---
title: "W5100S: auto-negotiation never completes on 1000BASE-T switch ports (ANLPAR stays 0x0000)"
url: "https://maker.wiznet.io/forum/16201"
markdown_url: "https://maker.wiznet.io/forum/16201/md"
type: "Forum topic"
category: "Ethernet Chips"
author: "LLS"
created: "2026-11-19T15:45:51+09:00"
last_activity: "2026-11-19T15:45:51+09:00"
language: "en"
views: 0
replies: 0
source: "WIZnet Makers forum (https://maker.wiznet.io/forum/)"
---

# W5100S: auto-negotiation never completes on 1000BASE-T switch ports (ANLPAR stays 0x0000)

## Question

Asked by LLS on 2026-11-19 in Ethernet Chips.

**Setup**

- W5100S, datasheet v1.2.8, SPI mode, custom STM32G0 board

- Magnetics: BT16B03 (10/100, 1CT:1CT, 350 µH), 49.9 Ω + 0.1 µF termination, centre tap to 3.3 V

- Peer: Aruba R8N86A (CX 6000), 1000BASE-T ports

- Poll `PHYSR0` for 20 s after reset; fail if `LNK = 0`

**Symptom** `CABOFF = 0` but `LNK = 0` for the whole 20 s window. Resetting PHY or the whole chip only helps occasionally.

**Register captures (MDIO via PHYRAR/PHYACR/PHYDOR)**

| Register | FAIL | OK |
| --- | --- | --- |
| ANAR (0x04) | 0x01E1 | 0x01E1 |
| ANLPAR (0x05) | **0x0000** | 0xDDE1 |
| ANER (0x06) | 0x0065 | 0x006F |
| PHYSR1 (0x003D) | **0x81** (RXP=0) | 0x85 (RXP=1) |

`ANAR = 0x01E1` → our advertisement has `NP = 0`, so we never request a next page.

**Control experiments**

1. Same board + cable on a 100BASE-TX port → **link OK**

2. Same board + cable on the same 1000BASE-T port, `PHYCR0 = 0x04` (forced 100FDX) → **link OK and stable**

3. W5500 on the same port and cable → **link OK**

4. `R50` on `XSCO` changed 100 Ω → 0 Ω → no change

So magnetics, crystal, termination, supply and the switch port are all fine. The only remaining variable is that the peer is a 1000BASE-T port, i.e. its base page has `NP = 1`.

**Questions**

1. Is this a known limitation of W5100S auto-negotiation?

2. Is the base page expected to be latched when the partner advertises `NP = 1`, even though our `ANAR` has `NP = 0`?

3. Is there any register to disable next page expectation, or to limit the advertised ability?

4. Has anyone reproduced this with other 1000BASE-T switches?

Any hint would be appreciated — we are deciding between forcing 100BASE-TX permanently and changing the PHY.

---

Source: https://maker.wiznet.io/forum/16201
