SATA Device IP on Artix-7: Make Your FPGA Look Like an SSD

SATA Device IP on Artix-7: Make Your FPGA Look Like an SSD

Industry update — At embedded world Germany 2026, Greenliant showcased its high-endurance industrial SSD portfolio, including SATA-interface EnduroSLC EX drives built for heavy-write, long-lifecycle systems, a reminder that SATA remains a mainstay of embedded storage design (Embedded Computing Design, March 2026).

Most storage-on-FPGA articles assume your FPGA is the computer driving a drive. This one is about the opposite job: making an Artix-7 FPGA look like a SATA SSD to some other computer. That is the province of a SATA Device IP core — the target side of the link — and it turns a low-cost 7-series part into a drive emulator, a transparent capture point, or a bridge that speaks SATA to legacy hardware. Below, the device role is drawn cleanly against the host role, then walked through the SATA layers involved and the use cases that make the target side worth building.

The short answer

A SATA Device IP core makes an Artix-7 FPGA present itself as a SATA SSD to a host computer: the host enumerates it as an ordinary drive, while your FPGA logic decides what every read and write does.The core drives the device end of the link through the on-chip GTP transceivers, entirely in fabric — no soft CPU, no operating system, deterministic response timing. It is the tool for drive emulation, storage interposition and SATA bridging, as distinct from a SATA Host core, which does the reverse and drives a real SSD.

Device vs host: which side are you building?

Every SATA link has exactly one host (the initiator) and one device (the target). The host issues commands; the device answers them. Both ends can live in an FPGA, and the two IP cores mirror those two roles. Getting this decision right is the single most important step, because the cores are not interchangeable and a design almost always needs one specific side.

SATA Host IP vs SATA Device IP
Aspect SATA Host IP SATA Device IP
Your FPGA acts as The computer / initiator The drive / target
It connects to An SSD or HDD An external host computer
Who issues commands Your FPGA The external host
Mental model “Store data on a drive” “Pretend to be a drive”
Also called SATA initiator core SATA target core
Typical use Recorder, logger, DAQ storage Emulation, capture, bridging, secure front-end

If you only remember one line: a host core lets you store data on a drive; a device core lets you become a drive. This article is about the second job. If you actually need the first, the host side is walked through in the companion piece on implementing SATA Host IP on Kintex UltraScale+, and the full role-by-role comparison lives in the SATA Host and Device selection guide.

How does an FPGA “become” a SATA SSD?

From the host’s point of view, a SATA drive is defined entirely by its behaviour on the wire — not by the flash chips inside it. If something answers the host’s low-level signalling, identifies itself as a drive, and services read and write commands correctly, the host cannot tell whether real NAND sits behind the connector or a state machine in an FPGA. A SATA Device core — a soft IP core loaded into the fabric — is exactly that state machine.

When a host is connected, a predictable sequence plays out, and the device core has to hold up its end of each step:

  1. Out-of-band signalling and reset. Before any high-speed data flows, host and device exchange out-of-band bursts (the COMRESET / COMINIT / COMWAKE handshake) to establish that both ends are alive and to agree the link is coming up.
  2. Speed negotiation. The two ends settle on a common generation — SATA I, II or III — based on what each side supports. The device advertises the rates it can run.
  3. Identification. The host issues an IDENTIFY DEVICE request. The device core answers with the parameter block that declares capacity, supported features and the reported model of the “drive.” This is the moment the FPGA declares itself to be an SSD.
  4. Command service. From then on the host sends read, write and control commands; the device core receives them, moves the data, and returns status. What the data is — real storage, a synthetic pattern, a live feed, another interface — is entirely up to your fabric logic.

The important idea is that “looking like an SSD” is a protocol behaviour, not a physical property. The device core supplies the behaviour; you supply the meaning behind the reads and writes.

The SATA layers a device core implements

SATA is defined by SATA-IO as a stack of layers, and a device core has to implement the target side of each one. Understanding the split helps you reason about where your application logic plugs in.

Physical (PHY) layer

The lowest layer is the serial electrical interface: differential signalling, 8b/10b line coding, clock recovery and the out-of-band signalling used to bring the link up. On an FPGA this maps onto a hardened gigabit transceiver — on Artix-7, a GTP transceiver — with the device core driving it. This is the layer that fixes the achievable line rate.

Link layer

The link layer frames the serial stream: primitives, flow control, scrambling and CRC generation and checking. It packages the payload into and out of the transport layer reliably, and it is where transmission errors are caught.

Transport layer

The transport layer assembles and disassembles Frame Information Structures (FIS) — the fundamental units SATA uses to move commands, data and status between host and device. On the device side, this layer parses inbound command FIS from the host and constructs the response and status FIS to send back.

Command / application layer

The top layer interprets the ATA command set: how the device responds to IDENTIFY, READ, WRITE and control commands. This is the boundary you design against. The device core presents your logic with decoded commands and a data path; your fabric decides what a “sector” means and where its bytes come from or go to. Because the whole stack runs in hardware, the response timing is deterministic — there is no operating system or driver latency in the loop.

Why is Artix-7 a natural fit?

Artix-7 is AMD’s (formerly Xilinx’s) cost- and power-optimised member of the 7-series. It carries GTP transceivers, the serial I/O blocks whose rate range covers the SATA generations, which is what makes SATA on a low-cost part practical in the first place. For a device-mode design, that combination is attractive for three reasons:

  • The transceiver is already there. The GTP handles the SATA physical layer, so the device core does not have to fight the FPGA for high-speed I/O — it uses the block the family was built to provide.
  • The economics suit an emulator or interposer. Drive emulators, test fixtures and bridges are often built in quantity or embedded in equipment where a large, expensive FPGA would be overkill. Artix-7 keeps the bill of materials sensible.
  • It reuses a part you may already have. Many mixed-signal and instrumentation designs already sit on Artix-7 for its logic and I/O. Adding a SATA target can mean staying on the same device rather than re-spinning to a bigger family.

LDS offers SATA Device IP across several FPGA families, Artix-7 among them, so the target you build can match the exact part you have already chosen for the rest of the system.

Where does a SATA Device core make sense?

The device role is rarer than the host role, and that rarity is precisely why it is useful — it unlocks products that a host core cannot build.

Drive emulation for test and validation

If you build storage-consuming equipment — a recorder, a set-top box, a diagnostic tool — you often need a “drive” whose behaviour you fully control. A SATA Device core lets an FPGA impersonate an SSD and inject exactly the responses, capacities or fault conditions your test plan calls for, without hunting for physical drives that happen to misbehave the right way.

Transparent capture and interposition

Because the device end sees every command and every byte the host sends, an FPGA in device mode can sit between a host and downstream logic as a capture point — observing, timestamping or filtering the traffic. This is valuable in instrumentation and in security work, where you want to inspect or transform what crosses a storage interface before it reaches media.

Secure and synthetic storage front-ends

A device core lets you present a “disk” whose contents are encrypted, access-controlled or generated on the fly. The host reads what looks like an ordinary drive; your fabric enforces policy or synthesises data behind it — useful for secure storage appliances and for feeding a host system that only knows how to talk to a disk.

Bridges to legacy hosts

Plenty of fielded systems — industrial controllers, avionics, older test equipment — can only boot from or log to a SATA drive. A device core lets an FPGA present a SATA face to that legacy host while doing something entirely different on the back end: exposing a network share, a modern flash array, or another protocol as if it were a local disk. It is a clean way to modernise behind an interface you are not allowed to change.

SATA I vs II vs III

SATA generations double the signalling rate each step. The device core negotiates whichever generation both ends support, bounded by the FPGA transceiver’s reach.

SATA generations and nominal signalling rates
Generation Common name Nominal signalling rate Notes
SATA I SATA 1.5 Gb/s 1.5 Gb/s First generation; still seen on legacy hosts and devices.
SATA II SATA 3 Gb/s 3 Gb/s Widely deployed; a common fallback when a link or part is rate-limited.
SATA III SATA 6 Gb/s 6 Gb/s Default target for full SATA-SSD throughput on new designs.

Two caveats worth keeping in mind. First, these are signalling rates: 8b/10b line coding and protocol overhead mean the useful application throughput lands well below the headline figure — which is the classic reason engineers move to NVMe instead of SATA when they genuinely need multiple gigabytes per second. Second, in device mode the generation you can offer is capped by the FPGA transceiver and by what the connected host will negotiate, so confirm both ends before assuming SATA III.

When a SATA Device is the wrong tool

Device mode is powerful within its envelope, but it is not the answer to every storage question:

  • You need to store data on a drive, not be one. If your FPGA is the recorder or logger writing to an SSD, that is a SATA Host core — the exact opposite role. Choosing device mode here is the most common early mistake.
  • You need NVMe-class throughput. If the host expects multiple gigabytes per second, a SATA target cannot deliver it; the interface itself is the ceiling, regardless of how fast your fabric is.
  • The host expects a non-SATA interface. If the system talks PCIe/NVMe, USB or Ethernet rather than SATA, a SATA Device core is the wrong protocol entirely — match the core to the host’s bus.

This article does not cover the internal design of any specific LDS core — resource utilisation, latency figures and configuration options are part of the product documentation and a technical exchange, not something to estimate here.

Need a SATA host or device IP core for a specific FPGA family? Talk to the team that builds them.

Talk to our IP team

Frequently asked questions

What does a SATA Device IP core actually do?

It implements the target (device) side of the SATA protocol in FPGA logic, so the FPGA presents itself as a SATA drive to an external host computer. The host enumerates it and issues read and write commands; your fabric logic decides what those reads and writes do.

How is SATA Device IP different from SATA Host IP?

A SATA Host core makes the FPGA the initiator that drives a real SSD or HDD. A SATA Device core makes the FPGA the target that a host reads — it emulates a drive. The host core stores data on a drive; the device core becomes a drive. They are separate cores for opposite roles; the selection guide compares them side by side.

Can an Artix-7 run a SATA Device core?

Yes. Artix-7 carries GTP transceivers that cover the SATA line rates, so a SATA Device core can drive the device end of the link on that family. LDS offers SATA Device IP across several FPGA families, Artix-7 among them.

Does a SATA Device core need a processor or operating system?

No. The core implements the SATA protocol stack — PHY, link, transport and command layers — entirely in FPGA fabric, so it runs with no soft CPU, no driver and deterministic response timing.

What can you build with a SATA Device core?

Drive emulators for test and validation, transparent capture or interposition points, secure and synthetic storage front-ends, and bridges that present a SATA face to a legacy host while doing something different on the back end.

Which SATA generation should a device design target?

SATA III (6 Gb/s) for full throughput on new designs, falling back to SATA II (3 Gb/s) or SATA I (1.5 Gb/s) when a legacy host, a link constraint or the FPGA transceiver fixes a lower rate. The device negotiates whatever both ends support.

LDS
Logic Design Solutions
FPGA design house and soft-IP core vendor — NVMe, SATA, ARINC 429 and Ethernet IP for AMD and Altera FPGAs since 1994.

Need a SATA Device core to make an Artix-7 — or another family — present as an SSD? A short technical exchange usually settles the role, the generation and the target part.

License or request the IP

  1. SATA-IO — Serial ATA Specification
  2. SATA-IO — SATA ecosystem and generations overview
  3. AMD (Xilinx) — Artix-7 and GTP transceiver documentation
  4. Logic Design Solutions — SATA Device IP product line