Use of External TCXO Without Bypass Mode – Clarification on Oscillator Behavior and Minimum Swing

[NirAlon]

Hello all,

I’m working with an STM32 microcontroller (specifically the STM32H743), and I have a question regarding the use of an external TCXO as the HSE clock source.

As per the datasheet, when using an external clock source (such as a TCXO), the HSE must be configured in bypass mode, and the input clock is expected to meet the digital I/O thresholds (e.g., 0.7 × VDD for logic high). However, many high-quality TCXOs have typical output swings of only 0.8–1 Vp-p (clipped sine wave), which do not meet CMOS thresholds when VDD is 3.3 V.

Therefore, directly driving OSC_IN with such a TCXO in bypass mode is not viable per the datasheet.

That said, I made an observation I’d like to confirm with you:

While using a crystal in crystal resonator mode (not bypass), I probed the OSC_IN pin using a low-capacitance active probe and measured a signal of approximately 1 Vp-p. Since this amplitude is generated internally by the oscillator circuit when using a crystal, it seems the STM32’s HSE input is designed to function with such low-level sine wave signals.

This raises two key questions:

If I connect a TCXO directly to OSC_IN, configure the HSE in crystal mode (not bypass), and leave OSC_OUT unconnected, will the internal oscillator accept and use the signal from the TCXO — even though it is technically not a crystal? The MCU would not be aware that a crystal isn't present, and the analog characteristics of the TCXO signal resemble those observed during normal crystal operation.

What is the actual minimum input voltage swing required at OSC_IN to guarantee reliable oscillator operation in this scenario (i.e., driving OSC_IN in crystal mode using an external sine wave signal)?

This approach would greatly simplify my design, as it would allow the use of a low-swing TCXO without needing level shifters or additional buffers.

Any insight you can provide into the feasibility and reliability of this configuration — even if unofficial or not guaranteed — would be much appreciated.

Best regards,

Nir

[RobK1]

A quick search on Digikey for TCXO's  with <= 0.5ppm stability shows there are more than 2200 different types with CMOS output in stock, compared to only 649 with clipped sine output.

That said, I have in the past succesfully used clipped sine oscillators on the crystal input, as you describe.This was a different MCU family, though, so YMMV.

[NirAlon]

Hi RobK1,

Funny enough, I actually had a bet with a colleague about the kinds of responses I’d get — and I predicted someone would suggest using a CMOS-output TCXO.

Of course, there's a specific reason I'm not using one in this case.

For the purpose of this discussion, please assume that the only available option is a clipped sine wave TCXO, and I’m trying to determine whether it's viable to use it directly with the HSE in crystal mode, without enabling bypass.

Thanks again for sharing your experience — it’s reassuring to know this approach has worked on other STM32 families, even if unofficially.

If anyone else has tried a similar approach (or has insight into the analog behavior of the HSE input in crystal mode), I’d really appreciate your input.

Thanks

Nir