Same physics, different packaging
A media converter converts one copper Ethernet port to fiber. An SFP module plugged into a switch does exactly the same conversion — the optics inside are functionally identical. The choice is therefore not about performance; both forward at line rate with negligible latency. It is about everything around the conversion: whose power supply, whose management, whose spare shelf, and how many links you are converting.
The standalone converter is a complete appliance: copper port, fiber port (fixed optic or SFP-based), its own small power supply, unmanaged and invisible to the network. The SFP-in-switch pattern makes the conversion a port on a managed device: visible, monitorable, powered by the switch, and standardised with the rest of your module stock.
The case for the converter
One link. That is the converter's whole argument, and it is a good one: joining two buildings, extending one camera run, landing fiber where no switch is wanted. A gigabit converter costs little, installs in minutes, and asks nothing of the existing network — either end can face any copper port. Converters also serve where the fiber side must exist independently of switch refreshes: the link stays when the switches are replaced. Chassis systems (ONV's catalogue includes 14/16-slot 2U racks) tidy the closet end once a site accumulates several converter links, giving them shared redundant power and one mounting point.
The converter is also the answer at the remote end of long links: a small box in a field cabinet, converting fiber back to copper for the local devices, is simpler than deploying a fiber-port switch to end one link.
The case for SFP in the switch
Multiply the links and the answer flips. Every converter is a separate power supply to fail, a separate box to mount, and a blind spot — when a converter dies, the network sees a dead link with no diagnostics. SFP ports on the switches aggregate those failure points into the switch's own (monitored, SNMP-visible, often redundant-powered) domain, and module standardisation makes spares trivial: one spare module covers every link of its type in the building.
The rule of thumb this site applies: a pattern of fiber (backbones, rings, multiple building links) belongs in SFP ports; an exception of fiber (one hop, one remote box) belongs in a converter. Most real estates end up with both — SFP at the core and aggregation, converters at the odd far end — which is why the ONV catalogue stocks deep ranges of each.
Power, heat and the closet nobody draws
Two practicalities complete the comparison. The first is power: a converter's wall adapter is the least reliable component in most links — a five-dollar supply deciding a five-figure camera system's fate. Chassis mounting exists largely to fix this, putting a row of converters on shared, redundant, rack-grade power. The SFP pattern inherits the switch's power engineering by construction. When a converter must be standalone, budget for a quality supply and a spare; the converter itself will likely outlive two of them.
The second is heat and density in the wiring closet. A dozen standalone converters with a dozen adapters is a thermal and cabling tangle that technicians curse; a chassis or a switch full of modules is a line in a rack elevation. Neither physics nor price separates the options at one link — but closets are where links accumulate, and the pattern that scales is the one that will look inevitable in three years. Choose per link if you must; choose the pattern if you can.
Worked example: a campus that runs both, on purpose
A school campus: a core room, six teaching buildings and a sports hall across a road. The six buildings are a pattern — each lands on twin SFP+ uplinks in the aggregation switches, modules standardised estate-wide (one multimode type in-building, one single-mode type between), one spare of each in the core drawer. The sports hall is an exception — one link, an awkward road crossing, and a single camera cluster at the far end: a converter pair does it, the far converter in the hall's cabinet next to a compact PoE switch that powers the cameras. Nobody proposed a seventh aggregation switch for the hall, and nobody proposed converters for the six buildings; the pattern/exception rule allocated each tool to its shape of problem.
Five years on, the arrangement's quiet virtues compound: the estate's fiber incidents are diagnosed from the NOC (the SFP links report themselves), the hall's converter has been swapped once — by a caretaker, from the photographed instructions taped inside the cabinet door — and the spare shelf holds exactly two module types and one converter. That is the real deliverable of the converter-vs-SFP decision: not today's link, but the estate's maintenance shape for a decade.
The decision, compressed
One link, one far end, no pattern in sight: converter. A pattern of fiber — backbones, rings, multiple buildings, anything an operator must see from a console: SFP ports on the switches. Both present: converters for the exceptions, modules for the pattern, and a ruthless standardisation of module types so the spare shelf stays at two or three items. Judge the choice not by the day it is installed but by the night, three years on, when someone who was not in the room has to keep the link alive with what is on that shelf.
One last practical note on documentation: whichever pattern a link uses, record it as a link — endpoints, fiber type, wavelength, module or converter model, and the date it was commissioned — in one register that lives with the network documentation, not in a departed installer's email. Converter links are the usual orphans, precisely because they were 'just one hop' on the day they were installed. Ten years later that one hop carries the gate cameras and nobody knows what it is. The register is five columns and ten minutes; the alternative is a tone generator, a ladder, and an afternoon you will not get back.
A closing observation on total cost: the two options price differently over time in a way quotations hide. The converter's costs arrive as small surprises — an adapter here, a blind swap there, one undocumented link becoming a project. The SFP pattern's costs arrive as one visible decision — the switch tier that carries the ports. Finance prefers the first pattern because it is cheaper this quarter; operations prefers the second because it is cheaper every quarter after. The estate's time horizon, stated honestly at the start, is the tiebreaker — and most estates outlive the horizon in the original business case.