Passive Optical LAN is based on the GPON – Gigabit Passive Optical LAN solution. In this article we will go over the most important concepts of GPON in order to understand how this influences the way we design and deploy GPON.
- What Is GPON?
- ODN Structure
- Splitters and Attenuation
- Other Sources of Attenuation
- GPON Specifications
- Data Transmission
- Glossary
What is GPON?
Passive Optical LAN doesn't just use optical fiber, it is based on a widely used technology called GPON or Gigabit Passive Optical Network.
The GPON technology is defined in the following standards:
- ITU-T G.984.1: Gigabit Passive Optical Networks (GPON): general characteristics
- ITU-T G.984.2: Gigabit Passive Optical Networks (GPON): Physical Media Dependent (PMD) layer specifications
- ITU-U G.984.3: Gigabit Passive Optical Networks (GPON): Transmission convergence layer specification
- ITU-U G.984.4: Gigabit Passive Optical Networks (GPON): ONT management and control interface specification (OMCI)
These are the basic standards, ITU-T G.984.5 through G.984.7 are additional standards that are out of the scope of this article.
GPON was originally designed and deployed in the carrier space: access providers connecting residential (and potentially business) users to the central office. The solution is typically called FTTH (Fiber To The Home), FTTU (Fiber To The User) or FTTP (Fiber To The Premise).
More recent GPON is also used by access or internet providers with solutions like FTTC (Fiber To The Curb), where a very short copper loop is still used (which would allow for VDLS2 or even G.Fast) from a modified ONT towards the end-user's house.
In the enterprise sector, GPON is used more and more as a superior alternative to LAN implementation. With proven use cases in many segments: hospitality, hospitals, corporate offices, banks, airports, age care facilities, ...
Below is the high-level overview of what a GPON system looks like, specifically used for a POL scenario.
(the above diagram originates from the Association for Passive Optical LAN)
It contains the following elements:
- OLT: Optical Line Termination: an access multiplexer, aggregating multiple dozens or hundreds of GPON terminations. Vendor specific active equipment.
- ONT: Optical Network Termination (unit): small end-user side device, which terminates GPON on the network side and offers several (typically copper based) end-user interfaces, like Ethernet (RJ45), POTS (RJ11) or Coax. Vendor specific active equipment.
- ODN: Optical Distribution Network: the passive part of the solution, a tree based optical structure, based on the GPON technology.
- The other elements of Core Switch and Workstation will be describe when we come to the POL specific concepts in a future article.
In this article we concentrate on the ODN part of the POL solution.
ODN Structure
The basic structure of GPON is a tree-based (point-to-multipoint) passive topology and consists of the following parts:
- Feeder cable : Single Mode Fiber (SMF) optical fiber, single core, going from the GPON port on the OLT towards a passive splitter
- Passive Splitter: splits the single feeder cabling into several outgoing SMF cores going towards the ONTs. The split ratio (number of outgoing SMF cores downstream) is always a power of 2. We talk about 1:2 splitters all the way up to 1:128 splitters.
- Drop Cables: SMF optical fibers connecting the passive splitter with the ONT
The structure as it is shown in the diagram above, with one feeder cable, splitter and several drop cables towards ONTs, is called a PON port.
 |
We are discussing GPON in a conceptual way in this article. Splitters do not need to be one centralised, they can be cascaded (as long as the total split ratio conforms to the standards).
|
Splitters and attenuation
The traditional procedure to create a splitter one fuses two fibers together and cuts one of the four legs away. This way you end up with a 1:2 splitter. See example below, if we cut away 'Arm 2' you have 'Arm 1' coming in as the feeder cable and 'Arm 3' and 'Arm 4' going downstream towards two ONTs.
You then repeat the process of fusing (hence splitting) two fibers at each of the arms ('Arm 3' and 'Arm 4') to end up with a 1:4 splitter, etc... until the desired split ratio is attained.
This original technique was called Fused Biconic Taper (FBT). These days, splitters are implemented as a Planar Lightwave Circuit (PLC), with the glass waveguides (SMF) attached to a substrate (similar to a circuitboard in modern day computers). But the design still looks similar (cascade of 1:2 splits)
The end result is a solid state (no moving parts), very compact and of a very high quality.
Within the ODN, the splitter represents the most important element of signal attenuation.
With each split (see above: I0 splitting into I1 and I2), half of the optical budget follows one downstream SMF, the other half of the budget the other SMF. This amounts to a loss of 3dB per split.
 |
Only the optical budget is split in two, the data flows through both the SMF leaving the splitter with no loss of data what so ever.
|
In other words, is you have a 1:32 split ratio, which is very common, you have internally in the splitter 5 levels of splits (since 32 == 2^5), and hence the splitter adds 5*3 == 15dB of attenuation.
 |
One special and interesting type of splitter occurs if you have two inputs (two feeder cables in other words coming from OLT to splitter). This special case is shown on the PLC example above.
|
Other sources of attenuation
While the splitter is by far the most important source of attenuation, in order to calculate the so-called optical budget, you will need to be aware of all sources of attenuation:
- connectors : at the very least at the both extreme ends of the fiber (connecting into the OLT and ONT), potentially also at splitter. The amount used for calculations may vary, in our example below we will be using 0.3dB per connector.
- splices: 0.1dB per splice
- ageing of the fiber: 1dB
- attenuation per km: maximum of 0.42dB/km (for upstream)
- WDM coupler (if any is present): can be used either when analogue video is being modulated over GPON or when a mix of GPON and XGS-PON is deployed (Both scenarios are beyond the scope of POL Basics)
- splitters: as mentioned earlier, for each 1:2 split, you have a theoretical loss (attenuation) of 3dB. However, most tools will take a slightly higher number to compensate for any inaccuracies. 3.5dB is a fair number to use.
There are multiple tools available to calculate this, some very sophisticated, some based on elaborate Excel worksheets.
In order to calculate the loss budget in GPON, you need to first know what class of SFP you are using. There are 5 types:
Attenuation Range
| |
|---|---|
CLASS A
|
5 dB – 20 dB
|
CLASS B
|
10 dB – 25 dB
|
CLASS B+
|
13 dB – 28 dB
|
CLASS C
|
15 dB – 30 dB
|
CLASS C+
|
17 dB – 32 dB
|
We only consider Class B+ and Class C+, both used for FTTH and POL deployments.
Let's consider a theoretical example, based on a FTTH deployment.
In the above diagram, we can easily confirm the optical budgets:
- downstream: 1.5 - (-27) - 0.5 = 28 dB
- upstream: 0.5 - (-28) – 0.5 = 28 dB
In our example we consider the following:
- budget: 28 dB (see above)
- 16 way splitter loss: 14 dB (theoretical: 12dB)
- Connector + splicing loss: 3 dB (24*0.1 dB + 2*0.3 dB)
- ageing: 1 dB
- attenuation:
- 0.30 dB/km – downstream
- 0.42 dB/km – upstream
This leads to a maximum distance : (28 – 14 – 1 – 3) / 0.42 = 23.8 km (for a 1:16 splitter)
In POL, the issues is sometimes that there is no enough attenuation since distances are very short (typically within the same building) and designers might calculate, correctly, that a certain floor (which would be connected to the same PON port on the OLT) only needs let's say 16 ONTs (which could still amount to 64 or more Ethernet ports for that floor).
Referring back to our optical budget diagram above, you saw that both the OLT and ONT have a range for the Rx and Tx budget. Consider the maximum Tx power at OLT for downstream and the maximum Rx power level for the ONT.
Further, the ONT will assume a margin of error on it's own measurement of 1dB
This gives us the minimum attenuation: 5 - (-8) – 0.5 + 1 = 13.5dB
Let's then consider:
- high quality splitter, hence we 'only' loose 12dB at the splitter
- short distance (few hundred meter) that will barely give any additional attenuation on the fiber
- short distance probably means fewer splices and connectors
- ageing factor is only a theoretical attenuation and might not be a factor in reality
We might barely get to the minimum attenuation that is required so that the ONT can correctly read the incoming signal from the OLT.
 |
This can easily be avoided by using a 1:32 splitter, even if only 16 ONTs are required. The price difference will be more than worth the trouble of having to check the ODN and potentially install optical attenuators afterwards (let alone the cost of delay in the deployment).
|
GPON Specifications
As you can see in the diagram below, the ODN between OLT and ONT consists of a single core, using a separate wavelength for upstream (ONT to OLT) and downstream (OLT to ONT) traffic.
As you can see on the diagram below, the following wavelengths are defined for standard GPON:
- 1490nm for downstream
- 1310nm for upstream
- 1550nm: a third wavelength is defined when analogue video is to be carried over GPON. We will come back to analogue video in another article.
The typical limit considered for standard GPON deployment is around 20km, although there are techniques available for extending this reach (which we might come back to in a future article).
Obviously this is important in a FTTH scenario but in any POL scenario, 20km is more than sufficient. Even a considerable campus environment will not require more.
To be more accurate, the limit of 20km applies to the difference in distance between ONT and OLT, when comparing the ONT that is closest to the OLT to the one that is furthest away, within the same PON system (which is a specific feeder cable, connected to a PON port on the OLT).
GPON (as defined by the G.984.x standards) was originally designed different rates in mind, however to my knowledge every GPON implementation worldwide has a limit of 2.5Gbps for downstream and 1.25Gbps for upstream. The asynchronous data rates capture the trend that was prevailing until recently that much more downstream capacity is required than upstream.
We will see that the next generation GPON allows for equal capacity (10Gbps) in both direction to accommodate for the recent trend of increased upstream traffic.
 |
GPON standards allow for encryption of the data in the downstream direction (as is considered a broadcast domain). This can be done using the relatively easy and light-weight symmetric encryption keys. See more about symmetric key algorithm here.
|
The maximum number of ONTs you typically can have on one PON port, is 64 in case you are using Class B+ SFPs in the OLT and 128 if you are using Class C+ SFPs.
However, I would strongly recommend not to maximise the number of ONT per PON port in order for example to save on SFPs and or LT cards in the OLT (see article on POL equipment).
All ONTs on the same PON port are competing for the available bandwidth. Even though there is a very flexible bandwidth allocation algorithm behind GPON, you need to be aware of what services are going to be running over the ONTs on a particular PON port. High resolution video services for example do consumer considerable amounts of bandwidth (depending however greatly on whether we are talking about a multicast or unicast video service).
IoT services typically consume a lot less.
We will come back to services and bandwidth requirements in another article.
Data Transmission
GPON allows for different types of data transmissions to be encapsulated:
- GEM – GPON Encapsulation Method: used for Ethernet and TDM traffic
- ATM – Asynchronous Transfer Mode: directly on top of GPON
For POL, we will focus on the Ethernet encapsulation using GEM. We will not go into details on the actual GEM encapsulation (GEM header, payload, …) in this article.
In downstream (OLT towards ONTs) GPON functions as a broadcast domain: the OLT decides which data, destined for which ONT (and Ethernet port on that ONT), will be put in each of the GEM packets. The packets are sent continuously downstream (even if there is no data to send), the ONTs use that continuous flow of packets as a way to time synchronise with the OLT. The passive splitter, as we have seen, has no intelligence of any kind. This means the GEM packet will be duplicated identically over each of the downstream SMF going from the splitter to the ONTs.
In case you are wondering, yes this means all ONTs receive all traffic. However, each compartment inside a GEM packet will have it's own identifier, called the GEM port ID. ONTs filter on GEM port IDs to allow only relevant information into the ONT.
An individual GEM Port ID defines not only the ONT or the Ethernet port on the ONT, but even the priority queue the packet is destined for.
An individual GEM Port ID defines not only the ONT or the Ethernet port on the ONT, but even the priority queue the packet is destined for.
Security in downstream: People often wonder just how secure the data transmission is in downstream due to the broadcasting nature of GPON. There are three main reasons why this is considered very secure:
- Symmetric encryption: mentioned earlier, a symmetric key is chosen by the ONT and shared with the OLT at regular intervals (depending on the manufacturer's implementation). Typically key length would be 128bits.
- GEM port ID filtered before Ethernet ports: as mentioned the ONT will have Ethernet ports (and possibly other types of interfaces) towards the end-user. The end-user however, if it would decide to use a tool like Ethereal or Wireshark, cannot see any data not destined for the ONT as the filtering is done on the network side of the ONT and the Ethernet interface is at the end-user side.
- ONT typically does not have management interface: the ONT is depending on the OLT to receive configuration and even software updates. There is no management interface on ONT for an end-user to exploit.
In upstream there are multiple sources (ONTs) and a passive splitter that does nothing but let the optical signal through.
Since we are still carrying the data in a GEM packet, each ONT that has something to send will occupy a specific area of the GEM packet.
For this reason, there is the need for a very accurate synchronisation between the ONTs so that they send out their signal at the right time in order for the different areas to get to the splitter at the correct time so that the resulting merger of all the incoming signals results in a correctly formatted GEM packet the OLT can read.
In the above scenario, we have 2 ONTs connected, at an equal distance of say 20km from the OLT (a scenario related to FTTH not POL). The first ONT (above) send a number of Bytes of data in order to occupy the first segment of the GEM packet, the second ONT will occupy the second segment.
In the below scenario, the 2 ONTs are at different distances from the OLT: ONT 1 at 20km and ONT 2 at 15km. Similar to the first scenario, ONT 1 will occupy the first segment of the GEM packet, ONT 2 will occupy the second segment of the GEM packet. However, since ONT 2 is closer to the splitter, it's data will arrive slightly earlier and in the merged GEM packet, it will overlap with the first segment (the date of ONT 1). As a result, both segments are lost.
For this reason, when an ONT is attached to a GPON (POL) system the OLT will initiate a ranging mechanism, to determine at what distance the ONT is.
In order to determine the distance between OLT and ONT, the ONT when starting up will not initiate communication with the OLT but awaits for the OLT to contact the ONT by sending a Ranging-Grant message. The ONT, when receiving the grant message, registers the timestamp and when the grant message is processed, the ONT sends out a Ranging-Grant-Acknowledgement message with the time difference (or Δt) between entering the ONT and leaving (which will be in the order of microseconds or μs).
The OLT can calculate the transition time (the time it took the Ranging-Grant message) to get to the ONT, but taking the difference between the time the message left the OLT and the time the response (Acknowledgement) arrived, minus the ONT processing time (Δt) and divided by 2.
The distance itself is then calculated by multiplying the time in seconds with the speed of light in fiber, or CFiber.
As an example, let's assume the transition time for the Ranging-Grant message is 75μs, since the speed of light through fiber is 200,000 km/s that gives us a distance of 15km.
In the example with ONT 1 at 20km of the OLT and ONT 2 at 15km of the OLT, you could synchronise the two by adding a delay of 25μs so that both ONTs seem to be at 20km. However that would make the delay of one ONT dependent on the distance of another.
In order to use the ranging information to synchronise all the ONTs in a way that the ONTs are not interdependent (in other words: that ONTs do not need to re-synchronise when one of them is either switched on or off), all ONTs receive a transition delay that places them on a virtual distance of 60km.
 |
Referring again to an earlier limit of GPON, the transition delay and ranging is why there cannot be more than 20km difference between the ONT closest to the OLT and the one furthest away from the OLT.
GPON sends 8000 frames per second, each GPON frame represents 125μs of data. Light travels roughly 25km in 125μs. If we apply an error margin of 20%, if two ONTs would have a difference in distance to the OLT of more than 20km, even with a transition delay applied, the data would simply not arrive in the same GPON frame and it would be impossible to synchronise the ONTs.
|
Glossary
- FTTH: Fiber To The Home
- GEM: GPON Encapsulation Method
- GPON : Gigabit Passive Optical Network
- ODN: Optical Distribution Network
- OLT: Optical Line Termination (unit)
- OMCI: ONT Management Control Interface
- ONT: Optical Network Termination (unit)
- POTS: Plain Old Telephone Service
- SFP: Small Form-factor Pluggable
