Hi On Jun 1, 2013, at 12:21 PM, Magnus Danielson <[email protected]> wrote:
> Attila, > > On 06/01/2013 05:11 PM, Attila Kinali wrote: >> On Wed, 29 May 2013 01:59:12 +0200 >> Magnus Danielson<[email protected]> wrote: >> >>> On 05/28/2013 07:55 PM, Attila Kinali wrote: >>> >>> http://www.pttimeeting.org/archivemeetings/1984papers/Vol%2016_10.pdf >>> >>> See also >>> http://www.pttimeeting.org/archivemeetings/1979papers/Vol%2011_25.pdf >> >> Interesting stuff. Thanks a lot! >> >>>> Juup. I just went back to Vig's tutorial and read up what he wrote. >>>> Misremembering things is not a good thing... >>>> But then, he explicitly writes that SC cut gives a higher stability >>>> over AT cut due to lower temperature dependence and less dips. >>> >>> That is true, but your discreditation of AT-cut was simply way off the >>> mark, so I wanted to bring it into context. There is a difference, but >>> it is not as huge as it sounded like. I like to think about it as such >>> that you better have done much of your homework in form of good >>> oscillator and oven before considering spending money on going from >>> AT-cut to SC-cut, but it does give that extra performance if you need >>> it. The OSA 8600 shows just how far you can take AT-cut. >> >> Oh.. Ok. Didn't want to sound that way. On the other hand, that conclusion >> is not far from the truth. I still have a lot to learn. Thanks for the >> correction. > > The benefits (as I recall it) is a somewhat higher Q and less thermal > dependence. Actually on a direct "same sized blank / same frequency" basis, the Q of an AT is often higher than the Q of an SC. Because the SC is higher impedance, it's phase slope will be higher. Bob > >> But then, the 8600 has a ~10dB higher noise then the 8607. Ie the noise >> is 10 times higher (it is power-dB, not voltage-dB isn't it?). Which >> makes me wonder what the noise contribution is. I would assume that the >> electronics are very similar if not the same (electronics are cheap compared >> to the crystal) and the mechanical construction seems to be very similar >> as well. > > You must be looking at the 1 Hz numbers. You really need to look at the > phasenoise at different offsets to understand what goes on. > > The electronics contribute a white phase noise, as well as a flicker noise > (1/f, a -10 dB/decade slope). The resonators Q-value and frequency will > define the break-up point, below witch you have a -20 dB/decade slope and > above it is flat. Depending on the resonator at hand, you then see a 1/f or > 1/f² noise between the flat white noise and the 1/f³ noise. The output amp > can then add white phase noise and flicker phase noise. > > The 10 dB improvement I would attribute to the improved Q value with SC cut > crystal, but... these are not the real values, it's the published values > which is guaranteed for a product. Actual values is different. > > Looking in Enrico Rubiolas book, he measures both OSA 8600 and OSA 8607 > (among others) and then the actual numbers is much closer. The 1 Hz values is > only 4 dB different, but then the wideband noise between these two samples > lets the 8600 be 2 dB quieter than the 8607 at -155 vs. -153 dB. > > So, we go back to actual measurements, we need to realize that each > oscillator is unique, and that data-sheets only intends to give some form of > guarantee of how bad they will be as they exit the factory, but not > necessarily reflect all aspects of the oscillator. > > I've seen AT-cut oscillators behave better than SC-cut within the same basic > conditions, but where the AT-cut was much better implemented. > > So, there is a difference, it's just not very large and the full performance > depends on so many other parameters of a design. The cost of SC-cut blanks is > higher, and it is not meaningful to use one unless the design have come so > far that it starts to become a limiting factor and other improvements is more > expensive. > >> In comparison the stability between 2 and 30s has only a factor of 2 >> inbetween. >> (maximum instability according to spec). >> >> Any ideas what the reason could be? > > #1 Datasheet specing difference. > > #2 Actual difference is due to Q-value difference, moving the break-up point > from the Q value. > > So, the Leeson model explains this pretty well, and the difference in Q-value > of the loaded crystal blank is then reflected in noise differences with > essentially the same buffer electronics. > > Cheers, > Magnus > _______________________________________________ > time-nuts mailing list -- [email protected] > To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts > and follow the instructions there. _______________________________________________ time-nuts mailing list -- [email protected] To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
