD-Byte is not aligned with the RXOUTCLK phase of the transceiver. The pulse (SFD-Byte) can be seen at low phase and high phase of RXOUTCLK.

Question:

Is the phase relationship of the SFD-Byte the cause for my ptp4l 8ns jumps?

Can I align the the received ethernet frame such that it is always aligned in the same way to the word (2-Byte) boundary? That way, the SFD-Byte would always be put to the low (or high) position within the 2-Byte word.

Best regards,

Torsten

Tags (6)
0 Kudos
1 Solution

Accepted Solutions
Highlighted
Participant
Participant
237 Views
Registered: ‎02-14-2019

Re: Byte alignment at synchronous Ethernet (SyncE) and GTH transceiver

Jump to solution

I did some more hardware measurements.

I triggered an oscilloscope when the SFD-byte left the MAC on the GMII interface of the transmitter and measured the delay when the SFD-byte entered the MAC on the GMII interface of the receiver.

The delay is always constant, regardless of the alignment (even, odd), there are no byte jumps (8ns) caused by preamble shrinkage.

Obviously, the byte alignment is not the cause for my ptp4l jumps (8ns).

For me it looks like the ptp4l application or the linux ptp support is responsible for this jitter.

I will close this thread  and might reopen another one with ptp4l/ptp/tsu question.

View solution in original post

0 Kudos
3 Replies
Highlighted
Participant
Participant
285 Views
Registered: ‎02-14-2019

Re: Byte alignment at synchronous Ethernet (SyncE) and GTH transceiver

Jump to solution

Hi,

Appendix C of pg047 (1G/2.5G PCS/PMA or SGMII v16.1) gives some explanation for Odd Transmission Case and Reception of the Odd Case. After that there is a paragraph about preamble shrinkage.

Could this be the cause for my 8ns jumps on ptp4l application?

If so, is there a way to synchronize the GEM to the 62.5MHz clock phase to get rid of the preamble shrinkage?

0 Kudos
Highlighted
Participant
Participant
238 Views
Registered: ‎02-14-2019

Re: Byte alignment at synchronous Ethernet (SyncE) and GTH transceiver

Jump to solution

I did some more hardware measurements.

I triggered an oscilloscope when the SFD-byte left the MAC on the GMII interface of the transmitter and measured the delay when the SFD-byte entered the MAC on the GMII interface of the receiver.

The delay is always constant, regardless of the alignment (even, odd), there are no byte jumps (8ns) caused by preamble shrinkage.

Obviously, the byte alignment is not the cause for my ptp4l jumps (8ns).

For me it looks like the ptp4l application or the linux ptp support is responsible for this jitter.

I will close this thread  and might reopen another one with ptp4l/ptp/tsu question.

View solution in original post

0 Kudos
Highlighted
Participant
Participant
234 Views
Registered: ‎02-14-2019

Re: Byte alignment at synchronous Ethernet (SyncE) and GTH transceiver

Jump to solution

-

0 Kudos