Why does my LSCO output for my LSE have a delay in matching my LSE input.

[LAdam.1]

I am developing on a STM32L4 product and was having issues with the CSS shutting down my LSE signal. I found this problem in the errata and turned the CSS off, but it forced me to look closer at the signal my LSE crystal was providing (using the LSCO output option), and the duty cycle was way off. I tried changing load capacitors and crystals without improvement, so I finally just changed it out for an oscillator instead. Now I know the signal for my LSE is good because I can directly measure it. What has me confused is that initially the LSCO signal still looked way off when I measured it after the hardware change, but upon returning in the morning the LSCO output is now tracking the oscillator input just fine. Why would there be a delay in the LSCO output changing to match the LSE?

[waclawek.jan]

> What has me confused is that initially the LSCO signal still looked way off when I measured it after the hardware change,

Delayed, like you said in the title? Can you estimate how much?

> but upon returning in the morning the LSCO output is now tracking the oscillator input just fine. Why would there be a delay in the LSCO output changing to match the LSE?

I can only speculate, but have you powered down completely the VBAT domain while changing crystal for oscillator (and by "completely" I mean deliberately shorted, to remove any residual voltage from charged capacitors)? If not, I can imagine that the original setting of "non-bypass" might have persisted, resulting possibly in some amplifier/reshaping stages being inserted in the way of the signal from the oscillator-in pin towards the logic.

JW

[LAdam.1]

By delayed, i don't mean the signal itself had a noticeable delay, just that the output didn't immediately match my Oscillator input until later in the day or the next morning. I do not have a battery installed in my VBAT connector at the moment. I did not try deliberately shorting the power to ground.

[waclawek.jan]

> output didn't immediately match my Oscillator input

So, might it be outputting LSI?

> I do not have a battery installed in my VBAT connector at the moment.

Is it just floating, or is there any capacitor?

From RM0351:

RCC_BDCR [...] reset by Backup domain Reset, except LSCOSEL, LSCOEN

and BDRST which are reset only by Backup domain power-on reset.

I can also imagine latches/register bits hidden to user, related to the CSS/LSI/LSE complex.

JW

PS.

> having issues with the CSS shutting down my LSE signal

Which 'L4? What erratum?