Reference Signal Design for ISAC

The design of reference signals (RS) has many caveats attached to it at this moment in time. One philosophy for the design of RS is to have a common RS for both communication and sensing, which can be used for channel estimation for both communication and sensing. This philosophy has one big advantage and one big disadvantage. And the advantage and disadvantage are so dividing that it’s almost impossible to digest both of them from a purist and practitioner point of view.

I will start with the disadvantage because it is more academic and less interesting. The disadvantage is that the design of a common RS for both communication and sensing is very difficult, because the requirements for communication and sensing are very different. For communication, the RS should be designed to acquire channel estimates that are good enough to provide the best communication performance within the limits of the designed codebooks, with the minimum CSI-RS overhead possible. If improvement in the quality of channel estimates does not lead to improvement in communication performance for the given codebook-based feedback schemes, then the overhead induced by this additional improvement in the quality of channel estimates is not justified. Typically, the CSI is acquired at the subband level in the frequency domain, which is at least 4 RB wide and at most 32 RB wide. On the other hand, for sensing, the CSI across time is typically acquired at the TDD configuration level, which is a minimum of 4 slots. The CSI-RS can be as wide as 4 symbols in time for 32-port sounding. If looked at analytically, the CSI-RS overhead in 5G was roughly 1% to 3%. But the CSI-RS is far from optimal for sensing due to both its time and frequency granularity. For sensing, the RS should be designed to have good range and velocity resolution, which means that it should have good autocorrelation properties and high PAPR. Therefore, it is very difficult to design a common RS that can satisfy both requirements.

On the other hand, the advantage is very interesting. If we agree to live with a common sub-optimal RS for communication and sensing, then the integration of the RS into the air interface is much easier, and if something fits organically into the air interface, the adoption of that feature is much faster. It makes it difficult for the vendors to sell it as an add-on feature, but it makes it much easier for the industry to adopt it. Hence, ISAC can become a mainstream feature in the air interface and not just a niche feature for specific use cases. This way ISAC will not suffer the same fate as positioning where a new reference signal was design to deliver accurate positioning information, but PRS was not very well intergrated into the TDD confirmation and air interface, hence was not widely adopted by the industry.

The other challenge with a common RS design is the logistics. The RS design for Sensing is being looked under the ISAC study item (10.8.2) where as the same for communication is being looked at under the CSI-RS enhancement study item (10.4). As per my understanding once the RS for MIMO is finalized, the ISAC RS design will be send as a recommendation to the CSI-RS enhancement study item, and the CSI-RS enhancement study item will decide how to adopt it. Furthermore, very nuanced changes tailer the RS for accurate sensing will be made in upcoming 3GPP releases.