STM32H73x: HSE / LSE oscillator levels with passive crystals

[LCE]

Heyho,

there's lots of info in the datasheets and the AN2867 about the specs of the HSE / LSE oscillators with passive crystals...

But I'd like to have some "measurable" *** specs, so that I can compare a real life custom board, which means:

voltage levels at the oscillator IN/OUT pins.

*** Okay, the scope's probe's capacitance probably falsifies the actual signal, but so would be ST's...

That would be some helpful info in the respective datasheet - and not some other app note.
IMO it's not really good documentation if there are specs used in the AN which you cannot find in the datasheet, but only in a table in the AN.

So, what can ST please tell me about "good, reliable voltage levels" / swing in Vpp at the HSE and LSE IO pins?

[Danish1 · Accepted solution]

Think about what it is you're proposing to measure, and what you want to measure.

The oscillator consists of an amplifier and the crystal.

What we need to ensure is that the gain of the amplifier is sufficient to overcome the losses of the crystal by a sufficiently high margin to start up sufficiently quickly from thermal noise or any residual kick from power-on.

That gain suffers from process variations (as do the losses in the crystal) so you need a good level of reserve.

Now think about what happens with a healthy level of gain. Each time round the loop, the signal gets bigger and bigger. Until a point where the amplifier saturates so much its effective gain falls and the overall loop gain drops to unity.

The problem is that the signal-level for this saturation is also subject to process variation. Measuring the signal level will only tell you the level at which saturation kicks in, not the amount of gain in reserve.

Ok so the saturation process is "soft" so with more surplus gain the amplifier output will be more saturated, but I'm not convinced that measuring the harmonic-distortion at the output of the amplifier will give results that are easy to analyse.

[LCE]

I don't care about harmonic distortion - or could that be a problem at this "digital clock input"?

I rather about some minimum required voltage swing that the STM32 oscillator should produce with a crystal.

Background:

I have a custom PCB here with 2 other ICs that use the same crystal (25 MHz, 10ppm, ESR 40R, 12pF, in 3225), and the voltage levels with the crystal is much higher, and the low voltage (400 mVpp, compared to the other 2 ICs with > 1 Vpp) at the STM32 makes me wonder what a good, robust, reliable voltage-swing for the STM32 might be.
The H7 is starting without problem, but for now this is tested only at room temperature.

[BarryWhit]

> The H7 is starting without problem, but for now this is tested only at room temperature.

I think what Danish is saying is that if you see *any* measurable oscillation, this indicates that the feedback loop has sufficient gain to ovecome losses, which means the circuit itself (once amplitude stabilizes) is saturating.

I don't think it works the way you imply (I'm no expert, but from what I know) - It's not the case that the oscillator may startup but not "reach" it's desired amplitude. There's either not enough gain, no startup and no waveform, or there are all these things and whatever amplitude you see ( once it has increased sufficiently to reach a stable amplitude), is the saturation level for the oscillator circuit.  So in effect, any sustained oscillation of stable amplitude indicates the circuit is in a "healthy" state, and that the amplitude is as high as the amplifier can handle.

Now, even if you see sustained oscillation at room temp. it's still possible that the crystal may not start up at a different temperature - but I don't see that measuring absolute amplitude will get you more insight/predictive power about that.

> I don't care about harmonic distortion 

Checking for harmonic distortion is a useful diagnostic tool. See here.

[LCE]

Ouch... thanks guys!

Next time, I start thinking first.