Low Jitter Clock Generation with a External Crystal

[neoflash]

Hi folks:

We are leaning to design a clock generator with a external crystal. What's the term people use for such kind of circuits?

Just want to better shoot at the searching.

Regards,
Neo

 

[JoannesPaulus]

crystal oscillator

 

[neoflash]

Which type of crystal oscillator is best fit for CMOS technology?

Pierce, Colpitts or some other form?

 

[JoannesPaulus]

I mostly designed pierce oscillators. The disadvantage is that they require 2 package pins but they are fairly easy to design. The colpitts oscillator could be designed for a one package pin but - I believe - it has worse jitter performance (if this is an issue for you).

 

[neoflash]

two types of pierce shown there,

one doesn't have a resistor, the other has R1. What's their difference?

Thanks.

 

[JoannesPaulus]

The Rbias is used to set the input impedance (it is usually a big resistor, like 500kohm). The amplifier based design will, most likely do that with some kind of feedback.

 

[neoflash]

Sorry for the confusion.

I mean R1 in figure 1. Some Application note stated that R1 will introduce some phase shift.

 

[JoannesPaulus]

it is used to adjust the phase shift in the loop and to set the crystal drive level.

 

[neoflash]

why phase shift need to be adjusted? What if we don't?

 

[JoannesPaulus]

Well, R1 is often a 0ohm resistor but if the oscillator does not work (because the parasitics on the board are different from what you expected), you might want to be able to tweak here and there and this is a very simple adjustment.

 

[neoflash]

New question about XO.

https://www.designers-guide.org/Forum/YaBB.pl?num=1256090508/0

 

[hock]

R1 is used to insure the startup of oscillator by providing the phase shift.
Rbisa is used to reduce the gain of internal amplifier.

hock

 

[JoannesPaulus]

you should try to kick start the oscillator (i.e. use a current pulse, say, 10nA for 10ns between ground and the gate) or try to ramp up the power supply.

 

[neoflash]

Tried to set initial condition. But the oscillation will die out.

Added after 3 hours 8 minutes:

I did some analysis and found that pierce and colpitts are exactly same in small signal.

More confused why Colpitts doesn't start while pierce always start easily.

 

[JoannesPaulus]

yes, they are even though they have a very different behavior.

Try to have a look at this paper (a must read for xtal oscillator design):
"High-performance crystal oscillator circuits: theory and application" E.A. Vittoz, M.G.R.Degrauwe, S. Bitz
IEEE Journal of Solid-State Circuits, Volume 23, Issue 3, June 1988 Page(s):774 - 783

It has some very helpful basic theory.