Hi,

I posted this on some communication forums but without success.

There is plenty talk about beam forming (e.g. 4G, 5G ...etc) in the context of downlink (base station to handsets). This is understandable.

But surely it does not apply to Uplink. Is there any useful resource how the uplink processes all those handset signals arriving at same time/frequency to all antenna elements. I am aware of some algorithms based on channel estimation such as Zero Forcing to reduce interference but I can't see any useful details as those flooding for the downlink.

Hi Kaz,

Yes, it applies to uplink data too. It’s the reciprocal processing so is covered by the same logic as the downlink.

Regards.

Sorry, but the way I see it is that:

logically there can't be uplink beams because transmission is always towards the same target no matter where the handset is. Just draw the beams from downlink and mirror that to see what I mean. In downlink it is "one to many" so beams apply but in uplink it is "many to one". I expect only the angle of arrival will differ depending on handset location. Even if the handset generates its beam it just generates more gain towards the same target.

I am aware the same terminology is used for uplink but it is more like a convenient misnomer that likely will be corrected sometime in future.

Reciprocity applies, so the basestation beam formed to transmit to the handset will also work for receiving the signal from the handset.

Other systems, like WiFi, can beamform simultaneously from both directions.

Thanks Slartibartfast, 

"reciprocity" makes more sense. I have also heard of "Directivity".

So just like how the downlink radiates a pattern so does this pattern reverses back at uplink.

by the way I got two more related questions:

1) The radiation pattern is controlled by beam weights (I/Q if digital) but I understand the channel state is also needed. Is that because the channel state is needed to modify the radiation pattern that is decided by I/Q by default.

2) Beam weights are said to be generated per PRB. I assume it implies per UE allocation. Does that mean signals like Prach needs weights per preamble (about one or more subframes for long formats) as it comes from same UE.

I am not exactly sure, what 5g,4g and the likes do, but I did a lot of beamforming in audio applications - and from your question I think you might be stuck with confusing terminology. The word beamforming implies there is some sort of beam radiating from a device. It is used in that case, but the term is also used for an array of microphones or antennas RECEIVING signals from a desired source or angle with higher sensitivity than from other directions/angles.

For most beamforming designs, the different angles of incident of the different sources are exploited to discriminate between them. In the near-field, however, you could also use the different amplitudes at different sensors.

Here's my 2 cents which at least provides some useful background if it does not answer your exact question.

Uplink beamforming is done by estimating the uplink channel between device and base station, and using that estimate to calculate beamforming weights.

A generic MIMO eigen-beamformer exploits multipath to achieve spatial diversity gain and typically aims to maximize the SINR at the receiver. This concept applies to non-line of sight (NLOS) or near NLOS channels as strict LOS channels do not, by definition, exhibit multipath. Angle of arrival beamforming is applicable in a LOS channel.

As others have mentioned, when channel reciprocity applies (TDD systems), downlink channel state information can be reported by a device and used by a base station to compute uplink weights.

In 4G and 5G FDD systems a sounding reference signal (SRS) is defined in the Standard for use by a base station to explicitly measure the uplink channel.

4G and 5G systems include multi-user MIMO (MU-MIMO), a special case of beamforming where the same RF resources are used simultaneously by multiple users. 4G systems support MU-MIMO in the downlink (DL) direction only, while 5G supports MU-MIMO on both downlink and uplink (UL). In both systems MU-MIMO applies to the physical shared channel.

One practical limitation of UL MU-MIMO is the small physical size typical of handsets (UE) which limits the physical size of the MIMO antenna array, which in turn limits the spatial granularity that can be achieved. This is not an issue in scenarios such as fixed wireless access where a separate fixed antenna structure can be used.

Also, UE size is less of a limitation in 5G systems operating in the FR2 millimeter wave band whose wavelengths are much smaller. However, these systems are typically limited to LOS channels.

Here's a study on 5G MU-MIMO that you may find useful as it includes both DL and UL benchmarks.

https://images.samsung.com/is/content/samsung/assets/global/business/networks/insights/white-paper/5g-multi-user-mimo-it-isnt-just-for-downlink-anymore/srg_samsung_mu_mimo_study_final.pdf

You can think of uplink beam forming as reciprocal to downlink. The channel estimation occurs between the UE and the basestation and based on this the beam forming weights are computed. For 5G as we expand into mmWave frequencies for increased BW, we need beamforming gain to compensate for the path losses by having highly directional signals from UE to basestation. At the receiver (basestation), we compensate for phase deltas due to different delays of the multi-path signals and hence coherent combining for higher processing gain than if you did not have beam forming.