
The fast answer
- 5×6 mm = MFF2 — the mainstream 8-pin soldered eSIM defined in ETSI TS 102 671, and the part most suppliers actually stock.
- 3×3 mm = MFF3 — a 10-pin package commonly described as JEDEC MO-229F-compatible. Smaller body, finer pitch, thinner vendor list.
- 2×2 mm = MFF4 — an 8-pin package now moving into volume production for wearables and dense modules.
One table, three decisions
| Package | Body size | Pins | Pitch | Thickness | Defined by | Best for |
|---|---|---|---|---|---|---|
| 1FF–4FF | 85.6×54 to 12.3×8.8 mm | 6–8 contacts | n/a | 0.67–0.76 mm | ISO/IEC 7810, ETSI TS 102 221 | Removable SIM |
| MFF1 | 5×6 mm | 8 (dual pads) | 1.27 mm | 0.50–0.65 mm | ETSI TS 102 671 | Socketed M2M |
| MFF2 | 5×6 mm | 8 | 1.27 mm | ≈0.75–0.9 mm | ETSI TS 102 671 | Mainstream soldered eSIM |
| MFF3 | 3×3 mm | 10 | 0.5 mm (typical) | 0.65 mm | Referenced to JEDEC MO-229F | Dense wearables, compact IoT |
| MFF4 | 2×2 mm | 8 | Confirm in datasheet | Per MO-229F | JEDEC MO-229F-compatible | Ultra-compact devices |
| WLCSP | Die-scale | Vendor-specific | Vendor-specific | Vendor-specific | Vendor-specific | Modules with no spare space |
One thing worth repeating: ETSI TS 102 671 defines only MFF1 and MFF2. MFF3 and MFF4 sit in the JEDEC MO-229F family and are supported by far fewer silicon vendors. If a datasheet calls a 3×3 mm part "small MFF2", treat that as marketing language, not a standard.
MFF2 (5×6 mm): the safe default
MFF2 is what most engineers picture when they hear "eSIM chip": an 8-lead DFN, 1.27 mm pitch, an exposed die pad in the centre, industrial temperature range of −40 °C to +105 °C, shipped 4,000 pieces to a 13-inch reel.
Why it keeps winning:
-
Second sources. Multiple silicon vendors and distributors stock it, so a shortage does not stop your line.
-
Assembly-friendly. 1.27 mm pitch is forgiving. Standard stencils, standard placement, standard reflow.
-
Test-friendly. DFN8 sockets are cheap and everywhere, which means you can validate a batch without soldering anything.
The exposed pad is not an electrical node. It is there for heat and mechanical anchoring. Tie it to ground and forget it.
Five pins, two mistakes
The pin mapping is consistent across ETSI TS 102 671 and the MFF2 datasheets from STMicroelectronics, Avnet and Soracom:
| Package pin | UICC contact | Signal | Notes |
|---|---|---|---|
| 1 | C5 | GND | Ground |
| 2 | C6 | Not connected | Reserved |
| 3 | C7 | I/O | Serial data, T=0 |
| 4 | C8 | Not connected | Reserved |
| 5 | C4 | Not connected | Reserved |
| 6 | C3 | CLK | Clock |
| 7 | C2 | RST | Reset |
| 8 | C1 | VCC | 1.62–5.5 V |
Only five signals matter. The other three pins stay isolated — that is correct, not a shortcut.
Most first-pass failures come from two omissions, both cheap to fix and expensive to discover late:
-
No pull-up on I/O. ST's reference design calls for a 20 kΩ external pull-up. Without it, some readers fail to activate the card reliably.
-
No local decoupling. A 100 nF capacitor placed as close as possible to VCC prevents the kind of intermittent behaviour that costs days of debugging.
Neither part costs more than a few cents. Both cost weeks if missing.

MFF3 (3×3 mm): the space saver with a documentation gap
MFF3 cuts the body to 3×3 mm and the profile to 0.65 mm — but adds two pins and halves the pitch. You gain board area and take on a tighter process window: finer stencil, more precise placement, a reflow profile that will not tombstone a 10-lead part.
The bigger risk is not assembly. It is documentation.
Generic DFN 3×3 10-lead test sockets are easy to buy and inexpensive. But a socket's pin numbers are mechanical positions, not SIM signals. Public documentation mapping MFF3 pins to UICC contacts is scarce, and very few MFF3 eUICCs are in volume today. The major vendors' published datasheets concentrate on MFF2, MFF4 and WLCSP.
How to de-risk a 3×3 design
-
Request the pin assignment table for the exact part number from the silicon vendor — not the package drawing, the pin table.
-
Buy a small batch of samples plus a DFN 3×3 10-lead socket.
-
Identify VCC and GND with a multimeter, then confirm CLK, I/O and RST by powering the part through a PC/SC reader and watching for the ATR.
-
Only after that, release the footprint. A five-piece prototype run is always cheaper than a respin.
If a supplier cannot give you the pin table, that is your answer: pick a different supplier.
MFF4 (2×2 mm) and WLCSP: when every millimetre counts
MFF4 shrinks the soldered eSIM to a 2×2 mm, 8-pin body for wearables, trackers and dense gateways, while keeping full remote provisioning. Some product lines label the same package USON8-6 — another reason to check the drawing, not the name.
WLCSP goes smaller by packaging at wafer scale. It lives inside modules where the SIM is one component among many, and pinouts are vendor-specific. Expect less flexibility and harder rework.
Adapters and test fixtures: the fastest way to validate
If you need to test a soldered eSIM against legacy equipment, qualification rigs or evaluation kits, an adapter card gets you there quickly. Because only five signals matter, the routing is short:
| MFF2 pin | Card contact | Signal |
|---|---|---|
| 8 | C1 | VCC |
| 7 | C2 | RST |
| 6 | C3 | CLK |
| 1 | C5 | GND |
| 3 | C7 | I/O |
| 2, 4, 5 | — | Leave open |
Practical notes:
-
Keep the card's contact pads flush and flat — a raised pad changes how the card sits in the reader.
-
Route VCC and GND with a solid ground return; SIM current spikes are brief but real.
-
Put the 20 kΩ pull-up and 100 nF decoupling on the adapter if the host does not provide them.
-
A 3×3 mm design needs its own PCB. The 10th pin has no counterpart on an 8-pin footprint.
Which package for your next build?
| Your priority | Pick | Why |
|---|---|---|
| Supply security and second sources | MFF2 (5×6 mm) | Broadest availability, multiple vendors, forgiving pitch |
| Smallest possible board area | MFF3 (3×3 mm) or MFF4 (2×2 mm) | Smaller body, but fewer vendors and tighter assembly |
| Fastest route to a working prototype | MFF2 + DFN8 socket | Cheap sockets, no soldering, quick validation |
| Absolute minimum footprint inside a module | WLCSP | Die-scale, vendor-specific, harder to rework |
The rule we give most teams: choose MFF2 unless the mechanical envelope truly cannot fit it. A 3×3 mm part can save a few square millimetres and cost you weeks in tooling, stencil adjustment and vendor qualification. That trade is rarely worth it unless the product's size depends on it.
FAQ
Is 3×3 mm just a smaller MFF2?
No. MFF3 has 10 pins at a finer pitch and a different pad layout. The five SIM signals behave the same, but the physical mapping must come from the chip datasheet.
Can I socket an eSIM for engineering work?
Yes. DFN test sockets exist for both 8-lead 5×6 mm and 10-lead 3×3 mm bodies, typically US$8–12 at low volume. Use them for validation and small-batch programming, not production.
Do I need to route all eight MFF2 pins?
No. Five pins carry signals. Pins 2, 4 and 5 are reserved and stay isolated.
Does the exposed pad need solder?
Yes, for thermal and mechanical reasons — but connect it to ground, never to VCC or I/O.
Building something that needs a soldered eSIM?
Tell us the mechanical envelope, the volume and the host interface. We can help you pick the package, check the footprint against the real datasheet, and supply adapter or fixture hardware so you can validate before you commit tooling.
Send us your drawing and your target volume — we will tell you which package survives your supply chain.
Packaging details vary by silicon vendor. Verify pin assignment, pitch and thickness against the datasheet for the exact part number before releasing a footprint.

