The TO-72 package is the four-lead member of the TO-18 metal can family, registered by JEDEC as TO-206AF. Body dimensions match TO-18 exactly; the extra lead connects to the can and gives you a shield you can ground. It handles 300 mW in free air, with a junction-to-ambient resistance near 417 °C/W. The pinout is not standardized, and that is where most TO-72 designs go wrong.

Key takeaways

• TO-72 is TO-206AF. TO-18 became TO-206AA, TO-46 became TO-206AB, TO-52 became TO-206AC. All share the 2.54 mm (0.100 in) lead circle.
• Four different pin assignments for this one case appear in current manufacturer documents. Central Semiconductor publishes two FET conventions on the same page and tells you to check the device datasheet.
• The fourth lead is the can. Ground it or it works as a coupling plate rather than a shield.
• The hermetic metal can is thermally worse than plastic. The 2N4416 in TO-72 is rated 300 mW; the same die as SST4416 in SOT-23 is rated 350 mW.
• Availability is thin and shrinking. Mouser listed five TO-72 transistor line items in August 2026, and DigiKey shows the Vishay 2N4416A as obsolete.
• Specify TO-72 for hermetic RF sustainment or where the case shield earns its keep. New designs belong in SOT-23 at roughly one fiftieth the price.

What the TO-72 package is

The TO-72 package is a hermetically sealed metal can holding a transistor die on a Kovar header, with four leads on a 2.54 mm (0.100 in) circle passing through glass-to-metal seals. Body dimensions are identical to TO-18. The fourth lead bonds to the can itself, providing an electrostatic shield for RF work. JEDEC registers it as TO-206AF.

JEDEC consolidated every outline with a 0.100 in lead circle into the TO-206 family in Publication 95, November 1982. TO-72 became TO-206AF alongside TO-18 as TO-206AA, TO-46 as TO-206AB and TO-52 as TO-206AC. Cap height is what separates the siblings: 4.83 mm on TO-18 and TO-72, 3.30 mm on TO-52, and 1.90 mm on TO-46. The eight-lead TO-71 uses the same header with a 45 degree minimum angle between adjacent leads.

Both names remain in circulation. Vishay heads its 2N4416 drawing TO-206AF (TO-72), Semelab prints TO-72 (TO-206AF), and DigiKey files the package as “TO-206AF, TO-72-4 Metal Can”. Treat the two strings as one part space when you search for stock, because distributor parametric filters do not.

Construction is worth knowing before you write an outgassing or temperature requirement. Central Semiconductor reports the header as Kovar, an iron-nickel-cobalt alloy making up 59.38 percent of device mass, the can as a steel alloy at 35.34 percent with nickel inner plating and matte tin outside, aluminum bond wire, and a silver epoxy die attach. That last detail matters: the die is glued, not soldered, so this package does not tolerate the die-attach temperatures a soldered power package does.

TO-72 package dimensions and footprint

Figures below come from the Semelab 2N4416A drawing, Document 3631 Issue 2, cross-checked against the JEDEC-derived TO-18 family reference data.

Feature	Value	Source
Cap diameter	4.52 mm to 4.95 mm (0.178 in to 0.195 in)	Semelab doc 3631 Issue 2
Cap height	4.32 mm to 5.33 mm (0.170 in to 0.210 in)	Semelab doc 3631 Issue 2
Lead circle diameter	2.54 mm (0.100 in) nominal	Semelab doc 3631; defines the TO-206 family
Lead diameter	0.41 mm to 0.48 mm (0.016 in to 0.019 in)	Semelab doc 3631 Issue 2
Lead length	12.7 mm (0.500 in) minimum	Semelab doc 3631 Issue 2
Base diameter	5.6 mm (0.22 in) typical	JEDEC-derived TO-18 family reference data
Orientation feature	Tab located 45 degrees from pin 1	JEDEC-derived TO-18 family reference data
Device mass	322.2 mg ±10 %	Central Semiconductor TO-72 package details

Table 1. Verified TO-72 (TO-206AF) outline dimensions.

Four leads at 90 degree spacing on a 2.54 mm circle put the pads on a 1.80 mm square once you rotate 45 degrees, which is why most TO-72 footprints are drawn as a diamond of four holes rather than a circle. The tab sits 45 degrees from pin 1, so on a four-lead part the tab points between two leads, not at one. Get that wrong and the board is rotated 90 degrees.

Lead length is specified as a minimum, not a range. You are free to cut the leads to suit the board, and on RF parts you should. What you must not do is bend them where they leave the glass seal.

The TO-72 pinout is not standardized

This is the single most useful thing to know about the package, and no general reference page says it. Four distinct pin assignments for the same case appear in documents that are current today.

Source document	Pin 1	Pin 2	Pin 3	Pin 4
Central, TO-72 package details, transistor	Emitter	Base	Collector	Case
Central, TO-72 package details, FET option A	Source	Gate	Drain	Case
Central, TO-72 package details, FET option B	Source	Drain	Gate	Case
Central 2N918, R1 (11 Sep 2012)	Emitter	Base	Collector	Case
Semelab 2N4416A, doc 3631 Issue 2	Case	Gate	Drain	Source

Table 2. Published TO-72 pin assignments. Central lists two FET conventions on one page.

Read Table 2 twice. Central prints two mutually exclusive FET pinouts side by side on its own package data sheet, separated by the word “or”, with a footnote directing you to the individual device datasheet. Gate and drain swap between them. On a JFET amplifier, wiring gate to drain does not degrade performance, it removes the amplifier.

The case lead is pin 4 in four of the five rows and pin 1 in the fifth. If your library part assumes the case is always pin 4, a Semelab-sourced 2N4416A will short the source to your ground plane and the stage will sit at zero gain with no obvious fault.

The practical rule: never populate a TO-72 footprint from a package drawing. Pull the pinout from the datasheet for the exact manufacturer part number you are buying, and re-check it if the second-source part comes from a different vendor. Put the pin function, not just the pin number, in the schematic symbol.

What the fourth lead actually buys you

The case lead exists to shield. Bonding the can to circuit ground puts a grounded metal enclosure between the die and everything around it, which is why the package survived in VHF and UHF front ends long after plastic took over elsewhere.

The parameter the shield protects is reverse transfer capacitance. Vishay specifies the 2N4416 at 0.8 pF maximum for Crss, against 4 pF for Ciss and 2 pF for Coss. At those levels, a few tenths of a picofarad of stray coupling around the outside of the device is not a rounding error, it is a meaningful fraction of the feedback path that sets stability.

The gain those numbers support is real. Vishay specifies common-source power gain of 18 dB minimum at 100 MHz and 10 dB minimum at 400 MHz, with noise figure at 2 dB and 4 dB maximum respectively into a 1 kΩ source. That is the performance envelope you are paying the package premium for.

A floating case is worse than no shield. Left unconnected, the can is a piece of metal capacitively coupled to the die on one side and to whatever is nearby on the other, which turns it into a coupling plate. Tie it to the ground the stage actually references, and keep that lead short. The leads are 12.7 mm long as supplied, and at 400 MHz a long ground lead has enough inductance to stop being a ground.

Thermal behavior, and why the metal can loses

Engineers reach for hermetic metal expecting better heat performance. In this package they get the opposite.

Parameter	2N4416 in TO-72	SST4416 in SOT-23	2N918 in TO-72
Power dissipation at 25 °C ambient	300 mW	350 mW	200 mW
Power dissipation at 25 °C case	not specified	not applicable	300 mW
Derating above 25 °C	2.4 mW/°C	2.8 mW/°C	not specified
Implied θJA	417 °C/W	357 °C/W	see note below
Storage temperature	−65 to 200 °C	−65 to 150 °C	−65 to 200 °C
Operating junction temperature	−55 to 150 °C	−55 to 150 °C	−65 to 200 °C

Table 3. TO-72 against the SOT-23 version of the same JFET die. Vishay document 70242 and Central 2N918 R1.

The plastic SOT-23 part dissipates 17 percent more than the hermetic can holding the same die. The reason is structural: the TO-72 die sits in a sealed gas-filled cavity with heat leaving only through the header and leads, while a SOT-23 conducts through molding compound and lead frame into board copper. The metal can buys hermeticity, a wider storage range and a shield. It does not buy cooling.

One warning on published thermal numbers for this package. The Central 2N918 datasheet, R1 dated 11 September 2012, prints θJA as 87.5 °C/W and θJC as 58.3 °C/W. Those figures do not reconcile with the same page: at a 200 mW ambient rating and a 200 °C junction limit, junction-to-ambient works out at (200 − 25) / 0.2 = 875 °C/W, and the 300 mW case rating gives 583 °C/W. Both printed values are exactly one tenth of the derived ones. Derive from the power ratings rather than trusting the thermal table.

Note also that the 2N918 is rated 200 mW in free air but 300 mW with the case held at 25 °C. Clamping the can to a clip or a chassis gains you half again as much dissipation, which is the practical argument for a TO-72 heat clip in a linear stage.

Worked example: gate leakage in a high-impedance JFET stage

Take a 2N4416 in a buffer at VDS = 15 V and ID = 5 mA, so PD = 75 mW, in a 70 °C enclosure.

Self-heating: ΔT = 0.075 W × 417 °C/W = 31 °C, giving TJ = 101 °C.

Vishay specifies gate reverse current at 100 pA maximum at 25 °C and 100 nA maximum at 150 °C. Interpolating between those two guaranteed points, the 1000-fold rise across 125 °C is close to one doubling every 12.5 °C. At 101 °C junction the limit lands near 6 nA, roughly sixty times the room-temperature figure.

With a 10 MΩ gate bias resistor, 6 nA of leakage develops 60 mV of offset across it. In a buffer with a few hundred millivolts of headroom margin, that is the difference between a design that works on the bench at 25 °C and one that drifts out of specification in a hot rack. Either drop the bias resistor to 1 MΩ or budget the offset.

This is interpolation between two datasheet limits, not a curve reading, so treat it as a design margin rather than a prediction. The point stands either way: in this package the junction runs 30 °C above ambient at ordinary bias currents.

What you can still buy in TO-72, and what it costs

Prices and availability below were checked at Mouser and DigiKey on 8 August 2026.

Part	Manufacturer	Type	Availability	Price
2N2857 PBFREE	Central	NPN VHF/UHF	1,556 in stock	$8.55 at 1, $3.46 at 2,000
2N918 PBFREE	Central	NPN VHF oscillator	579 in stock	$7.98 at 1, $3.16 at 2,000
2N5179 PBFREE	Central	NPN VHF amplifier	1,788 due 22 Sep 2026	$4.95 at 1, $2.63 at 4,000
2N3821	Microchip	N-channel JFET, 50 V	31 in stock	$26.40 at 1
2N3822	Microchip	N-channel JFET, 50 V	non-stocked, 24 week lead	$26.40 at 1
2N4416A	Vishay Siliconix	N-channel JFET, 35 V	obsolete, not manufactured	no longer priced
2N4416A TO-72 4L ROHS	Linear Integrated Systems	N-channel JFET, 35 V	listed at DigiKey	$7.52 at 1

Table 4. TO-72 availability snapshot, 8 August 2026.

Three points come out of that table. The Central RF bipolars are the healthy end of the package, genuinely stocked in four-figure quantities and dropping under $4 at volume. The Microchip JFETs cost roughly six times as much and one of the two is on a 24 week lead. And the classic part, the Vishay 2N4416A, is gone: DigiKey lists it as obsolete and no longer manufactured, offering a TO-92 device as its only substitute.

Linear Integrated Systems still builds a 2N4416A in TO-72 with RoHS termination, listed at $7.52, which makes it the practical second source when a design cannot leave the package. Central sells its TO-72 parts in bulk boxes of 2,000 in static-shielded bags, so treat 2,000 as the natural production buy quantity even when the price break sits lower.

Distributor naming is a trap here. Filtering Mouser on “TO-72” returned five transistor line items in August 2026 and no JFETs from Vishay at all, because those are filed under TO-206AF. Run both strings, on both distributors, before you conclude a part is unobtainable.

Compliance and qualification

Central offers both plating options through a part number suffix: PBFREE for RoHS matte tin, TIN/LEAD for the traditional finish. Microchip lists its TO-72 JFETs as non-RoHS. Linear Integrated Systems calls its TO-72 4L part RoHS in the ordering string. Because the suffix, not the base number, decides compliance, an ordering error here is invisible until the declaration is audited.

For qualified work, the 2N918 is covered by MIL-PRF-19500/301 with JAN, JANTX and JANTXV levels available, and Semicoa publishes 100 percent die visual inspection to MIL-STD-750 method 2072 for JANTXV together with optional total-dose radiation testing. That combination, hermetic cavity plus screening plus radiation data, is the reason the package still exists.

Failure modes worth designing against

• Bending leads at the body. The leads exit through individual glass-to-metal seals that take no bending or axial load. A cracked seal loses hermeticity without any electrical symptom, so the part passes test and fails in the field. Cut the leads to length instead, and support the can if the assembly sees vibration.
• Leaving the case lead floating. It becomes a coupling plate rather than a shield, and the symptom is unexplained instability at the top of the band.
• Assuming a pinout from the package. Table 2 is the whole argument. Gate and drain swap between two conventions published by the same manufacturer.
• ESD on JFET gates during hand assembly. The 2N4416 gate is specified at 100 pA of leakage; a damaged gate junction shows up as a leakage shift long before it shows up as a dead part.
• Designing in a part without checking lifecycle. The 2N4416A is the best-known TO-72 device in the industry and Vishay has stopped making it.

Should TO-72 stay in your design?

Work down this list and stop at the first line that matches.

• You need a JAN, JANTX or JANTXV part, a hermetic cavity, or radiation screening: keep TO-72 and buy the qualified device. The 2N918 route through MIL-PRF-19500/301 is the well-trodden one.
• You are repairing or rebuilding equipment with an existing TO-72 footprint: keep the package, but verify the pinout of the specific replacement part number against the original before you fit it.
• You are building an RF front end above roughly 100 MHz where a grounded case shield materially improves stability, and 0.8 pF of reverse capacitance is a design constraint: TO-72 still earns its place.
• None of the above: move to SOT-23. The same 2N4416 die is available as SST4416 with a higher power rating, and comparable JFETs sell around $0.45 to $0.49 against $7.52 to $26.40 for the metal can.
• You need TO-72 but your usual part is obsolete: search TO-206AF as well as TO-72, and check Linear Integrated Systems and Central before declaring the design dead.

Frequently asked questions

What is the TO-72 package?

TO-72 is a hermetically sealed four-lead metal can for small-signal transistors and JFETs, registered by JEDEC as TO-206AF. It shares the TO-18 body: a cap around 4.95 mm across and 5.33 mm tall, with four leads on a 2.54 mm circle. The fourth lead connects to the can and is normally grounded as a shield.

Is TO-72 the same as TO-206AF?

Yes. JEDEC re-registered TO-72 as TO-206AF when it folded all 0.100 in lead circle outlines into the TO-206 family in Publication 95, November 1982. TO-18 became TO-206AA, TO-46 became TO-206AB and TO-52 became TO-206AC. Manufacturers and distributors still print both names, sometimes on the same drawing.

What is the fourth pin on a TO-72 transistor?

The can. It is bonded to the metal case rather than to the die, so grounding it turns the package into an electrostatic shield. That is why the outline survived in VHF and UHF front ends. Leaving it floating is worse than useless, because an ungrounded can couples signal rather than blocking it.

What is the TO-72 pinout?

There is no single answer, which is the point. Central Semiconductor publishes emitter, base, collector, case for bipolars and two different gate and drain orders for FETs. Semelab puts the case on pin 1 for the 2N4416A. Always take the pinout from the datasheet for the exact part number you are buying.

Is TO-72 the same size as TO-18?

Yes, apart from lead count. TO-72 is defined as the four-lead package otherwise dimensionally identical to TO-18, sharing the 4.83 mm cap height and the 2.54 mm lead circle. TO-46 and TO-52 differ from both, with cap heights of 1.90 mm and 3.30 mm respectively.

What can replace a TO-72 JFET?

For the 2N4416 family, Vishay built the same die as SST4416 in SOT-23, rated 350 mW against 300 mW. If the package must stay, Linear Integrated Systems lists a 2N4416A in TO-72 with RoHS termination. DigiKey offers the onsemi J111 in TO-92 as a similar substitute, though it is not a drop-in.

The decision

Specify the TO-72 package when a hermetic cavity, a grounded case shield or MIL-PRF-19500 screening is a requirement you cannot argue away. In that case, budget $8 to $27 per device, buy Central bipolars in bulk quantities of 2,000, and treat the pinout as a per-part-number question rather than a package property.

Everywhere else, design it out. SOT-23 gives you the same die with a better power rating at roughly one fiftieth of the cost, and the parts are still in production. The one thing it does not give you is the grounded can, so the honest test is whether your stage needs that shield. Above 100 MHz with a 0.8 pF feedback budget, it might. Below that, it almost certainly does not, and the metal can is buying you nothing but a lifecycle problem.