One of the new and exciting features we added to the VCFMW ShadyTel exhibit this year that almost no-one would have seen or realized existed was an actual high precision master frequency reference for our TDM network. This post will dig into what this device is, what it does, and why we use it.
What is a Frequency Reference?
The ShadyTel core network is entirely TDM, also known as Real Telephony. Since this is a Time Division Multiplexing network, the timing on the network is fairly important. Its worth understanding that in this context, timing is not absolute wall clock time in the sense of NTP or PTP, but instead refers to a stable frequency reference used for clocking synchronous serial interfaces.
For this purpose, I have a Spectracom Securesync 1200 clock. These timing appliances are 1U devices that support multiple input/output timing cards and can supply both absolute time and frequency references over a wide range of active and defunct standards. For using the clock with the show network, I obtained a 1204-0A T1/E1 output card. In North America, a T1 is a common channelized interface type, with the E1 being its counterpart in Europe and Australia.
With the Securesync connected to a suitable GPS antenna, it not only obtains a very high quality timing reference, but also disciplines its internal clock from the GPS signal, resulting in true atomic accuracy without having to actually bring an atomic clock. The T1/E1 card then exports this reference as a precisely clocked T1 circuit configured for the typical B8ZS ESF circuit type. The circuit doesn’t contain any actual data, in fact it contains statically allocated frames that just contain all 1’s, but this has execellent properties for clock recovery at the far end. Again, clock here does not refer to absolute time, but refers to frequency used for clocking a synchronous serial line.
This clock source was passed back over the outside plant from a master clock at Jeff Geerling’s booth near the edge of the convention hall to the CO in pod group C4. Why was the clock at Jeff’s booth? Well, since it requires a GPS reference, it was advantagous to place the clock as close to the edge of the building as possible to minimize the cable runs. Since GPS was already being delivered to this location as a ShadyTel Engineered Service (SM), it made sense to colocate the clock there as well.
I’d like to give a special thanks to Meinberg for supplying a much better quality GPS antenna than the one I originally brought, which allowed us to get a cleaner signal over the roughly 250 feet of cable that was required to get out the loading dock door and into an area with clear sky.
Of Various Strata
Just like absolute time systems, frequency reference and distribution systems have strata that are used to specify the quality of the reference. Unlike absolute time references, however, frequency strata are not a strict hierarchy, nor are they a distance from the source. Frequency strata express the quality of the clock, both in terms of holdover, and how far its able to tune to establish a lock. The holdover is relatively easy to understand and has the same meaning as it does in absolute time: given an absence of the reference, how far off will this clock get over a given amount of time.
Consider the following illustration:
Stratum 1: |+|
Stratum 2: |-+--|
Stratum 3: |---------+---|
Stratum 4: |-+---------------------------------------------------------|
Notice how the Stratum 1 reference can’t really be tuned at all, and each step down has a wider and wider tuning range, but that each range still contains the known frequency that should eventually be tuned to. These are contrived ranges, but they illustrate that the ranges aren’t necessarily all overlapped. When bringing up a network that is in free run without an absolute reference, its possible for clocks to be in a state that they can’t tune to each other. Consider if the Stratum 3 and 4 devices in the above diagram were tuned to opposite ends of their respective ranges, and due to the variability in electronic components, it is possible that they wouldn’t actually be able to tune back to their overlapping range.
Basically, each stratum is designed with exactly the minimum tunability required to ensure that it can lock to a primary reference standard, but that each additional level of tunability comes at the cost of reducing the overall accuracy of the un-disciplined clock.
Caveat Emptor: Note that through this post, I have referred to the clock we have on-site at the show as a Stratum 1 clock. This is, in the strictest possible sense, not quite true because it is not able to initialize to a known frequency on boot without external references. Systems like Cesium beams on the other hand can boot from cold and self-start to a known frequency without external help. What we have on the show floor is more accurately referred to as a Primary Reference Source, or PRS.
The VCFMW Clocking Chain
From what we have above, you now know that the high quality GPS signal was brought in to the Spectracom Securesync 1200 and then backhauled over a T1 to the CO. The Securesync, being a GPS disciplined oscillator, is definitionally a stratum 1 reference source. In the CO however, we had a variety of old and mismatched hardware. We fed the high quality timing reference directly into our tandem switch which was a Cisco ISR. The ISR is only a stratum 4E device, which means that it has a wide tunability range, but a quite poor holdover characteristic. This, coupled with the source being a stratum 1 led to a very interesting network core. We then distributed the clock alongside all of our in-office and intra-office T1 lines to other devices, and this is where we did some design no-nos.
The primary switch serving subscriber lines at VCFMW is a Nortel Meridian 1 Option 11c. This is a modular small office PBX capable of serving several hundred lines. It is meant to be disciplined from the telephone company’s high quality clocks, but when you are the telephone company, the quality of your own internal clocks gets a little fuzzy. The Meridian is a Stratum 3 clock, but we were connecting it to a Stratum 4 source. This means that there is a chance when the system starts up that before the absolute time reference of the Stratum 1 pulls the ISR into range, the Meridian might not be able to tune to it. We did not observe this failure mode, but we did discuss that there might need to be some dancing of shut/no shut interfaces to get clocks to come up in the right order.
From the Cisco tandem at the core, and the Meridian serving most subscriber circuits, the principal clock is distributed to the rest of the network:
+---------------------------+
|Spectracom SecureSync 1200 |
| |
+---------------------------+
|
v
+-----------+
|Cisco 3845 |
| |
+-----------+
| |
| +------+
| |
v v
+----------------------+-----------+
|Meridian 1 Option 11c |Cisco 2921 |
| | |
+----------------------+-----------+
| |
v v
+----------------+ +---------------+
|BRI Subscribers | |T1 Subscribers |
| | | |
+----------------+ +---------------+
Through this clocking chain, we were able to deliver very high quality frequency references all the way down to subscriber equipment. We were only delivering frequency references, if you want to learn about absolute time and where NTP came from for the show network, check out Jeff Geerling’s writeup here.
Every year we try to add something new to the VCFMW network, and while this year we added a lot of broadly invisible internal technology and quality of life improvements, the team was very happy with how it turned out. You can expect the precision clocks to make a return next year at VCFMW22, and experience precision time by dialing with ShadyTel.