If you haven’t noticed already, I have something of a thing for everything time synchronization related. “Proper”, exact time is something that I find important, it’s probably OCD related.
As a part of my day job, I manage a quite large network, with hundreds of different devices. And all of these require some method to synchronize their clocks. Anyone who has at some point tried to diagnose a network problem knows how absolutely critical it is to have proper time available at the devices, so that you for example have any chance to compare logs etc.
Enter time servers! Time servers are a dedicated device of some sort, that either produces reliable time by itself (these master clocks are few and far between, think nation states as the owners and you might get an idea of what they cost), or synchronize time from some upstream device. They then also forward this accurate time to some downstream device of some sort (this would be the server part). There are many different ways of doing this, but for this example well look at the most common version, a NTP time-server with a GPS receiver. You might be thinking now, why would we want GPS, we need accurate time, not position!?! Well, GPS relies on accurate time to calculate where you are in relation to the GPS satellites, and use that distance to triangulate your position. As an effect of this, the satellites all contain a very accurate clock, called an atomic clock, and they send the time of this internal clock to anybody that bothers to listen. This basically mean that we can get very accurate time from a system composed of several satellites, as long as we have a receiver.
This article’s focus is on this GPS receiver, the time source in our time servers. Many network technicians are familiar with the inner workings of NTP, which is the method that we use to deliver time to client devices, but most haven’t come up close and personal with GPS technology. It’s an old system, and as the satellites are deployed in space, it hasn’t changed much since it was first introduced. This can make it a bit unfriendly to beginners, and this is what I aim to help with in this article.
Proper GPS antenna placement
Since GPS is a bit of an esoteric system, you actually need to know quite a bit of the GPS system to place your antenna correctly, and most manufacturers don’t really do a good job of explaining why.
GPS satellites orbit the earth mostly around the equator, approximately N 55 deg to S 55 deg. This means that, as you get further away from the equator, you need to pay more and more attention to how your antenna is placed. The optimal scenario everywhere is to have the antenna placed so that it has an unobstructed, 360 degree view of the sky, but in my experience this is actually quite hard to achieve, especially up north here in Finland, where we have snow and ice to worry about as well. Meinberg has done some interesting tests on this and actually proved that you need to be 11 meters from the nearest wall to achieve perfect coverage.
The next best thing is mounting it to a wall, and this is where you can get into a bit of trouble. Even though the satellites pass “over you”, the antenna will work better if it’s placed facing the equator. This means, that the further away you get from the equator, the more you can get away with “improper” wall mounting, as long as you mount it on a wall facing the equator (south wall in the northern hemisphere and vice versa in the southern hemisphere), as the satellites aren’t actually overhead, but more towards the horizon.
Here are examples of what the receiver sees depending on the mounting choice. Luckily for my readers, I can be a complete idiot at sometimes, and as such I’ve actually managed to mount a receiver on the wrong side of a building (guess why I did all the research for this article in the first place).
What you see here is the trajectories of the satellites that the receiver has “seen” over time. The orientation of the antenna here is for Finland, reverse the directions if you’re in Cape Town for example.
Firstly, this antenna is mounted on the south side of a building, with clear view of the sky above.
This antenna on the other hand is mounted on the north side, aka. the wrong wall, with a outcroping above it, further limiting its visibility of the sky.
This one is also mounted on a north-facing wall, this time with communications tower next to it as well, but the out-cropping is smaller and it’s mounted in the north-eastern corner of the building, so the impact isn’t as severe.
As you can see, in the first picture, the antenna seems to have a nice, 360 degree view of the satellites, and can track them with ease. In the other two, you can see how effectively the building “shades” the receivers satellite reception.
These are quite extreme examples, but you can use this tool (built into every Meinberg device) to diagnose reception problems that aren’t as obvious.
Now, the hard question, did I move the antennas? Actually, the antennas are still there, in the “wrong” spot. Even though the receivers have “poor” reception, they don’t actually need to see 12 satellites at once. The reason is that GPS was designed to pinpoint location, not give accurate time, and to do so it need many different satellite signals to be able to calculate your location. Time servers on the other hand only need a couple of satellites (remember, every satellite contains that coveted atomic clock with near-perfect time) to get an accurate time to sync its internal clock to. There is one caveat however, getting that initial lock on the satellites. The more satellites the receivers sees, the faster it’s receiver can get a lock on the time signal. You might think, well, it doesn’t matter, but remember, GPS is an old system. I’ve seen receivers that have spent half a day locking on to satellites, and until it can do that, it’ll be free-running, aka. without any reference clock to sync to.
I’ll go more into GPS troubleshooting in the next instalment of this guide!