32.789 kHz crystal with HCS08DZ60

[shadoooo]

Hello.
I'm working over a project employing a HCS08DZ60. We use a crystal oscillator with 32.789 kHz frequency, and then internally I multiply this frequency by 1024 with FLL. I connected this crystal with two 22p cap over ground, a 10M res in parallel to the crystal, and a 100k from XTAL and one of the crystal pins, with the classic pierce configuration.
In the micro I configured for external crystal oscillator, low power mode, and enabled bus clock output on PTA0 for debugging purposes.
The oscillator starts with no problem, but the uP works bad and continues to reset itself.
What I see is that the frequency on PTA0 is sometime stuttering, and when it happens, the uP restarts.
What is really strange is that it does in this way also if I disabled clock monitor feature.
I tried to change a little the configuration by changing the series resistor to 330k and to 47k, with no change.
I tried also to disconnect the crystal and connect in place an external generator coupled with an 10n capacitor on pin EXTAL (input of the uP), giving 1Vpp and a frequency of 32.789 kHz, or other frequencies.
What happens is the same, but also the stability seems to be better if the frequency is around 34 kHz or so, but it could be just a case.
I tried to increase also the level of the signal, with no difference at all.
I would exclude in this way that it is a matter of noise, unstable oscillator or wrong external components, because I used a very stable R&S generator, and in any case the crystal oscillator alone works good.
I tried to switch to internal oscillator, in this way all works good.
It seems to be a problem related to the FLL, I could also suppose that something happens when you use a frequency that in some way is near the internal generated frequency.
Any idea to solve this problem?

[RichC]

I'm also interested in making sure this works, as I intend to use an external 32k clock source...

It would be worth posting your initialization code to check the registers are set up for the correct clock mode/division etc.

Also, I don't understand why have you ac coupled your external clock with a capacitor rather than just connecting directly to EXTAL - Wouldn't ac coupling take it below 0v?

[shadoooo]

I did the initialization through the bean of processor expert. So the code is generated accordingly by it.
I used a capacitor as the EXTAL line as well as the XTAL is biased by circa 500mV when used in crystal mode. So using a crystal I see a sinusoidal signal that oscillates between 0V and 1V, so 1Vpp biased at 500mV. This means that connecting the pin to a generator biased with 0V causes a current drain on the pin.

[shadoooo]

I forgot to mention that I found an ERRATA on "mask set errata for mask 3M05C", particularly "MCGV1 Can Cause a Very Short Clock Pulse" paragraph, SE128C-MCGV1.
It seems to me that the problem descripted is the same I experience this days.
However I would to be really sure that this is the same problem I have, because I'm using the mask 2M05C (older ?!) and not the 3M05C, and of course that it will be solved in the final mask.
If the problem has been already solved, I would know in which mask the fix was introduced, in way to request from Freescale a sample of the corrected chip, to be sure that I have no other problem to solve for the final production.