CST328: setup & troubleshooting

CST328 at a glance

We did not use a Hynitron datasheet for these figures. The copy we found online is a reseller upload, so the values above come from driver source and board documentation.

The CST3530 swap

LilyGO's T-Deck Pro README says the CST328 supplier "has discontinued production", so newer T-Deck Pro v1.1 boards use a CST3530. It adds that "both chips use the same device address", so the firmware has to identify the chip during initialization. LilyGO uses Hynitron's cst66xx routine for the CST3530 and cst3xx for the CST328.

Waveshare's RP2350-Touch-LCD-2.8 made the same change. The RS-Key project documents V1 with a CST328 at 0x1A and V2 with a CST3530 at 0x58. On the T-Deck Pro the address does not change; on this board it does.

The CST3530 columns come from RS-Key's merged driver, which its author tested on a V2 board, and from LilyGO's README.

How the swap shows up

Both projects that hit this reported the same thing: the firmware runs and the display draws, but no tap registers. The reticulum-tdeck issue says some T-Deck Pro v1.1 units worked and others did not, which fits a change between production batches rather than a wiring fault.

Two more traps:

• An I2C scan does not identify the chip. On the T-Deck Pro both chips answer at 0x1A. The reticulum-tdeck authors probe for the CST3530 first, because a CST3530 "happily acknowledges CST328 register reads". They also say this code had not yet run on a CST3530 board. A meshcore-firmware bench test on a T-Deck Pro V1.1 shows the same thing: the CST328-style read was acknowledged but returned no chip type and no resolution, while the CST3530-style read returned a valid block with a 0xCA 0xCA marker and the panel's 240×320 resolution.
• Other drivers can claim 0x1A. An open Meshtastic pull request reports that its I2C scan took the T-Deck Pro touch chip for a DS2482-800 bridge, because a status-register test happened to match.

This is a common pattern with small touch controllers: the part number on the board stays the same while the chip changes. If you build in volume, ask your display supplier which touch IC is fitted, and to tell you before it changes. Our EOL and PCN glossary explains what a change notice should contain.

Bring-up with ESPHome

ESPHome's cst328 platform needs I2C configured and takes optional interrupt_pin and reset_pin. Its setup does the following:

• Pulses reset low for 5 ms, then waits.
• Reads the firmware CRC and expects 0xCACA. On any other value it marks the component failed.
• Reads the chip type, project ID, firmware revision and resolution, then switches the chip to normal mode.

The CRC check is the first place to look if ESPHome reports the touchscreen failed. ESPHome uses CST328 reads with 2-byte register addresses, and in the meshcore-firmware bench test that kind of read returned nothing useful from a CST3530. So on a board with a CST3530, expect this check to fail. That is our inference: we have not run ESPHome on a CST3530 board.

Driver support

• ESPHome: the cst328 platform. Its docs name the Waveshare ESP32-S3 Touch LCD 2.8 board. It has no CST3530 platform.
• ESP-IDF: Waveshare publishes esp_lcd_touch_cst328 in the Espressif component registry. Its compatibility table lists only the CST328.
• Rust: the cst328 crate on crates.io, and the RS-Key firmware for both chips.
• Hynitron's own driver: LilyGO links Hynitron's hyn_driver source for both chips.

Related controllers

For one-finger touch on round displays, see CST816S, also from Hynitron. For larger panels with up to 5 points, see GT911.