CableLabs sends its time lords to help small-cell mobile nets When you need parts-per-billion frequency accuracy, “Let's synchronise our watches” doesn't cut it. Take LTE and 5G for example: they need tight synchronisation, both in frequency and phase, and that makes time-signalling an important part of the network. CableLabs, the US cable industry's research arm, is proposing a …
Why does it need such tight timing? I mean, frequencies, yes, but why inter-base-station timing like that?
Surely any connected device (e.g. a dumb GSM chip) nowadays is operating in the nano second ranges, and there are handshakes and regular updates to the clients... so the clients aren't going to get out of sync with the base-station. And base-stations are presumably operating entirely independently and on different frequencies and handshaking to one another.
So why does one base station have to be so perfectly synchronised to the next within the microsecond ranges for what it basically a large wireless network on reserved frequencies? We don't have that for any other wireless technology that uses all the same tricks and phase-shifting and all sorts as anything that 5G will do.
"The R&D team's distinguished technologist for wireless Jennifer Andreoli-Fang said the rise of microcells poses problems for the of GPS as a synchronisation standard: .... ***and the devices are often deployed indoors where they can't see the satellites.***"
So if they're not seeing the satellites at the moment, what does 5G need them to see satellites for, and why in fact does 5G say that seeing the satellites perfectly isn't even sufficient? I'm confused. I can't imagine any signalling scenario on this kind of level where you need perfect timing synchronisation between two stations which couldn't work it out on their own as part of the signalling handshaking process and maintain it all the while they are in contact and re-handshake if recovering from a fault.
And, hell, surely 5G should just be nothing but protocols-over-IP by now? Haven't we learned yet?
For the higher throughput 4G and 5G situations the user device will often be receiving multiple streams of data from multiple sites.
Also in order to improve throughout at the edge of a cell's coverage it means that the interference from adjacent sites must be managed and coordinated. With LTE's OFDM structure it can mean that other sites will not transmit in certain time or frequency subgroups so that the air is clear for the mobile to hear from it's serving site, thus requiring incredibly tight timing to pull it off.
A macro cell - a single base station covering a decent sized population centre - might be talking to, what? a thousand or more handsets simultaneously. Each one needs to it's own specific frequency band that doesn't overspill into another ones (FDM). Each base-station needs to be sync'd so handsets can seamlessly handover from one to the other on the move (otherwise you get dropped calls. Telcos take anything that affects user perception as quite a serious performance metric).
Small cells (those covering small areas like shopping centres or stadia) are being rolled out more, to a. improve coverage in not-spots, and b. to improve the use of spectrum as 4G and 5G have higher bandwidth requirements. This increases the rate of cell handovers even more. All this requires tight, nano-second timing.
DAB, DTT and Mobile use SFN that need tight timing. For years they have stupidly been using GPS because it's actually very cheap than alternative. External aerials then and coax to pipe the 1.4GHz signal to receivers is possible.
However it's really stupid, because then you have a single point of failure for multiple terrestrial networks that can be turned off by a single person in a foreign country or go off line due to a solar flare or be locally jammed by vandals.
I know I shouldn’t feed the trolls but;
A. Yes I do.
B. Yes I do.
And C. Er, Yes I do.
See, for example, https://www.microsemi.com/existing-parts/parts/138178. If you look at a big nationwide telcoike BT or DT, they will have two or even possibly more in their network that all frequency synchronised nodes will be traceable back to. Source: I can put my hands on one right now.
"I know I shouldn’t feed the trolls but;
A. Yes I do.
B. Yes I do.
And C. Er, Yes I do."
As someone who can also answer 'yes' to A and B, I have also worked in a telco that had a pair of Cesium clocks at the top of its clock distribution - this was fairly common back in the day.
I know that the telco I worked for changed its top level clocks to GPS sourced some years ago (selling the Cesium clocks back to the manufacturer) - whether this is a good long term strategy or not I'm not sure.
Clock distribution should be a fairly serious subject in any telco worthy of the name and the clock distribution network must be treated as a separate design and build exercise to the actual traffic carrying, revenue earning, network. Clocking loops and/or avoidable clocking degradation are bad things.
There are stock traders with Cesium atomic standards in their networks because high frequency trading requires such a level of accuracy that nothing less is sufficient.
Hell, there are two atomic frequency standards in my home workshop, precisely because telcos use them in almost every cell base station and they replace them regularly, selling the old gear off for scrap which eventually makes its way onto eBay for cheap, you can pick one up for a little over £100, Cesium standards occasionally turn up from the same telcos and I've seen them sell for a shade over £300 at the cheapest.
There have even been Hydrogen MASERs sold off and bought by hobbyists, rich hobbyists admittedly because they're expensive to run, you need hydrogen rated plumbing, a lab grade supply of hydrogen, a lot of power, a temperature and pressure stable room which takes a lot of space and backup power because (like all frequency standards) they take time to come up to temperature and long term stability so you don't want to have to switch them off and on again too often.
But, if you want a cheap frequency/time reference that's potentially as accurate as a Cesium reference then you can't get much better value than a ten quid GPS receiver with an outdoor antenna.
Atomic 'clock' is a pop science term used to describe an atomic frequncy standard, electrons don't know when it's pub 'o' clock (more fool them) , they just have a very precise energy state transition period which can be used to measure time.
There are now single chip (well more like a crystal oscillator sized module) with an itty bitty atomic clock inside. Its not quite as good as a proper hydrogen maser or rubidium clock as it uses Cs and a vertical cavity surface emitting laser aka VCSEL as its light absorption source costing about £1.3K each.
The actual lamp itself isn't much larger than a grain of wheat, and advances in materials technology mean its lifetime is measured in decades even if run 24/7/365.25 due to the smaller size and solid state processors/RF drivers/etc.
The newer units are so accurate that if you used two of them, and went up a high rise building there is enough of a difference to detect in a few weeks from relativistic effects under the weaker gravity.
I have heard that they are new enough not to show up used on Ebay (yet!) but its only a matter of time, pun intended.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
