jakapoor
commented
6 years ago Pros and cons of GPSDO for synchronization:
Krueger's dissertation, describing a COTS approach to TDOA provides a good summary of the pros and cons of the use of GPS disciplined oscillators (GPSDO) in coordinating distributed receivers:
[p. 60] GPS Disciplined Oscillators (GPSDOs) are used by similar TDOA tracking systems [1, 16] to maintain precise synchronization between the receivers and to provide a frequency stability of 1 ppb or better. However, a GPSDO adds cost and complexity to the design of the receiver, and inexpensive off-the-shelf hardware does not provide such capabilities without modifications to the circuitry.
[p. 142] The greatest aspect that differentiates our design from the other two is that it works with unsynchronised receiver clocks that are independent. The other two systems both rely on GPSDOs at the receivers for precise clock synchronisation and clock stability. Precise synchronisation simplifies position estimation but adds to the cost and complexity of the hardware. Our system, on the other hand, makes use of beacon detections to estimate and compensate for clock drifts in software after data acquisition. ATLAS also makes use of beacons to compensate for clock offsets between the receivers, but they use the beacons differently and still rely on a GPSDO for clock stability...The total cost of the products that they [i.e. ATLAS] use[s] for the RF module, i.e. an Ettus Research USRP N200 radio, WBX RF daughterboard and supposedly a GPSDO kit from Ettus Research, amounts to approximately $2800.
He concludes on page 148:
Synchronising the receivers with traditional methods adds to the cost and complexity of the hardware; using a common clock is hard to achieve and expensive with large arrays or over long distances, and using a reference clock such as a GPSDO adds to the cost of the receivers. Instead, inexpensive receivers with unsynchronised clocks can be used together with one or more beacon transmitters.
Gabrielson et al in Ch.33 of the Handbook of Position Location explains the pros of the use of GPSDO:
Fortunately, GPS receivers are capable of providing very precise 1 - pulse per second (pps) and 10 - MHz signals, which supply the reference signals for the rest of the receiver. Each TOA receiver uses an independent GPS receiver to maintain very tight synchronization with UTC. GPS receivers that are specifically designed for timekeeping purposes are now available for embedded applications. These devices assume a fixed location in order to overdetermine a solution that yields very accurate 1 - pps edges. These edges are used to discipline a voltage - controlled, temperature - compensated crystal oscillator ( VCTCXO ), or in some cases, an oven - compensated crystal oscillator ( OCXO ). Each of these devices provides excellent short - term stability, and the GPS synchronization maintains their long-term accuracy.
Summary:
Given the discussions above, the benefits of the use of GPSDO are a generally more elegant and simplified solution WRT hardware, and no need for precisely localized beacons. The drawbacks are added complexity in integrating COTS-based GPS solutions (if they exist), and the high price of speciality GPSDO devices; a quick search of GPSDO devices provided by Ettus Research revealed a unit costing $1,200 CAD.
Conclusion
If other GPSDOs are even remotely close to the above quoted price there is little prospect of the use of GPSDO in our system. On the other hand, if low-cost (~$100 USD) GPSDO units can be found and integrated into the RTL-SDR/Raspberry Pi architecture, the benefits of not having to use calibration beacons may outweigh the time cost of learning to integrate the GPSDO into our design. Another potential drawback is that GPS-based solutions have the potential to draw significantly more current than a beacon-based approach, which may prove problematic for field-deployment.
A remaining issue regarding this GPSDO-vs.-beacon-debate is whether a multi-frequency approach is compatible with beacon-based calibration. In a GPSDO-based approach all distributed oscillators are "disciplined" [i.e. synchronized], meaning that receiving different frequencies is less likely to be a problem than in the case of beacon synchronization. In this latter case, free-running oscillators are not synchronized, but instead calibrated after data collection (i.e. post hoc) WRT time/frequency information provided by known location beacons. If there is any unpredictability in the behavior of each ground-node's LO between frequency hops (potentially requiring that calibration be repeated for each received frequency) then there is little prospect that a beacon-based approach is compatible with our system beyond the first stage of the project (i.e. the single-frequency approach). Such a problem, if it is a valid concern, may force us into either abandoning the multi-frequency approach or into the adoption of GPSDO.
Further discussion regarding this latter problem should be directed to issue #20.
All discussion of the pros and cons of using calibration beacons to achieve phase/frequency/time coherence between distributed ground-nodes should go here. *Note that the use of NTP vs. RTC vs. GPS for coarse timestamp coordination should be directed to issue #19.