Post: TO-3 Package: Metal Can Power Transistor Outline

Components Packages

TO-3 Package: Metal Can Power Transistor Outline

The TO-3 package is a hermetic diamond-shaped metal can for power semiconductors, registered by JEDEC as the TO-204 family. Two glass-sealed pins pass through the base on 10.92 mm centers, and the case itself forms the third terminal. It bolts to a heatsink through two 6-32 holes spaced 30.15 mm apart.

What follows is the geometry, the mounting hardware that actually determines your junction temperature, and the conditions under which a TO-247 or TO-264 is the better choice.

[IMAGE 1: photograph of a 2N3055 in a TO-3 package next to a TO-220 and TO-247 for scale, all three on a graph-paper background | alt: “TO-3 package metal can compared with TO-220 and TO-247 plastic power packages”]

Key takeaways

TO-3 is the legacy name. The registered outline is TO-204AA, with TO-204AD and TO-204AE differing only in pin diameter.
Datasheet RθJC is not the whole thermal path. On a TO-3, greased mica adds 0.36 °C/W and a dry bare joint adds 0.5 °C/W.
Use 6-32 pan-head screws at 6 in-lb (4–7 in-lb acceptable), with a conical Belleville washer. Split-ring washers flatten below the pressure the joint needs.
The TO-3’s power advantage over the same die in plastic comes from TJ(max) = 200 °C versus 150 °C, not from a lower junction-to-case resistance.
Never drill one large clearance hole for both pins. Drill each pin hole individually to the registered pattern.
Check lifecycle by orderable part number, not by package. The plastic MJL21194 went obsolete while the metal-can MJ21194 stayed active.

TO-3 is the common name; TO-204 is the registered one

JEDEC reorganized its outline registrations in late 1968, grouping packages by physical family and distinguishing members with suffix letters. The original TO-3 became TO-204AA, TO-41 became TO-204AB, and TO-204AC was added with a reduced maximum height of 7.62 mm. Two further variants exist purely to carry more current through the pins: TO-204AD uses 1.27 mm pins and TO-204AE uses 1.52 mm pins, against the original 0.040 in (1.02 mm).

The old designation never died. onsemi’s own application note observes that DO-4, DO-5, and TO-3 still carry enough meaning for verbal communication, and current datasheets hedge by printing both — the 2N3055 ordering table lists the package as “TO−204AA (TO−3), CASE 1−07”.

One trap catches people who assume the outline fixes the pinout. onsemi’s CASE 1-07 drawing registers nine pin styles on the identical mechanical outline. Style 1 is base / emitter / case-collector. Style 3 is gate / source / case-drain for a MOSFET. Style 4 is ground / input / case-output for a regulator. Style 7 is anode / open / case-cathode for a rectifier. Reading the outline number tells you where the screws go, not what the pins do — check the style number on the specific datasheet.

TO-3 package dimensions that constrain your design

These figures come from onsemi’s controlled outline drawing, document 98ASB42001B (CASE 1-07, issue Z), which governs the 2N3055, MJ15003, and MJ21193 families alike.

Dimension	Inches	Millimeters	What it constrains
A — long diagonal	1.550 REF	39.37 REF	Bracket and chassis cutout clearance
B — short diagonal	1.050 max	26.67 max	Device-to-device spacing on a sink
U — mounting hole centers	1.187 BSC	30.15 BSC	Heatsink drill pattern
G — pin spacing	0.430 BSC	10.92 BSC	Socket and PCB footprint
Q — mounting hole diameter	0.151–0.165	3.84–4.19	6-32 clearance
K — height above seating plane	0.440–0.480	11.18–12.19	Enclosure headroom
D — pin diameter	0.038–0.043	0.97–1.09	Socket contact and sleeve ID
C — pin length below seating plane	0.250–0.335	6.35–8.51	Maximum heatsink thickness

Two of these deserve comment. Dimension C means the pins protrude at most 8.51 mm below the flange, so anything thicker than about 6 mm under the package leaves nothing to connect to — TI caps the heatsink at 0.250 in for socket-mounted schemes for exactly this reason.

Dimension A is worth flagging because a figure of 40.13 mm circulates widely for the long diagonal. onsemi’s controlled drawing gives 39.37 mm as a reference dimension. Machine your bracket against the outline drawing shipped with the part you actually bought.

Where the TO-3 still wins, and why

The honest comparison is the same silicon in two different packages. onsemi builds its perforated-emitter audio die into both the MJ21194 (TO-204AA) and the MJL21194 (TO-264), which makes the package the only variable.

Parameter	MJ21194 — TO-204AA	MJL21194 — TO-264
RθJC, max	0.7 °C/W	0.7 °C/W
PD at TC = 25 °C	250 W	200 W
Derating slope	1.43 W/°C	1.43 W/°C
TJ, max	200 °C	150 °C
Is/b at VCE = 80 V, 1 s	2.5 A	2.25 A
VCEO / IC	250 V / 16 A	250 V / 16 A

The junction-to-case resistance is identical and so is the derating slope. The entire 50 W difference falls out of maximum junction temperature: 1.43 W/°C multiplied by the 175 °C span from 25 °C to 200 °C gives the metal can’s 250 W exactly. A hermetic package survives temperatures that crack and delaminate a moulded plastic body, and that — not a shorter thermal path — is what you are buying.

TI documents the same effect on the regulator side. The LM138/LM338 specifications hold for dissipation up to 50 W in the TO-3 (NDS) package and 25 W in the TO-220 (NDE) package, same die, factor of two.

Spreader area is the second axis. Measurements of the actual heatsink contact patch, corrected for bevels and mounting-hole cutouts, put the TO-3 at roughly 620 mm² against 313 mm² for TO-264 and 207 mm² for TO-3P. Under steady state this matters less than the numbers suggest, because the sink dominates. Under the transient peaks of a class-AB output stage or a disk-head positioner, spreading area governs how fast heat leaves the die, and three times the contact patch is a real advantage.

Mounting hardware sets your real thermal resistance

Designers budget carefully for RθJC and then throw the gain away at the mounting joint. onsemi measured interface resistance across package types at their recommended torques:

Interface at recommended torque	TO-204AA (TO-3), 6 in-lb	TO-220AB, 8 in-lb
Metal to metal, dry	0.5 °C/W	1.2 °C/W
Metal to metal, greased	0.1 °C/W	1.0 °C/W
Mica insulator, dry	1.3 °C/W (3 mil)	3.4 °C/W (2 mil)
Mica insulator, greased	0.36 °C/W (3 mil)	1.6 °C/W (2 mil)

Dry mica on a TO-3 costs you more thermal resistance than the transistor’s own junction-to-case path. Grease is not optional with a mica washer.

The fastener specification from Burr-Brown’s TO-3 mounting bulletin, now published by TI, is unusually prescriptive and worth following literally:

6-32 machine screws only. Pan head — low profile, wide bearing face, and it won’t ride up onto the lip of the welded cover.
Stainless steel preferred, plated steel acceptable. Brass and plastic are not recommended.
A conical Belleville washer, installed on the heatsink or board side of the fastener, large face toward the mounting surface. Split-ring and star lock washers bottom out below 50 lb, and the joint needs 150–300 lb to reach low thermal resistance.
6 in-lb of torque, with an acceptable range of 4–7 in-lb (0.45–0.79 N·m). Tighten slowly, alternate between the two screws, and reach final torque in at least two passes.
Keep thermal grease off the threads. Contaminated threads give inconsistent torque readings.
All-metal locknuts for permanent installation. Nylon-insert types are not suitable at temperature.
Sleeve the pins with #18 PTFE tubing, cut slightly shorter than the heatsink thickness.

Conformal pads relax over time and take your mounting pressure with them. A CHO-THERM 1688 joint measured at 0.90 °C/W dropped to 0.70 °C/W after 1000 hours as the material crept — but the mounting torque decayed from 6 in-lb to 3 in-lb in the process. With a conformal material the resistance improved; with a non-conformal insulator the same torque loss drives resistance up. Re-torque after burn-in or any thermal-cycling screen.

Drilling the heatsink to the TO-3 hole pattern

The single most damaging shortcut is drilling one large clearance hole for both pins. That removes contact area from directly beneath the die, which is the only area that matters.

Hole type	Quantity	Min / max diameter	Recommended drill
Mounting	2	0.147 / 0.157 in	#24
Pin	2 (8 for a socket)	0.073 / 0.094 in	#46 (#49 at minimum size)
Thermocouple (optional)	1	0.067 / 0.070 in	#51

Mounting hole centers sit 1.186 in apart on the long diagonal. Deburr every hole, but do not chamfer deeply — a chamfer trades contact area for cosmetics and increases mounting stress. Avoid punched holes in sheet heatsinks unless the supplier uses proper progressive dies; a punched hole craters the surface and the package will either bridge the crater or deform into it. Do not tap the mounting holes, since threading raises a mound around each hole and eliminates the Belleville-washer-and-locknut arrangement that controls your pressure.

[IMAGE 2: dimensioned drill drawing of the TO-3 heatsink hole pattern showing mounting, pin, and thermocouple holes with drill numbers | alt: “TO-3 package heatsink hole pattern with mounting, pin, and thermocouple drill sizes”]

JEDEC recommends surface flatness of 0.004 in/in maximum. Standard extruded heatsink stock is specified to that tolerance and typically measures nearer 0.002 in/in. Surface finish around 60 µin rms is adequate: Thermalloy found that finishes between 16 and 64 µin changed interface resistance by under ±2.5 % once the voids were filled with joint compound. If you are using a pad rather than grease, spot-face the mounting area to 0.001 in/in, because pads do not flow to conform the way grease does.

Choosing the thermal interface material

Interface	θCH (°C/W)	Isolating	Practical note
Bare joint	0.5–1.0	No	Low power only; highly sensitive to flatness and finish
Thermal grease	0.1–0.2	No	Tolerates standard 0.004 in/in flatness
Pre-coated aluminum foil	0.2–0.4	No	Wants 0.001 in/in for the low end of the range
Kapton with grease	0.3–0.5	Yes	Handles high temperature; wants 0.001 in/in
Silicone rubber pad	0.4–1.0	Yes	Settles over time; can flex the header
Mica, dry	1.0–1.5	Yes	Not recommended
Mica with grease	0.3–0.4	Yes	The default isolated choice

Where the design allows it, isolate the whole heatsink from chassis ground rather than inserting an insulator into the thermal path. That drops you from the 0.3–0.4 °C/W mica row to the 0.1–0.2 °C/W grease row, which on a 60 W device is worth roughly 15 °C of junction temperature.

If the sink must sit at chassis potential, hard anodizing per MIL-A-8625 Type III at 0.001 in thickness is typically rated for 200 VDC and resists scratches and punctures. The ordinary black anodize on most extruded sinks is Type II — corrosion protection and emissivity, not insulation. Do not rely on it electrically.

Worked example: a 60 W pass element

Take an MJ15003 in a TO-3 package: RθJC = 0.70 °C/W max, TJ(max) = 200 °C, PD = 250 W at TC = 25 °C. It dissipates 60 W continuously inside an enclosure that reaches 45 °C, isolated with greased 3 mil mica.

Budget to a junction temperature of 125 °C rather than the rated 200 °C. Field history shows the failure rate of silicon devices roughly halving when junction temperature drops from 160 °C to 135 °C, and military power-supply design guidelines cap TJ at 110 °C.

Allowable junction-to-ambient: (125 − 45) / 60 = 1.33 °C/W.

Subtract the fixed terms: 1.33 − 0.70 (RθJC) − 0.36 (greased mica) = 0.27 °C/W left for the heatsink.

Extruded heatsinks for a single TO-3 run 0.4–3 °C/W in still air, so 0.27 °C/W is not available. Three ways out:

Force air across it. Forced convection cuts θHA by one half to two thirds, bringing a 0.8 °C/W sink into range — but the design now fails on fan loss.
Drop the isolation. Mount the case directly to a floating sink and RθCS falls to 0.1 °C/W, raising the heatsink budget to 0.53 °C/W. Still tight, and it moves a live collector onto the heatsink.
Split across two devices. At 30 W each the allowable total per device is 2.67 °C/W; after RθJC and mica that leaves 1.61 °C/W per heatsink, comfortably inside the natural-convection range.

Option three is the one that survives design review. For scale on why the interface matters this much: at 50 W of dissipation, an extra 0.5 °C/W at the mounting joint raises the junction by 25 °C.

Five failure modes that show up in the field

Bent pins crack the glass seals. The pins are not designed to be bent. Excessive bending or twisting fractures the glass-to-metal seal and destroys hermeticity, which is the whole reason you chose the package. If a pin must be straightened, clamp it against the base with needle-nose pliers to strain-relieve the seal first.

Over-torque cracks the die or substrate. Bending the base — from over-torque, an out-of-flat surface, or tightening one screw fully before starting the other — flexes the header and fractures the internal substrate. This is the most common cause of early-life failures in metal-packaged parts.

A single large pin clearance hole starves the joint. It removes copper contact from directly under the die.

Anodize mistaken for insulation. Type II black anodize is not an insulator. Devices fail this way in production units where a prototype happened to pass hipot.

Case temperature measured in the wrong place. A thermocouple under the screw head reads the flange, and the case backside directly beneath the die can be up to 10 °C hotter. Drill the #51 thermocouple hole; simulation puts its thermal penalty below 0.02 °C/W, which is unmeasurable in practice.

Availability, price, and counterfeits

The TO-3 package is still in production. onsemi lists the 2N3055G as an active product — TO-204-2, case 1-07, 100 units per tray, package envelope 39.37 × 26.67 × 7.43 mm — at a catalogue price of $2.2473. The complementary audio pairs MJ21193/MJ21194 (250 V, 16 A, 250 W) and MJ15003/MJ15004 (140 V, 20 A, 250 W) both remain in the portfolio, and both datasheets were revised as recently as December 2024.

The lifecycle picture runs opposite to intuition. DigiKey lists MJL21194 — the TO-264 plastic version of the same die — as obsolete and no longer manufactured, while the metal-can MJ21194 stays orderable. Package format is a poor proxy for lifecycle risk. Check the specific orderable part number.

Counterfeiting is the real sourcing problem. The 2N3055 is among the most-faked power transistors in circulation, and the tell is die size: fakes carry a die far too small for a 115 W rating, sometimes adhesive-bonded rather than soldered to the header. Screen against a datasheet parameter rather than a multimeter diode check. onsemi specifies second-breakdown collector current with the base forward biased at 2.87 A minimum with VCE = 40 V for a 1 s non-repetitive pulse (2N3055, rev 7, October 2024). A part that will not survive that is not a 2N3055, whatever the lid says. Buy through authorized distribution, and destructively sample one unit from any lot that arrives with crooked marking or printing that rubs off.

TO-3 or a modern plastic package? A decision path

Specify a TO-3 package when the design needs a junction temperature above 150 °C; when hermeticity or MIL-PRF-19500 screening is a requirement; when you are servicing installed equipment that already has a drilled chassis, sockets, and a spares pipeline; or when large transient peaks make spreader area worth paying for.

Move to TO-247 or TO-264 when the board goes through automated assembly; when a single-screw mount and a fully isolated package variant would eliminate hardware and hipot risk; when unit cost dominates; or when the die you want simply isn’t offered in a metal can.

If you are retrofitting, note that the TO-3P outline was deliberately designed to drop into TO-3 mounting holes. It fits — but only about a third of the contact area lands on the sink. Solid copper adapter shims are sold to recover the spreading area, and independent measurements suggest they work. Validate the joint with a thermocouple in the case backside hole rather than assuming parity.

[IMAGE 3: exploded diagram of a TO-3 mounting stack — screw, flat washer, insulating bushing, package flange, mica insulator, heatsink, Belleville washer, locknut | alt: “Exploded view of TO-3 package mounting hardware stack with mica insulator and Belleville washer”]

Frequently asked questions

What is the TO-3 package used for?

Power semiconductors that dissipate tens of watts: bipolar power transistors, SCRs, and linear voltage regulators. Typical hosts are linear bench supplies, class-AB audio output stages, motor drives, and older industrial and military equipment. Devices in this outline handle currents on the order of tens of amperes and dissipation up to roughly 250 W at 25 °C case temperature.

Is TO-3 the same as TO-204?

Effectively yes. JEDEC re-registered the TO-3 as TO-204AA when it reorganized outline numbering in 1968. TO-204 is the family; AA is the original TO-3 with 0.040 in pins. TO-204AD and TO-204AE are the same body with 0.050 in and 0.060 in pins for higher current. Datasheets normally print both names.

How many pins does a TO-3 have?

Two pins on the standard transistor variant, with the case forming a third connection — usually the collector or drain. Multi-lead variants exist for integrated circuits, with 3-lead and 8-lead versions occasionally seen and a 15-lead version very rarely.

What torque should I use on a TO-3?

Six inch-pounds, with an acceptable range of 4 to 7 in-lb (0.45 to 0.79 N·m), using 6-32 pan-head screws and a conical Belleville washer. Tighten the two screws alternately in at least two passes, and keep thermal grease off the threads because it corrupts the torque reading.

Can I replace a TO-3 with a TO-247?

Electrically, often yes; mechanically, not directly. The TO-3 has two mounting holes on 30.15 mm centers and pins that pass through the heatsink; a TO-247 has one hole and leads that stay on the component side. TO-3P shares the TO-3 hole spacing but contacts only about a third of the area. Also check junction temperature: a hermetic TO-3 may be rated to 200 °C where the plastic equivalent stops at 150 °C.

What to do next

If your design needs a junction temperature above 150 °C, hermetic sealing, or a qualified military part number, keep the TO-3 package and mount it properly: 6-32 pan-head screws, Belleville washers, 6 in-lb applied alternately, greased mica if you must isolate, and a #51 thermocouple hole so you can prove the joint. If none of those three conditions applies, specify a TO-264 or TO-247 instead and spend the saved assembly cost on a larger heatsink — you will end up with a cooler junction and a cheaper build.

For anything that must survive a decade in the field, run the lifecycle check on the exact orderable part number before you release the drawing, and buy through authorized distribution. The counterfeit rate on popular TO-3 devices is high enough that a destructive sample from each lot is cheap insurance.

External references

Proposed internal links

[INTERNAL LINK: junction-to-case thermal resistance → thermal resistance fundamentals for power design]
[INTERNAL LINK: TO-247 package → TO-247 package guide]
[INTERNAL LINK: heatsink selection → heatsink sizing and selection for through-hole power devices]
[INTERNAL LINK: counterfeit component screening → detecting counterfeit semiconductors in the supply chain]
[INTERNAL LINK: JEDEC package outlines → JEDEC JEP95 outline naming explained]