GPON Technology

Hello, everyone.

I’m not sure if this is the right place to ask so let me know if it isn’t!

I work at an ISP space and I encountered a technology called GPON. I understand that the principle is the following:

  1. The router installed at the customers’ premises is called the ONT
  2. The device inside the ISP’s space which connects the infrastructure is called the OLT
  3. The principle works in a way where a point-to-multipoint topology is built. If a customer watches YouTube or something, it will send the packets through the GPON network to the OLT.
  4. Once the OLT replies, a splitter splits this signal and sends it to every ONT connected to the GPON network (and they use some form of identification to determine whether the traffic is for them)

My question is about the 4th point. Are responses from the OLT the only thing that is split to all customers? To clarify this point a bit more, this is the traffic flow form the OLT to the ONTs.

If one of the ONTs opens up YouTube, is the traffic also split or only sent to the OLT?

And one final question, if the OLT really does send something to all customers, what if there is a customer that is, say downloading a big file? Wouldn’t this saturate the links of the other customers, considering that everything is broadcasted there?

Thank you
David

Hello David

Passive Optical Networks, or PONs, are networks that allow you to use passive splitting to provide multiple connections. GPON is Gigabit PON, providing speeds on the order of several Gigabits per second.

Now the passive splitting means that there are no active electronics or powered devices at the “splitter” location in the diagram. The splitter performs only optical splitting of the light pulses.

Now the splitter works in both directions. From OLT to ONT, the signal is split and sent to all ONTs, and from ONT to OLT, the splitter actually combines the signals from multiple ONTs onto a single fiber.

No. ANY traffic from OLT–> ONT is split, and any traffic from ONT–>OLT is combined.

  • Traffic sent from OLT–> ONT is actually received by all ONTs, BUT, each ONT only reads the traffic that is encrypted/identified for itself.
  • Traffic sent from ONT–>OLT is NOT received by all ONTs, but only by the OLT. This is because upstream traffic is sent using Time Division Multiple Access (TDMA), so each ONT will transmit only during its own allocated time window. So other ONTs never see each other’s upstream traffic.

The total bandwidth available for all users on a single GPON OLT port is typically 2.488Gbps downlink, and 1.244 Gbps uplink. All ONTs on the port share this (typically 32 to 64 customers). The OLT intelligently manages how that bandwidth is shared, and it has specific parameters in place to ensure one user doesn’t cause problems for other users.

Each customer has a specific SLA, and users typically have 100, 200, or 300 Mbps based on their plan (or some other maximum speed). It uses specific algorithms (that are configurable) to guarantee minimum bandwidths, and enforce maximum limits. So you won’t always be able to reach your maximum purchased speed, but the bandwidth is shared generally fairly.

PON is a really innovative technology, and most FTTH applications in most of the world typically use this architecture, because passive splitters are cheaper than active devices, and you typically use only a single fiber strand rather than two.

Your job sounds very interesting!

I hope this has been helpful!

Laz

Hello Laz.

Thank you so much. I need some further clarification regarding certain points.

Traffic sent from ONT–>OLT is NOT received by all ONTs, but only by the OLT. This is because upstream traffic is sent using Time Division Multiple Access (TDMA), so each ONT will transmit only during its own allocated time window. So other ONTs never see each other’s upstream traffic.

Two questions regarding this.

If the black ONT at the top transmits something, will the splitter split this traffic and also send it to the other ONTs? In other words, is the splitter bidirectional? Or does it only send traffic to the OLT in this case?

This is because upstream traffic is sent using Time Division Multiple Access (TDMA), so each ONT will transmit only during its own allocated time window.

My second question is, TDMA assures that at any given time, only one ONT will transmit?

Why does GPON specifically rely on TDMA upstream? Wouldn’t just using WDM upstream solve this problem? If multiple customers transmitted something at the same time, wouldn’t the splitter be able to multiplex the signal and divide them based off their colour/wavelength so they can be easily interpreted at the OLT’s side?

//Edit: Turns out that it does exist and is called a WDM-based PON

WDM-based PONs allocate a separate wavelength for each individual subscriber, thus creating a point-to-point channel between the subscriber and the central office.

So with a WDM-based PON, each subscriber needs a unique wavelength (so they can be put on the same optical line and differentiated between). This is pretty much also defined when talking about WDM in general → each optical signal has a unique wavelength. Do we know how is this unique wavelength ensured? Do we have to configure anything or is it automatic?

Then with a GPON that uses TDMA, each subscriber can share the same wavelength but never transmit at the same time?

Thank you.

David

Hello David

Good question, and yes my answer needs clarification. The passive optical splitter is bidirectional. Physically, the splitter is a simple, passive optical component with no intelligence. When an ONT (like your black ONT at the top) transmits upstream at 1310 nm, that optical signal is physically divided among ALL ports of the splitter including the fiber to the OLT and the fibers to other ONTs. The splitter cannot selectively direct traffic, it simply splits optical power proportionally.

However, other ONTs never “see” this upstream traffic because GPON typically uses wavelength separation. Upstream traffic uses 1310 nm (1260-1360 nm range), while downstream uses 1490 nm for data and optionally 1550 nm for video. Each ONT has a diplexer that blocks 1310 nm from the ONT’s receiver path because ONTs are designed only to receive at downstream wavelengths. Only the OLT has a receiver tuned to 1310 nm, so only the OLT can decode upstream traffic.

So while upstream light physically propagates to all splitter ports, only the OLT can receive it.

Yes, exactly. TDMA’s entire purpose is to ensure collision-free upstream transmission. The OLT acts as the central coordinator and performs “ranging” to measure the round-trip delay to each ONT. Using this timing information, the OLT assigns each ONT a specific time slot in which to transmit. These slots are communicated in special synchronizing frames called GTC frames, or GPON Transmission Convergence frames. The OLT calculates delays so that transmissions from ONTs at different distances arrive in their scheduled windows without overlap.

So, under normal operation, only one ONT transmits at any instant on the shared 1310 nm upstream wavelength.

As you have already discovered, there are various types of GPON, including standard GPON that uses TDMA for upstream, and WDM-based PON that uses WDM. Standard GPON deliberately chooses TDMA for this purpose. There are cost advantages to using TDMA which include:

  • With TDMA, every ONT uses an identical, inexpensive laser transmitting at 1310 nm. ONTs are completely interchangeable.
  • With per-ONT WDM, each ONT would need a laser tuned to a specific, unique wavelength. This requires more complex electronics on the ONT that can tune to the right wavelength.
  • The OLT would need multiple receivers or wavelength-selective demultiplexing equipment

So the use of TDMA is more aligned wth PON’s philosophy of ultra low cost.

Your description of WDM-based PON is accurate however.

I hope this has been helpful!

Laz

Thank you!

What if we have a switch that has 3 optical connections connected to it and this switch is connected to another switch by a single optical connection?

Optika means “Fiber”, it’s just translated to my language :smiley:

SW1 will have to perform WDM, right? Does this mean that the three connections going to SW1 need to have their laser tuned to a specific unique wavelength?

David

Hello David

In your scenario, you are using switches. Switches have active transceivers that terminate the optical communication, so you wouldn’t need to use WDM at all to separate traffic. YOu could use the same wavelength on all links. Remember, WDM is used to separate data streams over the same physical optical strand. In your scenario, you have separate pairs of strands dealing with each connection, so WDM is unnecessary.

Could you mean splitters instead of switches? Can you rephrase your question so we can address it more correctly? Thanks so much!

I hope this has been helpful!

Laz

Hello Laz.

I might be getting a bit confused at where specifically is WDM used. If SW1 above did receive three separate data streams and had to transmit them over the optical link to SW2, it wouldn’t need to use WDM? Does it only apply to splitters? Or how would exactly SW2 upon receiving the traffic be able to differentiate between those 3 streams on the same optical link?

I originally thought that SW1 would use WDM and SW2 would be able to differentiate the three data streams by putting together their colours.

Not sure if my question is any clearer now, this part has me scratching my head.

David

Hello David

I understand the confusion. WDM is a technology that operates at Layer 1 of the OSI model: the physical layer. It is the method by which data (bits) is encoded onto an optical fiber using different wavelengths of light. This functionality is handled by the optical transceivers connected to a switch. In your example, each transceiver, and therefore each fiber link, functions independently, and the WDM characteristics of one link do not affect another.

WDM becomes relevant when you want to multiplex multiple data streams over a single fiber. In such cases, the transceivers at each end encode traffic onto different wavelengths so several streams can coexist on the same physical strand without interfering with each other.

Technologies like GPON use this concept to multiplex multiple channels at the physical layer, enabling bidirectional communication and multi-user access over a single fiber. This works in combination with passive optical splitters, which divide the light signals so that multiple ONTs can communicate through the same physical infrastructure.

Switches, however, do not perform multiplexing at the physical layer. By design, they operate at Layer 2, managing frames from multiple sources while remaining completely unaware of the underlying optical encoding taking place at Layer 1.

I hope this has been helpful!

Laz

Hello Laz

Thank you for taking your time to discuss all of this with me as the topic can get quite complex. Glad we managed to sort it out!

David

1 Like