Network ICs are easy to overlook. In a finished product, they may be just one small chip among hundreds of components. But when the network connection is unstable, too slow, or simply does not work, that small IC can become one of the most important components on the board.
For engineers, choosing a network IC is not simply a matter of finding a chip with the right data rate. You also need to consider the network standard, MAC and PHY architecture, host interface, power supply, clock, PCB layout, temperature range, software support, and sometimes automotive or industrial qualification.
This guide explains what network ICs are, how they work, the main types you will encounter, and how to select one for a real design.
What Is a Network IC?
A network IC (integrated circuit) is a semiconductor device designed to provide one or more functions required for communication between electronic devices.
The term is broad. Depending on the application, a network IC can be an Ethernet PHY, Ethernet controller, Ethernet switch, network interface controller, Wi-Fi IC, Bluetooth IC, cellular modem, or a more integrated networking SoC.
For Ethernet, a useful way to understand the architecture is:
CPU/MCU → MAC → PHY → Cable/Network → PHY → MAC → CPU/MCU
The MAC handles Ethernet frame and data-link functions, while the PHY handles the physical connection between the MAC and the network medium. TI describes a typical Ethernet interface as consisting of the CPU, MAC, and PHY, with DMA commonly used to move data efficiently between memory and the MAC.
This distinction is important because many engineers initially assume that an Ethernet PHY is the same thing as an Ethernet controller. It is not.
MAC vs. PHY: What Is the Difference?
The easiest way to remember it is:
MAC = handles Ethernet data.
PHY = handles Ethernet signals.
The MAC creates, processes, filters, and receives Ethernet frames. The PHY converts the digital data from the MAC into electrical signals suitable for the physical medium and converts received signals back into digital data.
A typical PHY has two sides:
- A digital interface toward the MAC
- An analog/physical interface toward the network cable or other medium
Common MAC-to-PHY interfaces include MII, RMII, GMII, RGMII, and SGMII. Their pin counts, supported speeds, timing requirements, and routing requirements are different. For example, RMII reduces the number of signals compared with MII, while RGMII supports Gigabit Ethernet with fewer pins than GMII.
This is one of the first things I check when reviewing a network IC for a new design: Does the IC actually match the interface provided by the processor?
Main Types of Network ICs
There is no single device called a "network IC." The category contains several different types of components.
1. Ethernet PHY ICs
An Ethernet PHY is probably the most common network IC you will encounter in embedded hardware.
It sits between the Ethernet MAC and the physical network connection. Depending on the device, it may support 10 Mbps, 100 Mbps, 1 Gbps, or higher data rates.
For example, the TI DP83826ERHBT is a 10/100-Mbps Ethernet PHY, while the DP83869HMRGZT supports Gigabit Ethernet and copper/fiber applications. TI also offers automotive and Single-Pair Ethernet PHYs such as the DP83TG720SWRHATQ1 and DP83TG720SWRHARQ1.
When selecting a PHY, don't look only at speed. Check:
- MAC interface
- Network standard
- Cable type
- Maximum cable length
- Clock requirements
- Supply voltage
- Temperature range
- ESD/EMC performance
- Diagnostics
- Package and pinout
Modern PHYs can also include useful diagnostic and timing functions. TI, for example, highlights cable diagnostics, low latency, deterministic operation, and support for industrial and automotive environments in its Ethernet PHY portfolio.
2. Ethernet Controllers
An Ethernet controller provides the host processor with an interface to Ethernet. Depending on the device, the controller may include the MAC, PHY, or both.
This can be particularly useful when the host processor does not have a suitable integrated Ethernet peripheral.
Microchip, for example, offers Ethernet controllers with integrated MAC and PHY functions as well as devices supporting external interfaces such as MII and PCI.
A good example of a more specialized architecture is Microchip's ENC28J60-I/SS. It combines a 10BASE-T1S MAC and PHY and connects to a host MCU through SPI. This allows an MCU without an integrated Ethernet MAC to participate in an Ethernet network.
3. Ethernet Switch ICs
If a system needs to connect several Ethernet devices, an Ethernet switch IC may be more appropriate than a single PHY.
A switch receives Ethernet frames on one port and forwards them to the appropriate port based on network information such as MAC addresses.
Switch ICs are widely used in:
- Industrial equipment
- Network switches
- Automotive gateways
- Factory automation
- Embedded networking equipment
- Communication infrastructure
Depending on the device, features can include VLAN, QoS, traffic management, diagnostics, and time-sensitive networking.
4. Network Interface Controller (NIC) ICs
A NIC provides network connectivity between a host system and the network.
In a desktop PC or server, a NIC may connect through PCI Express. In an embedded system, the architecture can be much simpler and may use SPI, a parallel interface, or a processor's integrated Ethernet peripheral.
The important point is that a NIC is normally concerned with the host-to-network connection, rather than simply being a physical-layer transceiver.
5. Wi-Fi, Bluetooth, and Cellular ICs
Network ICs are not limited to wired Ethernet.
Wireless networking products can include:
- Wi-Fi ICs
- Bluetooth/BLE ICs
- Wi-Fi + Bluetooth combo ICs
- LTE modem ICs
- 5G modem ICs
- Wireless communication SoCs
These devices can integrate much more functionality than a simple Ethernet PHY, including baseband processing, RF functions, security, memory interfaces, and processor cores.
For this reason, when someone says "network IC," it is always worth asking: What type of network are we talking about?
Single-Pair Ethernet Is Changing Network IC Design
One area that deserves special attention is Single-Pair Ethernet (SPE).
Traditional Ethernet often uses multiple twisted pairs. Single-Pair Ethernet reduces the physical connection to one balanced pair, making it attractive for industrial and automotive applications where cable weight, size, and installation complexity matter.
Current Ethernet PHY portfolios include standards such as:
- 10BASE-T1L
- 10BASE-T1S
- 100BASE-T1
- 1000BASE-T1
TI, for example, lists 10BASE-T1L, 100BASE-T1, and 1000BASE-T1 PHY solutions for industrial and automotive applications.
Microchip's LAN8670 family(like LAN8670B1T-E/LMX) provides 10BASE-T1S PHY functionality, while the LAN8650/LAN8651 combine the MAC and PHY and use SPI to connect to an MCU.
This architecture can be particularly useful for small sensors and actuators that need Ethernet connectivity but use a low-cost MCU without an integrated Ethernet MAC.
Important Network IC Specifications
When comparing network ICs, I recommend creating a simple specification table before looking at prices.
Network speed
Check the actual required speed:
- 10 Mbps
- 100 Mbps
- 1 Gbps
- 2.5 Gbps
- 5 Gbps
- 10 Gbps or higher
Don't automatically select the fastest device. A higher-speed IC may increase power consumption, PCB complexity, and system cost without providing a real benefit.
Network standard
Make sure the IC supports the exact standard required by your system.
For example, 100BASE-TX and 100BASE-T1 are both 100-Mbps Ethernet technologies, but they are designed for different physical media and applications.
Host interface
This is often a deal-breaker.
Check whether the IC supports the interface available from your MCU, MPU, FPGA, or processor:
- MII
- RMII
- GMII
- RGMII
- SGMII
- SPI
- PCIe
- USB
Never assume that two Ethernet ICs with the same network speed are interchangeable.
Supply voltage
Look carefully at the:
- Core voltage
- I/O voltage
- Analog supply
- Digital supply
- Internal regulators
A device that appears compatible at the protocol level may still require different power rails.
Temperature range
For office equipment, commercial temperature may be enough. Industrial and automotive products often require a much wider operating range.
For example, Microchip's LAN8650 is specified from -40°C to +125°C and is AEC-Q100 qualified.
Diagnostics
Diagnostics become very valuable when equipment is installed in the field.
Depending on the device, features may include:
- Cable diagnostics
- Link status
- Signal-quality indication
- Loopback
- Fault detection
- Error counters
These features can save significant troubleshooting time.
How to Choose the Right Network IC
I normally approach network IC selection from the system side rather than starting with a distributor search.
Step 1: Define the network
First determine whether you need Ethernet, Wi-Fi, Bluetooth, cellular, or another technology.
For Ethernet, identify the exact standard and physical medium.
Step 2: Define the required speed
Don't confuse link speed with actual application throughput.
A 1-Gbps link does not necessarily mean your application will transfer data at 1 Gbps. CPU performance, memory bandwidth, protocol overhead, DMA configuration, and software can all become bottlenecks.
Step 3: Check the processor interface
Look at the processor datasheet first.
If your MCU provides RMII, for example, selecting a PHY that only supports a different interface may force a redesign.
Step 4: Check electrical requirements
Compare:
- Supply rails
- I/O voltage
- Clock source
- Reset requirements
- Strapping pins
- Power consumption
Step 5: Check PCB requirements
High-speed network interfaces are not forgiving of poor PCB design.
Pay attention to:
- Controlled impedance
- Differential routing
- Signal length
- Return paths
- Grounding
- Power decoupling
- Clock routing
- Magnetics
- EMI/EMC
For Gigabit Ethernet, these details can make the difference between a stable link and a board that works only intermittently.
Step 6: Check software support
Hardware compatibility is only half of the job.
Before committing to a device, check:
- Linux driver support
- MCU SDK support
- Register documentation
- Device-tree support
- Firmware examples
- Configuration tools
- Reference designs
A technically excellent IC can still be a poor choice if your team cannot easily support it.
Common Network IC Problems
When an Ethernet design does not work, engineers often start by replacing the PHY. That is not always the right first step.
No Ethernet link
Check the basics first:
- Power rails
- Reference clock
- Reset timing
- PHY strap pins
- MDIO/MDC communication
- MAC-to-PHY interface
- Magnetics
- Cable
- PCB routing
A PHY that does not initialize may have a power or clock problem rather than a damaged IC.
Intermittent connection
If the link repeatedly goes up and down, investigate:
- Power noise
- EMI
- Poor signal integrity
- Cable quality
- Thermal issues
- Incorrect PHY configuration
Low network throughput
Don't immediately blame the PHY.
Check the complete data path:
Network → PHY → MAC → DMA → Memory → CPU → Software
The bottleneck could be the host processor, DMA configuration, memory bandwidth, driver, or application software.
Network ICs in Industrial and Automotive Designs
The requirements become much stricter outside consumer electronics.
Industrial systems may need low latency, long operating life, strong EMC performance, wide temperature operation, and deterministic communication.
Automotive systems add requirements such as AEC-Q100 qualification, functional safety considerations, and specialized Ethernet standards.
For example, TI's DP83TC815-Q1 and DP83TC816-Q1 are automotive 100BASE-T1 PHYs with interfaces and timing features aimed at automotive networking.
Microchip's LAN8650 is another example of how networking is moving closer to sensors and edge nodes: its integrated MAC-PHY architecture allows low-cost MCUs to connect to 10BASE-T1S networks through SPI.
How to Select a Replacement Network IC
Finding an "equivalent" network IC requires more than matching the part number family.
At minimum, compare:
- Network standard
- Data rate
- MAC/PHY function
- Host interface
- Pinout
- Package
- Supply voltage
- Clock
- Temperature range
- Register compatibility
- Software/driver support
A device with the same package and similar speed is not automatically a drop-in replacement.
For production equipment, I would also check lifecycle status, PCN history, EOL risk, manufacturer support, and the availability of qualified alternative sources.
Final Thoughts
Network ICs are much more than simple communication chips. They sit at the boundary between digital processing, high-speed signaling, software, and the physical network.
The most important lesson is to select the complete networking architecture, not just the IC.
Start with the network standard and required data rate. Then work backward through the physical medium, PHY, MAC, host interface, processor, software, power system, and PCB.
For a simple embedded Ethernet design, a PHY such as the DP83826E may be all you need when the MCU already contains an Ethernet MAC. For a Gigabit application, a device such as the DP83869HM may be more appropriate. For Single-Pair Ethernet applications, devices such as LAN8650/LAN8651, LAN8670, or automotive PHYs such as DP83TC815-Q1 can provide a very different architecture.
The right network IC is ultimately the one that fits the entire system: electrically, logically, mechanically, thermally, and from a software and supply-chain perspective. That is the approach that usually saves the most engineering time later.
