The TO-247-3 package is the three-lead member of the JEDEC TO-247 through-hole family: roughly 21 mm by 16 mm of molded body, leads on a 5.44 mm pitch, an exposed drain or collector tab, and an isolated mounting hole. Three vendors’ drawings of it disagree by nearly 2 mm. Here are the outline numbers, the insulation limit almost nobody quotes, and the point at which four leads beat three.

Key takeaways

• JEDEC registers one outline, TO-247. Vendors ship it under at least five names, and their drawings differ by up to 1.75 mm on body length.
• Lead pitch is 5.44 mm BSC on Wolfspeed’s drawing and 5.46 mm BSC on Vishay’s. Both are stated as basic dimensions.
• Pin-to-pin functional insulation binds before blocking voltage does. Clearance of 2.6 mm and creepage of 2.8 mm coordinate to 400 V RMS or DC under IEC 60664-1 at pollution degree 2.
• Published mounting torque differs by 2:1 between a generic TO-247 application note and a specific SiC datasheet. Follow the device datasheet.
• At the top of the 1200 V SiC range the three-lead version is reaching end of life first. Check lifecycle status before you commit a footprint.

What the TO-247-3 package is

The TO-247-3 package is a three-lead plastic through-hole outline for power semiconductors, based on the JEDEC-registered TO-247 drawing. The molded body measures about 21 mm by 16 mm by 5 mm, the leads sit on a 5.44 mm pitch, and the exposed metal tab behind the body carries the drain on a MOSFET or the collector on an IGBT.

Naming is where the confusion starts. Vishay publishes the three-lead variant as TO-247AC and notes on the drawing that the outline conforms to the JEDEC TO-247 outline apart from dimension c, the lead thickness. onsemi and Qorvo, which absorbed United Silicon Carbide, both call it TO-247-3L. Infineon indexes it internally as PG-TO247-3-41, PG-TO247-3-44 and similar codes. Wolfspeed’s own outline drawing PRD-05964 is headed TO-247-3L while the datasheet cover page says TO 247-3.

The tab is electrically live. It sits at drain or collector potential and needs either an insulating pad or a heatsink that is isolated from chassis.

[IMAGE 1: Front and rear photograph of a TO-247-3 device with pin 1, 2, 3 and the exposed tab labelled | alt: “TO-247-3 package showing three leads, exposed drain tab and isolated mounting hole”]

Designation	Used by (source document)	Leads	Note
TO-247	JEDEC JEP95 registered outline	2 to 4	The registered drawing all others reference
TO-247AC	Vishay, document 95223	3	Conforms to JEDEC TO-247 except dimension c
TO-247-3 / TO-247-3L	Wolfspeed PRD-05964; Qorvo (United SiC) Rev. C; onsemi	3	Most common current usage in SiC datasheets
PG-TO247-3-41 / -44 / -901	Infineon package index	3	Internal codes; several coexist
TO247-3-HCC	Infineon AN-2021-09	3	High clearance and creepage; PCB-compatible
TO-247-3 Advanced Isolation	Infineon AN2017-42	3	Fully isolated tab, no thermal grease needed
TO-247PLUS-3	Infineon AN2017-01	3	Mounting hole removed, larger thermal pad
TO-247-4 / TO-247-4L	Wolfspeed PRD-05966; onsemi CASE 340CJ	4	Adds Kelvin (driver) source pin

Table 1. What the same outline is called across vendors. Names are interchangeable in conversation, not in a mechanical drawing.

TO-247-3 dimensions: three drawings, three answers

Dimension (mm)	Vishay TO-247AC (95223)	Wolfspeed TO-247-3 (PRD-05964)	Qorvo / United SiC TO-247-3L (Rev. C)	Composite spread
Body length D	19.71 to 20.70	20.75 to 21.05	20.803 to 21.463	1.75
Body width E	15.29 to 15.87	15.75 to 16.13	15.494 to 16.256	0.97
Body thickness A	4.65 to 5.31	4.83 to 5.21	4.699 to 5.309	0.66
Lead width b	0.99 to 1.40	1.07 to 1.33	0.991 to 1.397	0.41
Lead thickness c	0.38 to 0.89	0.55 to 0.74	0.381 to 0.889	0.51
Lead pitch	5.46 BSC	5.44 BSC	5.44 BSC	0.02
Mounting hole diameter	3.56 to 3.66	3.18 to 4.06 (P)	3.556 to 3.658	0.88
Overall lead length L	14.20 to 16.10	19.73 to 20.48	19.812 to 20.32	Datum differs

Table 2. Published TO-247-3 outline limits from three manufacturer drawings. Overall lead length is measured from different datums and is not directly comparable.

The composite envelope across those three drawings runs from 19.71 mm to 21.46 mm on body length and 15.29 mm to 16.26 mm on width. That is a spread of 1.75 mm and 0.97 mm on a part every catalog lists under one package name. A clip, a bracket or a machined heatsink pocket cut to one vendor’s maximum will not necessarily take another’s part.

Lead pitch is the quieter problem. Vishay registers 5.46 mm BSC where Wolfspeed registers 5.44 mm. Across two gaps that is 0.04 mm of cumulative drift, which a 1.0 mm plated hole swallows without complaint but a press-fit fixture or a bed-of-nails test socket may not.

The rule that follows: qualify against a named manufacturer outline drawing at a named revision, not against the string TO-247-3. Put the drawing number and revision in the approved vendor list next to the part number.

Pinout: the four-lead part is not a drop-in

On Wolfspeed’s TO-247-3 drawing the sequence is pin 1 gate, pin 2 drain, pin 3 source, with the tab as a fourth drain node. On the same vendor’s TO-247-4 drawing the sequence is pin 1 drain, pin 2 source, pin 3 driver source, pin 4 gate. The gate has moved from one end of the row to the other and drain and source have swapped. Moving a design to a Kelvin-source device needs a new footprint, not a jumper.

The four-lead body is also longer. Wolfspeed lists 23.30 mm to 23.63 mm against 20.75 mm to 21.05 mm for the three-lead outline, so each device claims about 2.5 mm more board length.

[IMAGE 2: Side-by-side pin maps of TO-247-3 (G, D, S) and TO-247-4 (D, S, driver source, G) with arrows showing the reversal | alt: “TO-247-3 package pinout compared with TO-247-4 Kelvin source pinout”]

Insulation coordination: the 400 V that catches people out

Infineon’s application note AN-2021-09 puts standard TO-247-3 pin-to-pin clearance above 2.6 mm and creepage above 2.8 mm. Coordinated per IEC 60664-1, those distances support a 3.6 kV transient overvoltage on the clearance path at pollution degree 2, inhomogeneous field, below 2000 m altitude, and 400 V RMS or DC on the creepage path at pollution degree 2 with a material group II mold compound.

That second figure is the one that surprises people. A 1200 V blocking rating describes what the die withstands. It says nothing about the working voltage the package body will support between adjacent pins once condensation or dust arrives. In an 800 V DC-link converter the gate-to-drain working voltage sits far above 400 V, so a standard body needs help: conformal coating, potting, or a higher-creepage variant.

Variant	Clearance	Creepage	Transient (clearance)	RMS or DC (creepage)
TO-247-3 standard	> 2.6 mm	> 2.8 mm	3.6 kV	400 V
TO247-3-HCC	> 3.4 mm	> 4.8 mm	4.4 kV	676 V
TO-247-3 Advanced Isolation	Not stated	5.33 mm terminal-to-terminal	Tab isolation tested above 2.5 kV RMS for 60 s	Not stated

Table 3. Pin-to-pin insulation coordination at pollution degree 2, material group II (400 <= CTI < 600), altitude below 2000 m. Values from Infineon AN-2021-09 and AN2017-42, coordinated per IEC 60664-1.

Two variants exist for that. The TO247-3-HCC package raises clearance above 3.4 mm and creepage above 4.8 mm, which coordinates to 4.4 kV transient and 676 V RMS or DC, and Infineon states the body dimensions and pin pitch stay compatible with an existing TO-247-3 PCB layout. The TO-247-3 Advanced Isolation version encloses the terminals in plastic covers that raise creepage between them to 5.33 mm and is tested on every unit to more than 2.5 kV RMS for 60 s. Neither removes the need for coating or potting at pollution degree 3 and above, which covers welding supplies and outdoor PV inverters.

What the Kelvin pin buys, and how not to measure it

Parameter (same die, 1200 V, 16 mΩ)	C3M0016120D (TO-247-3)	C3M0016120K (TO-247-4)	Comment
Body length	20.75 to 21.05 mm	23.30 to 23.63 mm	Roughly 2.5 mm more board length
Lead pitch	5.44 mm BSC	5.08 mm BSC	Different pitch as well as different count
RθJC	0.27 °C/W	0.23 °C/W	Modest thermal gain
ID at TC = 100 °C	85 A	90 A	Follows the thermal path
PD at TC = 25 °C	556 W	483 W	Datasheet as printed
TJ range	-40 to +175 °C	-55 to +175 °C	Wider low end on the 4-lead part
Eon (body diode FWD)	4.64 mJ at 75 A, RG 5 Ω	2552 µJ at 80 A, RG 2.5 Ω	Conditions differ: not comparable
Eoff (body diode FWD)	2.93 mJ at 75 A, RG 5 Ω	788 µJ at 80 A, RG 2.5 Ω	Conditions differ: not comparable
Mounting torque	1 N·m (M3 or 6-32)	1 N·m (M3 or 6-32)	Identical

Table 4. Wolfspeed C3M0016120D against C3M0016120K, the same 1200 V die in three-lead and four-lead outlines. Datasheet revisions Rev. 02 (September 2024) and Rev. 5 (November 2025).

Do not subtract the switching energies in that table. Wolfspeed characterises the three-lead part at 75 A with 5 Ω of external gate resistance and the four-lead part at 80 A with 2.5 Ω. Different current, different drive impedance, different loop inductance. The two datasheet columns are not a controlled comparison, and treating them as one overstates the gain.

For a controlled figure, use matched-condition data. Wolfspeed’s SiC design guidance reports roughly 430 µJ of total switching loss for a TO-247-3 SiC MOSFET at 30 A with 12 nH of source inductance, against about 150 µJ for the same product in TO-247-4. Their product literature puts the general case at up to 30% lower switching loss.

The engineering rule underneath it: the Kelvin pin pays when the di/dt across the shared source inductance subtracts meaningfully from the applied gate-source voltage. On silicon superjunction parts below roughly 20 A and 20 kHz that subtraction is small, and the extra footprint work rarely returns its cost. On SiC at any useful slew rate the common source inductance in a three-lead package is the thing limiting your switching speed.

Thermally the two sit close. The same die measures 0.27 °C/W junction-to-case in TO-247-3 and 0.23 °C/W in TO-247-4.

Mounting, and the torque numbers that disagree

TI’s TO-247 application note AN-1364 caps mounting torque at 50 N·cm. Wolfspeed’s C3M0016120D datasheet specifies 1 N·m with an M3 or 6-32 screw. Both are correct for what they cover. The TI figure is a family-level ceiling written around older mold compounds and die-attach; the Wolfspeed figure is qualified against one specific case.

Decision rule: torque to the device datasheet. Where the datasheet is silent, take the lower value, 0.5 N·m, and use a calibrated driver rather than a wrist.

The documented failure modes are worth listing plainly.

• Over-torque warps the package, which reduces contact area against the heatsink, raises case-to-sink thermal resistance and pushes junction temperature up.
• Push further and the die cracks.
• An extruded fiber washer under the screw head spreads the load and prevents chipping. TI names SPC Technology FSW-04-018 as a specific example.
• Solder the leads to the board before final tightening of the heatsink screw, or the leads absorb the misalignment.
• TO-247 is not recommended for surface mounting, and its through-hole footprint follows the TO-220 family hole pattern.
• Lead bending is a tooled operation. Additional bends outside that tooling are not guaranteed.

Worked example: what 556 W actually means

The Wolfspeed three-lead part carries a 556 W dissipation rating. Check where it comes from: (175 °C − 25 °C) / 0.27 °C/W = 556 W. It is the arithmetic ceiling with the case pinned at 25 °C, a condition no converter ever meets.

Run it at a realistic case temperature. At 100 °C the same expression gives (175 − 100) / 0.27 = 278 W. Holding the case at 100 °C while dissipating 200 W into 40 °C ambient demands a case-to-ambient path of (100 − 40) / 200 = 0.30 °C/W including insulator and thermal interface material. That is forced air over a substantial extrusion, not a clip-on finger.

With no heatsink at all the datasheet quotes 40 °C/W junction-to-ambient, which permits (175 − 25) / 40 = 3.75 W. The same package, rated 556 W and capable of under 4 W, depending entirely on what you bolt it to. The TO-247-3 is a thermal interface, not a heatsink.

Lifecycle, second sourcing and counterfeits

Availability is moving. DigiKey lists Wolfspeed’s C3M0016120D, the three-lead 1200 V 16 mΩ part used throughout this article, as obsolete and no longer manufactured, while the four-lead C3M0016120K remains active on Wolfspeed’s site with a November 2025 datasheet revision. Of the three substitutes DigiKey proposes, two are four-lead devices: Infineon’s IMZA120R012M2HXKSA1 and onsemi’s NTH4L013N120M3S.

The same pattern shows in onsemi’s catalog. Its newer 1200 V M3S and M3P SiC parts appear in TO-247-4L and D2PAK-7L, while TO-247-3L carries the older M1 generation such as NTHL020N120SC1, NTHL080N120SC1A and NTHL160N120SC1. Three leads is not disappearing from the 650 V mainstream or from IGBTs and rectifiers, but at the fast end of 1200 V SiC it is now the legacy option.

Three practical consequences. First, design the footprint for the four-lead outline from the start if the topology is hard-switched SiC; the extra 2.5 mm is cheaper than a respin. Second, when a second source is genuinely needed, compare outline drawings before comparing electrical parameters, because body length and pitch differ. Third, buy through authorized distribution. A package this widely copied, this expensive per unit and this easy to remark is an obvious counterfeit target, and the cheapest incoming check that catches remarks is a thermal one: measure junction-to-case resistance on samples and reject anything far from the datasheet figure.

[IMAGE 3: Decision flowchart from device technology and current level through gate-loop sensitivity and pollution degree to TO-247-3, TO247-3-HCC or TO-247-4 | alt: “Decision path for choosing the TO-247-3 package versus TO-247-4 or a high-creepage variant”]

Frequently asked questions

Is TO-247-3 the same as TO-247AC?

Practically yes, with a caveat. Both describe the three-lead variant of the JEDEC TO-247 outline. TO-247AC is Vishay’s designation, and its drawing states that the outline conforms to JEDEC TO-247 except for dimension c, the lead thickness. Body length and width still vary by more than a millimetre between vendors, so treat the names as interchangeable in conversation but not in a mechanical drawing.

What is the lead pitch of a TO-247-3 package?

Nominally 5.44 mm, though it depends whose drawing you read. Wolfspeed’s TO-247-3 outline gives 5.44 mm BSC and Vishay’s TO-247AC gives 5.46 mm BSC, both as basic dimensions without tolerance. The difference is negligible for a soldered through-hole joint on a 1.0 mm plated hole and can matter for press-fit tooling or test fixtures.

Can I put a TO-247-4 device in a TO-247-3 footprint?

No. Beyond the extra lead, the pin order reverses. Wolfspeed’s TO-247-3 runs gate, drain, source, while its TO-247-4 runs drain, source, driver source, gate. The four-lead body is also about 2.5 mm longer. A Kelvin-source upgrade needs a new footprint and a rerouted gate loop.

How much torque should I use on a TO-247-3?

Use the figure in the device datasheet. Wolfspeed specifies 1 N·m with an M3 or 6-32 screw. TI’s generic TO-247 application note caps torque at 50 N·cm. If the datasheet gives no number, use the lower value and a calibrated driver, add a fiber washer under the screw head, and solder the leads only after the heatsink screw is set.

Is TO-247 the same as TO-3P?

They are similar in size and both use a 5.45 mm nominal lead pitch, but they are not the same registered outline. Toshiba’s part-numbering scheme assigns separate package codes to TO-247 and to TO-3P variants, which is a good indication that insulators, clips and sockets should be selected for the specific one you have.

Specify it, or step up

Choose TO-247-3 when the switch is an IGBT, a rectifier, a silicon superjunction MOSFET below roughly 20 A and 20 kHz, or any part where the gate loop is not what limits you, and when the pin-to-pin working voltage stays under 400 V RMS or the board is coated. Under those conditions the extra pin buys nothing and the shorter body saves space.

Move to TO-247-4 when the device is SiC and hard-switched, when di/dt through the source lead is measurably eroding gate drive, or when the part you want is only released in four leads, which is increasingly the case above 1200 V and below 20 mΩ. Move to a high-creepage variant such as TO247-3-HCC when the equipment sees pollution degree 2 with condensation and you would otherwise be coating the board to compensate.

Whichever you pick, do three things before release: pull the specific outline drawing and revision into your library rather than a generic footprint, confirm lifecycle status with the manufacturer, and set the mounting torque from the datasheet with a calibrated tool.

Internal links to place

• [INTERNAL LINK: TO-247 vs TO-220 thermal comparison -> through-hole power package selection]
• [INTERNAL LINK: SiC MOSFET gate driver layout -> gate loop inductance and Kelvin source]
• [INTERNAL LINK: creepage and clearance for power converters -> IEC 60664-1 insulation coordination]
• [INTERNAL LINK: heatsink selection and thermal resistance budgets -> thermal design for discrete power devices]
• [INTERNAL LINK: avoiding counterfeit power semiconductors -> component sourcing and authorized distribution]