4 ms·

All modern digital radio networks (cellular, police/fire, etc.) rely on GPS signals to discipline high-stability frequency oscillators. If you look at any cellu

All modern digital radio networks (cellular, police/fire, etc.) rely on GPS signals to discipline high-stability frequency oscillators. If you look at any cellular antenna you can always see the white mushroom-shaped GPS antennas, usually lower on the tower or sometimes just sticking up off the equipment building. Those oscillators are used as references to generate the transmit/receive frequencies and also for very tight timing between separate transmit locations (many digital networks actually broadcast the same signal on the same frequency from multiple places at the same time). Without GPS, these reference oscillators can free-run for a while but they will eventually drift out of tolerance. Basically, without GPS, most/all cellular/digital radio networks will fail in some manner after a while. Precise failure modes will depend on a lot of things including exactly what is wrong with GPS and the type of emission, how the base station software is written, etc.

tzs 7y ago
Do they discipline by essentially being a full GPS receiver that decodes the timing information from multiple satellites, derives the time, and then adjusts the local oscillator so it is running at the right rate based on that time? Or do they do something simpler, and just derive a frequency reference from the frequency of the carrier signals and sync the local to some fixed ratio of the carrier frequency? The former would be more accurate, but would break if something went wrong with the data. The latter would probably not be as accurate, due to Doppler effect (I'm getting +/-0.00041% shift from Doppler effect), but would keep working as long as the carrier is being broadcast. You could probably correct for a good part of the Doppler effect just from the last known orbital parameters of the satellite--even if the data updates have been broken for a long time that should work.
- makomk 7y ago
  GPS-disciplined oscillators are based around full GPS receivers that decode the GPS signals and compute time from them in the usual way, then use that to steer the local oscillator. I think the better ones have a fixed-position mode which makes use of the fact the ground station doesn't move to get slightly more accurate timing, but otherwise they're basically the same as any other GPS receiver.
- davidf560 7y ago
  Sorry, late response but... It's the first - the GPS modules used for these purposes decode the time and then generate a digital pulse on an output pin at the top of each second (this is called a 1PPS output). A hardware counter counts the cycles from a local oscillator between pulses. The count between pulses is latched (this is all in hardware so far) and then the software reads the latched value. If your local reference oscillator is supposed to be 10MHz, your counter should report 10,000,000 ticks exactly. If it's high or low, the software can adjust the temperature of a little mini-oven that the oscillator resides in to speed it up or slow it down (some oscillators just use a voltage to speed up or slow down instead of an oven). Usually this is done in a PID loop or similar feedback loop. At this stage you can apply some filtering logic to help ride out certain errors (if you're suddenly off by 50% something else is probably wrong, etc.). To ensure the GPS timing is as accurate as possible, you can have the GPS module survey its position for a lengthy period of time (since it is presumably mounted on an antenna mast somewhere and not moving) and average it out to establish a more accurate well-known location. The GPS can then use that known location to apply a correction to the GPS signals. Basically, if it's currently-computed position based on GPS signals is 50 meters away from the true well-known location of the receiver, it can figure out how far off the timing computation is and adjust. This is mostly handled by the GPS module itself, and is probably even more automated and accurate now than when I worked on it ~15 years ago. There's also ways to increase GPS accuracy using supplementary broadcasts that contain localized offset correction information, referred to as Differential GPS (DGPS) run by many groups including the US Coast Guard and a similar technology called WAAS (wide area augmentation system) run by the US FAA. The GPS we were using didn't have the capability to use those sources but modern units probably do.
lxgr 7y ago
This is true for S-CDMA networks (a.k.a. CDMA2000, primarily used in the US), but definitely not for the internationally more common asynchronous CDMA based UMTS (a.k.a. 3G or WCDMA) systems. I'm not too sure about LTE, but some initial googling suggests that while it does need a precision frequency and time source, there are ways of deriving that from the network, which makes sense if you think about indoor base stations that often don't have a clear view of the sky.

tzs 7y ago

Do they discipline by essentially being a full GPS receiver that decodes the timing information from multiple satellites, derives the time, and then adjusts the local oscillator so it is running at the right rate based on that time? Or do they do something simpler, and just derive a frequency reference from the frequency of the carrier signals and sync the local to some fixed ratio of the carrier frequency? The former would be more accurate, but would break if something went wrong with the data. The latter would probably not be as accurate, due to Doppler effect (I'm getting +/-0.00041% shift from Doppler effect), but would keep working as long as the carrier is being broadcast. You could probably correct for a good part of the Doppler effect just from the last known orbital parameters of the satellite--even if the data updates have been broken for a long time that should work.

makomk 7y ago
GPS-disciplined oscillators are based around full GPS receivers that decode the GPS signals and compute time from them in the usual way, then use that to steer the local oscillator. I think the better ones have a fixed-position mode which makes use of the fact the ground station doesn't move to get slightly more accurate timing, but otherwise they're basically the same as any other GPS receiver.
davidf560 7y ago
Sorry, late response but... It's the first - the GPS modules used for these purposes decode the time and then generate a digital pulse on an output pin at the top of each second (this is called a 1PPS output). A hardware counter counts the cycles from a local oscillator between pulses. The count between pulses is latched (this is all in hardware so far) and then the software reads the latched value. If your local reference oscillator is supposed to be 10MHz, your counter should report 10,000,000 ticks exactly. If it's high or low, the software can adjust the temperature of a little mini-oven that the oscillator resides in to speed it up or slow it down (some oscillators just use a voltage to speed up or slow down instead of an oven). Usually this is done in a PID loop or similar feedback loop. At this stage you can apply some filtering logic to help ride out certain errors (if you're suddenly off by 50% something else is probably wrong, etc.). To ensure the GPS timing is as accurate as possible, you can have the GPS module survey its position for a lengthy period of time (since it is presumably mounted on an antenna mast somewhere and not moving) and average it out to establish a more accurate well-known location. The GPS can then use that known location to apply a correction to the GPS signals. Basically, if it's currently-computed position based on GPS signals is 50 meters away from the true well-known location of the receiver, it can figure out how far off the timing computation is and adjust. This is mostly handled by the GPS module itself, and is probably even more automated and accurate now than when I worked on it ~15 years ago. There's also ways to increase GPS accuracy using supplementary broadcasts that contain localized offset correction information, referred to as Differential GPS (DGPS) run by many groups including the US Coast Guard and a similar technology called WAAS (wide area augmentation system) run by the US FAA. The GPS we were using didn't have the capability to use those sources but modern units probably do.

lxgr 7y ago

This is true for S-CDMA networks (a.k.a. CDMA2000, primarily used in the US), but definitely not for the internationally more common asynchronous CDMA based UMTS (a.k.a. 3G or WCDMA) systems. I'm not too sure about LTE, but some initial googling suggests that while it does need a precision frequency and time source, there are ways of deriving that from the network, which makes sense if you think about indoor base stations that often don't have a clear view of the sky.