Engineers keep describing the same situation to us. A switch or industrial gateway has a handful of SFP+ or SFP28 ports; once the unit is racked and running, the management software starts raising module temperature alarms, and in bad cases a port flaps. Pull the module out and the shell is too hot to hold. The first instinct is usually: fit a cage with a heat sink.
That can work on a single-row port. On the far more common 2×N stacked cage it usually cannot. We went through the selector fields for all 205 SFP, SFP+ and SFP28 cages and assemblies on the VOOHU website: only 13 come with a heat sink, and every one of them is a single-row 1×N part. Of the 135 stacked 2×N parts, not one has a heat-sink option. In other words, an overheating lower port on a stacked cage is not fixed by changing the cage. The fix lies in module power, port assignment and airflow. This article walks through it in troubleshooting order.
The temperature you see is reported by the module itself. Under SFF-8472, a module with digital diagnostics (DDM) reports its internal temperature in bytes 96–97 at I²C address A2h, and stores the high/low alarm and warning thresholds in bytes 0–7. Before touching hardware, read both from the host:
(1) the current temperature and how far it is from the high alarm; (2) whether the thresholds are factory values or have been changed by host software; (3) how much the temperature differs from port to port. Many “temperature alarms” turn out to be commercial-grade modules (typical case range 0 to 70°C) used in a warm cabinet. The module is fine; the temperature grade is wrong. If the host cannot read I²C at all, get the management signals working first, as covered in SFP Module Not Detected: MOD_ABS and I²C Checks.
The cage produces no heat; all of it comes from the module inside. SFF-8431 groups SFP+ modules into power levels, and the table below gives a rough guide. The figures are typical industry ranges; always check the datasheet of the module you actually use.
| Module type | Typical power | Thermal impact |
|---|---|---|
| SFP+ multimode short reach (SR) | About 1 W or less | A vented cage is usually enough |
| SFP+ single-mode long reach (LR/ER) | About 1–1.5 W | Watch airflow on dense ports |
| SFP28 short/long reach | About 1–1.5 W and up | Lower-port rise is noticeable at high density |
| 10GBASE-T copper SFP+ | Often 2.5 W or more | The hottest; place on a single row or upstream in the airflow |
In the field, 10GBASE-T copper modules cause the most trouble: they drive up to 100 m of copper from inside the module and draw roughly twice the power of an optical module. So start by checking what is plugged into the hottest port. Often, simply moving the copper module to another position relieves the problem.
Every cage in the VOOHU selector carries a “Heat dissipation” field with only two values: heat dissipation holes (vented) or heat sink. By family:
| Family | Total | Single row 1×N | With heat sink | Stacked 2×N (with heat sink) |
|---|---|---|---|---|
| SFP | 76 | 26 | 2 | 50 (0) |
| SFP+ | 66 | 30 | 8 | 36 (0) |
| SFP28 | 63 | 14 | 3 | 49 (0) |
| Total | 205 | 70 | 13 | 135 (0) |
Two things stand out. All 13 heat-sink parts are cage-only, supplied without a connector. All 135 stacked parts are cage-plus-connector assemblies, and every one is vented. This is not a gap in the data; it follows from the structure. On a stacked cage the lower port sits beneath the upper port, so there is no top surface on which to mount a heat sink.
| Part number | Family | Ports | Mounting | Notes |
|---|---|---|---|---|
| WHSFP00111W050 | SFP | 1×1 | Press-fit | — |
| WHSFP01014W036 | SFP | 1×4 | Solder | — |
| WHSFP10111W124 | SFP+ | 1×1 | Press-fit | Aluminum sink, -40 to 85°C |
| WH81-111-Y0018-1 | SFP+ | 1×1 | Solder | — |
| WHSFP10612W042 | SFP+ | 1×2 | Press-fit | — |
| WH81-112-00013-1 | SFP+ | 1×2 | Press-fit | — |
| WHSFP11014W085 | SFP+ | 1×4 | Press-fit | — |
| WH81-114-00008-1 | SFP+ | 1×4 | Press-fit | — |
| WH296T459D0S32 | SFP+ | 1×6 | Press-fit | — |
| WH296T219D8S32 | SFP+ | 1×6 | Not listed | Dual light pipes |
| WHSFP30211W018 | SFP28 | 1×1 | Press-fit | Black anodized Al, -40 to 105°C |
| WHSFP30714W004 | SFP28 | 1×4 | Press-fit | Dual light pipes |
| WH394T219D7S32 | SFP28 | 1×4 | Press-fit | Light-pipe field blank |
Heat-sink versions cover four single-row layouts: 1×1, 1×2, 1×4 and 1×6. If your design uses a single row and the temperature really is over the limit, look up the heat-sink version with the same port count in this table.
We pulled the drawings for two heat-sink cages and compared the key dimensions:
| Item | WHSFP10111W124 (SFP+) | WHSFP30211W018 (SFP28) |
|---|---|---|
| Height incl. heat sink | 13.2 mm (ref.) | 12.95 ±0.35 mm |
| Cage body height | 8.90 mm | 8.95 ±0.15 mm |
| Added by heat sink | About 4.3 mm | About 4.0 mm |
| Length × width | 48.80 mm max × 14.00 mm | 48.73 mm × 14.00 mm |
| Panel cutout | 15.45 × 10.40 mm | 15.45 × 10.40 mm |
| Heat sink material | Aluminum, stainless clip | Aluminum, black anodized, SUS301 clip |
| Operating temperature | -40 to 85°C | -40 to 105°C |
One point is easy to miss. The panel cutout does not change, because the heat sink sits behind the panel, so the front panel needs no rework. But you need roughly 4 mm of extra clearance above the cage. If a daughter board, a bracket or the enclosure lid sits there, the heat sink either will not fit or ends up pressed against something with no air around it, and then it can perform worse than a vented cage. Measure the space above the cage before you commit to a board change.
Since stacked 2×N cages have no heat-sink version, an overheating lower port has to be handled in other ways:
(1) Reassign ports. Put the high-power modules, such as 10GBASE-T copper and long-reach optics, in the upper row or at the ends of the row, and use the lower middle ports for short-reach optics or DAC cables. No board change, lowest cost.
(2) Use lower-power modules. At the same data rate, a short-reach multimode module often draws less than half the power of a copper module.
(3) Improve airflow. The vents on a stacked cage only help when air actually moves through them. Fans should push air along the length of the cage (front-to-back or back-to-front), not across it from the side.
(4) Move to a single row in the next revision. If the temperature is still too high after the first three steps, change the hottest ports to a single-row cage with a heat sink. The cost is half the port density on the panel, so evaluate it together with the mechanical design.
A heat sink does not create cooling. It spreads heat from the top of the module over a larger area, and moving air carries it away. So a heat-sink cage works only if two conditions hold: there is airflow, and the air runs along the fins. On both drawings the fins run along the length of the cage, which suits front-to-back airflow. With side-to-side airflow the fins block the air and the benefit drops.
The same applies to vented cages. Many small units are fanless and rely on natural convection. In that case module temperature depends mainly on the internal enclosure temperature and module power, whichever cage you use. Before changing the cage, confirm the module is industrial grade (typically -40 to 85°C) and check whether the enclosure vents allow convection.
When moving from a vented cage to a heat-sink cage, the PCB footprint follows the same SFP MSA layout, but still check each item against the datasheet:
Mounting: heat-sink cages come in both press-fit and solder versions. For SFP+ 1×1, for example, there is press-fit WHSFP10111W124 and solder-tail WH81-111-Y0018-1. If the original board uses press-fit holes, choose the press-fit version and match hole size and board thickness; the drawings specify a 3.00 mm PCB.
Connector: every heat-sink cage is cage-only and does not include the 20-pin connector, which must be ordered separately. Stacked parts ship as cage-plus-connector assemblies, and this difference is easy to miss when building the BOM.
Light pipes: some heat-sink cages include light pipes, such as WH296T219D8S32 and WHSFP30714W004. The LED positions must match the original design.
The data in this article comes from the VOOHU selector and product drawings. A few fields should be confirmed against the datasheet: WHSFP30211W018 shows “Nickel Plated Shell: None” online, while its drawing specifies nickel plating over the whole cage; the mounting field for WH296T219D8S32 is blank; the light-pipe field for WH394T219D7S32 is blank; and all 76 SFP-family parts list a maximum data rate of 5G. If anything is unclear during selection, ask our FAE team to confirm.
Step 1: Read DDM. Read the temperature and alarm thresholds for each port from the host, and identify the hottest port and the module in it.
Step 2: Separate a module problem from a position problem. Swap the hottest module with one from a port that runs cool. If the high temperature follows the module, the cause is module power or temperature grade. If it stays with the port, the cause is cooling at that position.
Step 3: Check the cage structure. A single-row cage can move to a heat-sink version; for a stacked cage, apply the port and airflow measures in Section 6.
Step 4: Validate on samples. After changing the cage or the airflow, run with all ports populated at the maximum ambient temperature for an extended period, then read DDM temperatures again and compare.
| No. | Check item |
|---|---|
| 1 | Model, power and temperature grade (commercial or industrial) of the module in the hottest port |
| 2 | DDM temperature and alarm thresholds; whether host software has changed the thresholds |
| 3 | After swapping modules, does the temperature follow the module or stay with the port? |
| 4 | Is the existing cage single-row 1×N or stacked 2×N? |
| 5 | Single row: does a heat-sink version with the same port count exist (see Section 4)? |
| 6 | Is there about 4 mm of clearance above the cage for the heat sink? |
| 7 | Does the mounting type (press-fit or solder) match the original hole size and board thickness? |
| 8 | Has a 20-pin connector been ordered separately for the heat-sink cage? |
| 9 | Does the enclosure airflow run along the heat-sink fins? |
| 10 | Temperature log at maximum ambient with all ports populated over an extended run |
The top of the upper port has space in principle, but the lower port does not, and the lower port is usually the hottest. A retrofit also has to deal with clip retention, EMI springs and the force applied when modules are inserted. We do not recommend it for production.
SFP parts follow the same MSA footprint, but mounting type (press-fit or solder), locating holes and board thickness still need checking. Compare the footprint drawings in both datasheets item by item before replacing.
It depends on the module's own temperature grade and the alarm thresholds stored in DDM. Commercial modules typically cover a 0 to 70°C case range, industrial modules -40 to 85°C. Keeping about 10°C of margin below the high alarm threshold is a safe target.
The most common cause is no airflow, or side-to-side airflow blocked by the fins. The next most common is that the space above the cage is closed off, so no air moves around the heat sink.
No. Only 3 of the 63 SFP28 products on our website have a heat sink; most SFP28 designs use vented cages. Whether you need one depends on module power, port density and airflow, not the data rate itself.
Light pipes sit at the front of the cage and mainly affect panel layout. Two of our heat-sink cages also include light pipes (WH296T219D8S32 and WHSFP30714W004), so the two can coexist; confirm details in the datasheet.
If your equipment is raising module temperature alarms, send us the port layout (single row or stacked), module models, enclosure airflow direction and the space available above the cage. VOOHU FAEs will help you decide between a heat-sink cage, reassigning ports or reworking the airflow, and can provide heat-sink cage samples and drawings.
——— VOOHU Engineering Team
Technical support: fae.thorne@voohu.cn | Tel: 400-1048-018
Suzhou VOOHU Electronic Technology Co., Ltd. | www.voohuele.com