The MEMS Testboard. With a buffer amplifier and a low drop voltage regulator.

The circuit is straightforward. (as usual). The SiT8008 is a Low Power type. We have chosen the largest size (7 x 5 mm) to allow for easy handling. A resistor divider at the output of the oscillator scales down the voltage, reduces the load and matches the input of the following amplifier. This approach was chosen to increase the isolation and reduce the effects of pulling. (in case the output gets shorted ;-) A voltage regulator (IFX25001MEV33HTSA1) produces a clean 3.3 V supply. Only for the oscillator. The buffer amplifier (GALI-6+) is powered by 5 V. The circuit consumes 80 mA when powered with 5 V. Output power is + 6 dBm (fundamental).

A common way to classify oscillators is by the type of resonator (XCO, VCXO, RC, DRO, ...). In our case MEMS, which stands for Micro Electro Mechanical System. Our resonator is therefore a mechanical resonator in the shape of a tuning fork. What is new is its size and material. First, it has been shrinked into the sub-millimeter range and second, it is manufactured on (in) the silicon die - like other cmos circuits. This has a lot of advantages. As those very small resonators are fully encapsulated in silicon, they are extremely stable and highly durable. The resonators are protected under a layer of silicon and can withstand a very high pressure such as 100 bar. (Conventional low temperature packaging, that uses ceramic packages or wafer bonding, leaves volatile organics and water residues in the package which cause mass loading of the resonator and frequency drift.) A PLL then does the frequency synthesis. We assume, that it is a fractional-n type.

Programming can be done just once (OTP). Therefore this is a "set-and-forget" approach. The proprietary programmer works beautiful, but € 250 is slightly above the homebrewers budget (think we). So you may use the programming service from e.g. digi-key - or wait, till someone shows up with an Arduino/Genuino based solution :-)




When programming those little pll's you have to think in part-numbers. But thats not a big deal, as they are only a string of characteristics. You may also use the part number generator, which breaks it down to setting (some) ticks and enter a numerical value for the frequency.

Last but not least a quick look at the performance. The question is : can it replace a crystal oscillator in rf circuits ? The manufacturer probably would say 'Yes, we can' because SiTime offers a broad range of types, which seem to be optimised for almost any application. You may choose between LVPECL and LVCMOS, singled ended or differential. The user may also select a Frequency Stability (±20, ±25, ±30, ±50 ppm). Besides that there are Low Power, Ultra Performance, High Performance, High Temperature, Automotive, VCXO and Spread Spectrum types available. We measured some of the 7 x 5 mm types which have been programmed to 10.000000 MHz in order to compare it with our HP Agilent 58503B GPS Time/Frequency Receiver as well as a Efratom LPRO-101 (from the bay) and a standard 10 MHz crystal oscillator.

SiT 8008	Low Power Programmable Oscillator
Range	1 MHz ... 110 MHz
Power	5.0 V, 90 mA (incl. amplifier)
Marking	B0AY6

SiT 8208	Ultra Performance Oscillator
Range	1 MHz ... 80 MHz
Power	5.0 V, 110 mA (incl. amplifier)
Marking	B0DF4

SiT 8209	High Frequency, Ultra Performance Oscillator
Range	80.000001 MHz ... 220 MHz
Power	5.0 V, 110 mA (incl. amplifier)
Marking	B06WN

SiT 3808	High Performance MEMS VCXO
Range	1 MHz ... 80 MHz
Power	5.0 V, 110 mA (incl. amplifier)
Marking	B08SM
Remarks	• Vin connected to Vdd results in f = 10.000604 MHz • We assume this is a SiT8208 with a Varicap build in.

Possible "Reference" Oscillators

In order to to place those measurements in relation to other synthesizers, we measured the spectra of some "professional" synthesizers. See screenshots below. One was persuaded to retire afterwards :-)

The Type SiT8008 is not suiteable for rf mixing applications, but it mayst serve as a clock source for a dds. Due to its 1 Hz resolution, a frequency of 67.108864 MHz is attainable and can be used to clock an AD9834 (75 MHz max.)

The webmaster does not read these comments regularely. Urgent questions should be send via email. Ads or links to completely uncorrelated things will be removed.