Power quality metering across ten of JEA’s substations: voltage sags, swells and harmonics measured by 7700 ION meters at each station and carried back for aggregation into the PIE SCADA platform.
One design, drawn once and then built ten times. Every station got the same kit — a router, a switch, and the meters behind them — and its own build document, so a site was a page rather than a project.
The thing being measured is worth stating precisely, because the word invites confusion. A voltage sag is an electrical event — a brief drop in RMS voltage, usually from a fault or a large motor start — not a conductor drooping into a tree. Harmonics are distortion injected back into the system by non-linear loads. Both are invisible to anything that only reads energy consumed, and both damage equipment belonging to customers who then ask the utility why.
The first hop is the whole design
The link leaves the substation as 10Base-FL fiber and hits a fiber-to-copper repeater before anything electrical touches it.
That is not a bandwidth decision. Ten megabits over fiber in 2001 was slower and more expensive than copper. It is galvanic isolation. A substation ground grid rises to kilovolts during a fault, and a copper Ethernet run between that grid and anything outside it is a conductor for that rise — into the equipment at the far end, and into whoever happens to be touching it. Glass carries light and no current, so the fault stays in the yard where it belongs.
Everything downstream is ordinary: a 10/100 switch, a Cisco 3620, a T1.
A numbering plan, not a set of addresses
Each station was allocated a contiguous block before anything was configured: a /30 for its point-to-point link back to downtown, and a /27 for the meter network behind the router. The /30s step by four and the /27s step by thirty-two, in order, so the address of a station tells you which station it is and how much room is left.
The /27 is the part with judgment in it. It leaves room for twenty-eight meters behind a router that was serving a handful, which is the difference between a design and an installation — the second meter at a station is a form to fill in, not a renumbering exercise.
The same discipline runs through the naming. Each station’s link lands on its own subinterface at the downtown aggregation router, and the subinterface number is derived from the subnet it carries, so the number identifies the circuit without a lookup. Frame Relay carries a DLCI per station on the same scheme, allocated in sequence. None of this is clever; all of it is the reason a tenth site took an afternoon.
One build, ten times
Every station document describes the same thing: the WAN subinterface and its DLCI, the meter network behind the Ethernet interface, the switch on that network, and the address left free for the meters. Behind them the routers run a single interior routing process with a default back toward downtown, and the meters reach central services through helper addresses rather than needing anything local.
Management is uniform too, and uniformly enforced rather than uniformly intended — central authentication for login and privileged access, monitoring against a central collector, clock from a central source, and the same access banner on every device. A fleet configured to one standard can be audited against that standard; a fleet configured ten ways cannot be audited at all.
Whose transport it is
The T1 does not go to a carrier. It goes into a DDM-2000 and rides JEA’s own SONET — dual OC-3, with OC-12 north-south between the downtown node and the System Operations Control Center.
A utility that owns its transport gets sub-50-millisecond protection switching on a self-healing ring and answers to nobody’s service level agreement. Frame Relay encapsulation over a DS1 that is really a timeslot on your own fiber is a frugal piece of reuse: the circuit costs the utility a slot rather than a monthly bill.
The asymmetry is deliberate
The core is a redundant ring. The substation edge is single-threaded — one repeater, one switch, one router, no second path.
On a control system that would be a finding. Here it is the right answer, because this is measurement rather than control: losing the link loses visibility, not the ability to operate the plant. Resilience belongs in the transport, where it is shared across every station on the ring, rather than duplicated at each edge where it would be paid for many times over to protect data nobody trips a breaker on.
What is not on the drawing
There is no security boundary between the metering network and the utility network. Nothing sits between them.
That is period-accurate rather than an oversight — this predates NERC CIP, and the industry had not yet named the problem. It is also the reason the water treatment SCADA estate built later separates the control network from the corporate one by address range and by design: same author, same class of estate, after the question had been asked properly. The honest way to read the two together is as a before and an after.
Outcome
Power quality visible from the substation without putting a copper path across a station ground grid — and a pattern that could be repeated rather than redesigned per site.