I've built an Arduino-based VFO using the Si5351 clockgen and 3.5" TFT touchscreen following KW5GP's design. Works great for SSB/CW switching and step control, but I'm seeing about +200Hz drift over the first 20 minutes of operation on 20m. Temperature compensated crystal oscillator helps but doesn't eliminate it completely.

Running off 12V regulated supply, adequate heatsinking on the Si5351. Anyone else experienced this thermal drift with clock generators? Wondering if I need to implement software compensation or if there's a hardware fix I'm missing.

Current sketch uses Etherkit Si5351 library with default 25MHz crystal correction — should I be measuring and adjusting this value more precisely?

The 200Hz drift is typical for Si5351 without proper calibration. You definitely need to measure your actual crystal frequency and set the correction factor. I use a frequency counter on 10MHz output and calculate the error — usually need corrections around 15000-18000 range rather than default.

Had similar issue with my Arduino antenna analyzer project. Adding a small fan to keep the Si5351 cool made a huge difference. Also, let it warm up for 30 minutes before calibrating the correction factor.

Check your power supply regulation too. Si5351 is very sensitive to supply voltage changes. Even 0.1V drift can cause frequency instability. I ended up adding a dedicated 3.3V LDO regulator just for the clock chip.