G.fast: Time Domain Duplex (TDD) vs Frequency Domain Duplex (FDD)

In September 2015, the UNH-IOL opened a new consortium in the Broadband testing services called G.fast. G.fast is an emerging technology which leverages existing telephone copper networks, similar to DSL, to deliver incredibly high bandwidth. Optical networks are gaining popularity due to their high throughput capacity. Despite this, fibre is still not being integrated into residential homes due to the high costs involved as well as keeping the more fragile glass fibers protected. G.fast is the solution; offering close to fibre speeds without installing fiber into the home. Fibre networks are placed between internet service providers and neighborhoods or buildings, then G.fast is used to link homes to the nearby fibre. It boasts speeds of up to 1Gbps and has the potential to develop beyond that speed.

G.fast is often confused as being a newer form of the DSL technology. Although they have similarities in physical layer functionality, they are very different technologies. A key feature that sets them apart is Time Domain Duplex (TDD) versus Frequency Domain Duplex (FDD).

Consider DSL to start, which has evolved over the past years from ADSL (8 Mbps) through VDSL2 Vectoring (200 Mbps). DSL modulates its data communications as upstream (uploaded data) and downstream (downloaded data) and will use two (or more) separate frequency bands to send these bits at the same time, defined as FDD. This affects the speed of traffic and the amount of bits that can be sent physically over DSL.

TDD is another common scheme to support duplex data transport. It’s often described as a “phone conversation,” where each node takes turns sending data, but neither “talk” at the same time. On the phone, you’ll say “hello” and listen for the person you’re speaking to to say “Hello” back. And the conversation continues as such;, listening and then speaking when it’s your turn. This is how G.fast supports upstream and downstream data transfer.

G.fast faces a challenge with this mechanism since the DPU (Distribution Point Unit) and the CPE’s have to accurately control the time in which they transfer data. One of the DPU’s functions is to support up to 64 lines and has to manage the time each line sends its data via TDD. Similarly, CPE’s have to manage how they individually send upstream symbols in time. When deployed residentially or commercially the DPU and CPE have to be prepared for real-world instances of improper shutdown and that’s where TDD can be most affected, especially since the lines are also simultaneously using a function called vectoring.

G.fast lines are grouped together in a binder which causes crosstalk because the copper pairs are grouped so close together. This makes it easy for the signal to jump from one line to another, appearing as noise. Vectoring is a mechanism used in DSL to cancel out crosstalk by sending symbols down the line that already “seem” affected by crosstalk. It then uses the crosstalk on the line to their advantage by allowing it to affect the symbol as it would, yet making the symbol normal and ready to receive. An oversimplified comparison to vectoring is noise cancelling headphones - where microphones on the outside of the headphones capture the environmental sounds and turn them 180 degrees out of phase so that when it plays along with your music, it is totally cancelled out.

Vectoring is so important in the DSL and G.fast world because it keeps speeds on the copper line fast and transmission less corrupted. Since broadband is expanding into higher frequency technologies like G.fast, the lines suffer from increasing crosstalk, making vectoring more effective. When combined with TDD, time syncing is so sensitive that if it were to mess up only slightly, it would ruin the transmission of the symbols between multiple lines.

The DPU will send symbols downstream on multiple lines, no matter the length, at the same time to keep them synced in time. The challenge with TDD and vectoring comes with sending symbols in sync, upstream on lines with different lengths. To solve this challenge the CPE will learn the time it takes for each of the loops to transmit / receive their symbols during initialization, using that time to delay transmitting symbols on shorter loops, allowing longer loops to “catch up”.  This ensures the symbols are in sync across each of the lines. This time offset is known as ‘Tg1’ and is a mandatory function for both the CPE and DPU

When power is shut off the challenge to relearn that time delay is harder. If the CPE is manually shut down it has to relearn that time. If its power is completely cut (like losing power in your house and therefore to the CPE), it will let out a ‘dying gasp’. The point of the dying gasp is to send small bit data down the line to the DPU, this allows the DPU to quickly make the required changes in the vectoring calculations to account for the “missing” CPE.

Among other things, this is just one example of the challenges vendors will face when implementing a G.fast chipset, CPE and/or DPU as well as switching from FDD to TDD. G.fast tries to push the boundaries of data transmission over a copper line, meeting demands of the market and taking Broadband back into competition with fibre. G.fast also refreshes DSL technology with functions and mechanisms that make it more adaptable to the growing needs of the consumer.