UNB stands for Ultra Narrow Band and it is one of LTN radio access technology defined in ETSI GS LTN 003. It is not described in very detail in this spec. For now, following is the only description in terms of PHY described in LTN 003. I will keep updating as find more detailed information.

UNB is built mainly to carry uplink messages from a large number of LTN End Points (LEPs) to LTN Access Points (LAPs). So this page takes the uplink radio first and the downlink radio second. It then closes with the way the two directions are timed against each other.

Followings are the topics to be covered in this page.

• What does the UNB uplink radio look like?
• What does the UNB downlink radio look like?
• How are the uplink and the downlink timed against each other?
• Reference

What does the UNB uplink radio look like?

Why would a radio link use a channel only 100 Hz wide? The reason is range. A LEP runs on a small battery, so it cannot reach a distant LAP by raising its transmit power. UNB narrows the channel instead, and the uplink parameters below show how narrow it becomes.

Following is Radio requirement for uplink specified in 5.2.2.1 of LTN 003.

Let's connect the numbers in the table. BPSK carries one bit per symbol, so 100 baud gives an uplink bit rate of 100 bit/s, or 600 bit/s in the USA. The channelization mask of 100 Hz is equal to the baud rate. So the signal occupies about as much spectrum as its symbol rate needs, and no more.

The narrow channel is what makes the sensitivity target possible. Thermal noise is -174 dBm/Hz, so the noise power in a 100 Hz channel is about -154 dBm. A signal at the -135 dBm sensitivity level is still about 19 dB above that thermal noise, before the noise figure of the receiver is counted.

LTN 003 adds one more uplink requirement that the table does not show. Central frequency accuracy is not relevant, provided there is no significant frequency drift within an uplink packet. This matters for a low-cost LEP. At 868 MHz, a crystal error of only 1 ppm is already 868 Hz, which is wider than eight uplink channels. So the LAP must find the signal wherever it appears, and only a drift during the packet breaks the demodulation.

The note under 5.2.2.1 gives the European limits. The UNB uplink band is 868.00 to 868.60 MHz, and the maximum output power is 25 mW, or about 14 dBm. The maximum mean transmission time is 1%. UNB implements neither LBT (Listen Before Talk) nor AFA (Adaptive Frequency Agility), so this 1% limit is what bounds how often a LEP can send.

In time, 1% is 36 seconds of transmission per hour. A frame with the full 12 Byte payload is at least 24 Bytes long before its authentication field is added. That is 192 bits, or about 1.9 seconds at 100 baud. So a European LEP can send fewer than 19 such frames per hour. The frame fields themselves are on the MAC : UNB page.

• UNB uses a narrow channel to gain sensitivity : a 100 Hz channel has a thermal noise floor near -154 dBm, which makes a -135 dBm sensitivity target realistic.
• The uplink bit rate equals the baud rate : BPSK carries one bit per symbol, so the uplink runs at 100 bit/s, or 600 bit/s in the USA.
• Frequency accuracy is relaxed, but frequency drift is not : the LAP accepts a carrier offset from the LEP, but not a drift inside one packet.
• In Europe the duty cycle limits the traffic : 1% mean transmission time in 868.00 to 868.60 MHz allows fewer than 19 full-payload frames per hour.

What does the UNB downlink radio look like?

The downlink does not copy the uplink settings. The LAP is not short of energy, so it can transmit with much more power. The LEP receiver, on the other hand, must stay simple and cheap. The downlink parameters below reflect both facts.

Following is Radio requirement for downlink specified in 5.2.3.1 of LTN 003.

Let's compare the two directions. The uplink uses BPSK at 100 baud in a 100 Hz channel. The downlink uses GFSK at 600 baud, and its channelization mask is a dynamic selection rather than a fixed value. GFSK can be demodulated with a simple non-coherent receiver, which suits a low-cost LEP.

The downlink power is 500 mW, which is about 27 dBm. That is 13 dB more than the 25 mW European uplink limit. The extra power offsets part of the higher baud rate. A 600 baud signal needs about six times the receive bandwidth of a 100 baud signal, so the LEP receiver collects about 8 dB more noise.

The note under 5.2.3.1 places the European downlink in 869.40 to 869.65 MHz. This is the 500 mW band, with a 10% duty cycle. So one LAP can transmit for at most 6 minutes in each hour, and every LEP served by that LAP shares this time. This is one reason UNB treats the downlink as the secondary direction.

• The downlink is faster and stronger than the uplink : GFSK at 600 baud with 500 mW, against BPSK at 100 baud on the uplink.
• The downlink channel is chosen dynamically : the downlink channelization mask is a dynamic selection, not a fixed width.
• In Europe a LAP has 6 minutes of downlink per hour : the 869.40 to 869.65 MHz band allows a 10% duty cycle, shared by all LEPs under that LAP.

How are the uplink and the downlink timed against each other?

A LEP is asleep for most of the time. So the network cannot send a downlink message at any moment it chooses, because the LEP must be listening when the message arrives. Clause 5.2.4 of LTN 003 solves this problem by linking each downlink opportunity to an uplink transmission.

The UNB solution is mainly dedicated to carrying uplink messages. The downlink is still available, and LTN 003 describes two ways to deliver it.

The first way is piggy-backing. A LEP that wants to receive downlink messages opens a fixed reception window after each uplink transmission. The delay and the duration of this window are fixed values. The LTN network sends the downlink message for that LEP inside this window. The network also decides which LAP is the best one to send it.

This design has a direct consequence. A LEP can receive a downlink message only after it sends an uplink message. So the downlink latency depends on how often the LEP transmits. The 1% uplink duty cycle in Europe therefore limits the downlink as well.

The second way is broadcasting. Broadcast is possible in the UNB radio interface, but LTN 003 V1.1.1 states that it is not yet implemented. So piggy-backing is the only downlink method that the specification actually describes.

• The downlink follows the uplink : the LEP opens a reception window with a fixed delay and a fixed duration after each uplink transmission.
• The network picks the LAP : the LTN network chooses the best LAP to send the downlink message in that window.
• Downlink latency depends on uplink traffic : a LEP that sends rarely can be reached only rarely.
• Broadcast is possible but not implemented : LTN 003 V1.1.1 describes it as not yet implemented.

Reference

[1] ETSI GS LTN 003 V1.1.1 (2014-09) - Low Throughput Networks (LTN); Protocols and Interfaces, clause 5.2