MF-TDMA or Dynamic BoD BM-FDMA: what jitter really costs

Every shared-inbound satellite system has to answer the same question: how do many remote terminals transmit into one piece of spectrum without colliding? The two common answers produce networks that feel completely different to use, even when the datasheet throughput looks similar.

MF-TDMA: share the time

In multi-frequency time division multiple access, the inbound spectrum is cut into channels with predefined centre frequencies and fixed bandwidths. Each channel is then divided into time slots, and several terminals are assigned to the same channel. A terminal may only transmit during its own slots.

That design has three consequences that follow directly from the mechanism.

• Variable delay is built in. A packet arriving just after the terminal’s slot has closed waits for the next one. The wait varies, and variable delay is jitter.
• Everything must be synchronised. All terminals need a common time reference accurate enough that bursts do not overlap, which adds complexity at every site.
• The outdoor unit is sized for the widest channel. Because any terminal may be assigned to the widest channel in the system, every terminal needs an amplifier and antenna capable of it — including the small site that will never need the capacity.

Dynamic BoD BM-FDMA: share the spectrum

Bandwidth on demand with burst mode frequency division multiple access takes the opposite approach. Multichannel demodulators process the whole inbound band at once. When a terminal asks for capacity, the system allocates it a dedicated channel, with a centre frequency and a width chosen for that request, from anywhere in the band.

The terminal transmits continuously on its own channel for as long as it needs it. There is no slot to wait for, so packet transfer is jitter free. No common time reference is required. And because each terminal only ever transmits at the rate assigned to it, its outdoor unit can be sized for that rate rather than for the busiest site in the network.

Which applications notice

Average throughput hides this difference; the applications do not.

• Voice degrades with jitter before it degrades with bandwidth. Jitter buffers hide it by adding delay, which is the thing voice can least afford.
• Videoconferencing shows it as freezing and re-syncing rather than as lower resolution.
• Live contribution video is the least forgiving: a transport stream with variable arrival times needs buffering that a live broadcast has no time for.
• Industrial telemetry and SCADA often assume bounded latency. Unbounded jitter shows up as spurious timeouts, not as slowness.
• Bulk file transfer and web browsing genuinely do not care. If that is the whole traffic profile, this choice matters much less.

Reading a datasheet honestly

Three questions separate marketing from mechanism. Is the inbound channel width fixed or allocated per request? Do terminals require time synchronisation? Must every outdoor unit support the maximum system rate, or is it sized per site? The answers tell you what the network will feel like long before a field trial does.

Our EvmarSAT system uses Dynamic BoD BM-FDMA for exactly these reasons, and the page sets the two technologies side by side in detail.