The low-cost nature of NB-IoT modems encourages them to deploy uncompensated crystal oscillators (XOs) as frequency references. The frequency offset of an uncompensated XO, however, makes network acquisition inefficient under low network coverage. In the worst case, the bulk of an NB-IoT modem's power is consumed in network acquisition. This work demonstrates a discrete frequency synthesizer prototype that employs its phase-locked loop to compensate for the frequency offset of its reference XO. We propose an accurate crystal model and a compensation logic that are suitable for 32-bit microprocessors, commonly available in NB-IoT modems. Alternatively, when synthesized in a commercial 65-nm process in 24-bit precision, the proposed compensation logic is simulated to require a total area of 0.029 mm2 and power of 290 nW at a refresh rate of 1.4 kHz. Moreover, the XO model requires only 0.16 kB of RAM. The prototype achieves a compensation error level of down to 27 ppb (3Σ) over the temperature range from-40 to 85 °C. The dominant error sources of the prototype are discussed in detail. In addition, we show that an NB-IoT modem can use the information acquired from sequential network acquisitions to compose and update its XO model in the field. This support for field calibration removes the need for XO characterization in production and ensures the validity of the model over the NB-IoT modem's lifespan. The prototype achieves a compensation error level of down to 50 ppb (3Σ) in an emulated in-field calibration test, sufficient to ensure energy-efficient low-latency network acquisition under low network coverage.
|Julkaisu||IEEE Transactions on Circuits and Systems I: Regular Papers|
|DOI - pysyväislinkit|
|Tila||Julkaistu - heinäkuuta 2020|
|OKM-julkaisutyyppi||A1 Julkaistu artikkeli, soviteltu|
Energiaomavaraiset langattomat anturijärjestelmät teholähteenään painettavat aurinkokennot ja energiavarastot
01/09/2018 → 31/08/2021
Projekti: Academy of Finland: Other research funding