[time-nuts] sand9 TCMO

[Magnus Danielson]

On 06/01/14 20:13, Attila Kinali wrote:
> On Mon, 06 Jan 2014 06:24:18 -0800
> Jim Lux <jimlux at earthlink.net> wrote:
>
>>> MEMS might be good for certain tasks, but for closer in noise I've only
>>> seen some progress recently, but not measured it myself. Close-in noise
>>> seems to have been pretty bad for all MEMS so far.
>>>
>>
>> I think that's probably related to the physically small size. It's hard
>> to get a high Q in something that's smaller than a gnat's eyelash.
>
> It's not only the Q (although i have not seen any Q values yet), but
> also that (almost?) all of those MEMS oscillators have a fixed frequency
> oscillator structure and use a fractional-N PLL together with a standard
> CMOS VCO. The spurs of the PLL are clearly visible if you go down to 1kHz
> (i haven't seen any spectrums going further down than 1kHz for MEMS oscillators).

Well, you want to make sure you do not have multiple resonant modes 
oscillating at the same time. The Q of each of those modes will 
naturally be of interest. In quartz oscillators you kill of nearby modes 
to avoid oscillating in modes with unwanted behaviour, such as high TC 
dependence. You can modify the oscillator circuit of 10811A in order to 
have it oscillate in a more temperature dependent mode, as it helps to 
identify the stability of the oven. MEMS is just another acoustical 
resonator, but implemented in silicon rather than silicon dioxide.

Q values of 75000 mentioned in this little article, which is early in 
the MEMS oscillator industrialization:
http://www.memsjournal.com/2006/02/timing_with_mem.html

This is a good read:
http://www.ifcs-eftf2011.org/sites/ifcs-eftf2011.org/files/editor-files/Slides_Piazza.pdf
Notice that on page 46 there is a phase-noise plot with close-in noise 
down to 10 Hz shown.

> The big promise of MEMS oscillators of having very low power consumption
> is not fullfiled yet. About half a year ago i checked all the 32kHz
> MEMS oscillators i could find and got numbers for power consumption that
> were about the same as an MSP430 would use, with its 32kHz crystal
> oscillator running.

I think their main advantage is size rather than power-consumption. 
Turns out that low power-consumption comes at a performance reduction, 
that have never happen before :)
I've noticed that there now seems to be a split, depending on what needs 
there is.

For good performance, I still want a good crystal, for many auxillary 
oscillators MEMS can be useful, or locked to a good crystal.

Cheers,
Magnus