[TowerTalk] LMR-400 LMR-600 Equivalent (and stuff)

[Jim Lux]

-----Original Message-----
>From: James Wolf <jbwolf at comcast.net>
>Sent: Jun 4, 2008 4:45 PM
>To: 'Kevin' <rkstover at mchsi.com>, "'Roger (K8RI)'" <K8RI-on-TowerTalk at tm.net>
>Cc: towertalk at contesting.com
>Subject: Re: [TowerTalk] LMR-400 LMR-600 Equivalent (and stuff)
>
>Kevin,
>
>I agree.  We tried using Cat5/e on a 1 GHz link that had to run at 1 GHz or
>it would lose sync. We couldn't get anything even close to reliable beyond
>about 12 ft, with fresh straight runs.
> 
>The reason for the 330 ft maximum length is that each pair of conductors is
>twisted with a slightly different turn rate, at about 330 ft down the line
>the velocity factor is such that there is too much delay on the slower pair
>and the bits start getting confused as to which frame they actually belong
>to.
>

And keeping the whole discussion on things RF.. the manufacturers of networking gear (at least in non price sensitive applications, like 10Gbit E, and GigE for right now) recognize that the physical plant might not be up to snuff, so they have a fair amount of margin in their adaptive equalization algorithms.

Over the years, there's been a trend to accepting the fact that the medium is non-ideal, and dealing with it. Back in 10baseT or even older, thicknet, days, the signal processing was expensive and the medium could be specified and installed "good enough" for it to work. But when you get up to 1Gbps kinds of data rates (which, by the way is signaled at a much lower rate, 125 Mbaud.. they use 5 level PAM and use all 4 pairs with a hybrid, echo cancellation and equalization, along with trellis coding to get 6dB of coding gain)

Once you've bought into the all the DSP to do this, then accommodating a few lumps and bumps in the cable's crosstalk (near and far end) and transmission properties is not a big deal.

In fact, running at 10Mbps or 100Mbps might actually be harder over a given poorly installed cable than running at 1Gbps, because your average 10Mbps interface is unlikely to have any sort of adaptive equalization.. it depends on the cable meeting the spec.


Personally, I'm amazed that they can run 40 Gbps or more over basically run of the mill twisted pair cables with connectors, with 1E-12 kinds of error rates.  This is really cutting edge RF design.. it's a long way from a RS-422 line driver at one end and a RS-422 receiver at the other.