Laser Diode Packages: TO-Can, C-Mount, CoS, and Butterfly

Scope

The same laser chip behaves like four different products depending on the metal around it. This guide covers the standard package families: TO-can, C-mount / chip-on-submount, coaxial/TOSA, and butterfly. It covers what's inside each, the thermal path, the electrical traps, and the selection logic for lab and product use. Companion procedures: junction temperature (the thermal consequence of the choice) and the datasheet guide (reading the package section before power-up).

TO-can (5.6 mm / 9 mm / TO-3)

The lightbulb-socket of laser diodes: chip on a header post, monitor photodiode on the base catching the back facet, sealed cap with window. Cheap, rugged, everywhere from CD pickups to instrument seeders.

The famous trap is internal wiring variants. With three pins serving a laser diode and a monitor photodiode plus the case, vendors wire them at least three ways: LD and PD sharing a common anode to case, sharing a common cathode, or PD floating. These are historically labeled styles like A/B/C, though not consistently across manufacturers. A driver configured for the wrong style reverse-biases the laser. The two-minute ritual before first power: datasheet pinout + DMM diode-mode check of both junctions against the case. Thermal reality: the only heat path is the header into whatever you clamp it into. A bare TO-can dangling from clip leads at 100 mW dissipation is running tens of degrees above ambient. Use a collet mount with real contact area; add TEC control for anything wavelength-sensitive.

C-mount and chip-on-submount (CoS)

The chip soldered to a copper (C-mount) or AlN/CuW submount, facets in open air. This is the lowest-thermal-resistance, highest-access format: the standard for high-power single emitters and for device characterization, where you need the facet for far-field or LIV work without a window in the way.

The costs are exposure. No hermeticity: facets collect dust and moisture, and un-passivated high-power facets degrade in humid air. No monitor, no TEC: you supply temperature control (mount on a TEC'd stage, thermistor close to the submount) and power monitoring externally. And absolute ESD discipline: an unprotected bare junction on an exposed carrier is the most static-fragile object in the lab. Grounded wrist strap, shorting clip during handling, drivers with slow-start.

Coaxial and TOSA

A TO-can heart with a fiber pigtail or receptacle aligned and welded on, in a cylindrical body. This is the transmit optical subassembly (TOSA) inside every pluggable module. Uncooled by design, so the laser inside is specified over the full case-temperature range (the uncooled-DFB and FP territory of access networks). For lab use, treat as a fiber-coupled black box: the alignment inside is laser-welded and permanent. There is nothing to adjust, which is the point.

Butterfly (14-pin) and mini-DIL

The instrument-grade answer: hermetic body and fiber feedthrough. Inside: chip on submount, TEC, thermistor, monitor PD, and often an isolator between chip and fiber. Everything a stable, wavelength-controlled source needs, hence its home in DWDM transmitters, pump lasers (fitted with a FBG stabilizer in the pigtail), and narrow-linewidth modules.

Butterfly layout is broadly standardized (TEC pair on the corner pins, thermistor pair, PD pair, LD pair, case grounds), but polarity conventions and NC-pin usage differ by vendor, and a TEC driven backwards heats precisely when commanded to cool, a failure that diagnoses as "thermal runaway" until you flip two wires. Verify pins against the specific datasheet drawing; check thermistor resistance (~10 kΩ at 25 °C typically) and both diode drops before applying drive. Mount the package flange to a heatsink able to absorb the TEC's worst-case draw (several watts). The internal TEC only moves heat to the case; you still have to remove it.

Choosing

Bench characterization of bare devices → CoS/C-mount on a TEC'd fixture. Wavelength-critical or narrow-linewidth work → butterfly, no debate: the internal TEC, monitor, and isolator are each worth their pins. Cost-driven products at loose specs → TO-can or TOSA, uncooled, with the temperature behavior (T₀, Δλ/ΔT) designed around rather than fought. And across all of them, the two universal rules: confirm the pinout empirically, and respect ESD handling as if every part were the last one, because at the wrong moment, it is.