The concept and definition of Antenna Ports are same as LTE Antenna Ports. It is defined as follows (38.211-4.4.1).

An antenna port is defined such that the channel over which a symbol on the antenna port is conveyed can be inferred from the channel over which another symbol on the same antenna port is conveyed.

What does this mean ? I interpret this as follows :

Each Antenna ports carries its own resource grid and a specific set of reference signal in the grid. The channel properties for RE(resource element) for the reference signal is assumed to be same (or very close to same) as the resource elements for other data(e.g, REs for PDSCH). Due to this facts, we can help demodulate the data by using the channel information obtained by the anaysis of reference channel.

Let me add a little bit more on top of this, because the wording of the specification is unusual and that is what makes it hard to read at the first time.

Look at the definition once more. It never says what an antenna port is. It does not say that it is a connector, a cable, an antenna element or any other piece of hardware. It only says what you are allowed to do with it : if you know the channel of one symbol, then you also know the channel of another symbol sent on the same port. In other words, the standard defines an antenna port by a property, not by an object. Once you notice this, the sentence stops being strange.

The reason for writing it in this way becomes clear as soon as you look at it from the receiver side. Every time a UE receives a symbol, it has to answer one question before it can decode anything : "which channel estimate am I allowed to apply to this symbol ?". The antenna port is exactly the answer to that question. Everything carried on the same antenna port shares one channel, so one channel estimate is enough for all of it. Anything carried on a different port needs a different estimate. The definition is therefore written from the point of view of the receiver, not the transmitter, and this is the key to reading it.

A simple way to picture it is a voice in a large hall. The hall may have one loudspeaker or fifty loudspeakers, and they may be spread all over the ceiling, but if all of them play the same voice, the listener still hears one voice and has to tune his ear only once. If a second, different voice starts, the listener has to tune to it separately. An antenna port is one voice. How many loudspeakers are producing it is a completely different question, and the listener never has to know the answer.

The concept itself is not new, as mentioned above it is exactly the same as in LTE. What has changed in NR is how much weight it carries. In LTE the number of ports was small and it usually matched what you would intuitively call an antenna, so you could work with it for years without thinking too hard about the definition. In NR, with beamforming, massive MIMO and a much richer set of reference signals, the number of antenna ports and the number of physical antennas have almost nothing to do with each other any more. That is why a topic which was almost a footnote in LTE deserves a page of its own in NR.

Followings are topics I am going to look into in this note

• Mapping between Antenna port and Physical channel/Signal
• Practical Implications and Insights
• More to consider

Mapping between Antenna port and Physical channel/Signal

In 5G New Radio (NR), the concept of antenna ports serves as the foundation for transmitting and receiving data, control signals, and reference signals across the air interface. Each antenna port is logically associated with a specific physical channel or signal, enabling efficient communication and robust channel estimation. This mapping between antenna ports and physical channels ensures clear separation of functionalities, allowing the system to handle diverse tasks such as data demodulation, control signaling, and initial access procedures.

Before going into the numbers, there is one thing that has to be clear, because this is where almost every misunderstanding about this topic starts. An antenna port is not a physical antenna. You cannot count antenna ports by counting the connectors on the back of a radio unit. An antenna port is a purely logical thing, and the easiest way to remember it is this : one antenna port = one channel that the receiver is able to estimate on its own.

Once you accept this definition, the relations that look strange at first become natural. One physical antenna can carry several antenna ports at the same time, for example by sending different reference signals on different resource elements. And the opposite is also true : one antenna port can be radiated by a hundred physical antenna elements, which is exactly what happens when a massive MIMO array forms a single beam. The UE does not know how many real antennas are behind the port, and it does not need to know. All it knows is "this is port 1000, and I can estimate one channel for it".

Then why does each channel and signal need its own port number ? Because the receiver has to know which reference signal belongs to which data. If PDSCH and PDCCH shared the same numbering, the UE would be holding a channel estimate without knowing what it may be applied to. So 3GPP gave each channel and signal its own number range, in the same way a company gives each department its own range of phone extensions. Just by looking at the number, you already know what it belongs to.

Two more things are worth keeping in mind before you read the port numbers.

• The numbers are not counting anything. 1000, 2000, 3000 and 4000 are only the starting points of each range. The real ports are 1000, 1001, 1002 and so on, and how many of them actually exist depends on the configuration - for example, on how many DM-RS ports are configured for a PDSCH transmission.
• The same number appears twice with a completely different meaning. Port 1000 is PDSCH in the downlink but SRS in the uplink, and port 4000 is SS/PBCH in the downlink but PRACH in the uplink. A port number is therefore meaningful only together with the direction. When you see a port number in a log, always ask "downlink or uplink ?" before you interpret it.

In NR, a certain range of antenna port number is assigned for each channel and signal as follows.

<38.211 - 6.2, 7.2>

A few notes from 38.211 - 6.2 are :

• When using PUSCH repetition Type B, the PUSCH symbols on the same uplink antenna port can infer each other if the two symbols correspond to the same actual repetition of a PUSCH transmission with repetition Type B.
• If there's no intra-slot frequency hopping and PUSCH repetition Type B isn't applied, symbols on the uplink antenna port can infer each other if the two symbols correspond to the same slot.
• With intra-slot frequency hopping enabled, symbols on the uplink antenna port can infer each other only only if the two symbols correspond to the same frequency hop, regardless of whether the frequency hop distance is zero or not.
• For DM-RS bundling on PUSCH and/or PUCCH repetition and/or multi-slot transport-block processing, the symbols on the uplink antenna port can infer each other if the two symbols are transmitted within the same actual time-domain window.

A few notes from 38.211 - 7.2 are :

• Unless otherwise specified, the UE shouldn't assume that two antenna ports are quasi co-located with respect to any QCL type unless specified
• For DM-RS linked to PDSCH: A PDSCH symbol on an antenna port can infer information from a DM-RS symbol on the same port if both symbols fall within the same scheduled PDSCH resource, are in the same slot, and belong to the same PRG
• For DM-RS linked to PDCCH: A PDCCH symbol on an antenna port can gather details from a DM-RS symbol on the same port only when both symbols are in resources where the UE presumes identical precoding is used
• For DM-RS associated with PBCH: A PBCH symbol on one antenna can deduce data from a DM-RS symbol on the same antenna if both symbols belong to a SS/PBCH block sent within the same slot and share the same block index

The implication of the statement above specifications are :

• PUSCH Symbols on the Same Antenna Port:
• Symbols on the same antenna port can infer each other’s channel properties under certain conditions:
• When no intra-slot frequency hopping is applied, symbols in the same slot share channel properties.
• For Type-B repetition, symbols within the same time-domain repetition window are assumed to have similar channel characteristics.
• DM-RS Usage:
• DM-RS for PDSCH:
• PDSCH symbols can infer channel properties from DM-RS if both symbols are in the same slot and belong to the same PRG (Physical Resource Group).
• DM-RS for PDCCH:
• Precoding assumptions are crucial; only when identical precoding is applied can PDCCH symbols infer channel properties from the DM-RS.
• SS/PBCH Block:
• A PBCH symbol can infer information from the DM-RS of the same SS/PBCH block if they share the same block index and slot.

Practical Implications and Insights

Antenna ports play a crucial role in enabling efficient communication in 5G networks by facilitating key processes like channel estimation, data demodulation, and beam management. Each antenna port transmits unique demodulation reference signals (DM-RS), which help the UE measure channel characteristics such as signal strength and phase, allowing it to correct for issues like fading and noise. Additionally, the concept of quasi co-location (QCL) ensures that relationships between antenna ports are clearly defined when necessary, which is vital for beamforming and efficient use of MIMO systems. Antenna ports also support advanced features like intra-slot frequency hopping and repeated transmissions, ensuring reliable performance even under complex conditions. By structuring these processes, antenna ports ensure seamless and robust communication across a variety of scenarios in 5G networks.

If you want to take away only one sentence from this whole page, take this one, because everything else in this section is derived from it.

If two symbols are sent on the same antenna port, the receiver is allowed to use the channel estimated from one of them to demodulate the other. If they are sent on different antenna ports, it is not allowed to do that, unless the network explicitly tells it that the two ports are quasi co-located.

This sounds abstract, but it is in fact a very practical rule, and it quietly decides a lot of things in the system design. It decides how many reference signals must be transmitted, because every port needs its own. It decides how many independent data streams can be sent, because every MIMO layer needs its own port. And it decides where a channel estimate stops being valid, because a frequency hop or the end of a repetition window breaks the assumption that the channel is still the same.

The same rule also explains a whole family of problems that engineers meet in real testing. When a UE fails to demodulate a channel although the received power looks perfectly good, the cause is very often not the radio condition at all. It is a broken port assumption : a wrong DM-RS port configuration, a missing or wrong TCI state, or a QCL relation that the UE did not have at the moment it needed it. The signal is there, but the UE simply does not know which channel estimate it is allowed to apply to it. This is why a port related mistake usually looks like a coverage problem at first sight, and why looking only at the power measurement will never find it.

With that single rule in mind, the areas where antenna ports matter in practice can be grouped as follows.

• Reference Signal-Based Demodulation:
• Each antenna port carries demodulation reference signals (DM-RS), enabling the UE to infer the channel characteristics. This inference facilitates:
• Channel estimation: Measuring amplitude, phase, and other parameters.
• Data demodulation: Correcting channel effects like fading and phase noise.
• For example, the DM-RS linked to PDSCH allows the UE to demodulate PDSCH data using channel information derived from the DM-RS transmitted on the same port.
• Quasi Co-Location (QCL):
• Unless specified otherwise, the UE does not assume that two antenna ports are quasi co-located (QCL). QCL is important for identifying relationships between ports for:
• Beam management: Beamforming and beam tracking depend on QCL Type-D information.
• Precoding: Efficient MIMO operation requires QCL assumptions between antenna ports.
• Specific rules for QCL are outlined for:
• DM-RS linked to PDSCH, PDCCH, and PBCH.
• Multiple Antenna Configurations:
• MIMO systems use multiple antenna ports for transmitting different data streams. Each port typically corresponds to one logical stream, enabling spatial multiplexing.
• Beamforming relies on precise control of signals at different antenna ports, with QCL information used to manage beam relationships.
• Frequency and Time Domain Behavior:
• Intra-slot frequency hopping: Symbols on an antenna port can only infer channel information from other symbols on the same frequency hop, even when the frequency hop distance is zero.
• PUSCH Repetition Type B: Specific conditions allow channel information sharing across repeated transmissions within a slot or across multiple slots.

More to consider

What has been described so far is the antenna port as the specification defines it. But when you work with a real system - reading a log, writing a test case, or comparing two products - a few more questions keep coming back. Almost all of them come from the same place : the gap between the logical port in the specification and the physical antenna in the hardware.

Massive MIMO is the clearest example of this gap. A 64T64R radio unit has 64 physical transmit chains, but the UE may only ever see 8 or 16 CSI-RS ports. The remaining dimensions are hidden inside the beamforming that the gNB does internally, and they never appear in any message toward the UE. So the question "how many antenna ports does this cell have ?" has two different answers depending on whether you ask the hardware team or the UE, and both answers are correct. Whenever a port count does not match your expectation, this is usually the first place to look.

The other questions come from the physical side leaking into the logical one. The polarization of the antenna never appears in the port definition, yet it is one of the main reasons why two ports can carry two independent streams from the same position in space. And the quality of the channel estimate on a port - which is what finally decides whether a high order MCS can be used - depends on the noise and interference sitting on the reference signals of that particular port, not on the total received power of the cell. The points below are the ones I find worth keeping in mind.

When analyzing antenna ports in 5G networks, there are some additional  aspects to consider that significantly impact performance and system design. Followings are some of those aspects worth notice

• Channel State Information (CSI) Estimation:
• CSI-RS, mapped to ports starting at 3000, is critical for advanced beamforming and link adaptation. UEs use CSI-RS to estimate channel quality, enabling:
• Link adaptation: Adjusting modulation and coding schemes.
• Beam management: Supporting massive MIMO systems in 5G.
• Massive MIMO and Port Scaling:
• Massive MIMO systems, with hundreds of antenna elements, require efficient mapping of ports to physical antennas. Logical antenna ports often represent groups of physical antennas.
• PRACH-Specific Port Usage:
• Port 4000 is uniquely assigned to PRACH, ensuring separation from data and control channels. PRACH transmissions use this port for initial access and random access procedures.
• Impact of Noise and Interference:
• Noise and interference on antenna ports impact channel estimation accuracy. Advanced techniques, such as minimum mean square error (MMSE) filtering, improve channel estimation.
• Cross-Polarized Antenna Ports:
• In dual-polarized systems, two antenna ports may transmit on orthogonal polarizations. This configuration enhances diversity and capacity.

Reference

• 5G NR Antenna Ports – Logical and Physical Antenna Ports - LinkedIn