Hello,
We are working with an STM32F03 and an ECS-320-8-37B-CKY-TR 32 MHz crystal (CL = 8 pF).
Our earlier design used an 8 MHz crystal (CL = 8 pF) with 4 pF load capacitors and a 5.1 kΩ series resistor. After switching to the 32 MHz crystal, the HSE no longer starts. This sparked criticism about the correct component values depending on the crystal used.
Here are some results from my testing with 32 MHz + 5.1 kΩ:
- Startup works reliably at ~0 Ω and ~1 kΩ
- Marginal at ~1.5 kΩ
- Fails at ≥2 kΩ
Based on the ECS-320 datasheet (CL = 8 pF, and estimated ~3 pF stray), the correct load capacitors should be ~10 pF each, not 4 pF.
Where we are unsure is the interpretation of AN2867:
- Harmonic suppression (§3.5.3): Formula Rext = 1/(2πfCL²) → ~1.2 kΩ at 32 MHz with 4 pF loads, which matches our *failing* results.
- Drive level trimming (§3.3): Says to use Rext = 0 Ω unless the drive exceeds the crystal spec. We don’t have low-C probes to measure the drive directly.
- Empirical: Anything in the hundreds of ohms to kΩ range makes startup unreliable; 0 Ω is robust.
This leaves us unsure what to prioritize — the harmonic suppression formula points to high Rext values, while both ST’s drive-level guidance and our tests show that only very low values (0 Ω - 1 kΩ for our crystal) ensure reliable startup.
For comparison, the Nucleo-F030R8 (MB1136) uses an 8 MHz, CL = 12 pF crystal with 20 pF caps and no Rext.
Questions:
- Is there a practical way to approximate drive level (e.g., Vpp across the crystal + ESR) without specialized probes?
- Should §3.5.3’s Rext formula be seen as *theoretical only* (not a BOM design rule) for MHz crystals?
- Is ST’s effective guidance: default to Rext = 0 Ω, adding only small values (tens of ohms) if measured drive is too high?
We’d like to finalize our BOM for the 32 MHz crystal (10 pF caps + a clear Rext strategy) to ensure reliable startup and safe drive.
Thanks kindly,
Andreas
