Pierce Oscillator - no output on XTAL

[blewis]

We have a Pierce oscillator on a 9s12dg128 part and we have the XCLKS tied to gnd permanently, but I think we are in the Colpits configuration. There is no signal on the XTAL clock output line. There should be a DC offset (sort of) sine wave type signal when the oscillator is working normally, but we see nothing at all. So either the signal is very weak or the processor is not working correctly in the full swing Pierce mode. We have 22pf cap's on the two legs to gnd, and a 1M resistor across the crystal, which is a standard 22pf ECS through hole 16Mhz H49U crystal. The pll is connected but not enabled in the software. Has anyone any ideas about how to determine why there is no output on the XTAL line? The cct is pretty much exactly as in the design spec and app note 2727.

We have VDD1, and VDD2 bypassed with 220nf caps and not connected incorrectly to VCC.

The track and lock bits are set in the CRGFLG register so the oscillator is not in self clock mode but is locked in and tracking. The serial port is working correctly so we know the part is running at the crystal freq of 16Mhz.

Any advice would be appreciated. Thanks inadvance.

Bob

[Steve]

Bob,

  Since it sounds like everything is working I'm wondering if the problem is a measurement one. When you probe the crystal pins does the part lose lock? The reason I ask is because it sounds like you are adding a big load on the crystal that is killing the oscillation that is otherwise working properly.

[imajeff]

I find it helpful to make my scope lead work like an antenna, so I can measure frequency without touching. Of course I can't measure voltage that way.

[blewis]

Steve, imajeff

Thank you for your suggestions, they helped me in searching out the problem.

What I found was a cold solder joint on the crystal. I had used a through hole part to solder onto the pads of an smd crystal. One of the leads had a cold solder joint. I found it when I had to bend the part slightly to access the micro's pins and it didn't feel right. A bit more movement and it broke free. Once it was soldered I could see the signal. Why the micro wasn't in self clock mode is a mystery, but I guess the resistance added enough of a shift of freq to mess up the serial connection. I wouldn't have found this without the serial connecton because all the spi parts worked fine... The baud rate couldn't be set with the crystal freq off.

Well a lesson learned but what a time sink... thanks again.

Bob