A few months ago, we looked at Sand 9’s initial announcement. They had laid out 3 families at that point: a basic resonator (TM061), a temperature-sensing resonator (TM361), and a “roadmap” family for temperature-sensing oscillators. Well, they recently announced a new device that goes in yet a new family: temperature-compensated oscillators – the TM651. When chip-scale packaged, they claim it’s the smallest oscillator in the world (although an LGA is also available).
They’ve come out swinging at their performance versus quartz, in particular the latter’s susceptibility to so-called “activity dips,” which we covered in the prior piece. But they’re also comparing themselves to other MEMS – and, in particular, electrostatic – oscillators. They say:
- The filter and noise are better than any MEMS oscillator and competitive with quartz;
- They have 250-ps edge rates as compared to about 1 ns for quartz;
- Their vibration immunity is an order of magnitude better than quartz;
- They have 1/15th the drift of other silicon MEMS due to their analog compensation, which is smoother than digital;
- They can achieve higher frequencies without a DLL, which quartz needs for frequencies above 50 MHz;
- They don’t have quartz’s finicky start-up time (and, apparently, start-up can occasionally fail outright with quartz);
- They have better electromechanical coupling than electrostatic MEMS devices because they’re piezoelectric (with the presumed effect that the coupling happens intrinsically within the material as opposed to being between two distinct mechanical members);
- They have larger transduction area, which, counter-intuitively, reduces die area (presumably because of better intrinsic sensitivity);
- They have no air gaps, vs. those used with electrostatic devices (which goes to coupling efficiency);
- They operate off of a lower voltage;
- They have linear power vs. non-linear for electrostatic, giving them better noise performance;
- They can work with a customer to have the customer’s electronics co-packaged with their resonator for better integration, which isn’t possible with quartz.
That’s a lot of claims.
Their electronics are in the cap wafer. The bonding is done wafer-to-wafer; both the MEMS and ASIC see very high yields (in fact, wafer probing on the MEMS die is done only on a wafer sample basis to see if it looks like there’s a problem with the lot). GlobalFoundries does this for them.
You can read more in their announcement.
