re: Seeking to Determine Who Will Enroll in SMA or Gigabit Ethernet Evaluation Kits

Hello!

Is there a decision which one is going to be roadtested (Ethernet or SMA)?

I think it is an interesting subject - the copper Gigabit Ethernet seems easier to test, but the results could be somewhat limited by copper PHY media converters. I have (long time ago) done a project about the possibility of use Gigabit Ethernet in High Performance Cluster interconnects in place of Infiniband (which provided significantly lower propagation times - there was no RDMA over Ethernet possibilities then - but was very expensive).  

During my work I have started to wonder if copper Gigabit Ethernet physical interface introduces some additional delay - these times I was unable to find any solid information about this, but today it is clearly present in the Ethernet switch documentation (this one from one model of Netgear - different models may have different values, maybe it is also dependent on the chipset used):

Latency (64-byte packets)
1G Copper < 3.5us
1G Fiber < 2.5us

which difference may (or may not) introduce some artificial limit for the evalkit

I think that SMA version can be tested (at hobbyist level of lab) using some sort of passive DAC (TWINAX) cable breakout (copper cable connecting two SFP sockets - commonly used for optical interfaces, thus probably directly connecting TX and RX pairs) and two Gigabit Ethernet switches - but the signalling level should be verified. At first glance it seems compatible, but only after about of hour of research.

On the other hand, evalkit complete with Gigabit Ethernet media converters can be tested as optically isolated Gigabit Ethernet link - something very desirable to all kinds of Extreme Environment type scenarios (if power consumption is reasonable and it can work through the typical transparent lid [or dirty transparent lid] - but this remains to be tested)

Hello!

Is there a decision which one is going to be roadtested (Ethernet or SMA)?

I think it is an interesting subject - the copper Gigabit Ethernet seems easier to test, but the results could be somewhat limited by copper PHY media converters. I have (long time ago) done a project about the possibility of use Gigabit Ethernet in High Performance Cluster interconnects in place of Infiniband (which provided significantly lower propagation times - there was no RDMA over Ethernet possibilities then - but was very expensive).  

During my work I have started to wonder if copper Gigabit Ethernet physical interface introduces some additional delay - these times I was unable to find any solid information about this, but today it is clearly present in the Ethernet switch documentation (this one from one model of Netgear - different models may have different values, maybe it is also dependent on the chipset used):

Latency (64-byte packets)
1G Copper < 3.5us
1G Fiber < 2.5us

which difference may (or may not) introduce some artificial limit for the evalkit

I think that SMA version can be tested (at hobbyist level of lab) using some sort of passive DAC (TWINAX) cable breakout (copper cable connecting two SFP sockets - commonly used for optical interfaces, thus probably directly connecting TX and RX pairs) and two Gigabit Ethernet switches - but the signalling level should be verified. At first glance it seems compatible, but only after about of hour of research.

On the other hand, evalkit complete with Gigabit Ethernet media converters can be tested as optically isolated Gigabit Ethernet link - something very desirable to all kinds of Extreme Environment type scenarios (if power consumption is reasonable and it can work through the typical transparent lid [or dirty transparent lid] - but this remains to be tested)