Not something I had heard of, but I see...
so I dig for the specs...
Table 5-4. 48-MHz Bulk Acoustic Wave (BAW) Resonator ( Measured on a Texas Instruments reference design with Tc = 25°C, VDDS = 3.0 V, unless otherwise noted. )
Frequency 48 MHz PPM –40 ~ +40
Start-up time 8 µs
No temperature curves, but the white paper http://www.ti.com/lit/wp/sway027/sway027.pdf
" As Figure 1 illustrates, the BAW resonator is fully integrated into the 7-mm-by-7-mm quad flat nolead (QFN) CC2652RB device package. It services the digital phase-locked loop (PLL) by providing
an accurate reference frequency, enabling the PLL to operate at 48 MHz reliably. In order to achieve superior frequency stability over temperature and battery (voltage) conditions, the BAW resonator is
actively compensated. This active compensation is carried out in the integrated radio core without impacting the application’s MCU bandwidth, and enables the CC2652RB device to provide a tight frequency error of 40 ppm across the full
operating voltage (1.8 V–3.6 V) and temperature (–40°C to 85°C) range. In comparison, external 48-MHz crystals exhibit reasonable frequency error (~10 ppm typical) at room temperature but tend to vary by 10s of ppm across the full operating range."
Seems it does need correction via temperature + SW + some-fix-applied (trimcap, or nudge PLL? ), but ±40ppm over Voltage and Temperature is covering a lot of applications.
Wonder how specific that correction is - do they need to calibrate every single device ?
If it can start and be within spec in 8 µs, that's well ahead of a Crystal solution.
This white paper http://www.ti.com/lit/wp/snoaa34/snoaa34.pdf
covers BAW used at 2.5GHz in a VCO/PLL. Very high Q, and low phase noise, but it also has a very
narrow pull range (2.5 GHz ± 50 ppm), so they include a LC-VCO 5.5~6.25GHz in the same package, for less critical clocks out.
BAW VCO -> 312.5MHz output typical total rms jitter is around 47fs.
LC-VCO, ->155.52MHz rms jitter is around 118fs.
At $11.44/1k the LMK05318RGZT Ultra-low jitter single channel network synchronizer clock with BAW, is not super cheap, but if you need
those numbers, price matters less.
Some P2 users may need those levels of very-high-spec jitter, so this part could be a solution ?