DEV Community: geoffery Bob

How to Select 2.5G/5G Ethernet Magnetics: A Speed Rating Is Only the Starting Point

geoffery Bob — Fri, 09 Oct 2026 02:02:44 +0000

Selecting 2.5G/5G Ethernet magnetics starts with the target PHY’s magnetics requirements and the signal integrity of the complete Ethernet interface. A speed rating gets a component onto the shortlist. Qualification depends on how the component, PCB, connector, and cable perform together.

1. Moving beyond Gigabit Ethernet means reviewing the entire signal path

As wireless access capacity increases, the wired backhaul can become a bottleneck. This creates a practical role for 2.5GBASE-T and 5GBASE-T: increasing copper Ethernet link rates where the installed cabling supports the intended application. That capability does not come from replacing a single component in an existing Gigabit Ethernet design.

Start by defining the system boundary. The PHY handles physical-layer transmission and reception, while the Ethernet transformer provides signal coupling and galvanic isolation. PCB traces, pads, connectors, and cables all contribute to the transmission path. Impedance discontinuities along that path can increase reflections. Focusing exclusively on the magnetics can leave the actual problem unresolved.

Do not carry over the magnetics requirements of a Gigabit PHY without checking them. Obtain the reference design for the target 2.5G or 5G PHY, and confirm the winding configuration, center-tap connections, and frequency-domain requirements. Then screen components against the relevant interface and test conditions. Center-tap connections depend on the specific PHY and power architecture; there is no universal wiring rule.

Figure 1. Validation boundaries from the PHY to the link partner. This conceptual diagram is not a schematic; use the selected design’s connection and termination requirements.

2. Check the return-loss convention before comparing frequency responses

Return loss describes reflection, while insertion loss describes transmission loss. They are different quantities. For a differential interface, the mixed-mode S-parameter Sdd11 describes differential reflection at the input port.

The sign convention is easy to misread. With return loss expressed as a positive quantity, RL = −20 log10|Sdd11|: a higher value means less reflection. If the instrument displays 20 log10|Sdd11| instead, a more negative trace means less reflection. Confirm the convention before comparing reports.

Consider an illustrative calculation at the same frequency, with the same real reference impedance and consistent terminations. A return loss of 10 dB corresponds to differential reflected power equal to 10% of the incident differential power; at 20 dB, the proportion is 1%. The relationship is 10^(−RL/10). These figures do not mean throughput falls by 10% or 1%, and they do not describe mode conversion or other losses.

Figure 2. Calculated examples, not measured product data. A single frequency point explains the relationship; it does not demonstrate compliance across the required frequency range.

Frequency coverage matters just as much. A good result at one low-frequency point does not establish performance across the band. When requesting curves, also confirm the reference impedance, port definitions, fixture calibration, and de-embedding conditions. Otherwise, the two reports may describe different measurement boundaries.

3. Use operating requirements to narrow the shortlist

For example, VOOHU’s approved catalog lists WHSQ24002G for 2.5G/5G operation with an operating temperature range of 0 to 70 °C. If the equipment’s environmental requirements fall within that range, the component can proceed to further technical review.

For an application requiring −40 to +85 °C, prioritize candidates with the corresponding temperature rating, such as WHSQ24002TG. The temperature range is a screening criterion. It does not establish better return loss on your PCB or demonstrate that one part can directly replace the other.

Both parts are listed as non-PoE. If the port will also carry power, select candidates explicitly specified for the relevant PoE conditions. A data-rate rating does not establish power-handling capability. Following that sequence makes VOOHU Ethernet magnetics information useful as an input to the design process.

4. Engineering defines the test coverage; procurement controls the approved revision

Validation should extend from the component to the assembled board. Check insertion loss, return loss, and other requirements specified for the target PHY, then cover the intended cables, link partners, and operating temperatures. Record negotiated link speed, link stability, and traffic-test results. Include applicable interface compliance and EMC testing according to the project requirements.

A link LED is insufficient evidence. A connection that negotiates a lower speed can still appear to work. The report should state the actual link rate, traffic load, test duration, and environment. Application throughput is also different from the nominal physical-layer rate; a single speed-test result cannot settle the qualification decision.

For procurement, include the engineering-approved full part number, datasheet revision, and sample revision in the request for quotation. Request confirmation of lot traceability and the change-notification process instead of asking only whether a “5G transformer” is available. Release depends on the evaluation record; lead time and supply terms require written confirmation for the specific order.

Why Wi-Fi 7 Needs a 2.5G Port: Seeing the Wired Link Behind the Wireless Through a NAS Transfer

geoffery Bob — Thu, 08 Oct 2026 06:24:20 +0000

Wi-Fi 7 devices come with a 2.5G Ethernet port for a reason: once the wireless side can push much more throughput, the old gigabit wired link stops being enough and becomes the bottleneck. But whether your files actually move faster depends on which ports the data passes through, and on what the endpoints, the storage, and the network are truly capable of.

When you're shooting video, organizing photos, or building out a media library, you run into a very practical question. Your laptop connects to the new-generation Wi-Fi, yet copying files to the NAS doesn't feel any faster. The wireless logo has been upgraded — so why is the wait still the same?

Let's trace the data path first.

My take: the most valuable thing about a network upgrade is learning to find where the performance limit actually lives. If you only chase the biggest number on a single device, it's easy to spend your budget on a segment that was already fast enough.

1. Once the wireless gets faster, you still have to see where the data leaves

Wi-Fi 7 brings higher-order modulation and Multi-Link Operation, among other improvements. Multi-Link Operation, usually shortened to MLO, lets supporting devices use multiple wireless links at once; exactly how you can use them depends on the terminal, the access point, and how the feature is implemented.

What it improves is the *wireless segment.

Say your laptop joins the network over Wi-Fi through an access point (AP), and the NAS is plugged into a switch. The link between the AP and the switch is a single gigabit pipe. Even if the wireless side now has more throughput to offer, the data heading for the NAS still has to flow through that one wired connection.

MLO doesn't make that cable any faster.

This wired segment is what people call the backhaul or uplink. Once the wireless side scales up, the backhaul has to be re-evaluated too — and that's the entry point to understanding why multi-gigabit Ethernet like 2.5G and 5G matters. These standards solve a capacity-matching problem between segments; they aren't just a fancier label slapped on a new Wi-Fi generation.

Also, the combined "multi-band" speed quoted in product marketing is not the file-transfer speed of a single computer. How many spatial streams the terminal supports, how wide a channel it negotiates — all of that affects the result, and so do the radio environment and how traffic is actually scheduled. Before you compare two devices, make sure you know exactly what the number is describing.

2. One file transfer is enough to draw the line on what the upgrade is worth

Picture this: you need to push a 20 GB media file to the NAS. Using decimal units, 20 GB equals 160 Gb. We'll ignore protocol overhead for now, and assume the storage on both ends is fast enough to keep the link fully saturated.

In this idealized model, a 1 Gbps link takes 160 seconds, while a 2.5 Gbps link takes 64 seconds. That's a 96-second difference — a 60% cut in transfer time. This is the theoretical gap that comes purely from the change in link capacity; it is not a benchmark result from any particular router.

In the real world, there's another layer to account for.

If the NAS's sustained write speed for this task is only 180 MB/s, that works out to an effective data rate of about 1.44 Gbps. Even after you upgrade the network link to 2.5 Gbps, this write simply can't jump to 312.5 MB/s. Just from that one constraint, 20 GB divided by 180 MB/s is already around 111 seconds — and it could well take longer in practice.

The bottleneck has moved somewhere else.

That doesn't mean the upgrade is pointless. It means you have to evaluate in two steps: first, decide whether the gigabit link is actually capping this task; then, once you remove that cap, figure out which segment becomes the new ceiling. Don't treat the multiplier on the network as if it were the multiplier on your file transfers.

For creators, archiving large files and syncing lots of small files shouldn't be lumped into the same test. Small-file workloads involve far more metadata and file-system operations. The copy speed of one big file tells you very little about how the other kind of task will feel.

3. Inside the multi-gigabit port, there's a layer of magnetics you shouldn't ignore

Keep looking, this time inside the circuit board. An Ethernet port isn't just a socket — it's a whole subsystem.

The PHY handles physical-layer transmit and receive; the Ethernet magnetics take care of signal coupling and electrical isolation; and the RJ45 provides the physical connector interface. When you move up to multi-gigabit, the PHY, the magnetics, the connector, and the board layout all have to be re-checked against the target speed.

Looking similar on the outside doesn't prove they're equivalent.

If you're studying a single-port multi-gigabit device that runs on its own power (no power delivered over the cable), the VOOHU WHSQ24002G makes a good starting point for evaluating the Ethernet magnetics. It supports 2.5G/5G, and it lets you ground the abstract question of "how is a multi-gigabit port actually built" in a real part number and a real connection diagram.

If your design instead calls for a single-port setup that needs 4PPoE, then the VOOHU WHSQ24015G is worth a closer look. The data-rate requirements can be identical across the two scenarios, but the power-delivery requirements are different. Just because both are called "multi-gigabit ports" doesn't mean you can reuse the same selection criteria.

VOOHU's 2.5G/5G Ethernet magnetics give you distinct options for non-PoE and 4PPoE ports. When you actually get into the design, you still have to line each candidate model up against the PHY's requirements one by one — winding connections, turns ratio, and the loss across the target frequency band, to name a few.

Among those specs, insertion loss measures how much the signal attenuates as it passes through the component, while return loss reflects the reflection caused by impedance discontinuities. When you compare these numbers, you have to align the frequency band and the test conditions. A lone "lower" or "higher" without that context can't replace a real match-up judgment.

That's also the real value of learning hardware: you start understanding how the components relate to one another, instead of just memorizing a few part numbers. The fact that a given Ethernet magnetics chip supports multi-gigabit doesn't mean the whole device has already cleared compatibility, signal-quality, and EMC validation.

4. Test the wired path first, then decide which segment to upgrade

To figure out why a NAS transfer is slow, start with a wired baseline. Connect the computer and the NAS over a known wired path, confirm the negotiated link speed of every segment, and then run a large-file transfer.

Rule out the wireless variables first.

If the wired transfer is already limited by storage or host performance, changing your wireless equipment probably won't fix it. If the wired side clearly delivers higher effective throughput, then connect the same computer over Wi-Fi — keeping the files and the NAS unchanged — and compare the difference.

You can also run a memory-to-memory network throughput test to help separate network limits from disk limits. But CPU, the software stack, and the test configuration all still skew the numbers, so it carries its own caveats.

Pay attention to what each port is actually used for. A device with only one 2.5G port, once that port is assigned to your broadband uplink, may leave the NAS on a gigabit port. On the flip side, traffic between devices on the same LAN usually never touches your ISP link at all — so you can't dismiss the value of a multi-gigabit internal network just because your home broadband is gigabit.

Upgrade follows the data.

When I read a deep dive on a networking product, what I value most is a clear connection diagram, a set of tests run under matched conditions, and an explanation of where the limits actually sit. The device generation tells you what it might be capable of; the actual data path tells you whether any of that capability is actually helping the task in front of you.

Should an Ethernet Transformer Center Tap Connect to Power or Ground? Identify the Side First

geoffery Bob — Wed, 30 Sep 2026 06:53:42 +0000

Whether an Ethernet transformer center tap should connect to power or ground depends on which side of the transformer it is on and the requirements of the PHY reference circuit. On the PHY side, the center tap may connect to a specified supply rail or to ground through a capacitor. On the cable side, the connection also depends on the termination network and Power over Ethernet (PoE) circuitry. The label “center tap” alone does not determine how the pin should be connected.

1. Distinguish the PHY Side from the Cable Side

A center tap is a connection brought out from the midpoint of a winding. The windings on the two sides of an Ethernet transformer have separate electrical connections. Even if both center taps are labeled CT, they are not the same electrical node.

The easiest way to identify the two sides is to trace the differential signal paths in the schematic. The side connected to the PHY transmit and receive pins is the PHY side; the side connected to the RJ45 cable contacts is the cable side. Do not rely solely on whether a winding appears on the left or right of the schematic, as transformer symbols may be mirrored.

The two sides serve different purposes. The PHY-side center-tap connections must be compatible with the PHY’s analog transceiver circuitry. Cable-side center taps may form part of a termination network or, in an applicable PoE implementation, connect to the power sourcing or receiving path. Do not copy a cable-side power connection onto the PHY side.

For a single-port 100 Mbps design that does not require power delivery over the Ethernet cable, the VOOHU WHD12401G can serve as a starting point for evaluation. First establish that PoE is not required, then verify the center-tap connections against the PHY reference schematic. This gives component selection and schematic design a common basis. If PoE is added later, select components to meet the new power requirements rather than simply adding a supply connection to the existing schematic.

2. Connecting to Ground Through a Capacitor Is Not the Same as Connecting Directly to Ground

Why do some schematics connect the center tap to a supply rail while others connect it to a capacitor? The difference comes from the PHY’s output-driver architecture and internal circuitry.

In a typical current-mode transmit-driver implementation, current flows from the specified supply through the center tap into the winding, with the PHY’s driver branches completing the transmit-current path. This type of implementation therefore requires the center tap to connect to the designated supply node with appropriate decoupling. Use the supply voltage specified in the documentation for the particular PHY; do not connect the tap to an arbitrary supply rail on the board.

The reference circuits for some voltage-mode PHYs connect each PHY-side center tap to ground through its own capacitor. The capacitor provides a path for AC components while blocking DC. Removing the capacitor changes the electrical connection.

These connections are not equivalent.

The rule of thumb “current-mode drivers use a supply connection; voltage-mode drivers use a capacitor to ground” can help explain common implementations, but it is not sufficient on its own for PCB design. Individual PHYs have specific requirements for whether center taps must remain separate, how they should be decoupled, and which external networks are needed. Follow the datasheet and reference design for the PHY you are actually using, rather than copying another board simply because it has a working Ethernet connection.

In particular, do not tie multiple center taps together without checking the requirements. Some reference circuits explicitly require each center tap to be decoupled independently. Joining them into a common node and using a single shared capacitor does not preserve the original circuit.

3. Document Every Center-Tap Connection During Schematic Review

Start with the PHY reference schematic and compare it with the Ethernet transformer’s internal circuit diagram. Do this before changing capacitor values or connections.

For each center tap, identify the winding and transformer side it belongs to, and check whether it is internally connected to any other nodes. Then verify its external connection: does it connect to a specified supply, to a specified ground through a capacitor, or to a cable-side network? Check the ground net names carefully as well. Signal ground and chassis ground must not be treated as the same node simply because they use similar ground symbols.

Repeat this review whenever a component is substituted. A similar package only indicates a similar physical form; it does not establish that the winding orientation, center-tap grouping, or pinout is identical. A useful review record includes “component pin — associated winding — external net — documentation page number.” This provides a clear reference for future changes to the PHY or Ethernet transformer.

The same review method applies when selecting VOOHU Ethernet transformers. Shortlist candidates based on the Ethernet port speed and power requirements, then compare the component datasheet with the PHY reference design. Considering component selection and center-tap connections together provides a clear basis for subsequent board revisions. At this stage, center-tap pin numbers, winding connections, and external component values should all be defined in the actual schematic.

4. Frequently Asked Questions

4.1 Can the Center-Tap Connections Be Copied Between Two 100 Mbps PHYs?

Not on the basis of data rate alone. The same speed does not imply the same analog driver architecture or internal termination. Check the new PHY’s reference schematic first, then verify the magnetics pinout and winding connections. If the PHY is replaced while the existing external circuitry is retained, recheck the center-tap supply, decoupling arrangement, and whether any center taps may be tied together.

4.2 Does Adding PoE Mean the PHY-Side Center Taps Must Connect to the PoE Supply Voltage?

No. In PoE implementations that carry power over the data pairs, the power sourcing or receiving connections are on the cable side of the Ethernet transformer. The PHY side must still follow the PHY’s requirements. Clearly distinguish the two sides during schematic review, then verify the magnetics’ PoE ratings. Do not change the PHY-side biasing simply because the port needs to carry power.

4.3 Does a Working Ethernet Link Prove That the Center Taps Are Connected Correctly?

A working link is an initial result, not a substitute for checking the connections. First confirm that the actual circuit meets the PHY’s reference-design requirements, then verify communication performance and electromagnetic compatibility (EMC) according to the system test plan. A successful test using one short cable does not establish that the product will meet its requirements with other cables or under different temperature and interference conditions.

Differences between RJ11, RJ12 and RJ45

geoffery Bob — Thu, 20 Aug 2026 10:13:29 +0000

There are currently several common variants of the Registered Jack (RJ), which are generally distinguished by specific numbers following 'RJ'. These numbers correspond to the wire sequence and the number of wires within the connector. As the number of wires varies, so does their size.

RJ11 Connector(RJ11 jack)

The RJ11 falls into the 4P4C category of modular connectors; there is also a 6P4C specification.Its primary function is to transmit audio and control signals, connecting telephones to the Public Switched Telephone Network (PSTN), and it is also used for Asymmetric Digital Subscriber Line (ADSL) and modem connections
The four wires are arranged in two pairs, allowing connection to two telephone lines: one pair for normal operation and one pair as a backup

RJ12 Connector(RJ12 jack)

The RJ12 is a 6P6C connector, with six pins and six wires; the only difference between it and the RJ11 is that it has two additional wires,the RJ12 is the same size as the 6P4C RJ11.
The RJ12 connector establishes three connections via three pairs of wires. It is primarily used in system telephones and can also be used as a replacement for RJ11 connectors; however, RJ11 connectors cannot be used as a substitute for RJ12 connectors.

RJ45 Connector(RJ45 jack)

The RJ45 is a well-known connector widely used for Ethernet and data connections. As such, they are also known as Ethernet connectors. It is an 8P8C connector. Due to the higher number of pins, the RJ45 is slightly larger than the RJ12. The RJ45 is used to terminate coaxial cables that comply with the Cat6 and Cat5e standards.

SFP crimping or soldering? Pitfalls encountered in several real-world projects

geoffery Bob — Wed, 19 Aug 2026 08:14:33 +0000

Last week, I spoke to a friend who works on switches; for their new project involving 25G ports, they’d opted for solder-type SFP cages. When the boards came back, they couldn’t get them to work—the eye diagrams wouldn’t open.

It took a long time to investigate before they realised the problem was residual solder. During the PCB design phase, they hadn’t allowed for back drilling; once the boards were manufactured and the cages soldered on, the through-holes were filled with solder, and it was too late to drill them—you simply can’t drill after soldering.

In fact, it’s not that soldered cages cannot be back-drilled at all; it’s simply that back-drilling must be carried out during the PCB manufacturing stage. Once soldering is complete, it is indeed impossible to rectify the issue. Furthermore, even if back-drilling is provided for during the PCB stage, the thick leads of soldered cages make it more difficult to achieve proper solder penetration during the soldering process, which leads to a lower yield rate. This is why most 25G+ designs still opt for crimp-type connectors—it is not that soldered connections are physically impossible, but rather that production yield rates are too low and the cost of rework is too high, making it uneconomical overall.

Let’s move on to repairs.

Welded cages are easy to attach, but should a problem arise, the cost of removing the welds is exorbitant—with so many thick pins, forcing them off will almost inevitably damage the PCB through-holes. Ultimately, the entire board must be scrapped.

Crimped cages can be removed and refitted using a specialised crimping tool. As components often need to be replaced during the R&D phase, this advantage is particularly evident in practical applications. When we were building prototypes previously, we had to replace the SFP cages several times due to design revisions; had they been soldered in place, those boards would have been scrapped long ago.

In terms of EMI, the soldered type is indeed slightly better.

The metallurgical bond formed by solder provides lower ground impedance than mechanical contact; in theory, therefore, soldered shielding offers better performance. A well-designed crimped connection can also meet Class B requirements, but in high-reliability applications there is a risk that the ground impedance of the mechanical contact may drift under prolonged vibration and temperature fluctuations, whereas soldered connections do not suffer from this issue.

A few points that are easily overlooked in practice:

It is essential to keep a close eye on the tolerance for crimping hole diameters; many PCB manufacturers may not be able to consistently meet the ±0.05 mm requirement. Remember to confirm this clearly with the manufacturer before dispatching the boards.

Crimping cages must be handled using hydraulic or pneumatic fixtures; under no circumstances should they be tapped by hand. Manual tapping can lead to inconsistent deformation of the ‘fish-eye’ pins, resulting in loose connections on some pins. This may subsequently cause occasional packet loss or intermittent signal issues, which are extremely difficult to troubleshoot.

Soldered cages have relatively thick pins; the stencil design for through-hole reflow soldering must be determined based on the pin diameter, pad size and stencil thickness—a one-size-fits-all approach is not acceptable. It is recommended to refer to the stencil specifications recommended in the component datasheet.

Ground vias should be densely spaced, with a pitch of no more than 5 mm, and should connect directly to the ground plane; avoid using fine traces.

As for how to choose, my personal experience is as follows:

For speeds of 25G and above, crimp-type connectors are the safer option. For 10G and below, either type is acceptable, depending on the production line’s capabilities. If components are frequently swapped during the R&D phase, crimp-type connectors are less hassle. If the production line only has wave soldering and no crimping equipment, soldered connectors are the more practical choice.

Wohu offers both crimped and soldered SFP cages, covering SFP, SFP+ and SFP28. Please check the official website for specific models.

SFP Connector Cage Assembly: Press-Fit vs Solder - A Comprehensive Selection Guide for High-Speed Designs

geoffery Bob — Tue, 18 Aug 2026 07:32:56 +0000

SFP Connector Cage Assembly: Press-Fit vs Solder - A Comprehensive Selection Guide for High-Speed Designs
SFP (Small Form-factor Pluggable) cages are the core mechanical and electrical interface components used to mount and secure optical modules in communication equipment. They consist of a metal shielding shell and a connector core, and can be integrated with accessories such as heat sinks or light pipes.

They come with either press-fit or solder tails, and choosing between the two is one of the first decisions engineers face when selecting an SFP connector. The choice affects not only assembly cost but also signal integrity at high speeds, production yield, and field reliability.

How Each Process Works

**Press-Fit **relies on the compliant pin - often called a "fish-eye" pin - that elastically deforms when pressed into a PCB plated through-hole. The interference fit creates both mechanical retention and electrical contact. No solder is involved. The critical parameter here is hole diameter tolerance, which needs to stay within ±0.05mm. Too loose, and the pin won't make reliable contact; too tight, and you risk cracking the via barrel or buckling the pin.

Solder, by contrast, is exactly what it sounds like: the cage pins are inserted into plated through-holes and soldered - typically by wave soldering, though some variants are surface-mount and reflow-soldered. The process is straightforward and compatible with standard PCB assembly lines. The main practical concern is getting enough solder fill into those large through-holes to achieve solid mechanical strength.

What Really Matters in Practice

Signal integrity is where the two diverge most significantly, especially above 10G. With soldered cages, once the pin is soldered, the via is filled with solder - and that means you can't back-drill it. The leftover stub creates reflections that become increasingly problematic as data rates climb. At 25G NRZ or 50G PAM4, an un-back-drilled stub can ruin an eye diagram.

Press-Fit leaves the via empty, so you can back-drill it after assembly. If you're working on a 25G or higher design, this alone often makes the decision: Press-Fit is the only practical choice.
Reliability and rework also point in opposite directions depending on your perspective. Soldered joints are metallurgical bonds - strong and stable, but nearly impossible to rework if something goes wrong. A damaged cage often means scrapping the entire board. Press-Fit, on the other hand, can be pressed out and re-installed. For R&D prototypes or small-batch production, that reworkability is a genuine advantage.

EMI performance tends to favor soldering, at least on paper. Solder provides a lower DC resistance path to ground compared to the mechanical contact of press-fit pins. That said, a well-designed press-fit cage with properly plated through-holes and good grounding spring fingers can still meet Class B EMI requirements. The difference matters most in designs with very tight EMC margins.

Making the Call

There's no universal right answer. Here's how I typically break it down:
Pick Press-Fit when:
Your design runs at 25G or faster (back-drill is non-negotiable)
You're doing high-volume production where per-unit cost matters (press-fit is faster once tooling is in place)
You value reworkability and want to avoid scrapping boards from connector damage

Pick Solder when:
Your data rate is 10G or below
Your production line already has wave/reflow capability but no press-fit equipment
You're in a cost-sensitive, medium-volume project where the simplicity of soldering outweighs other considerations

VOOHU offers both mounting options across its SFP, SFP+, and SFP28 connector series, with 1×N and 2×N configurations. Visit www.voohuele.com for product specifications.
FAQ
Q1: What's the essential difference between Press-Fit and Solder?
Press-Fit uses compliant pins to form a mechanical interference fit with PCB through-holes. No soldering, no thermal stress. Solder relies on wave or reflow soldering to fix the cage. Press-Fit requires tighter hole tolerance (±0.05mm) but offers reworkability; Solder is cheaper up front but nearly impossible to rework.
Q2: Why is Press-Fit almost mandatory for 25G+ designs?
Because of back-drill. Soldered vias are filled with solder, making back-drilling impossible. The remaining stub creates reflections that kill signal integrity at 25G NRZ or 50G PAM4. Press-Fit vias stay empty - you can back-drill after assembly.
Q3: What's the PCB hole tolerance requirement for Press-Fit?
VOOHU specifies ±0.05mm. Too loose, the pin won't make reliable contact. Too tight, you risk cracking the via barrel or damaging the pin.

The RJ45 Selection Decision-Making Process

geoffery Bob — Thu, 13 Aug 2026 09:57:04 +0000

Now that we’ve looked at the design of the RJ45, let’s discuss the steps involved in selecting the right model.

Step 1: Determine the speed rating
The speed category determines the transmission bandwidth that an RJ45 connector must support, and directly affects the connector’s insertion loss and return loss specifications. The following selection recommendations apply:

10/100M → Select an RJ45 connector that supports a bandwidth of 100 MHz
1G (Gigabit) → Select an RJ45 connector that supports a bandwidth of 100 MHz and meets the return loss requirements of IEEE 802.3ab
2.5G/5G → Select an RJ45 connector that supports a bandwidth of 200 MHz/400 MHz
10G → Select an RJ45 connector that supports a bandwidth of 500 MHz

Step 2: Determine the PHY driver type

Refer to the PHY data sheet to confirm whether it is voltage-mode or current-mode.
Plan the centre-tap connection according to the drive type:
Current-mode → Connect the centre tap to VDD (3.3V/2.5V/1.8V).
Voltage-mode → Connect the centre tap to earth via a 100 nF capacitor.
If the PHY data sheet does not explicitly specify the drive type, you can make an informed judgement by measuring the DC voltage between the TXD± pins and ground: for voltage-mode, this is approximately VDD/2; for current-mode, it is close to 0 V.

Step 3: Determine PoE power requirements

No PoE required → Select a standard model (non-PoE).
PoE (15.4W) → Select a model supporting 350mA (compliant with IEEE 802.3af).
PoE+ (30W) → Select a model supporting 720mA (compliant with IEEE 802.3at; WoHu models with the suffix ‘DP’).
PoE++ (60W/90W) → Select a model supporting 1.2A/2A (compliant with IEEE 802.3bt; WoHu models with the suffix ‘PG’ or ‘PTG’).

Step 4: Determine the mounting method

DIP (Through-hole) → High mechanical strength; suitable for industrial equipment and vibration-prone environments; requires space on both sides of the PCB.
SMT (Surface Mount Technology) → Suitable for automated reflow soldering production; lower profile, suitable for high-density equipment; mechanical strength is lower than that of DIP.
Low-Profile → Part of the housing is recessed into a slot in the PCB; height can be as low as 8 mm; suitable for 1U/half-U slim-line equipment.
Wire-Bonded (Cable Type) → Wires are soldered directly; suitable for non-board-mounted applications.

Step 5: Determine the environmental rating

Commercial grade (0–70 °C): Offices, homes, controlled environments.
Industrial grade (-40–85 °C): Factories, outdoor environments, in-vehicle applications, rail transport.
Automotive grade (-40–105 °C): Extreme temperature environments such as engine compartments.

Step 6: Identify any special requirements

Waterproof and dustproof: Is an IP67/IP68 rating required (e.g. for outdoor base stations or marine equipment)?
Vibration-resistant locking: Is a metal spring, Cannon self-locking mechanism or threaded locking required (e.g. for rail transport or vehicle-mounted applications)?
LED indicators: Are specific colours (green/orange/yellow), brightness levels or common-cathode/common-anode configurations required?
Shielding requirements: Is a fully shielded enclosure required (for EMI-sensitive environments)?

How does a DIY alarm panel achieve single-wire PoE power supply and data transmission?

geoffery Bob — Tue, 11 Aug 2026 08:08:44 +0000

Today, a novice engineer approached me with a question: he wants to develop an alarm panel and is looking to use PoE to enable a single Ethernet cable to provide both power and data transmission. He hasn’t been able to find a PoE module and would like to know if this is feasible.

His key requirement is for the alarm panel to operate using just a single Ethernet cable, which is, of course, entirely possible; this is a standard application of PoE technology.

Returning to his question, this solution can be integrated onto the PCB using a PoE module, though I would recommend using an integrated RJ45 connector (a version with PoE functionality) in conjunction with an external PD chip, as this reduces PCB layout work and, once soldered, allows power and signals to be routed to the main control chip via the connector's pins.

The PoE Power Supply Process

The PSE first transmits direct current via the network cable.
The PD device (alarm panel) feeds the power into the PoE power transformer at the rear via an integrated RJ45 connector.
The PoE power transformer, in conjunction with the controller, converts the 48V high voltage into the low voltage (5V/12V) required by the alarm panel, thereby supplying power to the main controller and sensors.

Given that the alarm panel does not have high speed requirements, I have compiled a list of a few components as examples for your reference, based on the requirements.

100 Mbps integrated RJ45: SYT111B032DB1A1DPQ

Depending on the device's voltage and power requirements, two PoE power transformers are recommended: (1) 5V/12W: ST-0880WHB (2) 12V/13W: WH13P-12L If your aim is to learn how to design a PoE circuit, VOOHU's reference design is an excellent starting point. If your aim is to quickly build a functional alarm panel, a PoE-enabled development board or a ready-made PoE splitter remains the more straightforward option.

Typical RJ45 Circuit Designs – PHY Driver Types and Centre-Tap Connections

geoffery Bob — Mon, 10 Aug 2026 10:10:11 +0000

Ethernet PHY chips are classified into voltage-mode and current-mode types based on the structure of their output stages. The two types have entirely different requirements regarding the connection of the network transformer’s centre tap; incorrect connection can result in reduced signal amplitude, waveform distortion or even failure to establish a link.

Voltage-Mode PHY

The output stage of a voltage-driven PHY employs a voltage-source configuration and incorporates an internal impedance-matching resistor (typically 50 Ω). Its output impedance is matched to the characteristic impedance of the transmission line (100 Ω differential), and the signal amplitude is determined by the internal voltage source.

Centre-tapped connection:
Must be earthed via a capacitor (typically 100 nF) to isolate the DC component.
**Principle: **The output of a voltage-driven PHY already has a defined DC offset point; the centre-tapped ground capacitor provides an AC coupling path, ensuring that the signal across the primary winding of the transformer is symmetrical with respect to ground.
**Typical chips: **Realtek RTL8211F, Intel LXT972, Jinglue JL2101C.

(2) Current-Mode PHY

The output stage of a current-mode PHY employs a current-source configuration, requiring an external resistor to convert the current into a voltage. The centre tap of the transformer provides a DC bias path, enabling the differential current to flow through the primary winding of the transformer to form a complete AC loop.

Centre-tapped connection: **This must be connected directly to the PHY’s supply voltage (VDD, typically 3.3 V, 2.5 V or 1.8 V), and a capacitor to ground (0.01 μF to 0.1 μF) must be used for high-frequency decoupling.
**Principle: **Connecting the centre tap to VDD provides a DC path for the differential current; the resistor (typically two 50Ω resistors in series or integrated within the chip) and the primary winding of the transformer together form a current-to-voltage conversion network.
**Typical chips: TI DP83848, Microchip LAN8720, Jinglue JL2201B/JL1111B.

The Function and Design of the Bob-Smith Circuit

The Bob-Smith circuit is an impedance-matching and common-mode rejection network widely used in the Ethernet physical layer. It consists of resistors and capacitors connected across a pair of differential lines near the transformer (each pair of lines is configured independently).

(1) Circuit Configuration

A standard Bob-Smith circuit consists of two resistors (typically 75 Ω) and a high-voltage capacitor (1000 pF/2 kV) connected in series between them, with its terminals connected to the differential lines (TX+ and TX-) respectively; common-mode surge energy is dissipated via the capacitor to the protective earth (PGND).

(2) Main Functions

Impedance matching: Provides a defined AC termination impedance for differential signals, reducing signal reflection.
Common-mode noise suppression: Provides a discharge path to ground for common-mode currents, reducing radiated emissions from the cable.
Surge energy absorption: External surge energy can be coupled via the capacitor to protective earth for dissipation.

(3) Design Considerations

The resistance value is typically 75 Ω (for a differential impedance of 100 Ω, 75 Ω single-ended is a common choice for impedance matching).
A resistance tolerance of 1 per cent is recommended to ensure impedance matching accuracy.
The capacitance withstand voltage must be ≥2 kV to ensure compliance with the IEC 61000-4-5 surge test requirements (2 kV line-to-ground).

The RJ45 production process

geoffery Bob — Thu, 06 Aug 2026 10:12:44 +0000

The production process described here reflects only the procedures followed at our factory; it does not represent the manufacturing processes and procedures of other factories on the market. We simply wish to provide some insight into just how complex the production process behind a simple network port actually is.

(1) Automatic terminal insertion (see the diagram below for details)

Terminal insertion is the first step in the entire production process. It mainly involves arranging the eight gold-plated copper wires neatly and inserting them into the plastic body of the RJ45 connector, whilst ensuring the correct insertion angle.

(2) Terminal bending/shaping (see figure below)

Bend the inserted gold pins to the angle specified in the drawing (90° or 180°); this step creates the contact surface for the crystal connector.

(3) LED assembly (see figure below)

Insert the LEDs into the corresponding grooves in the plastic housing, sorting them by colour and polarity.

(4) Internal Pin Inspection (see figure below)

This step primarily involves checking whether the gold pins installed inside the housing have any faults, focusing on four key issues:

**1. Missing pins

Bent pins
Gold pins slipping out of the guide slots
Scratched gold pins**

(5) LED Moulding/Hot-Melting (see figure below)

Secure the LED leads and trim them to the same height as the leads

(6) External Pin Inspection (see figure below)

Check the external pins for defects, focusing on three main categories:

**1. Missing or bent pins

Deformation or scratches
For SMT-type RJ45s, check for coplanarity deviation **

(7) LED Testing (see figure below)

This step verifies whether the LED is functioning correctly, focusing on four key aspects:

On/off status
Forward voltage drop
Correct colour display and uniform brightness
Polarity matches the schematic

(8) Continuity and High-Voltage Testing (see figure below)

**Continuity test: **Primarily to confirm that each gold pin is connected to its corresponding lead
**High-voltage test: **Primarily to verify that the insulation withstand voltage meets the standard

Now that we have looked at the manufacturing process, let's discuss a few key quality factors that determine the quality of an RJ45 connector.

Firstly: the dimensional accuracy of the plastic body

The precision of the body's opening - that is, the socket - directly affects the tightness and stability of the connector's insertion. Generally speaking, manufacturers strictly adhere to the FCC Part 68 specifications. Our (VOOHU) solution is to use high-precision moulds and automated injection moulding machines to reduce tolerances to within 0.25 mm. Typically, issues such as loose connections, difficulty inserting or removing the connector, and poor contact are all caused by excessive dimensional tolerances at the interface.

Second: Co-planarity of gold pins

The solder pins on SMT-style RJ45 connectors must be perfectly horizontal; otherwise, this will result in poor soldering.

Third: Consistency of LED parameters

Whether the LED's brightness, forward voltage drop and polarity conform to the specification sheet is crucial to whether the panel's LED indicator lights up and whether the brightness is consistent.

Generally speaking, connector products undergo 100 per cent inspection prior to dispatch. The main test items for this comprehensive inspection are:

**1. Continuity/short-circuit testing

Contact resistance testing
Insulation resistance testing
High-voltage withstand testing
LED electrical parameter testing That concludes ** what I wanted to share in this instalment.

I'll leave it at that for now. If there are any errors or omissions in my explanation, please feel free to point them out; I shall be grateful for your feedback.

Are you designing an Ethernet solution? VOOHU utilises its own-brand industrial connectors, complementary magnetic components and chips to provide eight solutions for key industrial applications. Our services cover every stage from component selection and customised design through to mass production; every order undergoes an engineering review to verify the feasibility of the components and solutions.

Common electrical parameters of RJ45 connectors

geoffery Bob — Wed, 05 Aug 2026 09:12:17 +0000

The most common specification for an RJ45 connector is its data rate. Whilst the data rate is influenced by the product’s materials and structure, this applies only to integrated RJ45 connectors; in theory, discrete RJ45 connectors can support any data rate. The actual limitations on data rate are imposed by the network transformer and the PHY or switch downstream of the RJ45 connector. The figure below shows the standards and applicable scenarios for each data rate, which may serve as a reference.

Electrical performance refers to the reliable parameter standards that ensure the RJ45 meets transmission speed requirements. Let’s take a look at what these are:

1. Contact resistance

Put simply, this is a parameter that measures the quality of contact between the RJ45 connector and the contact terminals. Generally speaking, in industrial applications, the contact resistance must remain at ≤20 mΩ after 1,000 insertion and removal cycles.

2. Insulation resistance

This parameter primarily ensures the insulating properties of the connector’s insulating material

3. Dielectric Strength

Also known as a Hi-Pot test, this test is designed to ensure that the connector remains insulated at the specified voltage without breakdown.

Rated Current and Rated Voltage This is fairly straightforward to understand; it refers to the parameters for the maximum permissible current and voltage.

5. Operating Temperature

These are generally categorised as commercial grade, industrial grade, automotive grade, military grade and aerospace grade.
Commercial grade: 0°C to +70°C
Industrial grade: -40°C to +85°C
Automotive grade: -40°C to +125°C
Military grade: -55°C to +150°C
Aerospace grade: -55°C to +150°C

The Relationship Between Plating and Reliability

1. Gold Plating

Gold plating refers to the layer of gold applied to contact components. Why is gold plating used?
Gold plating is primarily used to ensure signal stability; the thickness of the gold plating determines the service life of the gold-plated pins during insertion and removal, as well as their environmental resistance. I have compiled typical parameters for different gold plating thicknesses, as shown in the figure below

2. Nickel Plating

Nickel plating usually refers to the nickel layer on the exterior of the shielding casing and on the pins. Nickel plating effectively resists environmental salt spray corrosion; when the nickel layer exceeds 3 μm, it effectively prevents ion diffusion between the copper and gold.

That concludes what I wanted to share in this instalment. I’ll leave it at that for now. If there are any errors or omissions in my explanation, please feel free to point them out; I shall be grateful for your feedback.
For more detailed information, please see this article I wrote previously: https://www.voohuele.com/news/2025-06-03-1/

Types of RJ45 Connectors

geoffery Bob — Tue, 04 Aug 2026 10:32:12 +0000

Types of RJ45 connectors
In the previous few articles, we provided a brief introduction to the history, standards and construction of the RJ45. In this instalment, let's take a look at the different types of RJ45 connectors.

1. Types of RJ45 Connectors

Over more than half a century of development, RJ45 connectors have evolved into a wide variety of types to suit different application scenarios, making classification rather difficult. We can categorise them based on various parameters, and the process of identifying the specific product we require by considering multiple parameters is known as 'product selection'. Below, I shall outline VOOHU's classification of RJ45 connectors as a reference for understanding this category:

1.1 Whether magnetic components are integrated

This is the most common classification method, comprising discrete and integrated types
** (1) Discrete RJ45 (Discrete Jack): **Consists solely of the RJ45 housing; the network transformer must be designed, laid out and soldered separately onto the PCB

**(2) Integrated RJ45 (Mag Jack): **This involves soldering the common-mode choke and transformer together onto a single PCB, which is then encapsulated with the RJ45 to form a single module; hence the name 'Mag Jack'.

1.2 Mounting Methods

** (1) DIP: **DIP-type RJ45 connectors have round pins that can be inserted into pre-drilled holes in the PCB and then soldered using wave soldering, resulting in high mechanical strength

(2) SMT: SMT-type RJ45 connectors are mounted directly onto the PCB; they typically feature flat pins that lie parallel to the PCB surface and are soldered using reflow soldering. As they are soldered to the surface of the PCB, their mechanical strength is slightly lower than that of DIP connectors

**(3) Low-Profile: **The port on this type of RJ45 is usually recessed below the pins; installation requires cutting a corresponding rectangular slot in the PCB into which it is embedded.

(4) Cable-Type: Cable-type RJ45 connectors have no pins; wires must be soldered directly to them or crimped onto the cable.

1.3 Port Angle

The port angle refers to the angle between the port and the PCB, and is primarily categorised as 90°, 180° or 45°.
(1) 90° (Right-Angle): See the figure below.

(2) 180° (Vertical): See the figure below.

**(3) 45° (Angled): **See the figure below.

1.4 Number of Ports
(1) Single-port: See the figure below.

**(2) Multi-port: **See the figure below

**(3) Stacked: **See the figure below

There are many other classification methods, such as the presence or absence of a shielding housing or LED indicators, which we will not list here. If you are interested in learning more, please visit this website (WWW.VOOHUELE.COM);

experiencing the selection process first-hand will help you better understand the information above Well, the content in this instalment is relatively basic. In the next instalment, we'll take a look at the electrical characteristics and manufacturing processes of VOOHU's RJ45 connectors.