TO18, TO46, TO56, TO60: Which Laser Diode Header for Your Application?

Q: What is the difference between TO56 and TO60?

Both are 5.6 mm-class telecom cans for high-speed lasers. TO56 is the 10G standard with MPD pin and RF-friendly leads. TO60 is optimized for 25G+ impedance-controlled routing and EML/coherent TOSA builds. Not a drop-in swap — PCB and thermal design change.

Q: Is there a TO9 laser package?

There is no mainstream telecom "TO9" laser header. Engineers often mean TO-39 (9 mm photodiode/laser cans), TO-92 (non-hermetic transistor outline), or a vendor-specific 9 mm outline. Specify base diameter and pinout on the drawing.

Q: When do I need cooled TO56/TO60 vs uncooled?

Add TEC when wavelength stability must stay below ~0.01 nm/°C (DWDM, coherent) or Tj must be held low. Most 10G datacom tolerates uncooled TO56. See our cooled vs uncooled TO guide.

Q: Where is the peer-reviewed depth?

Expand the technical review — hermetic sealing, getters, coupling efficiency, thermal Rth, reliability qualification, and extended TO family notes with DOI links.

The complete TO package selection guide — transmitter vs detector cans, data-rate limits, optical windows, thermal path, and when to move beyond standard TO to butterfly or TEC-cooled builds, plus an expandable literature review.

Start here: bandwidth, then can

A TO (transistor outline) header is a hermetic metal can with glass-to-metal sealed leads and an optical window or lens cap. It is still the default envelope for telecom DFBs, datacom EMLs, pumps, and many industrial emitters — because the ecosystem of caps, sockets, and TOSA assembly tooling is mature.

Pick the can after you know data rate, coupling optics, and thermal budget. The wrong TO forces a board respin or a coupling redesign that costs more than the header itself. Inside the can, the laser die still sits on a ceramic submount — header choice and AlN/SiC spreading are one stack.

Companions: optical window selection · FMCW vs pulsed packaging · FMCW phase noise & Tj · AuSn attach.

Quick selection (transmitters)

≤ 2.5G single-mode DFB/FP, cost-sensitive → TO18 (TO46 only if space forces 4.7 mm).
10G SFP+, CATV return path → TO56 with MPD pin and RF layout discipline.
25G+ SFP28/QSFP28, EML, PAM4 → TO60 or vendor proprietary high-speed can.
DWDM / coherent λ lock → TO56/TO60 + TEC; uncooled only if channel plan allows ± nm class drift.
Detector or slow sensor → TO5 / TO8 / TO-39-class — not TO56/TO60.

Full TO family at a glance

Package	Base ⌀ (typ.)	Speed class	Role	Primary use
TO46	4.7 mm	≤ ~2G	Emitter	VCSEL, ToF, compact sensing
TO18	5.6 mm	≤ 2.5G	Emitter	DFB/FP telecom, pumps, industrial
TO56	5.6 mm	~10G	Emitter	10G SFP+, CATV TOSA
TO60	5.6 mm*	25G+	Emitter	25G EML, coherent TOSA
TO5	9.0 mm	DC–low MHz	Detector	PIN/APD modules, sensors
TO8	12.7 mm	DC–low MHz	Detector	Large-area PD, arrays
TO-39†	9.0 mm	Varies	Either	Often PD; verify hermetic spec

*TO60 lead-frame outline is sometimes quoted as 6.0 mm by vendors — confirm mechanical drawing. †TO-39 is a common 9 mm can designation; do not confuse with non-existent "TO9."

TO18 — the workhorse

Lowest cost, widest supplier base, 2–3 pins, proven −40°C to +85°C field history. Hard ceiling near 2.5G from lead parasitics — not a path to 10G without external equalization fiction.

Standard part: FL-TO18-2G — telecom CWDM/DWDM, pumps, industrial CW.

TO46 — compact VCSEL and sensing

Smallest common hermetic emitter can. Strong for VCSEL arrays, proximity/ToF, and multimode short reach — weak for single-mode 10G+ because of lead inductance and aperture limits.

TO56 — 10G industry standard

Fourth pin for monitor photodiode (MPD), RF-aware lead routing, same 5.6 mm socket ecosystem as TO18 for mechanical tooling overlap. Requires controlled-impedance PCB — not a wire-bond-afterthought layout.

Standard part: FL-TO56-10G.

TO60 — 25G and coherent-class TOSA

Impedance-controlled transmission for 25G NRZ and advanced formats (PAM4, external modulator/coherent builds). Higher unit cost and tighter assembly yield than TO56. Many 100G lanes are four 25G optical paths — TO60 is the baseline hermetic can, though some vendors use proprietary cans above 50G aggregate.

Standard part: FL-TO60-25G.

TO5 / TO8 — detectors, not datacom lasers

Large-base cans optimized for photodiodes and sensors: wide temperature (−40°C to +100°C class), low dissipation, many pins on TO8. Using TO5/TO8 for 10G transmitters is a category error — bandwidth and RF design are not there.

“TO9” and other outlines engineers ask about

TO9: Not a standard telecom laser designation — clarify on drawings (often TO-39 or a custom 9 mm can).
TO-39: 9 mm hermetic can common for photodiodes and some lasers; specify leak rate, window, and pinout explicitly.
TO-52 / TO-72: Appear in legacy and specialty modules — validate against your socket vendor, not blog shorthand.
Butterfly / BOX: Beyond TO when you need TEC, fiber pigtail strain relief, and 40G+ coherent — not replaced by TO60 alone.

Optical window and coupling

Window / optic	Transmission (typ.)	Best for
Flat glass	~92%	Low-cost, non-critical coupling
Flat glass AR	> 99% at λ₀	Telecom — specify angle
Sapphire flat	> 96%	Harsh env, scratch resistance
Ball lens cap	Focused beam	Fiber pigtail, collimation
Aspheric cap	High NA	Premium SMF coupling, 25G+

Coupling method sets tolerance: butt ~10% efficiency and ±2 µm class alignment; ball lens ~30–40%; GRIN/microlens ~55–75%; aspheric up to ~80% with cost. Full window spec depth: optical window guide.

Thermal path inside the can

TO18/TO46: junction-to-case R_th often 200–300 K/W class on copper post — fine for sub-watt CW. TO56/TO60: lower R_th with optimized lead-frame and CuW spreaders in high-power TOSA builds. At 10–25G, die dissipation plus poor submount attach raises T_j and wavelength drift — size SiC or AlN and AuSn void limits before upgrading the can alone.

Common mistakes

Mistake	Outcome	Fix
TO60 for 2.5G CWDM	3–5× BOM, RF test burden	TO18 + AR window
TO18 for 10G SFP+	BER failures, heat	TO56 + RF PCB
TO56 for 25G PAM4	Eye closure	TO60 or proprietary can
TO8 for 10G TX	Wrong category	TO56/TO60 emitter can
Uncooled for 100 GHz-spaced DWDM	λ drift	TEC + λ lock loop

FerraLink supply

FerraLink supplies TO headers from ISO9001-qualified manufacturing partners. Glass-to-metal seal qualification data available; caps and windows per drawing. Samples from $25; 10-piece mixed TO18/TO56/TO60 evaluation box $250.

Specifying a TOSA or pump module? Expand the technical review — complete TO family pros/cons, hermetic/getter science, coupling economics, and qualification context with citations.

For experienced packaging engineers

Literature-backed TO package selection review

Peer-reviewed sources, interface data, and packaged-device literature — written by FerraLink materials engineering to support submount and attach decisions, not as neutral survey copy.

+26 min

Expand literature-backed review ↓

FerraLink publishes this expanded review because TO header selection is the highest-traffic packaging decision on our site — and because the full TO family (emitters and detector cans) is often collapsed to four part numbers. Below: mechanical roles, optical coupling physics, thermal and hermetic reliability, and when to exit standard TO entirely.

1. Standardization and hermetic architecture

TO packages evolved from discrete transistor outlines to optoelectronic cans requiring optical apertures, fiber-alignment tolerances, and controlled internal atmosphere. IEC/JEDEC-style outlines define base diameter and lead geometry; Telcordia-class qualification adds thermal cycling, humidity bias, and accelerated aging for carrier-grade modules.

Modern headers use kovar leads with glass-to-metal seals; cap weld completes hermeticity. Getters absorb residual H₂O, H₂, and organics that otherwise degrade facet coatings and power stability^{[Mauri 2024]}. Black-coated getter formulations reduce scattered light in high-power cans^{[Mauri 2024]}.

2. Transmitter packages — pros and cons

TO18 remains the cost and ecosystem leader for ≤2.5G: mature supply, wide pigtail tooling, acceptable R_th for sub-watt lasers^{[Verma 2008]} ^{[Yan 2021]}. Limits: lead inductance blocks 10G+ without redesign; no integrated MPD pin on standard 3-pin configs.

TO46 trades RF performance for 4.7 mm footprint — ideal for VCSEL and sensing, not for SM 10G TOSA. TO56 adds MPD feedback and RF-aware routing for 10G SFP+ ecosystems^{[Komanec 2023]}. TO60 targets 25G impedance-controlled paths and EML/coherent TOSA — higher cost, stricter PCB SI, and thermal margin at full modulation^{[Wu 2020]}.

3. Detector packages — TO5, TO8, TO-39

TO5 (9 mm) and TO8 (12.7 mm) prioritize environmental range and detector mechanics over bandwidth. TO-39 is a parallel 9 mm hermetic family often used for photodiodes — specify whether the drawing means TO5-style sensor cans or TO-39 pinout. None replace TO56/TO60 for datacom transmitters.

4. Optical coupling efficiency

Butt coupling (~10% theoretical best-case) demands ±2 µm class alignment; ball lens raises efficiency to ~30–40% with relaxed tolerance; GRIN rods and fiber-tip microlenses reach ~55–75%; aspheric caps target ~65–80% for premium 25G links. A 10 dB coupling margin at the transmitter can dominate datacenter power budgets — efficiency is a system-cost variable, not a lens afterthought^{[Wu 2020]}.

Window materials: borosilicate + AR for telecom volume; sapphire for abrasion and thermal shock; quartz when UV–NIR broadband matters. Laser-direct soldered optic micro-assemblies are an emerging packaging path for hybrid integration^{[Stenchly 2022]}. Drawing detail: optical window guide.

5. Thermal management and TEC

Junction temperature sets wavelength, threshold, and reliability. Early TO18 designs used centered Cu posts (200–300 K/W class). High-power TOSA work pushes CuW spreaders and multi-path lead frames toward ~1 K/W class in advanced pump modules^{[King 2023]} ^{[Verma 2008]}.

Failure modes: T_j drift, coating damage under localized flux, wire-bond lift under thermal cycling^{[Yan 2021]}. DWDM and coherent programs add TEC when uncooled λ drift exceeds channel plan — budget 2–3× laser electrical power for automotive-grade stabilization in harsh ambients^{[Verma 2008]}. DWDM/coherent programs: plan TEC and λ lock in the module BOM — see FMCW thermal and optical window guides.

6. Application mapping

Application	Typical TO	Notes
2.5G CWDM/DWDM	TO18	Cost leader; λ stability may need TEC
10G Ethernet SFP+	TO56	RF layout + MPD required
25G SFP28 / QSFP28	TO60	PAM4 stresses thermal + SI
CATV return / DTT	TO56	Proven 10G ecosystem
Pump / industrial CW	TO18	Submount Rth often first lever
VCSEL / ToF / auto	TO46	Compact; multimode friendly
PIN receiver slow link	TO5 / TO-39	Not TO56
Coherent / FMCW TX	TO60 + TEC or butterfly	See FMCW thermal guide

7. Reliability and qualification

Hermetic leak rate (He fine leak, <10⁻⁹ atm·cc/s class targets), humidity bias (85°C/85% RH), and thermal cycling (−40/+85°C, 100–500 cycles) gate telecom release. Wire-bond cross-section after cycling is mandatory for aerospace-tier programs. Optical degradation ties to moisture and contamination — getter and clean cap assembly are process requirements, not options^{[Mauri 2024]}.

8. Total cost of ownership

Over-specifying TO60 on a 2.5G line adds 30–50% system cost (package, PCB, RF lab). Under-specifying TO18 on 10G drives rework, equalization hacks, and field BER cost. Correct-first-time selection: TO18 ≤2.5G, TO56 at 10G, TO60 at 25G+, detector cans for receive-only paths.

9. Stack below the header — FerraLink view

The TO can is the hermetic envelope; yield and T_j are won on the submount and AuSn attach inside. Pre-deposited AuSn on DPC AlN/SiC reduces void fraction versus preforms for sub-mm padspreform vs predep void inspection. We supply headers, submounts, and evaluation kits so you qualify the full stack — not the catalog line item alone.

References

A. Verma et al. (2008). Thermal management issues in laser diode packaging. IEEE SMELEC. DOI
Y. Yan et al. (2021). Review of high-power semiconductor laser packaging. Front. Phys.. DOI
L. Mauri et al. (2024). Gaseous impurities in packaged laser diodes. IEEE LS24. DOI
M. Komanec et al. (2023). 1310 nm laser TOSA for radio-over-fiber. IEEE Access. DOI
H.-S. Wu et al. (2020). Optical transmitter module for O-band Si photonic engine. ECTC. DOI
B. King et al. (2023). High-power GaAs emitters — packaging thermal context. CLEO/Europe-EQEC. DOI
V. Stenchly & W. Reinert (2022). Opto-packaging with laser direct soldering. ESTC. DOI

FerraLink selects citations for packaging relevance; verify against your program requirements before qualification sign-off.

The part that depends on your die

The rules above hold for most edge-emitter modules. What changes from program to program is geometry, duty cycle, and how hard you are pushing junction temperature — those inputs decide material, thickness, and whether catalog samples are enough.

Target Gbps, modulation (NRZ vs PAM4), and transceiver form factor (SFP+, butterfly).
Coupling architecture (butt, ball lens, GRIN, aspheric) and required insertion loss.
Submount material, AuSn attach, and whether TEC is in the module BOM.

Go deeper — Package context

These guides answer adjacent questions teams ask while choosing a submount. Each ends the same way: what you can decide in general, then what needs your die and power.

More topics coming — thermal path, attach yield, qualification, and packaging context.

Frequently asked questions

What is the difference between TO56 and TO60?expand_more